Original Research

Prostate cancer will be diagnosed in 12.5% of men during their lifetime. It is the most commonly diagnosed solid organ cancer in men.¹ However, prostate cancer screening for prostate-specific antigen (PSA) remains controversial due to concerns about overdiagnosis, as the overall risk of dying of prostate cancer is only 2.4%.¹

To address the risk and benefits of PSA testing, in 2012 the US Preventive Services Task Force (USPSTF) recommended against routine PSA testing.² Updated 2018 recommendations continued this recommendation in men aged > 70 years but acknowledged a small potential benefit in men aged 55 to 69 years and suggested individualized shared decision making between patient and clinician.³ In addition, American Urological Association (AUA) guidelines for the early detection of prostate cancer recommend against PSA screening in men aged < 40 years or those aged > 70 years, shared decision making for individuals aged 55 to 70 years or in high-risk men aged 40 to 55 years (ie, family history of prostate cancer or African American race).⁴ PSA screening is not recommended for men with a life expectancy shorter than 10 to 15 years aged > 70 years.⁴

The Veterans Health Administration (VHA) is the largest integrated health care system in the US.⁵ In addition, the US Department of Veterans Affairs (VA) Spinal Cord Injury and Disorders System of Care operates 25 centers throughout the US.⁶ Life expectancy following spinal cord injury (SCI) increased significantly through the 1980s but has since plateaued, with life expectancy being impacted by age at injury, completeness of injury, and neurologic level.^7,8 As part of a program of uniform care, all persons with SCI followed at the Spinal Cord Injury and Disorders System of Care centers are offered comprehensive annual evaluations, including screening laboratory tests, such as PSA level.⁹

Patients with SCI present a unique challenge when interpreting PSA levels, given potentially confounding factors, including neurogenic bladder management, high rates of bacteriuria, urinary tract infections (UTIs), testosterone deficiency, and pelvic innervation that differs from the noninjured population.^10,11 Unfortunately, the literature on prostate cancer prevalence and average PSA levels in patients with SCI is limited by the small scope of studies and inconsistent data.^10-16 Therefore, the purpose of the current investigation was to quantify and analyze the rates of annual PSA testing for all men with SCI in the VHA.

 

Methods

Approval was granted by the Richmond VA Medical Center (VAMC) Institutional Review Board in Virginia, and by the VA Informatics and Computing Infrastructure (VINCI) data access request tracker system for extraction of data from the VA Corporate Data Warehouse. Microsoft Structured Query Language was used for data programming and query design. Statistical analysis was conducted using Stata version 15.1 with assistance from professional biostatisticians.

Only male veterans with a nervous system disorder affecting the spinal cord or with myelopathy were included, based on International Classification of Diseases (ICD) version 9 and 10 codes, corresponding to traumatic and nontraumatic myelopathy. Veterans diagnosed with myelopathy based on ICD codes corresponding to progressive or degenerative myelopathies, such as multiple sclerosis or amyotrophic lateral sclerosis, were excluded.

For each veteran, extracted data included the unique identification number, date of birth, ICD code, date ICD code first appeared, race, gender, death status (yes/no), date of death (when applicable), date of each PSA test, PSA test values, and the VAMC where each test was performed. Only tests for total PSA were included. The date that the ICD code first appeared served as an approximation for the date of SCI. The time frame for the study included all PSA tests in the VINCI database for 2000 through 2017. However, only post-SCI PSA tests were included in the analysis. Duplicate tests (same date/time) were eliminated.

Race is considered a risk factor for prostate cancer only for African American patients, likely due to racial health disparities.¹⁷ Given this, we chose to categorize race as either African American or other, with a third category for missing/inconsistent reporting. Age at time of the PSA test was categorized into 4 groups (≤ 39, 40-54, 55-69, and ≥ 70 years) based on AUA guidelines.⁴ The annual PSA testing rate was calculated for each veteran with SCI as the number of PSA tests per year. A mean annual PSA test rate was then calculated as the weighted (by exposure time) mean value for all annual PSA testing rates from 2000 through 2017 for each age group and race. Annual exposure was calculated for each veteran and defined as the number of days a veteran was eligible to have a PSA test. This started with the date of SCI diagnosis and ended with either the date of death or the date of last PSA. If a veteran moved from one age group to another in 1 year, the first part of this year’s exposure was included in the calculation of the annual PSA testing rate for the younger group and the second part was included for the calculation of the older group. For deceased veterans, the death date was excluded from the exposure period, and their exposure period ended on the day before death.

Statistical Analysis

To compare PSA testing rates between African American race and other races, Poisson regression was used with exposure treated as an offset (exposures were summed across years for each veteran). An indicator (dummy) variable for African American race vs other races was coded, and statistical significance was set at P < .05. To check sensitivity for the Poisson assumption that the mean was equal to the variance, negative binomial regression was used. To assess for geographic PSA testing rate variability, the data were further analyzed based on the locations where PSA tests were performed. This subanalysis was limited to veterans who had all PSA tests in a single station. For each station, the average PSA testing rate was calculated for each veteran, and the mean for all annual PSA testing rates was used to determine station-specific PSA testing rates.

 

Results

A total of 45,274 veterans were initially identified of which 367 females were excluded (Figure 1).

Moreover, 1688 male veterans were excluded for ICD codes that were less relevant, yielding 43,219 male veterans with relevant ICD codes. From this group, an additional 5976 were excluded because no PSA test was found after the SCI date. The racial makeup of the remaining 37,243 male veterans included 6327 African American patients, 25,277 of other races, and 5639 with missing/inconsistent race data. The included sample received care in ≥ 1 of 129 VAMCs. The final cohort yielded 261,125 PSA tests.

The Table shows PSA tests categorized by age group and race.

The PSA testing rate rose for veterans in the age groups ≤ 39, 40 to 54, and 55 to 69 years (Figure 2A).

The PSA testing rate dropped for the oldest age group (≥ 70 years), for the entire population, and the other race and missing/inconsistent race groups; however, PSA testing rates continued to rise in the African American group aged ≥ 70 years. For the entire population, average PSA testing rates in tests per year for the age groups were 0.46 (aged ≤ 39 years), 0.78 (aged 40-54 years), 1.0 (aged 55-69 years), and 0.91 (aged ≥ 70 years). However, PSA testing rates were significantly higher for the African American vs other races group at all ages (0.47 vs 0.46 tests per year, respectively, aged ≤ 39 years; 0.83 vs 0.77 tests per year, respectively, aged 40-54 years; 1.04 vs 1.00 tests per year, respectively, aged 55-69 years; and 1.08 vs 0.90 tests per year respectively, aged ≥ 70 years; P < .001) (Figure 2B).

Of the cohort of 37,243 veterans, 28,396 (76.2%) had their post-SCI tests done at a single facility, 6770 (18.1%) at 2 locations, and 2077 (5.5%) at > 2 locations. Single-station group data were included in a subanalysis to determine the mean (SD) PSA testing rates, which for the 123 locations was 0.98 (0.36) tests per veteran per year (range, 0.2-3.0 tests per veteran per year).

Figure 3 shows a heat map of the US: each dot represents a specific VAMC and shows PSA testing rate variability between stations.

To assess the impact of the 2012 USPSTF recommendations on PSA testing rates in veterans with SCI, mean PSA testing rates were calculated for 5 years before the recommendations (2007-2011) and compared with the average PSA testing rate for 5 years following the updated recommendations (2013-2017). The USPSTF updated its recommendation again in 2018 and acknowledged the potential benefit for PSA screening in certain patient populations.^2,3 Surprisingly, and despite recommendations, the results show a significant increase in PSA testing rates in all age groups for all races (P < .001) (Figure 4).

For the entire population, the average PSA testing rates for 2007 to 2011 in tests per year were 0.39, 0.76, 1.03, and 0.89 for the ≤ 39 years, 40 to 54 years, 55 to 69 years, and ≥ 70 years age groups, respectively. Likewise, the average PSA testing rates for years 2013 to 2017 in tests per year were 0.75, 0.96, 1.13, and 0.98 for the ≤ 39 years, 40 to 54 years, 55 to 69 years, and ≥ 70 years age groups, respectively, with an increased rate of testing of 0.92, 0.26, 0.10, and 0.11, respectively, from years 2007-2011 to 2013-2017 (P < .001).

 

Discussion

The goal of this study was to establish testing rates and analyze PSA testing trends across races and age groups in veterans with SCI. This is the largest cohort of patients with SCI analyzed in the literature. The key findings of this study were that despite clear AUA guidelines recommending against PSA testing in patients aged ≤ 39 years and ≥ 70 years, there are high rates of testing in veterans with SCI in these age groups (0.46 tests per year in those aged ≤ 39 years and 0.91 tests per year in those aged ≥ 70 years). In terms of race, as expected based on increased risk, African American veterans with SCI had higher PSA test rates.¹⁸ However, the continued increase in PSA testing rate for African American veterans aged ≥ 70 years was unexpected and not seen in other racial groups. As racial disparities are known to affect prostate cancer outcomes in African American men, it is reassuring that PSA testing was actually higher among African American men with SCI in our population, suggesting this vulnerable population is not being left behind in terms of screening.¹⁷ In contrast to other studies that show a lower rate of PSA screening in patients with SCI, our study suggests general PSA overtesting in veterans with SCI and a need for improved education for both veterans and their health care practitioners.¹⁹

Prostate Cancer Incidence

Although the exact mechanism behind alterations in prostate function in the SCI population have yet to be fully elucidated, research suggests that the prostate behaves differently after SCI. Animal models of prostate gland denervation show decreased prostate volume and suggest that SCI may lead to a reduction in prostatic secretory function associated with autonomic dysfunction. Shim and colleagues hypothesized that impaired autonomic prostate innervation alters the prostatic volume and PSA in patients with SCI.¹⁰

Additional studies looking at actual PSA levels in men with SCI reveal conflicting data.^10-15,20 Toricelli and colleagues retrospectively studied 140 men with SCI, of whom 34 had PSA levels available and found that mean PSA was not significantly different for patients with SCI compared with controls, but patients using clean intermittent catheterization had 2-fold higher PSA levels.²¹ In contrast, Konety and colleagues found that mean PSA was not significantly different from uninjured controls in their cohort of 79 patients with SCI, though they did find a correlation between indwelling catheter use and a higher PSA.²²

Studies have shown an overall decreased risk of prostate cancer in patients with SCI, though the mechanism remains unclear. A large cohort study from Taiwan showed a lower risk of prostate cancer for 54,401 patients with SCI with an adjusted hazard ratio of 0.73.²³ Patel and colleagues found the overall rate of prostate cancer in the population of veterans with SCI was lower than the general uninjured VA population, though this study was limited by scope with only 350 patients with SCI.²⁴ A more recent systematic review and meta-analysis of 9 studies evaluating the prevalence of prostate cancer in men with SCI found a reduction of up to 65% in the risk of prostate cancer in men with SCI, and PSA was found to be a poor screening tool for prostate cancer due to large study heterogeneity.¹⁶

 

PSA Screening

This study identified widespread overscreening using the PSA test in veterans with SCI, which is likely attributable to many factors. Per VHA Directive 1176, all eligible veterans are offered yearly interdisciplinary comprehensive evaluations, including laboratory testing, and as such veterans with SCI have high rates of annual visit attendance due to the complexity of their care.⁹ PSA testing is included in the standard battery of laboratory tests ordered for all patients with SCI during their annual examinations. Additionally, many SCI specialists use the PSA level in patients with SCI for identifying cystitis or prostatitis in patients with colonization who may not experience typical symptoms. Everaert and colleagues demonstrated the clinical utility for localizing UTIs to the upper or lower tract, with elevated PSA indicating prostatitis. They found that serum PSA has a sensitivity of 68% and a specificity of 100% in the differential diagnosis of prostatitis and pyelonephritis.²⁵ As such, the high PSA screening rates may be reflective of diagnostic use for infection rather than for cancer screening.

Likely as a response to the USPSTF recommendations, there has been a national slow decline in overall PSA screening rates since 2012.^26-28 A study from Vetterlein and colleagues examining changes in the PSA screening trends related to USPSTF recommendations found an 8.5% decline in overall PSA screening from 2012 to 2014.²⁹ However, the increase in PSA testing across all ages and races in the VA population with SCI over the same period is not entirely understood and suggests the need for further research and education in this area. Additionally, as factors associated with SCI impact the life expectancy of these patients, further shared decision making is needed in deciding whether to pursue PSA screening in this population to minimize unnecessary screening in patients with a life expectancy of < 10 to 15 years.

Limitations

This study is limited by the use of data identified by ICD codes rather than by review of individual health records. This required the use of decision algorithms for data points, such as the date of SCI. In addition, analysis was not able to capture shared decision making that may have contributed to PSA screening outside the recommended age ranges based on additional risk factors, such as family history of lethal malignancy. Furthermore, a detailed attempt to define specific age-adjusted PSA levels was beyond the scope of this study but will be addressed in later publications. In addition, we did not exclude individuals with a diagnosis of prostate adenocarcinoma, prostatitis, or recurrent UTIs because the onset, duration, and severity of disease could not be definitively ascertained. Finally, veterans with SCI are unique and may not be reflective of individuals with SCI who do not receive care within the VA. However, despite these limitations, this is, to our knowledge, the largest and most comprehensive study evaluating PSA testing rates in individuals with SCI.

Conclusions

Currently, PSA screening is recommended following shared decision making for patients at average risk aged 55 to 70 years. Patients with SCI experience many conditions that may affect PSA values, but data regarding normal PSA ranges and rates of prostate cancer in this population remain sparse. The study demonstrated high rates of overtesting in veterans with SCI, higher than expected testing rates in African American veterans, a paradoxical increase in PSA testing rates after the 2012 publication of the USPSTF PSA guidelines, and wide variability in testing rates depending on VA location.

African American men were tested at higher rates across all age groups, including in patients aged > 70 years. To balance the benefits of detecting clinically significant prostate cancer vs the risks of invasive testing in high-risk populations with SCI, more work is needed to determine the clinical impact of screening practices. Future work is currently ongoing to define age-based PSA values in patients with SCI.

Acknowledgments

This research was supported in part through funding from the Center for Rehabilitation Science and Engineering, Virginia Commonwealth University Health System.

Prostate cancer will be diagnosed in 12.5% of men during their lifetime. It is the most commonly diagnosed solid organ cancer in men.¹ However, prostate cancer screening for prostate-specific antigen (PSA) remains controversial due to concerns about overdiagnosis, as the overall risk of dying of prostate cancer is only 2.4%.¹

To address the risk and benefits of PSA testing, in 2012 the US Preventive Services Task Force (USPSTF) recommended against routine PSA testing.² Updated 2018 recommendations continued this recommendation in men aged > 70 years but acknowledged a small potential benefit in men aged 55 to 69 years and suggested individualized shared decision making between patient and clinician.³ In addition, American Urological Association (AUA) guidelines for the early detection of prostate cancer recommend against PSA screening in men aged < 40 years or those aged > 70 years, shared decision making for individuals aged 55 to 70 years or in high-risk men aged 40 to 55 years (ie, family history of prostate cancer or African American race).⁴ PSA screening is not recommended for men with a life expectancy shorter than 10 to 15 years aged > 70 years.⁴

The Veterans Health Administration (VHA) is the largest integrated health care system in the US.⁵ In addition, the US Department of Veterans Affairs (VA) Spinal Cord Injury and Disorders System of Care operates 25 centers throughout the US.⁶ Life expectancy following spinal cord injury (SCI) increased significantly through the 1980s but has since plateaued, with life expectancy being impacted by age at injury, completeness of injury, and neurologic level.^7,8 As part of a program of uniform care, all persons with SCI followed at the Spinal Cord Injury and Disorders System of Care centers are offered comprehensive annual evaluations, including screening laboratory tests, such as PSA level.⁹

Patients with SCI present a unique challenge when interpreting PSA levels, given potentially confounding factors, including neurogenic bladder management, high rates of bacteriuria, urinary tract infections (UTIs), testosterone deficiency, and pelvic innervation that differs from the noninjured population.^10,11 Unfortunately, the literature on prostate cancer prevalence and average PSA levels in patients with SCI is limited by the small scope of studies and inconsistent data.^10-16 Therefore, the purpose of the current investigation was to quantify and analyze the rates of annual PSA testing for all men with SCI in the VHA.

 

Methods

Approval was granted by the Richmond VA Medical Center (VAMC) Institutional Review Board in Virginia, and by the VA Informatics and Computing Infrastructure (VINCI) data access request tracker system for extraction of data from the VA Corporate Data Warehouse. Microsoft Structured Query Language was used for data programming and query design. Statistical analysis was conducted using Stata version 15.1 with assistance from professional biostatisticians.

Only male veterans with a nervous system disorder affecting the spinal cord or with myelopathy were included, based on International Classification of Diseases (ICD) version 9 and 10 codes, corresponding to traumatic and nontraumatic myelopathy. Veterans diagnosed with myelopathy based on ICD codes corresponding to progressive or degenerative myelopathies, such as multiple sclerosis or amyotrophic lateral sclerosis, were excluded.

For each veteran, extracted data included the unique identification number, date of birth, ICD code, date ICD code first appeared, race, gender, death status (yes/no), date of death (when applicable), date of each PSA test, PSA test values, and the VAMC where each test was performed. Only tests for total PSA were included. The date that the ICD code first appeared served as an approximation for the date of SCI. The time frame for the study included all PSA tests in the VINCI database for 2000 through 2017. However, only post-SCI PSA tests were included in the analysis. Duplicate tests (same date/time) were eliminated.

Race is considered a risk factor for prostate cancer only for African American patients, likely due to racial health disparities.¹⁷ Given this, we chose to categorize race as either African American or other, with a third category for missing/inconsistent reporting. Age at time of the PSA test was categorized into 4 groups (≤ 39, 40-54, 55-69, and ≥ 70 years) based on AUA guidelines.⁴ The annual PSA testing rate was calculated for each veteran with SCI as the number of PSA tests per year. A mean annual PSA test rate was then calculated as the weighted (by exposure time) mean value for all annual PSA testing rates from 2000 through 2017 for each age group and race. Annual exposure was calculated for each veteran and defined as the number of days a veteran was eligible to have a PSA test. This started with the date of SCI diagnosis and ended with either the date of death or the date of last PSA. If a veteran moved from one age group to another in 1 year, the first part of this year’s exposure was included in the calculation of the annual PSA testing rate for the younger group and the second part was included for the calculation of the older group. For deceased veterans, the death date was excluded from the exposure period, and their exposure period ended on the day before death.

Statistical Analysis

To compare PSA testing rates between African American race and other races, Poisson regression was used with exposure treated as an offset (exposures were summed across years for each veteran). An indicator (dummy) variable for African American race vs other races was coded, and statistical significance was set at P < .05. To check sensitivity for the Poisson assumption that the mean was equal to the variance, negative binomial regression was used. To assess for geographic PSA testing rate variability, the data were further analyzed based on the locations where PSA tests were performed. This subanalysis was limited to veterans who had all PSA tests in a single station. For each station, the average PSA testing rate was calculated for each veteran, and the mean for all annual PSA testing rates was used to determine station-specific PSA testing rates.

 

Results

A total of 45,274 veterans were initially identified of which 367 females were excluded (Figure 1).

Moreover, 1688 male veterans were excluded for ICD codes that were less relevant, yielding 43,219 male veterans with relevant ICD codes. From this group, an additional 5976 were excluded because no PSA test was found after the SCI date. The racial makeup of the remaining 37,243 male veterans included 6327 African American patients, 25,277 of other races, and 5639 with missing/inconsistent race data. The included sample received care in ≥ 1 of 129 VAMCs. The final cohort yielded 261,125 PSA tests.

The Table shows PSA tests categorized by age group and race.

The PSA testing rate rose for veterans in the age groups ≤ 39, 40 to 54, and 55 to 69 years (Figure 2A).

The PSA testing rate dropped for the oldest age group (≥ 70 years), for the entire population, and the other race and missing/inconsistent race groups; however, PSA testing rates continued to rise in the African American group aged ≥ 70 years. For the entire population, average PSA testing rates in tests per year for the age groups were 0.46 (aged ≤ 39 years), 0.78 (aged 40-54 years), 1.0 (aged 55-69 years), and 0.91 (aged ≥ 70 years). However, PSA testing rates were significantly higher for the African American vs other races group at all ages (0.47 vs 0.46 tests per year, respectively, aged ≤ 39 years; 0.83 vs 0.77 tests per year, respectively, aged 40-54 years; 1.04 vs 1.00 tests per year, respectively, aged 55-69 years; and 1.08 vs 0.90 tests per year respectively, aged ≥ 70 years; P < .001) (Figure 2B).

Of the cohort of 37,243 veterans, 28,396 (76.2%) had their post-SCI tests done at a single facility, 6770 (18.1%) at 2 locations, and 2077 (5.5%) at > 2 locations. Single-station group data were included in a subanalysis to determine the mean (SD) PSA testing rates, which for the 123 locations was 0.98 (0.36) tests per veteran per year (range, 0.2-3.0 tests per veteran per year).

Figure 3 shows a heat map of the US: each dot represents a specific VAMC and shows PSA testing rate variability between stations.

To assess the impact of the 2012 USPSTF recommendations on PSA testing rates in veterans with SCI, mean PSA testing rates were calculated for 5 years before the recommendations (2007-2011) and compared with the average PSA testing rate for 5 years following the updated recommendations (2013-2017). The USPSTF updated its recommendation again in 2018 and acknowledged the potential benefit for PSA screening in certain patient populations.^2,3 Surprisingly, and despite recommendations, the results show a significant increase in PSA testing rates in all age groups for all races (P < .001) (Figure 4).

For the entire population, the average PSA testing rates for 2007 to 2011 in tests per year were 0.39, 0.76, 1.03, and 0.89 for the ≤ 39 years, 40 to 54 years, 55 to 69 years, and ≥ 70 years age groups, respectively. Likewise, the average PSA testing rates for years 2013 to 2017 in tests per year were 0.75, 0.96, 1.13, and 0.98 for the ≤ 39 years, 40 to 54 years, 55 to 69 years, and ≥ 70 years age groups, respectively, with an increased rate of testing of 0.92, 0.26, 0.10, and 0.11, respectively, from years 2007-2011 to 2013-2017 (P < .001).

 

Discussion

The goal of this study was to establish testing rates and analyze PSA testing trends across races and age groups in veterans with SCI. This is the largest cohort of patients with SCI analyzed in the literature. The key findings of this study were that despite clear AUA guidelines recommending against PSA testing in patients aged ≤ 39 years and ≥ 70 years, there are high rates of testing in veterans with SCI in these age groups (0.46 tests per year in those aged ≤ 39 years and 0.91 tests per year in those aged ≥ 70 years). In terms of race, as expected based on increased risk, African American veterans with SCI had higher PSA test rates.¹⁸ However, the continued increase in PSA testing rate for African American veterans aged ≥ 70 years was unexpected and not seen in other racial groups. As racial disparities are known to affect prostate cancer outcomes in African American men, it is reassuring that PSA testing was actually higher among African American men with SCI in our population, suggesting this vulnerable population is not being left behind in terms of screening.¹⁷ In contrast to other studies that show a lower rate of PSA screening in patients with SCI, our study suggests general PSA overtesting in veterans with SCI and a need for improved education for both veterans and their health care practitioners.¹⁹

Prostate Cancer Incidence

Although the exact mechanism behind alterations in prostate function in the SCI population have yet to be fully elucidated, research suggests that the prostate behaves differently after SCI. Animal models of prostate gland denervation show decreased prostate volume and suggest that SCI may lead to a reduction in prostatic secretory function associated with autonomic dysfunction. Shim and colleagues hypothesized that impaired autonomic prostate innervation alters the prostatic volume and PSA in patients with SCI.¹⁰

Additional studies looking at actual PSA levels in men with SCI reveal conflicting data.^10-15,20 Toricelli and colleagues retrospectively studied 140 men with SCI, of whom 34 had PSA levels available and found that mean PSA was not significantly different for patients with SCI compared with controls, but patients using clean intermittent catheterization had 2-fold higher PSA levels.²¹ In contrast, Konety and colleagues found that mean PSA was not significantly different from uninjured controls in their cohort of 79 patients with SCI, though they did find a correlation between indwelling catheter use and a higher PSA.²²

Studies have shown an overall decreased risk of prostate cancer in patients with SCI, though the mechanism remains unclear. A large cohort study from Taiwan showed a lower risk of prostate cancer for 54,401 patients with SCI with an adjusted hazard ratio of 0.73.²³ Patel and colleagues found the overall rate of prostate cancer in the population of veterans with SCI was lower than the general uninjured VA population, though this study was limited by scope with only 350 patients with SCI.²⁴ A more recent systematic review and meta-analysis of 9 studies evaluating the prevalence of prostate cancer in men with SCI found a reduction of up to 65% in the risk of prostate cancer in men with SCI, and PSA was found to be a poor screening tool for prostate cancer due to large study heterogeneity.¹⁶

 

PSA Screening

This study identified widespread overscreening using the PSA test in veterans with SCI, which is likely attributable to many factors. Per VHA Directive 1176, all eligible veterans are offered yearly interdisciplinary comprehensive evaluations, including laboratory testing, and as such veterans with SCI have high rates of annual visit attendance due to the complexity of their care.⁹ PSA testing is included in the standard battery of laboratory tests ordered for all patients with SCI during their annual examinations. Additionally, many SCI specialists use the PSA level in patients with SCI for identifying cystitis or prostatitis in patients with colonization who may not experience typical symptoms. Everaert and colleagues demonstrated the clinical utility for localizing UTIs to the upper or lower tract, with elevated PSA indicating prostatitis. They found that serum PSA has a sensitivity of 68% and a specificity of 100% in the differential diagnosis of prostatitis and pyelonephritis.²⁵ As such, the high PSA screening rates may be reflective of diagnostic use for infection rather than for cancer screening.

Likely as a response to the USPSTF recommendations, there has been a national slow decline in overall PSA screening rates since 2012.^26-28 A study from Vetterlein and colleagues examining changes in the PSA screening trends related to USPSTF recommendations found an 8.5% decline in overall PSA screening from 2012 to 2014.²⁹ However, the increase in PSA testing across all ages and races in the VA population with SCI over the same period is not entirely understood and suggests the need for further research and education in this area. Additionally, as factors associated with SCI impact the life expectancy of these patients, further shared decision making is needed in deciding whether to pursue PSA screening in this population to minimize unnecessary screening in patients with a life expectancy of < 10 to 15 years.

Limitations

This study is limited by the use of data identified by ICD codes rather than by review of individual health records. This required the use of decision algorithms for data points, such as the date of SCI. In addition, analysis was not able to capture shared decision making that may have contributed to PSA screening outside the recommended age ranges based on additional risk factors, such as family history of lethal malignancy. Furthermore, a detailed attempt to define specific age-adjusted PSA levels was beyond the scope of this study but will be addressed in later publications. In addition, we did not exclude individuals with a diagnosis of prostate adenocarcinoma, prostatitis, or recurrent UTIs because the onset, duration, and severity of disease could not be definitively ascertained. Finally, veterans with SCI are unique and may not be reflective of individuals with SCI who do not receive care within the VA. However, despite these limitations, this is, to our knowledge, the largest and most comprehensive study evaluating PSA testing rates in individuals with SCI.

Conclusions

Currently, PSA screening is recommended following shared decision making for patients at average risk aged 55 to 70 years. Patients with SCI experience many conditions that may affect PSA values, but data regarding normal PSA ranges and rates of prostate cancer in this population remain sparse. The study demonstrated high rates of overtesting in veterans with SCI, higher than expected testing rates in African American veterans, a paradoxical increase in PSA testing rates after the 2012 publication of the USPSTF PSA guidelines, and wide variability in testing rates depending on VA location.

African American men were tested at higher rates across all age groups, including in patients aged > 70 years. To balance the benefits of detecting clinically significant prostate cancer vs the risks of invasive testing in high-risk populations with SCI, more work is needed to determine the clinical impact of screening practices. Future work is currently ongoing to define age-based PSA values in patients with SCI.

Acknowledgments

This research was supported in part through funding from the Center for Rehabilitation Science and Engineering, Virginia Commonwealth University Health System.

Prostate cancer will be diagnosed in 12.5% of men during their lifetime. It is the most commonly diagnosed solid organ cancer in men.¹ However, prostate cancer screening for prostate-specific antigen (PSA) remains controversial due to concerns about overdiagnosis, as the overall risk of dying of prostate cancer is only 2.4%.¹

To address the risk and benefits of PSA testing, in 2012 the US Preventive Services Task Force (USPSTF) recommended against routine PSA testing.² Updated 2018 recommendations continued this recommendation in men aged > 70 years but acknowledged a small potential benefit in men aged 55 to 69 years and suggested individualized shared decision making between patient and clinician.³ In addition, American Urological Association (AUA) guidelines for the early detection of prostate cancer recommend against PSA screening in men aged < 40 years or those aged > 70 years, shared decision making for individuals aged 55 to 70 years or in high-risk men aged 40 to 55 years (ie, family history of prostate cancer or African American race).⁴ PSA screening is not recommended for men with a life expectancy shorter than 10 to 15 years aged > 70 years.⁴

The Veterans Health Administration (VHA) is the largest integrated health care system in the US.⁵ In addition, the US Department of Veterans Affairs (VA) Spinal Cord Injury and Disorders System of Care operates 25 centers throughout the US.⁶ Life expectancy following spinal cord injury (SCI) increased significantly through the 1980s but has since plateaued, with life expectancy being impacted by age at injury, completeness of injury, and neurologic level.^7,8 As part of a program of uniform care, all persons with SCI followed at the Spinal Cord Injury and Disorders System of Care centers are offered comprehensive annual evaluations, including screening laboratory tests, such as PSA level.⁹

Patients with SCI present a unique challenge when interpreting PSA levels, given potentially confounding factors, including neurogenic bladder management, high rates of bacteriuria, urinary tract infections (UTIs), testosterone deficiency, and pelvic innervation that differs from the noninjured population.^10,11 Unfortunately, the literature on prostate cancer prevalence and average PSA levels in patients with SCI is limited by the small scope of studies and inconsistent data.^10-16 Therefore, the purpose of the current investigation was to quantify and analyze the rates of annual PSA testing for all men with SCI in the VHA.

 

Methods

Approval was granted by the Richmond VA Medical Center (VAMC) Institutional Review Board in Virginia, and by the VA Informatics and Computing Infrastructure (VINCI) data access request tracker system for extraction of data from the VA Corporate Data Warehouse. Microsoft Structured Query Language was used for data programming and query design. Statistical analysis was conducted using Stata version 15.1 with assistance from professional biostatisticians.

Only male veterans with a nervous system disorder affecting the spinal cord or with myelopathy were included, based on International Classification of Diseases (ICD) version 9 and 10 codes, corresponding to traumatic and nontraumatic myelopathy. Veterans diagnosed with myelopathy based on ICD codes corresponding to progressive or degenerative myelopathies, such as multiple sclerosis or amyotrophic lateral sclerosis, were excluded.

For each veteran, extracted data included the unique identification number, date of birth, ICD code, date ICD code first appeared, race, gender, death status (yes/no), date of death (when applicable), date of each PSA test, PSA test values, and the VAMC where each test was performed. Only tests for total PSA were included. The date that the ICD code first appeared served as an approximation for the date of SCI. The time frame for the study included all PSA tests in the VINCI database for 2000 through 2017. However, only post-SCI PSA tests were included in the analysis. Duplicate tests (same date/time) were eliminated.

Race is considered a risk factor for prostate cancer only for African American patients, likely due to racial health disparities.¹⁷ Given this, we chose to categorize race as either African American or other, with a third category for missing/inconsistent reporting. Age at time of the PSA test was categorized into 4 groups (≤ 39, 40-54, 55-69, and ≥ 70 years) based on AUA guidelines.⁴ The annual PSA testing rate was calculated for each veteran with SCI as the number of PSA tests per year. A mean annual PSA test rate was then calculated as the weighted (by exposure time) mean value for all annual PSA testing rates from 2000 through 2017 for each age group and race. Annual exposure was calculated for each veteran and defined as the number of days a veteran was eligible to have a PSA test. This started with the date of SCI diagnosis and ended with either the date of death or the date of last PSA. If a veteran moved from one age group to another in 1 year, the first part of this year’s exposure was included in the calculation of the annual PSA testing rate for the younger group and the second part was included for the calculation of the older group. For deceased veterans, the death date was excluded from the exposure period, and their exposure period ended on the day before death.

Statistical Analysis

To compare PSA testing rates between African American race and other races, Poisson regression was used with exposure treated as an offset (exposures were summed across years for each veteran). An indicator (dummy) variable for African American race vs other races was coded, and statistical significance was set at P < .05. To check sensitivity for the Poisson assumption that the mean was equal to the variance, negative binomial regression was used. To assess for geographic PSA testing rate variability, the data were further analyzed based on the locations where PSA tests were performed. This subanalysis was limited to veterans who had all PSA tests in a single station. For each station, the average PSA testing rate was calculated for each veteran, and the mean for all annual PSA testing rates was used to determine station-specific PSA testing rates.

 

Results

A total of 45,274 veterans were initially identified of which 367 females were excluded (Figure 1).

Moreover, 1688 male veterans were excluded for ICD codes that were less relevant, yielding 43,219 male veterans with relevant ICD codes. From this group, an additional 5976 were excluded because no PSA test was found after the SCI date. The racial makeup of the remaining 37,243 male veterans included 6327 African American patients, 25,277 of other races, and 5639 with missing/inconsistent race data. The included sample received care in ≥ 1 of 129 VAMCs. The final cohort yielded 261,125 PSA tests.

The Table shows PSA tests categorized by age group and race.

The PSA testing rate rose for veterans in the age groups ≤ 39, 40 to 54, and 55 to 69 years (Figure 2A).

The PSA testing rate dropped for the oldest age group (≥ 70 years), for the entire population, and the other race and missing/inconsistent race groups; however, PSA testing rates continued to rise in the African American group aged ≥ 70 years. For the entire population, average PSA testing rates in tests per year for the age groups were 0.46 (aged ≤ 39 years), 0.78 (aged 40-54 years), 1.0 (aged 55-69 years), and 0.91 (aged ≥ 70 years). However, PSA testing rates were significantly higher for the African American vs other races group at all ages (0.47 vs 0.46 tests per year, respectively, aged ≤ 39 years; 0.83 vs 0.77 tests per year, respectively, aged 40-54 years; 1.04 vs 1.00 tests per year, respectively, aged 55-69 years; and 1.08 vs 0.90 tests per year respectively, aged ≥ 70 years; P < .001) (Figure 2B).

Of the cohort of 37,243 veterans, 28,396 (76.2%) had their post-SCI tests done at a single facility, 6770 (18.1%) at 2 locations, and 2077 (5.5%) at > 2 locations. Single-station group data were included in a subanalysis to determine the mean (SD) PSA testing rates, which for the 123 locations was 0.98 (0.36) tests per veteran per year (range, 0.2-3.0 tests per veteran per year).

Figure 3 shows a heat map of the US: each dot represents a specific VAMC and shows PSA testing rate variability between stations.

To assess the impact of the 2012 USPSTF recommendations on PSA testing rates in veterans with SCI, mean PSA testing rates were calculated for 5 years before the recommendations (2007-2011) and compared with the average PSA testing rate for 5 years following the updated recommendations (2013-2017). The USPSTF updated its recommendation again in 2018 and acknowledged the potential benefit for PSA screening in certain patient populations.^2,3 Surprisingly, and despite recommendations, the results show a significant increase in PSA testing rates in all age groups for all races (P < .001) (Figure 4).

For the entire population, the average PSA testing rates for 2007 to 2011 in tests per year were 0.39, 0.76, 1.03, and 0.89 for the ≤ 39 years, 40 to 54 years, 55 to 69 years, and ≥ 70 years age groups, respectively. Likewise, the average PSA testing rates for years 2013 to 2017 in tests per year were 0.75, 0.96, 1.13, and 0.98 for the ≤ 39 years, 40 to 54 years, 55 to 69 years, and ≥ 70 years age groups, respectively, with an increased rate of testing of 0.92, 0.26, 0.10, and 0.11, respectively, from years 2007-2011 to 2013-2017 (P < .001).

 

Discussion

The goal of this study was to establish testing rates and analyze PSA testing trends across races and age groups in veterans with SCI. This is the largest cohort of patients with SCI analyzed in the literature. The key findings of this study were that despite clear AUA guidelines recommending against PSA testing in patients aged ≤ 39 years and ≥ 70 years, there are high rates of testing in veterans with SCI in these age groups (0.46 tests per year in those aged ≤ 39 years and 0.91 tests per year in those aged ≥ 70 years). In terms of race, as expected based on increased risk, African American veterans with SCI had higher PSA test rates.¹⁸ However, the continued increase in PSA testing rate for African American veterans aged ≥ 70 years was unexpected and not seen in other racial groups. As racial disparities are known to affect prostate cancer outcomes in African American men, it is reassuring that PSA testing was actually higher among African American men with SCI in our population, suggesting this vulnerable population is not being left behind in terms of screening.¹⁷ In contrast to other studies that show a lower rate of PSA screening in patients with SCI, our study suggests general PSA overtesting in veterans with SCI and a need for improved education for both veterans and their health care practitioners.¹⁹

Prostate Cancer Incidence

Although the exact mechanism behind alterations in prostate function in the SCI population have yet to be fully elucidated, research suggests that the prostate behaves differently after SCI. Animal models of prostate gland denervation show decreased prostate volume and suggest that SCI may lead to a reduction in prostatic secretory function associated with autonomic dysfunction. Shim and colleagues hypothesized that impaired autonomic prostate innervation alters the prostatic volume and PSA in patients with SCI.¹⁰

Additional studies looking at actual PSA levels in men with SCI reveal conflicting data.^10-15,20 Toricelli and colleagues retrospectively studied 140 men with SCI, of whom 34 had PSA levels available and found that mean PSA was not significantly different for patients with SCI compared with controls, but patients using clean intermittent catheterization had 2-fold higher PSA levels.²¹ In contrast, Konety and colleagues found that mean PSA was not significantly different from uninjured controls in their cohort of 79 patients with SCI, though they did find a correlation between indwelling catheter use and a higher PSA.²²

Studies have shown an overall decreased risk of prostate cancer in patients with SCI, though the mechanism remains unclear. A large cohort study from Taiwan showed a lower risk of prostate cancer for 54,401 patients with SCI with an adjusted hazard ratio of 0.73.²³ Patel and colleagues found the overall rate of prostate cancer in the population of veterans with SCI was lower than the general uninjured VA population, though this study was limited by scope with only 350 patients with SCI.²⁴ A more recent systematic review and meta-analysis of 9 studies evaluating the prevalence of prostate cancer in men with SCI found a reduction of up to 65% in the risk of prostate cancer in men with SCI, and PSA was found to be a poor screening tool for prostate cancer due to large study heterogeneity.¹⁶

 

PSA Screening

This study identified widespread overscreening using the PSA test in veterans with SCI, which is likely attributable to many factors. Per VHA Directive 1176, all eligible veterans are offered yearly interdisciplinary comprehensive evaluations, including laboratory testing, and as such veterans with SCI have high rates of annual visit attendance due to the complexity of their care.⁹ PSA testing is included in the standard battery of laboratory tests ordered for all patients with SCI during their annual examinations. Additionally, many SCI specialists use the PSA level in patients with SCI for identifying cystitis or prostatitis in patients with colonization who may not experience typical symptoms. Everaert and colleagues demonstrated the clinical utility for localizing UTIs to the upper or lower tract, with elevated PSA indicating prostatitis. They found that serum PSA has a sensitivity of 68% and a specificity of 100% in the differential diagnosis of prostatitis and pyelonephritis.²⁵ As such, the high PSA screening rates may be reflective of diagnostic use for infection rather than for cancer screening.

Likely as a response to the USPSTF recommendations, there has been a national slow decline in overall PSA screening rates since 2012.^26-28 A study from Vetterlein and colleagues examining changes in the PSA screening trends related to USPSTF recommendations found an 8.5% decline in overall PSA screening from 2012 to 2014.²⁹ However, the increase in PSA testing across all ages and races in the VA population with SCI over the same period is not entirely understood and suggests the need for further research and education in this area. Additionally, as factors associated with SCI impact the life expectancy of these patients, further shared decision making is needed in deciding whether to pursue PSA screening in this population to minimize unnecessary screening in patients with a life expectancy of < 10 to 15 years.

Limitations

This study is limited by the use of data identified by ICD codes rather than by review of individual health records. This required the use of decision algorithms for data points, such as the date of SCI. In addition, analysis was not able to capture shared decision making that may have contributed to PSA screening outside the recommended age ranges based on additional risk factors, such as family history of lethal malignancy. Furthermore, a detailed attempt to define specific age-adjusted PSA levels was beyond the scope of this study but will be addressed in later publications. In addition, we did not exclude individuals with a diagnosis of prostate adenocarcinoma, prostatitis, or recurrent UTIs because the onset, duration, and severity of disease could not be definitively ascertained. Finally, veterans with SCI are unique and may not be reflective of individuals with SCI who do not receive care within the VA. However, despite these limitations, this is, to our knowledge, the largest and most comprehensive study evaluating PSA testing rates in individuals with SCI.

Conclusions

Currently, PSA screening is recommended following shared decision making for patients at average risk aged 55 to 70 years. Patients with SCI experience many conditions that may affect PSA values, but data regarding normal PSA ranges and rates of prostate cancer in this population remain sparse. The study demonstrated high rates of overtesting in veterans with SCI, higher than expected testing rates in African American veterans, a paradoxical increase in PSA testing rates after the 2012 publication of the USPSTF PSA guidelines, and wide variability in testing rates depending on VA location.

African American men were tested at higher rates across all age groups, including in patients aged > 70 years. To balance the benefits of detecting clinically significant prostate cancer vs the risks of invasive testing in high-risk populations with SCI, more work is needed to determine the clinical impact of screening practices. Future work is currently ongoing to define age-based PSA values in patients with SCI.

Acknowledgments

This research was supported in part through funding from the Center for Rehabilitation Science and Engineering, Virginia Commonwealth University Health System.

Most cancer treatment regimens include the administration of several chemotherapeutic agents. Drug-drug interactions (DDIs) can increase the risk of fatal adverse events and reduce therapeutic efficacy.^1,2 Erlotinib, gefitinib, afatinib, osimertinib, and icotinib are epidermal growth factor receptor–tyrosine kinase inhibitors (EGFR-TKIs) that have proven efficacy for treating advanced non–small cell lung cancer (NSCLC). Erlotinib strongly inhibits cytochrome P450 (CYP) isoenzymes CYP 1A1, moderately inhibits CYP 3A4 and 2C8, and induces CYP 1A1 and 1A2.² Gefitinib weakly inhibits CYP 2C19 and 2D6.² CYP 3A4 inducers and inhibitors affect metabolism of both erlotinib and gefitinib.^3,4

Erlotinib and gefitinib are first-generation EGFR-TKIs and have been approved for NSCLC treatment by the US Food and Drug Administration (FDA). These agents have been used since the early 2000s and increase the possibility of long-term response and survival.^2,5,6 EGFR-TKIs have a range of potential DDIs, including interactions with CYP-dependent metabolism, uridine diphosphate-glucuronosyltransferase, and transporter proteins.² Few retrospective studies have focused on the therapeutic efficacy of erlotinib, gefitinib, or the combination of these agents.^7-14

DDIs from cancer and noncancer therapies could lead to treatment discontinuation and affect patient outcomes. The goals for this study were to perform a broad-scale retrospective analysis focused on investigating prescribed drugs used with erlotinib and gefitinib and determine patient outcomes as obtained through several Military Health System (MHS) databases. Our investigation focused on (1) the functions of these drugs; (2) identifying adverse effects (AEs) that patients experienced; (3) evaluating differences when these drugs are used alone vs concomitantly, and between the completed vs discontinued treatment groups; (4) identifying all drugs used during erlotinib or gefitinib treatment; and (5) evaluating DDIs with antidepressants.

This retrospective study was performed at the Department of Research Programs at Walter Reed National Military Medical Center (WRNMMC) in Bethesda, Maryland. The WRNMMC Institutional Review Board approved the study protocol and ensured compliance with the Health Insurance Portability and Accountability Act as an exempt protocol. The Joint Pathology Center of the US Department of Defense (DoD) Cancer Registry and MHS data experts from the Comprehensive Ambulatory/Professional Encounter Record (CAPER) and the Pharmacy Data Transaction Service (PDTS) provided data for the analysis.

 

Methods

The DoD Cancer Registry Program was established in 1986 by the Assistant Secretary of Defense for Health Affairs. The registry currently contains data from 1998 to 2023. CAPER and PDTS are part of the MHS Data Repository/Management Analysis and Reporting Tool database. Each observation in the CAPER record represents an ambulatory encounter at a military treatment facility (MTF). CAPER records are available from 2003 to 2023.

Each observation in the PDTS record represents an outpatient prescription filled for an MHS beneficiary at MTFs through the TRICARE mail-order program or a retail pharmacy in the United States. Missing from this record are prescriptions filled at civilian pharmacies outside the United States and inpatient pharmacy prescriptions. The MHS Data Repository PDTS record is available from 2002 to 2023. The Composite Health Care System—the legacy system—is being replaced by GENESIS at MTFs.

Data Extraction Design

The study design involved a cross-sectional analysis. We requested data extraction for erlotinib and gefitinib from 1998 to 2021. Data from the DoD Cancer Registry were used to identify patients who received cancer treatment. Once patients were identified, the CAPER database was searched for diagnoses to identify other health conditions, while the PDTS database was used to populate a list of prescription medications filled during chemotherapy treatment.

Data collected from the Joint Pathology Center included cancer treatment (alone or concomitant), cancer information (cancer types and stages), demographics (sex, age at diagnosis), and physicians’ comments on AEs. Collected data from the MHS include diagnosis and filled prescription history from initiation to completion of the therapy period (or a buffer of 6 months after the initial period). We used all collected data in this analysis. The only exclusion criterion was a provided physician’s note commenting that the patient did not use erlotinib or gefitinib.

Data Extraction Analysis

The Surveillance, Epidemiology, and End Results Program Coding and Staging Manual 2016 and the International Classification of Diseases for Oncology (ICD-O) were used to decode disease and cancer types.^15,16 Data sorting and analysis were performed using Microsoft Excel. The percentage for the total was calculated by using the total number of patients or data available within the gefitinib and erlotinib groups divided by total number of patients or data variables. The subgroup percentage was calculated by using the number of patients or data available within the subgroup divided by the total number of patients in that subgroup.

In alone vs concomitant and completed vs discontinued treatment groups, a 2-tailed, 2-sample z test was used to calculate P to determine statistical significance (P < .05) using a statistics website.¹⁷ Concomitant was defined as erlotinib or gefitinib taken with other medication(s) before, after, or at the same time as cancer therapy. For the retrospective data analysis, physicians’ notes with “.”, “,”, “/”, “;”, (period, comma, forward slash, semicolon) or space between medication names were interpreted as concurrent, while “+”, “-/+” (plus, minus/plus), or and between drug names were interpreted as combined. Completed treatment was defined as erlotinib or gefitinib as the last medication the patient took without recorded AEs; switching or experiencing AEs was defined as discontinued treatment.

 

Results

Erlotinib

The Joint Pathology Center provided 387 entries for 382 patients aged 21 to 93 years (mean, 65 years) who were treated systemically with erlotinib from January 1, 2001, to December 31, 2020. Five patients had duplicate entries because they had different cancer sites. There were 287 patients (74%) with lung cancer, 61 (16%) with pancreatic cancer, and 39 (10%) with other cancers. For lung cancer, there were 118 patients (30%) for the upper lobe, 78 (20%) for the lower lobe, and 60 (16%) not otherwise specified (NOS). Other lung cancer sites had fewer patients: 21 (5%) middle lobe lung, 6 (2%) overlapping lung lesion(s), and 4 (1%) main bronchus of the lung. For pancreatic cancer, there were 27 patients (7%) for the head of the pancreas, 10 (3%) pancreas NOS, 9 (2%) body of the pancreas, 9 (2%) tail of the pancreas, 4 (1%) overlapping lesions of the pancreas, 1 (< 1%) pancreatic duct, and 1 (< 1%) other specified parts of the pancreas

. Thirty-nine patients (10%) received erlotinib for indications that were not for FDA-approved indications, which included 9 (2%) for kidney NOS, 8 (2%) for the unknown primary site, 5 (1%) for liver cancer, 2 (1%) for intrahepatic bile duct, 2 (1%) for tonsil, and 1 (< 1%) for 13 disease sites (Table 1).

There were 342 patients (88%) who were aged > 50 years; 186 male patients (48%) and 201 female patients (52%). There were 293 patients (76%) who had a cancer diagnosis of stage III or IV disease and 94 (24%) who had a cancer diagnosis of stage ≤ II (combination of data for stage 0, 1, and 2, not applicable, and unknown). For their systemic treatment, 161 patients (42%) were treated with erlotinib alone and 226 (58%) received erlotinib concomitantly with additional chemotherapy.

Of these patients, 287 (74%) were diagnosed with lung cancer (Table 2).

Patients were more likely to discontinue erlotinib for chemotherapy if they received concomitant treatment. Among the patients receiving erlotinib monotherapy, 5% stopped the treatment, whereas 51% of patients treated concomitantly discontinued (P < .001).

The comparisons for lung cancer vs other cancer and those aged ≤ 50 years vs > 50 years were significant (P = .005 and .05, respectively) while other comparisons were not significant (Table 3).

Among the 123 patients who discontinued their treatment, 101 switched treatment with no AEs notes, 22 died or experienced fatigue with blurry vision, constipation, nonspecific gastrointestinal effects, grade-4 diarrhea (as defined by the Common Terminology Criteria for Adverse Events), or developed a pleural fluid, pneumonitis, renal failure, skin swelling and facial rash, and unknown AEs of discontinuation. Patients who discontinued treatment because of unknown AEs had physicians’ notes that detailed emergency department visits, peripheral vascular disease, progressive disease, and treatment cessation, but did not specify the exact symptom(s) that led to discontinuation. The causes of death are unknown because they were not detailed in the available notes or databases. The overall results in this retrospective review cannot establish causality between taking erlotinib or gefitinib and death.

 

Gefitinib

In September 2021, the Joint Pathology Center provided 33 entries for 33 patients who were systemically treated with gefitinib from January 1, 2002, to December 31, 2017. The patient ages ranged from 49 to 89 years with a mean age of 66 years. There were 31 (94%) and 2 (6%) patients with lung and other cancers, respectively. The upper lobe, lower lobe, and lung NOS had the most patients: 14 (42%), 8 (24%), and 6 (18%), respectively.

There were 31 patients (94%) who were aged > 50 years; 15 were male (45%) and 18 were female (55%). There were 26 patients (79%) who had a cancer diagnosis of stage III or IV disease. Nineteen patients (58%) were treated with gefitinib alone, and 14 (42%) were treated with gefitinib concomitantly with additional chemotherapy. Thirty-one patients (94%) were treated for lung cancer (Table 2). Thirty-three patients are a small sample size to determine whether patients were likely to stop gefitinib if used concomitantly with other drugs. Among the patients treated with gefitinib monotherapy, 5% (n = 1) stopped treatment, whereas 29% (n = 4) of patients treated concomitantly discontinued treatment (P = .06). All comparisons for gefitinib yielded insignificant P values. Physicians’ notes indicated that the reasons for gefitinib discontinuation were life-altering pruritis and unknown (progressive disease outcome) (Table 3).

Management Analysis and Reporting Tool Database

MHS data analysts provided data on diagnoses for 348 patients among 415 submitted, with 232 and 112 patients completing and discontinuing erlotinib or gefitinib treatment, respectively. Each patient had 1 to 104 (completed treatment group) and 1 to 157 (discontinued treatment group) unique health conditions documented. The MHS reported 1319 unique-diagnosis conditions for the completed group and 1266 for the discontinued group. Patients with additional health issues stopped chemotherapy use more often than those without; P < .001 for the completed group (232 patients, 1319 diagnoses) vs the discontinued group (112 patients, 1266 diagnoses). The mean (SD) number of diagnoses was 19 (17) for the completed and 30 (22) for the discontinued treatment groups (Figure).

The 5 most recorded diagnoses with erlotinib among 358 patients were malignant neoplasm of bronchus and lung for 225 patients, unspecified essential hypertension for 120 patients, encounters for antineoplastic chemotherapy for 113 patients, dietary surveillance and counseling for 102 patients, and unspecified administrative purposes for 97 patients.

MHS data was provided for patients who filled erlotinib (n = 240) or gefitinib (n = 18). Among the 258 patients, there were 179 and 79 patients in the completed and discontinued treatment groups, respectively. Each patient filled 1 to 75 (for the completed treatment group) and 3 to 103 (for the discontinued treatment group) prescription drugs. There were 805 unique-filled prescriptions for the completed and 670 for the discontinued group. Patients in the discontinued group filled more prescriptions than those who completed treatment; P < .001 for the completed group (179 patients,805 drugs) vs the discontinued group (79 patients, 670 drugs).

The mean (SD) number of filled prescription drugs was 19 (11) for the completed group and 29 (18) for the discontinued treatment group. The 5 most filled prescriptions with erlotinib from 258 patients with PDTS data were ondansetron (151 prescriptions, 10 recorded AEs), dexamethasone (119 prescriptions, 9 recorded AEs), prochlorperazine (105 prescriptions, 15 recorded AEs), oxycodone (99 prescriptions, 1 AE), and docusate (96 prescriptions, 7 recorded AEs).

 

Discussion

The difference between erlotinib and gefitinib data can be attributed to the FDA approval date and gefitinib’s association with a higher frequency of hepatotoxicity.^18-20 The FDA designated gefitinib as an orphan drug for EGFR mutation–positive NSCLC treatment. Gefitinib first received accelerated approval in 2003 for the treatment of locally advanced or metastatic NSCLC. Gefitinib then was voluntarily withdrawn from the market following confirmatory clinical trials that did not verify clinical benefit.

The current approval is for a different patient population—previously untreated, metastatic EGFR exon 19 or 21 L858R mutation—than the 2003 approval.^4,6 There was no record of gefitinib use after 2017 in our study.

Erlotinib is a reversible EGFR-TKI that is approved by the FDA as first-line (maintenance) or second-line treatment (after progression following at least 1 earlier chemotherapy regimen) for patients with metastatic NSCLC who harbor EGFR exon 19 deletions or exon 21 L858R substitution mutations, as detected by an FDA-approved test.³ Since 2005, the FDA also approved erlotinib for first-line treatment of patients with locally advanced, unresectable, or metastatic pancreatic cancer in combination with gemcitabine.³ Without FDA indication, erlotinib is used for colorectal, head and neck, ovarian carcinoma, pancreatic carcinoma, and breast cancer.²¹

Erlotinib and gefitinib are not considered first-line treatments in EGFR exon 19 or 21–mutated NSCLC because osimertinib was approved in 2018. Targeted therapies for EGFR mutation continue to advance at a fast pace, with amivantamab and mobocertinib now FDA approved for EGFR exon 20 insertion–mutated NSCLC.

Erlotinib Use

Thirty-nine patients (10%) in this study were prescribed erlotinib for off-label indications. Erlotinib was used alone or in combination with bevacizumab, capecitabine, cisplatin, denosumab, docetaxel, gemcitabine, and the MEK-inhibitor selumetinib. Erlotinib combined with cisplatin, denosumab, docetaxel, and gemcitabine had no recorded AEs, with 10 data entries for gemcitabine and 1 for other drugs. Three patients received bevacizumab and erlotinib, and 1 patient (diagnosed with kidney NOS) showed rash or facial swelling/erythema and diffuse body itching then stable disease after 2 cycles.

One patient (diagnosed with cancer located at the pancreas head) was bridged with capecitabine and erlotinib when going on a vacation, then received FOLFIRINOX (a combination chemotherapy regimen containing folinic acid [leucovorin], fluorouracil, irinotecan, and oxaliplatin), which led to significant fatigue, blurry vision, and constipation. One patient was treated for lung NOS with the MEK-inhibitor selumetinib plus erlotinib and developed pneumonitis following treatment.

Because oncologists followed guidelines and protocols in systemic treatment, DDIs of erlotinib concurrently (before or after) and in combination with cancer drugs were unlikely. Further investigation is needed for several 1:1:1 DDIs with noncancer drugs. A retrospective overview is not a randomized clinical study; therefore, analysis is limited. Data from the MHS were obtained solely from notes from physicians who treated the patients; therefore, exact information explaining whether a patient completed treatment or had to withdraw could not be extrapolated (ie, blood/plasma samples were not obtained to confirm).

Discontinued Treatment

The reasons for treatment discontinuation with erlotinib or gefitinib varied among patients, with no consistent AE or cause. Most data were for switching treatments after discontinuing treatment with erlotinib (101 of 123 patients) and gefitinib (2 of 5 patients). This is not surprising given the widely recognized pillars of therapy for NSCLC: chemotherapy, target therapy, and immunotherapy.²² From the MHS records, the reasons patients switched treatment of erlotinib or gefitinib were not listed or listed as due to negative EGFR testing, lack of responsiveness, or enrollment in a different treatment.

 

Physicians’ notes on AEs were not detailed in most cases. Notes for gastrointestinal effects, life-altering pruritis, intolerance, peripheral vascular disease, pneumonitis, and progressive disease described the change in status or appearance of a new medical condition but did not indicate whether erlotinib or gefitinib caused the changes or worsened a pre-existing condition.

The causes of AEs were not described in the available notes or the databases. This retrospective data analysis only focused on identifying drugs involved with erlotinib and gefitinib treatment; further mapping of DDIs among patients experiencing AEs needs to be performed, then in vitro data testing before researchers can reach a conclusion.

DDIs With Antidepressants

We used the PDTS database to evaluate patients who experienced AEs, excluding patients who switched treatment. Thirteen patients filled a prescription for erlotinib and reported taking 220 cancer and noncancer prescription drugs. One patient (pruritis) was taking gefitinib along with 16 noncancer prescription drugs.

Table 4 details CYP information for cancer drugs, antidepressants, and noncancer drugs (top 11 drugs) among patients who recorded AEs with erlotinib.^3-4,23-47

Selective serotonin reuptake inhibitors and other antidepressants have been implicated in CYP 2D6 inhibition and DDIs.^48,49 Losartan is a widely used antihypertensive drug with a favorable DDI profile.⁵⁰ Erlotinib and gefitinib are primarily metabolized via CYP 2D6 and 3A4 pathways. DDIs from in vitro human hepatocytes assays revealed that gefitinib had significant metabolic changes in a 1:1 (P < .05) combination with paroxetine or sertraline, and a 1:1:1 combination with losartan and fluoxetine, fluvoxamine, paroxetine, or sertraline. Citalopram and venlafaxine seemed to be unaffected by any combination (P ≥ .05).⁵¹ Erlotinib with fluoxetine or losartan 1:1 yielded insignificant differences in metabolism for all drugs (P ≥ .05). Three drug combinations of 1:1:1 involving fluoxetine and losartan with erlotinib yielded significant degrees of inhibition of fluoxetine and losartan metabolism (P < .05) but not erlotinib.⁵²

Our data showed that 16 antidepressants (amitriptyline, bupropion, citalopram, desvenlafaxine, duloxetine, escitalopram, imipramine, fluoxetine, fluvoxamine, mirtazapine, nortriptyline, paroxetine, phenelzine, sertraline, trazodone, and venlafaxine) were recorded with concomitant erlotinib or gefitinib from initiation to completion of therapy or a buffer of 6 months from the first diagnosis date. Based on the date dispensed and days’ supply, only escitalopram could be used in combination with gefitinib treatment. The one patient who filled a prescription for gefitinib and escitalopram completed treatment without recorded AEs. PDTS database confirmed that patients experienced AEs with 5 antidepressants (amitriptyline, mirtazapine, paroxetine, trazodone, and venlafaxine) with concomitant erlotinib use.

Based on the date dispensed and days’ supply, only trazodone could be used in combination with erlotinib. PDTS database showed that cancer drugs (erlotinib and megestrol) and 39 noncancer drugs (including acetaminophen, azithromycin, dexamethasone, hydrocortisone, and polyethylene glycol) were filled by 1 patient whose physician noted skin rash. Another limitation of using databases to reflect clinical practice is that although megestrol is listed as a cancer drug by code in the PDTS database, it is not used for nonendometrial or gynecologic cancers. However, because of the PDTS database classification, megestrol is classified as a cancer drug in this retrospective review.

This retrospective review found no significant DDIs for erlotinib or gefitinib, with 1 antidepressant taken by 1 patient for each respective treatment. The degree of inhibition and induction for escitalopram and trazodone are categorized as weak, minimal, or none; therefore, while 1:1 DDIs might be little or no effect, 1:1:1 combination DDIs could have a different outcome. This retrospective data collection cannot be linked to the in­ vitro hepatocyte DDIs from erlotinib and gefitinib in previous studies.^51,52

 

Conclusions

This retrospective study describes erlotinib and gefitinib use in the MHS and their potential for DDIs. Because of military service requirements, people who are qualified to serve must be healthy or have either controlled or nonactive medical diagnoses and be physically fit. Consequently, our patient population had fewer common medical illnesses, such as diabetes and obesity, compared with the general population. Most noncancer drugs mentioned in this study are not known CYP metabolizers; therefore, recorded AEs alone cannot conclusively determine whether there is a DDI among erlotinib or gefitinib and noncancer drugs. Antidepressants generally are safe but have boxed warnings in the US for increased risk of suicidal ideation in young people.^53,54 This retrospective study did not find statistically significant DDIs for erlotinib or gefitinib with antidepressants. Based on this retrospective data analysis, future in vitro testing is needed to assess DDIs for erlotinib or gefitinib and cancer or noncancer drugs identified in this study.

Acknowledgments

The Department of Research Program funds at Walter Reed National Military Medical Center supported this protocol. We sincerely appreciate the contribution of data extraction from the Joint Pathology Center teams (Francisco J. Rentas, John D. McGeeney, Kimberly M. Greenfield, Beatriz A. Hallo, and Johnny P. Beason) and the MHS database personnel (Maj Ryan Costantino, Lee Ann Zarzabal, Brandon Jenkins, and Alex Rittel). We gratefully thank you for the protocol support from the Department of Research programs: CDR Wesley R. Campbell, CDR Ling Ye, Yaling Zhou, Elizabeth Schafer, Robert Roogow, Micah Stretch, Diane Beaner, Adrienne Woodard, David L. Evers, and Paula Amann.

Most cancer treatment regimens include the administration of several chemotherapeutic agents. Drug-drug interactions (DDIs) can increase the risk of fatal adverse events and reduce therapeutic efficacy.^1,2 Erlotinib, gefitinib, afatinib, osimertinib, and icotinib are epidermal growth factor receptor–tyrosine kinase inhibitors (EGFR-TKIs) that have proven efficacy for treating advanced non–small cell lung cancer (NSCLC). Erlotinib strongly inhibits cytochrome P450 (CYP) isoenzymes CYP 1A1, moderately inhibits CYP 3A4 and 2C8, and induces CYP 1A1 and 1A2.² Gefitinib weakly inhibits CYP 2C19 and 2D6.² CYP 3A4 inducers and inhibitors affect metabolism of both erlotinib and gefitinib.^3,4

Erlotinib and gefitinib are first-generation EGFR-TKIs and have been approved for NSCLC treatment by the US Food and Drug Administration (FDA). These agents have been used since the early 2000s and increase the possibility of long-term response and survival.^2,5,6 EGFR-TKIs have a range of potential DDIs, including interactions with CYP-dependent metabolism, uridine diphosphate-glucuronosyltransferase, and transporter proteins.² Few retrospective studies have focused on the therapeutic efficacy of erlotinib, gefitinib, or the combination of these agents.^7-14

DDIs from cancer and noncancer therapies could lead to treatment discontinuation and affect patient outcomes. The goals for this study were to perform a broad-scale retrospective analysis focused on investigating prescribed drugs used with erlotinib and gefitinib and determine patient outcomes as obtained through several Military Health System (MHS) databases. Our investigation focused on (1) the functions of these drugs; (2) identifying adverse effects (AEs) that patients experienced; (3) evaluating differences when these drugs are used alone vs concomitantly, and between the completed vs discontinued treatment groups; (4) identifying all drugs used during erlotinib or gefitinib treatment; and (5) evaluating DDIs with antidepressants.

This retrospective study was performed at the Department of Research Programs at Walter Reed National Military Medical Center (WRNMMC) in Bethesda, Maryland. The WRNMMC Institutional Review Board approved the study protocol and ensured compliance with the Health Insurance Portability and Accountability Act as an exempt protocol. The Joint Pathology Center of the US Department of Defense (DoD) Cancer Registry and MHS data experts from the Comprehensive Ambulatory/Professional Encounter Record (CAPER) and the Pharmacy Data Transaction Service (PDTS) provided data for the analysis.

 

Methods

The DoD Cancer Registry Program was established in 1986 by the Assistant Secretary of Defense for Health Affairs. The registry currently contains data from 1998 to 2023. CAPER and PDTS are part of the MHS Data Repository/Management Analysis and Reporting Tool database. Each observation in the CAPER record represents an ambulatory encounter at a military treatment facility (MTF). CAPER records are available from 2003 to 2023.

Each observation in the PDTS record represents an outpatient prescription filled for an MHS beneficiary at MTFs through the TRICARE mail-order program or a retail pharmacy in the United States. Missing from this record are prescriptions filled at civilian pharmacies outside the United States and inpatient pharmacy prescriptions. The MHS Data Repository PDTS record is available from 2002 to 2023. The Composite Health Care System—the legacy system—is being replaced by GENESIS at MTFs.

Data Extraction Design

The study design involved a cross-sectional analysis. We requested data extraction for erlotinib and gefitinib from 1998 to 2021. Data from the DoD Cancer Registry were used to identify patients who received cancer treatment. Once patients were identified, the CAPER database was searched for diagnoses to identify other health conditions, while the PDTS database was used to populate a list of prescription medications filled during chemotherapy treatment.

Data collected from the Joint Pathology Center included cancer treatment (alone or concomitant), cancer information (cancer types and stages), demographics (sex, age at diagnosis), and physicians’ comments on AEs. Collected data from the MHS include diagnosis and filled prescription history from initiation to completion of the therapy period (or a buffer of 6 months after the initial period). We used all collected data in this analysis. The only exclusion criterion was a provided physician’s note commenting that the patient did not use erlotinib or gefitinib.

Data Extraction Analysis

The Surveillance, Epidemiology, and End Results Program Coding and Staging Manual 2016 and the International Classification of Diseases for Oncology (ICD-O) were used to decode disease and cancer types.^15,16 Data sorting and analysis were performed using Microsoft Excel. The percentage for the total was calculated by using the total number of patients or data available within the gefitinib and erlotinib groups divided by total number of patients or data variables. The subgroup percentage was calculated by using the number of patients or data available within the subgroup divided by the total number of patients in that subgroup.

In alone vs concomitant and completed vs discontinued treatment groups, a 2-tailed, 2-sample z test was used to calculate P to determine statistical significance (P < .05) using a statistics website.¹⁷ Concomitant was defined as erlotinib or gefitinib taken with other medication(s) before, after, or at the same time as cancer therapy. For the retrospective data analysis, physicians’ notes with “.”, “,”, “/”, “;”, (period, comma, forward slash, semicolon) or space between medication names were interpreted as concurrent, while “+”, “-/+” (plus, minus/plus), or and between drug names were interpreted as combined. Completed treatment was defined as erlotinib or gefitinib as the last medication the patient took without recorded AEs; switching or experiencing AEs was defined as discontinued treatment.

 

Results

Erlotinib

The Joint Pathology Center provided 387 entries for 382 patients aged 21 to 93 years (mean, 65 years) who were treated systemically with erlotinib from January 1, 2001, to December 31, 2020. Five patients had duplicate entries because they had different cancer sites. There were 287 patients (74%) with lung cancer, 61 (16%) with pancreatic cancer, and 39 (10%) with other cancers. For lung cancer, there were 118 patients (30%) for the upper lobe, 78 (20%) for the lower lobe, and 60 (16%) not otherwise specified (NOS). Other lung cancer sites had fewer patients: 21 (5%) middle lobe lung, 6 (2%) overlapping lung lesion(s), and 4 (1%) main bronchus of the lung. For pancreatic cancer, there were 27 patients (7%) for the head of the pancreas, 10 (3%) pancreas NOS, 9 (2%) body of the pancreas, 9 (2%) tail of the pancreas, 4 (1%) overlapping lesions of the pancreas, 1 (< 1%) pancreatic duct, and 1 (< 1%) other specified parts of the pancreas

. Thirty-nine patients (10%) received erlotinib for indications that were not for FDA-approved indications, which included 9 (2%) for kidney NOS, 8 (2%) for the unknown primary site, 5 (1%) for liver cancer, 2 (1%) for intrahepatic bile duct, 2 (1%) for tonsil, and 1 (< 1%) for 13 disease sites (Table 1).

There were 342 patients (88%) who were aged > 50 years; 186 male patients (48%) and 201 female patients (52%). There were 293 patients (76%) who had a cancer diagnosis of stage III or IV disease and 94 (24%) who had a cancer diagnosis of stage ≤ II (combination of data for stage 0, 1, and 2, not applicable, and unknown). For their systemic treatment, 161 patients (42%) were treated with erlotinib alone and 226 (58%) received erlotinib concomitantly with additional chemotherapy.

Of these patients, 287 (74%) were diagnosed with lung cancer (Table 2).

Patients were more likely to discontinue erlotinib for chemotherapy if they received concomitant treatment. Among the patients receiving erlotinib monotherapy, 5% stopped the treatment, whereas 51% of patients treated concomitantly discontinued (P < .001).

The comparisons for lung cancer vs other cancer and those aged ≤ 50 years vs > 50 years were significant (P = .005 and .05, respectively) while other comparisons were not significant (Table 3).

Among the 123 patients who discontinued their treatment, 101 switched treatment with no AEs notes, 22 died or experienced fatigue with blurry vision, constipation, nonspecific gastrointestinal effects, grade-4 diarrhea (as defined by the Common Terminology Criteria for Adverse Events), or developed a pleural fluid, pneumonitis, renal failure, skin swelling and facial rash, and unknown AEs of discontinuation. Patients who discontinued treatment because of unknown AEs had physicians’ notes that detailed emergency department visits, peripheral vascular disease, progressive disease, and treatment cessation, but did not specify the exact symptom(s) that led to discontinuation. The causes of death are unknown because they were not detailed in the available notes or databases. The overall results in this retrospective review cannot establish causality between taking erlotinib or gefitinib and death.

 

Gefitinib

In September 2021, the Joint Pathology Center provided 33 entries for 33 patients who were systemically treated with gefitinib from January 1, 2002, to December 31, 2017. The patient ages ranged from 49 to 89 years with a mean age of 66 years. There were 31 (94%) and 2 (6%) patients with lung and other cancers, respectively. The upper lobe, lower lobe, and lung NOS had the most patients: 14 (42%), 8 (24%), and 6 (18%), respectively.

There were 31 patients (94%) who were aged > 50 years; 15 were male (45%) and 18 were female (55%). There were 26 patients (79%) who had a cancer diagnosis of stage III or IV disease. Nineteen patients (58%) were treated with gefitinib alone, and 14 (42%) were treated with gefitinib concomitantly with additional chemotherapy. Thirty-one patients (94%) were treated for lung cancer (Table 2). Thirty-three patients are a small sample size to determine whether patients were likely to stop gefitinib if used concomitantly with other drugs. Among the patients treated with gefitinib monotherapy, 5% (n = 1) stopped treatment, whereas 29% (n = 4) of patients treated concomitantly discontinued treatment (P = .06). All comparisons for gefitinib yielded insignificant P values. Physicians’ notes indicated that the reasons for gefitinib discontinuation were life-altering pruritis and unknown (progressive disease outcome) (Table 3).

Management Analysis and Reporting Tool Database

MHS data analysts provided data on diagnoses for 348 patients among 415 submitted, with 232 and 112 patients completing and discontinuing erlotinib or gefitinib treatment, respectively. Each patient had 1 to 104 (completed treatment group) and 1 to 157 (discontinued treatment group) unique health conditions documented. The MHS reported 1319 unique-diagnosis conditions for the completed group and 1266 for the discontinued group. Patients with additional health issues stopped chemotherapy use more often than those without; P < .001 for the completed group (232 patients, 1319 diagnoses) vs the discontinued group (112 patients, 1266 diagnoses). The mean (SD) number of diagnoses was 19 (17) for the completed and 30 (22) for the discontinued treatment groups (Figure).

The 5 most recorded diagnoses with erlotinib among 358 patients were malignant neoplasm of bronchus and lung for 225 patients, unspecified essential hypertension for 120 patients, encounters for antineoplastic chemotherapy for 113 patients, dietary surveillance and counseling for 102 patients, and unspecified administrative purposes for 97 patients.

MHS data was provided for patients who filled erlotinib (n = 240) or gefitinib (n = 18). Among the 258 patients, there were 179 and 79 patients in the completed and discontinued treatment groups, respectively. Each patient filled 1 to 75 (for the completed treatment group) and 3 to 103 (for the discontinued treatment group) prescription drugs. There were 805 unique-filled prescriptions for the completed and 670 for the discontinued group. Patients in the discontinued group filled more prescriptions than those who completed treatment; P < .001 for the completed group (179 patients,805 drugs) vs the discontinued group (79 patients, 670 drugs).

The mean (SD) number of filled prescription drugs was 19 (11) for the completed group and 29 (18) for the discontinued treatment group. The 5 most filled prescriptions with erlotinib from 258 patients with PDTS data were ondansetron (151 prescriptions, 10 recorded AEs), dexamethasone (119 prescriptions, 9 recorded AEs), prochlorperazine (105 prescriptions, 15 recorded AEs), oxycodone (99 prescriptions, 1 AE), and docusate (96 prescriptions, 7 recorded AEs).

 

Discussion

The difference between erlotinib and gefitinib data can be attributed to the FDA approval date and gefitinib’s association with a higher frequency of hepatotoxicity.^18-20 The FDA designated gefitinib as an orphan drug for EGFR mutation–positive NSCLC treatment. Gefitinib first received accelerated approval in 2003 for the treatment of locally advanced or metastatic NSCLC. Gefitinib then was voluntarily withdrawn from the market following confirmatory clinical trials that did not verify clinical benefit.

The current approval is for a different patient population—previously untreated, metastatic EGFR exon 19 or 21 L858R mutation—than the 2003 approval.^4,6 There was no record of gefitinib use after 2017 in our study.

Erlotinib is a reversible EGFR-TKI that is approved by the FDA as first-line (maintenance) or second-line treatment (after progression following at least 1 earlier chemotherapy regimen) for patients with metastatic NSCLC who harbor EGFR exon 19 deletions or exon 21 L858R substitution mutations, as detected by an FDA-approved test.³ Since 2005, the FDA also approved erlotinib for first-line treatment of patients with locally advanced, unresectable, or metastatic pancreatic cancer in combination with gemcitabine.³ Without FDA indication, erlotinib is used for colorectal, head and neck, ovarian carcinoma, pancreatic carcinoma, and breast cancer.²¹

Erlotinib and gefitinib are not considered first-line treatments in EGFR exon 19 or 21–mutated NSCLC because osimertinib was approved in 2018. Targeted therapies for EGFR mutation continue to advance at a fast pace, with amivantamab and mobocertinib now FDA approved for EGFR exon 20 insertion–mutated NSCLC.

Erlotinib Use

Thirty-nine patients (10%) in this study were prescribed erlotinib for off-label indications. Erlotinib was used alone or in combination with bevacizumab, capecitabine, cisplatin, denosumab, docetaxel, gemcitabine, and the MEK-inhibitor selumetinib. Erlotinib combined with cisplatin, denosumab, docetaxel, and gemcitabine had no recorded AEs, with 10 data entries for gemcitabine and 1 for other drugs. Three patients received bevacizumab and erlotinib, and 1 patient (diagnosed with kidney NOS) showed rash or facial swelling/erythema and diffuse body itching then stable disease after 2 cycles.

One patient (diagnosed with cancer located at the pancreas head) was bridged with capecitabine and erlotinib when going on a vacation, then received FOLFIRINOX (a combination chemotherapy regimen containing folinic acid [leucovorin], fluorouracil, irinotecan, and oxaliplatin), which led to significant fatigue, blurry vision, and constipation. One patient was treated for lung NOS with the MEK-inhibitor selumetinib plus erlotinib and developed pneumonitis following treatment.

Because oncologists followed guidelines and protocols in systemic treatment, DDIs of erlotinib concurrently (before or after) and in combination with cancer drugs were unlikely. Further investigation is needed for several 1:1:1 DDIs with noncancer drugs. A retrospective overview is not a randomized clinical study; therefore, analysis is limited. Data from the MHS were obtained solely from notes from physicians who treated the patients; therefore, exact information explaining whether a patient completed treatment or had to withdraw could not be extrapolated (ie, blood/plasma samples were not obtained to confirm).

Discontinued Treatment

The reasons for treatment discontinuation with erlotinib or gefitinib varied among patients, with no consistent AE or cause. Most data were for switching treatments after discontinuing treatment with erlotinib (101 of 123 patients) and gefitinib (2 of 5 patients). This is not surprising given the widely recognized pillars of therapy for NSCLC: chemotherapy, target therapy, and immunotherapy.²² From the MHS records, the reasons patients switched treatment of erlotinib or gefitinib were not listed or listed as due to negative EGFR testing, lack of responsiveness, or enrollment in a different treatment.

 

Physicians’ notes on AEs were not detailed in most cases. Notes for gastrointestinal effects, life-altering pruritis, intolerance, peripheral vascular disease, pneumonitis, and progressive disease described the change in status or appearance of a new medical condition but did not indicate whether erlotinib or gefitinib caused the changes or worsened a pre-existing condition.

The causes of AEs were not described in the available notes or the databases. This retrospective data analysis only focused on identifying drugs involved with erlotinib and gefitinib treatment; further mapping of DDIs among patients experiencing AEs needs to be performed, then in vitro data testing before researchers can reach a conclusion.

DDIs With Antidepressants

We used the PDTS database to evaluate patients who experienced AEs, excluding patients who switched treatment. Thirteen patients filled a prescription for erlotinib and reported taking 220 cancer and noncancer prescription drugs. One patient (pruritis) was taking gefitinib along with 16 noncancer prescription drugs.

Table 4 details CYP information for cancer drugs, antidepressants, and noncancer drugs (top 11 drugs) among patients who recorded AEs with erlotinib.^3-4,23-47

Selective serotonin reuptake inhibitors and other antidepressants have been implicated in CYP 2D6 inhibition and DDIs.^48,49 Losartan is a widely used antihypertensive drug with a favorable DDI profile.⁵⁰ Erlotinib and gefitinib are primarily metabolized via CYP 2D6 and 3A4 pathways. DDIs from in vitro human hepatocytes assays revealed that gefitinib had significant metabolic changes in a 1:1 (P < .05) combination with paroxetine or sertraline, and a 1:1:1 combination with losartan and fluoxetine, fluvoxamine, paroxetine, or sertraline. Citalopram and venlafaxine seemed to be unaffected by any combination (P ≥ .05).⁵¹ Erlotinib with fluoxetine or losartan 1:1 yielded insignificant differences in metabolism for all drugs (P ≥ .05). Three drug combinations of 1:1:1 involving fluoxetine and losartan with erlotinib yielded significant degrees of inhibition of fluoxetine and losartan metabolism (P < .05) but not erlotinib.⁵²

Our data showed that 16 antidepressants (amitriptyline, bupropion, citalopram, desvenlafaxine, duloxetine, escitalopram, imipramine, fluoxetine, fluvoxamine, mirtazapine, nortriptyline, paroxetine, phenelzine, sertraline, trazodone, and venlafaxine) were recorded with concomitant erlotinib or gefitinib from initiation to completion of therapy or a buffer of 6 months from the first diagnosis date. Based on the date dispensed and days’ supply, only escitalopram could be used in combination with gefitinib treatment. The one patient who filled a prescription for gefitinib and escitalopram completed treatment without recorded AEs. PDTS database confirmed that patients experienced AEs with 5 antidepressants (amitriptyline, mirtazapine, paroxetine, trazodone, and venlafaxine) with concomitant erlotinib use.

Based on the date dispensed and days’ supply, only trazodone could be used in combination with erlotinib. PDTS database showed that cancer drugs (erlotinib and megestrol) and 39 noncancer drugs (including acetaminophen, azithromycin, dexamethasone, hydrocortisone, and polyethylene glycol) were filled by 1 patient whose physician noted skin rash. Another limitation of using databases to reflect clinical practice is that although megestrol is listed as a cancer drug by code in the PDTS database, it is not used for nonendometrial or gynecologic cancers. However, because of the PDTS database classification, megestrol is classified as a cancer drug in this retrospective review.

This retrospective review found no significant DDIs for erlotinib or gefitinib, with 1 antidepressant taken by 1 patient for each respective treatment. The degree of inhibition and induction for escitalopram and trazodone are categorized as weak, minimal, or none; therefore, while 1:1 DDIs might be little or no effect, 1:1:1 combination DDIs could have a different outcome. This retrospective data collection cannot be linked to the in­ vitro hepatocyte DDIs from erlotinib and gefitinib in previous studies.^51,52

 

Conclusions

This retrospective study describes erlotinib and gefitinib use in the MHS and their potential for DDIs. Because of military service requirements, people who are qualified to serve must be healthy or have either controlled or nonactive medical diagnoses and be physically fit. Consequently, our patient population had fewer common medical illnesses, such as diabetes and obesity, compared with the general population. Most noncancer drugs mentioned in this study are not known CYP metabolizers; therefore, recorded AEs alone cannot conclusively determine whether there is a DDI among erlotinib or gefitinib and noncancer drugs. Antidepressants generally are safe but have boxed warnings in the US for increased risk of suicidal ideation in young people.^53,54 This retrospective study did not find statistically significant DDIs for erlotinib or gefitinib with antidepressants. Based on this retrospective data analysis, future in vitro testing is needed to assess DDIs for erlotinib or gefitinib and cancer or noncancer drugs identified in this study.

Acknowledgments

The Department of Research Program funds at Walter Reed National Military Medical Center supported this protocol. We sincerely appreciate the contribution of data extraction from the Joint Pathology Center teams (Francisco J. Rentas, John D. McGeeney, Kimberly M. Greenfield, Beatriz A. Hallo, and Johnny P. Beason) and the MHS database personnel (Maj Ryan Costantino, Lee Ann Zarzabal, Brandon Jenkins, and Alex Rittel). We gratefully thank you for the protocol support from the Department of Research programs: CDR Wesley R. Campbell, CDR Ling Ye, Yaling Zhou, Elizabeth Schafer, Robert Roogow, Micah Stretch, Diane Beaner, Adrienne Woodard, David L. Evers, and Paula Amann.

Most cancer treatment regimens include the administration of several chemotherapeutic agents. Drug-drug interactions (DDIs) can increase the risk of fatal adverse events and reduce therapeutic efficacy.^1,2 Erlotinib, gefitinib, afatinib, osimertinib, and icotinib are epidermal growth factor receptor–tyrosine kinase inhibitors (EGFR-TKIs) that have proven efficacy for treating advanced non–small cell lung cancer (NSCLC). Erlotinib strongly inhibits cytochrome P450 (CYP) isoenzymes CYP 1A1, moderately inhibits CYP 3A4 and 2C8, and induces CYP 1A1 and 1A2.² Gefitinib weakly inhibits CYP 2C19 and 2D6.² CYP 3A4 inducers and inhibitors affect metabolism of both erlotinib and gefitinib.^3,4

Erlotinib and gefitinib are first-generation EGFR-TKIs and have been approved for NSCLC treatment by the US Food and Drug Administration (FDA). These agents have been used since the early 2000s and increase the possibility of long-term response and survival.^2,5,6 EGFR-TKIs have a range of potential DDIs, including interactions with CYP-dependent metabolism, uridine diphosphate-glucuronosyltransferase, and transporter proteins.² Few retrospective studies have focused on the therapeutic efficacy of erlotinib, gefitinib, or the combination of these agents.^7-14

DDIs from cancer and noncancer therapies could lead to treatment discontinuation and affect patient outcomes. The goals for this study were to perform a broad-scale retrospective analysis focused on investigating prescribed drugs used with erlotinib and gefitinib and determine patient outcomes as obtained through several Military Health System (MHS) databases. Our investigation focused on (1) the functions of these drugs; (2) identifying adverse effects (AEs) that patients experienced; (3) evaluating differences when these drugs are used alone vs concomitantly, and between the completed vs discontinued treatment groups; (4) identifying all drugs used during erlotinib or gefitinib treatment; and (5) evaluating DDIs with antidepressants.

This retrospective study was performed at the Department of Research Programs at Walter Reed National Military Medical Center (WRNMMC) in Bethesda, Maryland. The WRNMMC Institutional Review Board approved the study protocol and ensured compliance with the Health Insurance Portability and Accountability Act as an exempt protocol. The Joint Pathology Center of the US Department of Defense (DoD) Cancer Registry and MHS data experts from the Comprehensive Ambulatory/Professional Encounter Record (CAPER) and the Pharmacy Data Transaction Service (PDTS) provided data for the analysis.

 

Methods

The DoD Cancer Registry Program was established in 1986 by the Assistant Secretary of Defense for Health Affairs. The registry currently contains data from 1998 to 2023. CAPER and PDTS are part of the MHS Data Repository/Management Analysis and Reporting Tool database. Each observation in the CAPER record represents an ambulatory encounter at a military treatment facility (MTF). CAPER records are available from 2003 to 2023.

Each observation in the PDTS record represents an outpatient prescription filled for an MHS beneficiary at MTFs through the TRICARE mail-order program or a retail pharmacy in the United States. Missing from this record are prescriptions filled at civilian pharmacies outside the United States and inpatient pharmacy prescriptions. The MHS Data Repository PDTS record is available from 2002 to 2023. The Composite Health Care System—the legacy system—is being replaced by GENESIS at MTFs.

Data Extraction Design

The study design involved a cross-sectional analysis. We requested data extraction for erlotinib and gefitinib from 1998 to 2021. Data from the DoD Cancer Registry were used to identify patients who received cancer treatment. Once patients were identified, the CAPER database was searched for diagnoses to identify other health conditions, while the PDTS database was used to populate a list of prescription medications filled during chemotherapy treatment.

Data collected from the Joint Pathology Center included cancer treatment (alone or concomitant), cancer information (cancer types and stages), demographics (sex, age at diagnosis), and physicians’ comments on AEs. Collected data from the MHS include diagnosis and filled prescription history from initiation to completion of the therapy period (or a buffer of 6 months after the initial period). We used all collected data in this analysis. The only exclusion criterion was a provided physician’s note commenting that the patient did not use erlotinib or gefitinib.

Data Extraction Analysis

The Surveillance, Epidemiology, and End Results Program Coding and Staging Manual 2016 and the International Classification of Diseases for Oncology (ICD-O) were used to decode disease and cancer types.^15,16 Data sorting and analysis were performed using Microsoft Excel. The percentage for the total was calculated by using the total number of patients or data available within the gefitinib and erlotinib groups divided by total number of patients or data variables. The subgroup percentage was calculated by using the number of patients or data available within the subgroup divided by the total number of patients in that subgroup.

In alone vs concomitant and completed vs discontinued treatment groups, a 2-tailed, 2-sample z test was used to calculate P to determine statistical significance (P < .05) using a statistics website.¹⁷ Concomitant was defined as erlotinib or gefitinib taken with other medication(s) before, after, or at the same time as cancer therapy. For the retrospective data analysis, physicians’ notes with “.”, “,”, “/”, “;”, (period, comma, forward slash, semicolon) or space between medication names were interpreted as concurrent, while “+”, “-/+” (plus, minus/plus), or and between drug names were interpreted as combined. Completed treatment was defined as erlotinib or gefitinib as the last medication the patient took without recorded AEs; switching or experiencing AEs was defined as discontinued treatment.

 

Results

Erlotinib

The Joint Pathology Center provided 387 entries for 382 patients aged 21 to 93 years (mean, 65 years) who were treated systemically with erlotinib from January 1, 2001, to December 31, 2020. Five patients had duplicate entries because they had different cancer sites. There were 287 patients (74%) with lung cancer, 61 (16%) with pancreatic cancer, and 39 (10%) with other cancers. For lung cancer, there were 118 patients (30%) for the upper lobe, 78 (20%) for the lower lobe, and 60 (16%) not otherwise specified (NOS). Other lung cancer sites had fewer patients: 21 (5%) middle lobe lung, 6 (2%) overlapping lung lesion(s), and 4 (1%) main bronchus of the lung. For pancreatic cancer, there were 27 patients (7%) for the head of the pancreas, 10 (3%) pancreas NOS, 9 (2%) body of the pancreas, 9 (2%) tail of the pancreas, 4 (1%) overlapping lesions of the pancreas, 1 (< 1%) pancreatic duct, and 1 (< 1%) other specified parts of the pancreas

. Thirty-nine patients (10%) received erlotinib for indications that were not for FDA-approved indications, which included 9 (2%) for kidney NOS, 8 (2%) for the unknown primary site, 5 (1%) for liver cancer, 2 (1%) for intrahepatic bile duct, 2 (1%) for tonsil, and 1 (< 1%) for 13 disease sites (Table 1).

There were 342 patients (88%) who were aged > 50 years; 186 male patients (48%) and 201 female patients (52%). There were 293 patients (76%) who had a cancer diagnosis of stage III or IV disease and 94 (24%) who had a cancer diagnosis of stage ≤ II (combination of data for stage 0, 1, and 2, not applicable, and unknown). For their systemic treatment, 161 patients (42%) were treated with erlotinib alone and 226 (58%) received erlotinib concomitantly with additional chemotherapy.

Of these patients, 287 (74%) were diagnosed with lung cancer (Table 2).

Patients were more likely to discontinue erlotinib for chemotherapy if they received concomitant treatment. Among the patients receiving erlotinib monotherapy, 5% stopped the treatment, whereas 51% of patients treated concomitantly discontinued (P < .001).

The comparisons for lung cancer vs other cancer and those aged ≤ 50 years vs > 50 years were significant (P = .005 and .05, respectively) while other comparisons were not significant (Table 3).

Among the 123 patients who discontinued their treatment, 101 switched treatment with no AEs notes, 22 died or experienced fatigue with blurry vision, constipation, nonspecific gastrointestinal effects, grade-4 diarrhea (as defined by the Common Terminology Criteria for Adverse Events), or developed a pleural fluid, pneumonitis, renal failure, skin swelling and facial rash, and unknown AEs of discontinuation. Patients who discontinued treatment because of unknown AEs had physicians’ notes that detailed emergency department visits, peripheral vascular disease, progressive disease, and treatment cessation, but did not specify the exact symptom(s) that led to discontinuation. The causes of death are unknown because they were not detailed in the available notes or databases. The overall results in this retrospective review cannot establish causality between taking erlotinib or gefitinib and death.

 

Gefitinib

In September 2021, the Joint Pathology Center provided 33 entries for 33 patients who were systemically treated with gefitinib from January 1, 2002, to December 31, 2017. The patient ages ranged from 49 to 89 years with a mean age of 66 years. There were 31 (94%) and 2 (6%) patients with lung and other cancers, respectively. The upper lobe, lower lobe, and lung NOS had the most patients: 14 (42%), 8 (24%), and 6 (18%), respectively.

There were 31 patients (94%) who were aged > 50 years; 15 were male (45%) and 18 were female (55%). There were 26 patients (79%) who had a cancer diagnosis of stage III or IV disease. Nineteen patients (58%) were treated with gefitinib alone, and 14 (42%) were treated with gefitinib concomitantly with additional chemotherapy. Thirty-one patients (94%) were treated for lung cancer (Table 2). Thirty-three patients are a small sample size to determine whether patients were likely to stop gefitinib if used concomitantly with other drugs. Among the patients treated with gefitinib monotherapy, 5% (n = 1) stopped treatment, whereas 29% (n = 4) of patients treated concomitantly discontinued treatment (P = .06). All comparisons for gefitinib yielded insignificant P values. Physicians’ notes indicated that the reasons for gefitinib discontinuation were life-altering pruritis and unknown (progressive disease outcome) (Table 3).

Management Analysis and Reporting Tool Database

MHS data analysts provided data on diagnoses for 348 patients among 415 submitted, with 232 and 112 patients completing and discontinuing erlotinib or gefitinib treatment, respectively. Each patient had 1 to 104 (completed treatment group) and 1 to 157 (discontinued treatment group) unique health conditions documented. The MHS reported 1319 unique-diagnosis conditions for the completed group and 1266 for the discontinued group. Patients with additional health issues stopped chemotherapy use more often than those without; P < .001 for the completed group (232 patients, 1319 diagnoses) vs the discontinued group (112 patients, 1266 diagnoses). The mean (SD) number of diagnoses was 19 (17) for the completed and 30 (22) for the discontinued treatment groups (Figure).

The 5 most recorded diagnoses with erlotinib among 358 patients were malignant neoplasm of bronchus and lung for 225 patients, unspecified essential hypertension for 120 patients, encounters for antineoplastic chemotherapy for 113 patients, dietary surveillance and counseling for 102 patients, and unspecified administrative purposes for 97 patients.

MHS data was provided for patients who filled erlotinib (n = 240) or gefitinib (n = 18). Among the 258 patients, there were 179 and 79 patients in the completed and discontinued treatment groups, respectively. Each patient filled 1 to 75 (for the completed treatment group) and 3 to 103 (for the discontinued treatment group) prescription drugs. There were 805 unique-filled prescriptions for the completed and 670 for the discontinued group. Patients in the discontinued group filled more prescriptions than those who completed treatment; P < .001 for the completed group (179 patients,805 drugs) vs the discontinued group (79 patients, 670 drugs).

The mean (SD) number of filled prescription drugs was 19 (11) for the completed group and 29 (18) for the discontinued treatment group. The 5 most filled prescriptions with erlotinib from 258 patients with PDTS data were ondansetron (151 prescriptions, 10 recorded AEs), dexamethasone (119 prescriptions, 9 recorded AEs), prochlorperazine (105 prescriptions, 15 recorded AEs), oxycodone (99 prescriptions, 1 AE), and docusate (96 prescriptions, 7 recorded AEs).

 

Discussion

The difference between erlotinib and gefitinib data can be attributed to the FDA approval date and gefitinib’s association with a higher frequency of hepatotoxicity.^18-20 The FDA designated gefitinib as an orphan drug for EGFR mutation–positive NSCLC treatment. Gefitinib first received accelerated approval in 2003 for the treatment of locally advanced or metastatic NSCLC. Gefitinib then was voluntarily withdrawn from the market following confirmatory clinical trials that did not verify clinical benefit.

The current approval is for a different patient population—previously untreated, metastatic EGFR exon 19 or 21 L858R mutation—than the 2003 approval.^4,6 There was no record of gefitinib use after 2017 in our study.

Erlotinib is a reversible EGFR-TKI that is approved by the FDA as first-line (maintenance) or second-line treatment (after progression following at least 1 earlier chemotherapy regimen) for patients with metastatic NSCLC who harbor EGFR exon 19 deletions or exon 21 L858R substitution mutations, as detected by an FDA-approved test.³ Since 2005, the FDA also approved erlotinib for first-line treatment of patients with locally advanced, unresectable, or metastatic pancreatic cancer in combination with gemcitabine.³ Without FDA indication, erlotinib is used for colorectal, head and neck, ovarian carcinoma, pancreatic carcinoma, and breast cancer.²¹

Erlotinib and gefitinib are not considered first-line treatments in EGFR exon 19 or 21–mutated NSCLC because osimertinib was approved in 2018. Targeted therapies for EGFR mutation continue to advance at a fast pace, with amivantamab and mobocertinib now FDA approved for EGFR exon 20 insertion–mutated NSCLC.

Erlotinib Use

Thirty-nine patients (10%) in this study were prescribed erlotinib for off-label indications. Erlotinib was used alone or in combination with bevacizumab, capecitabine, cisplatin, denosumab, docetaxel, gemcitabine, and the MEK-inhibitor selumetinib. Erlotinib combined with cisplatin, denosumab, docetaxel, and gemcitabine had no recorded AEs, with 10 data entries for gemcitabine and 1 for other drugs. Three patients received bevacizumab and erlotinib, and 1 patient (diagnosed with kidney NOS) showed rash or facial swelling/erythema and diffuse body itching then stable disease after 2 cycles.

One patient (diagnosed with cancer located at the pancreas head) was bridged with capecitabine and erlotinib when going on a vacation, then received FOLFIRINOX (a combination chemotherapy regimen containing folinic acid [leucovorin], fluorouracil, irinotecan, and oxaliplatin), which led to significant fatigue, blurry vision, and constipation. One patient was treated for lung NOS with the MEK-inhibitor selumetinib plus erlotinib and developed pneumonitis following treatment.

Because oncologists followed guidelines and protocols in systemic treatment, DDIs of erlotinib concurrently (before or after) and in combination with cancer drugs were unlikely. Further investigation is needed for several 1:1:1 DDIs with noncancer drugs. A retrospective overview is not a randomized clinical study; therefore, analysis is limited. Data from the MHS were obtained solely from notes from physicians who treated the patients; therefore, exact information explaining whether a patient completed treatment or had to withdraw could not be extrapolated (ie, blood/plasma samples were not obtained to confirm).

Discontinued Treatment

The reasons for treatment discontinuation with erlotinib or gefitinib varied among patients, with no consistent AE or cause. Most data were for switching treatments after discontinuing treatment with erlotinib (101 of 123 patients) and gefitinib (2 of 5 patients). This is not surprising given the widely recognized pillars of therapy for NSCLC: chemotherapy, target therapy, and immunotherapy.²² From the MHS records, the reasons patients switched treatment of erlotinib or gefitinib were not listed or listed as due to negative EGFR testing, lack of responsiveness, or enrollment in a different treatment.

 

Physicians’ notes on AEs were not detailed in most cases. Notes for gastrointestinal effects, life-altering pruritis, intolerance, peripheral vascular disease, pneumonitis, and progressive disease described the change in status or appearance of a new medical condition but did not indicate whether erlotinib or gefitinib caused the changes or worsened a pre-existing condition.

The causes of AEs were not described in the available notes or the databases. This retrospective data analysis only focused on identifying drugs involved with erlotinib and gefitinib treatment; further mapping of DDIs among patients experiencing AEs needs to be performed, then in vitro data testing before researchers can reach a conclusion.

DDIs With Antidepressants

We used the PDTS database to evaluate patients who experienced AEs, excluding patients who switched treatment. Thirteen patients filled a prescription for erlotinib and reported taking 220 cancer and noncancer prescription drugs. One patient (pruritis) was taking gefitinib along with 16 noncancer prescription drugs.

Table 4 details CYP information for cancer drugs, antidepressants, and noncancer drugs (top 11 drugs) among patients who recorded AEs with erlotinib.^3-4,23-47

Selective serotonin reuptake inhibitors and other antidepressants have been implicated in CYP 2D6 inhibition and DDIs.^48,49 Losartan is a widely used antihypertensive drug with a favorable DDI profile.⁵⁰ Erlotinib and gefitinib are primarily metabolized via CYP 2D6 and 3A4 pathways. DDIs from in vitro human hepatocytes assays revealed that gefitinib had significant metabolic changes in a 1:1 (P < .05) combination with paroxetine or sertraline, and a 1:1:1 combination with losartan and fluoxetine, fluvoxamine, paroxetine, or sertraline. Citalopram and venlafaxine seemed to be unaffected by any combination (P ≥ .05).⁵¹ Erlotinib with fluoxetine or losartan 1:1 yielded insignificant differences in metabolism for all drugs (P ≥ .05). Three drug combinations of 1:1:1 involving fluoxetine and losartan with erlotinib yielded significant degrees of inhibition of fluoxetine and losartan metabolism (P < .05) but not erlotinib.⁵²

Our data showed that 16 antidepressants (amitriptyline, bupropion, citalopram, desvenlafaxine, duloxetine, escitalopram, imipramine, fluoxetine, fluvoxamine, mirtazapine, nortriptyline, paroxetine, phenelzine, sertraline, trazodone, and venlafaxine) were recorded with concomitant erlotinib or gefitinib from initiation to completion of therapy or a buffer of 6 months from the first diagnosis date. Based on the date dispensed and days’ supply, only escitalopram could be used in combination with gefitinib treatment. The one patient who filled a prescription for gefitinib and escitalopram completed treatment without recorded AEs. PDTS database confirmed that patients experienced AEs with 5 antidepressants (amitriptyline, mirtazapine, paroxetine, trazodone, and venlafaxine) with concomitant erlotinib use.

Based on the date dispensed and days’ supply, only trazodone could be used in combination with erlotinib. PDTS database showed that cancer drugs (erlotinib and megestrol) and 39 noncancer drugs (including acetaminophen, azithromycin, dexamethasone, hydrocortisone, and polyethylene glycol) were filled by 1 patient whose physician noted skin rash. Another limitation of using databases to reflect clinical practice is that although megestrol is listed as a cancer drug by code in the PDTS database, it is not used for nonendometrial or gynecologic cancers. However, because of the PDTS database classification, megestrol is classified as a cancer drug in this retrospective review.

This retrospective review found no significant DDIs for erlotinib or gefitinib, with 1 antidepressant taken by 1 patient for each respective treatment. The degree of inhibition and induction for escitalopram and trazodone are categorized as weak, minimal, or none; therefore, while 1:1 DDIs might be little or no effect, 1:1:1 combination DDIs could have a different outcome. This retrospective data collection cannot be linked to the in­ vitro hepatocyte DDIs from erlotinib and gefitinib in previous studies.^51,52

 

Conclusions

This retrospective study describes erlotinib and gefitinib use in the MHS and their potential for DDIs. Because of military service requirements, people who are qualified to serve must be healthy or have either controlled or nonactive medical diagnoses and be physically fit. Consequently, our patient population had fewer common medical illnesses, such as diabetes and obesity, compared with the general population. Most noncancer drugs mentioned in this study are not known CYP metabolizers; therefore, recorded AEs alone cannot conclusively determine whether there is a DDI among erlotinib or gefitinib and noncancer drugs. Antidepressants generally are safe but have boxed warnings in the US for increased risk of suicidal ideation in young people.^53,54 This retrospective study did not find statistically significant DDIs for erlotinib or gefitinib with antidepressants. Based on this retrospective data analysis, future in vitro testing is needed to assess DDIs for erlotinib or gefitinib and cancer or noncancer drugs identified in this study.

Acknowledgments

The Department of Research Program funds at Walter Reed National Military Medical Center supported this protocol. We sincerely appreciate the contribution of data extraction from the Joint Pathology Center teams (Francisco J. Rentas, John D. McGeeney, Kimberly M. Greenfield, Beatriz A. Hallo, and Johnny P. Beason) and the MHS database personnel (Maj Ryan Costantino, Lee Ann Zarzabal, Brandon Jenkins, and Alex Rittel). We gratefully thank you for the protocol support from the Department of Research programs: CDR Wesley R. Campbell, CDR Ling Ye, Yaling Zhou, Elizabeth Schafer, Robert Roogow, Micah Stretch, Diane Beaner, Adrienne Woodard, David L. Evers, and Paula Amann.

The United States’ initial public health response to the COVID-19 pandemic included containment measures that varied by state but generally required closing or suspending schools, nonessential businesses, and travel (commonly called lockdown).¹ During these periods, hospitalizations for serious and common conditions declined.^2,3 In Massachusetts, a state of emergency was declared on March 10, 2020, which remained in place until May 18, 2020, when a phased reopening of businesses began.

Although the evidence on the mental health impact of containment periods has been mixed, it has been suggested that these measures could lead to increases in alcohol-related hospitalizations.⁴ Social isolation and increased psychosocial and financial stressors raise the risk of relapse among patients with substance use disorders.^5-7 Marketing and survey data from the US and United Kingdom from the early months of the pandemic suggest that in-home alcohol consumption and sales of alcoholic beverages increased, while consumption of alcohol outside the home decreased.^8-10 Other research has shown an increase in the percentage—but not necessarily the absolute number—of emergency department (ED) visits and hospitalizations for alcohol-related diagnoses during periods of containment.^11,12 At least 1 study suggests that alcohol-related deaths increased beginning in the lockdown period and persisting into mid-2021.¹³

Because earlier studies suggest that lockdown periods are associated with increased alcohol consumption and relapse of alcohol use disorder, we hypothesized that the spring 2020 lockdown period in Massachusetts would be associated temporally with an increase in alcohol-related hospitalizations. To evaluate this hypothesis, we examined all hospitalizations in the US Department of Veterans Affairs (VA) Boston Healthcare System (VABHS) before, during, and after this lockdown period. VABHS includes a 160-bed acute care hospital and a 50-bed inpatient psychiatric facility.

 

Methods

We conducted an interrupted time-series analysis including all inpatient hospitalizations at VABHS from January 1, 2017, to December 31, 2020, to compare the daily number of alcohol-related hospitalizations across 3 exposure groups: prelockdown (the reference group, 1/1/2017-3/9/2020); lockdown (3/10/2020-5/18/2020); and postlockdown (5/19/2020-12/31/2020).

The VA Corporate Data Warehouse at VABHS was queried to identify all hospitalizations on the medical, psychiatry, and neurology services during the study period. Hospitalizations were considered alcohol-related if the International Statistical Classification of Diseases, Tenth Revision (ICD-10) primary diagnosis code (the main reason for hospitalization) was defined as an alcohol-related diagnosis by the VA Centralized Interactive Phenomics Resource (eAppendix 1, available online at doi:10.1278/fp.0404). This database, which has been previously used for COVID-19 research, is a catalog and knowledge-sharing platform of VA electronic health record–based phenotype algorithms, definitions, and metadata that builds on the Million Veteran Program and Cooperative Studies Program.^14,15 Hospitalizations under observation status were excluded.

To examine whether alcohol-related hospitalizations could have been categorized as COVID-19 when the conditions were co-occurring, we identified 244 hospitalizations coded with a primary ICD-10 code for COVID-19 during the lockdown and postlockdown periods. At the time of admission, each hospitalization carries an initial (free text) diagnosis, of which 3 had an initial diagnosis related to alcohol use. The population at risk for alcohol-related hospitalizations was estimated as the number of patients actively engaged in care at the VABHS. This was defined as the number of patients enrolled in VA care who have previously received any VA care; patients who are enrolled but have never received VA care were excluded from the population-at-risk denominator. Population-at-risk data were available for each fiscal year (FY) of the study period (9/30-10/1); the following population-at-risk sizes were used: 38,057 for FY 2017, 38,527 for FY 2018, 39,472 for FY 2019, and 37,893 for FY 2020.

The primary outcome was the daily number of alcohol-related hospitalizations in the prelockdown, lockdown, and postlockdown periods. A sensitivity analysis was performed using an alternate definition of the primary outcome using a broader set of alcohol-related ICD-10 codes (eAppendix 2, available online at doi:10.1278/fp.0404).

Statistical Analysis

To visually examine hospitalization trends during the study period, we generated a smoothed time-series plot of the 7-day moving average of the daily number of all-cause hospitalizations and the daily number of alcohol-related hospitalizations from January 1, 2017, to December 31, 2020. We used multivariable regression to model the daily number of alcohol-related hospitalizations over prelockdown (the reference group), lockdown, and postlockdown. In addition to the exposure, we included the following covariates in our model: day of the week, calendar date (to account for secular trends), and harmonic polynomials of the day of the year (to account for seasonal variation).¹⁶

We also examined models that included the daily total number of hospitalizations to account for the reduced likelihood of hospital admission for any reason during the pandemic. We used generalized linear models with a Poisson link to generate rate ratios and corresponding 95% CIs for estimates of the daily number of alcohol-related hospitalizations. We estimated the population incidence of alcohol-related hospitalizations per 100,000 patient-months for the exposure periods using the population denominators previously described. All analyses were performed in Stata 16.1.

 

Results

During the study period, 27,508 hospitalizations were available for analysis. The 7-day moving average of total daily hospitalizations and total daily alcohol-related hospitalizations over time for the period January 1, 2017, to December 31, 2020, are shown in the Figure.

Compared with the prelockdown period, the 7-day average of hospitalizations per day for all hospitalizations and alcohol-related hospitalizations decreased substantially during the lockdown and did not return to the prelockdown baseline during the postlockdown period.

The incidence of alcohol-related hospitalizations in the population dropped from 72 per 100,000 patient-months to 10 per 100,000 patient-months during the lockdown period and increased to 46 per 100,000 patient-months during the postlockdown period (Table).

Compared with the 3-year prelockdown period, the rate ratio for daily alcohol-related hospitalizations during the lockdown period decreased to 0.20 (95% CI, 0.10-0.39). In the postlockdown period, the rate ratio for daily alcohol-related hospitalizations increased, but to only 0.72 (95% CI, 0.57-0.92) compared with the prepandemic baseline.

Our results were not substantially different when we ran a sensitivity analysis that excluded the total daily number of admissions from our model. Compared with the prelockdown period, the rate ratio for the number of alcohol-related hospitalizations during the lockdown period was 0.16 (95% CI, 0.08-0.30), and the rate ratio for the postlockdown period was 0.65 (95% CI, 0.52-0.82). We conducted an additional sensitivity analysis using a broader definition of the primary outcome to include all alcohol-related diagnosis codes; however, the results were unchanged.

Discussion

During the spring 2020 COVID-19 lockdown period in Massachusetts, the daily number of VABHS alcohol-related hospitalizations decreased by nearly 80% compared with the prelockdown period. During the postlockdown period, the daily number of alcohol-related hospitalizations increased but only to 72% of the prelockdown baseline by the end of December 2020. A similar trend was observed for all-cause hospitalizations for the same exposure periods.

These results differ from 2 related studies on the effect of the COVID-19 pandemic on alcohol-related hospitalizations.^10,11 In a retrospective study of ED visits to 4 hospitals in New York City, Schimmel and colleagues reported that from March 1 to 31, 2020 (the initial COVID-19 peak), hospital visits for alcohol withdrawal increased while those for alcohol use decreased.¹⁰ However, these results are reported as a percentage of total ED visits rather than the total number of visits, which are vulnerable to spurious correlation because of concomitant changes in the total number of ED visits. In their study, the absolute number of alcohol-related ED visits did not increase during the initial 2020 COVID-19 peak, and the number of visits for alcohol withdrawal syndrome declined slightly (195 in 2019 and 180 in 2020). However, the percentage of visits increased from 7% to 10% because of a greater decline in total ED visits. This pattern of decline in the number of alcohol-related ED visits, accompanied by an increase in the percentage of alcohol-related ED visits, has been observed in at least 1 nationwide surveillance study.¹⁷ This apparent increase does not reflect an absolute increase in ED visits for alcohol withdrawal syndrome and represents a greater relative decline in visits for other causes during the study period.

Sharma and colleagues reported an increase in the percentage of patients who developed alcohol withdrawal syndrome while hospitalized in Delaware per 1000 hospitalizations during consecutive 2-week periods during the pandemic in 2020 compared with corresponding weeks in 2019.¹¹ The greatest increase occurred during the last 2 weeks of the Delaware stay-at-home order. The Clinical Institute Withdrawal Assessment of Alcohol Scale, revised (CIWA-Ar) score of > 8 was used to define alcohol withdrawal syndrome. The American Society of Addiction Medicine does not recommend using CIWA-Ar to diagnose alcohol withdrawal syndrome because the scale was developed to monitor response to treatment, not to establish a diagnosis.¹⁸

Although the true population incidence of alcohol-related hospitalizations is difficult to estimate because the size of the population at risk (ie, the denominator) often is not known, the total number of hospitalizations is not a reliable surrogate.¹⁹ Individuals hospitalized for nonalcohol causes are no longer at risk for alcohol-related hospitalization.

In our study, we assume the population at risk during the study period is constant and model changes in the absolute number—rather than percentage—of alcohol-related ED visits. These absolute estimates of alcohol-related hospitalizations better reflect the true burden on the health care system and avoid the confounding effect of declining total ED visits and hospitalizations that could lead to artificially increased percentages and spurious correlation.²⁰ The absolute percentage of alcohol-related hospitalizations also decreased during this period; therefore, our results are not sensitive to this approach.

Several factors could have contributed to the decrease in alcohol-related hospitalizations. Our findings suggest that patient likelihood to seek care and clinician threshold to admit patients for alcohol-related conditions are influenced by external factors, in this case, a public health lockdown. Although our data do not inform why hospitalizations did not return to prelockdown levels, our experience suggests that limited bed capacity and longer length of stay might have contributed. Other hypotheses include a shift to outpatient care, increased use of telehealth (a significant focus early in the pandemic), and avoiding care for less severe alcohol-related complications because of lingering concerns about exposure to COVID-19 in health care settings reported early in the pandemic. Massachusetts experienced a particularly deadly outbreak of COVID-19 in the Soldiers’ Home, a long-term care facility for veterans in Holyoke.²¹

Evidence suggests that in-home consumption of alcohol increased during lockdowns.^8-10 Our results show that during this period hospitalizations for alcohol-related conditions decreased at VABHS, a large urban VA medical system, while alcohol-related deaths increased nationally.¹³ Although this observation is not evidence of causality, these outcomes could be related.

In the 2 decades before the pandemic, alcohol-related deaths increased by about 2% per year.²² From 2019 to 2020, there was a 25% increase that continued through 2021.¹³ Death certificate data often are inaccurate, and it is difficult to determine whether COVID-19 had a substantial contributing role to these deaths, particularly during the initial period when testing was limited or unavailable. Nonetheless, deaths due to alcohol-associated liver disease, overdoses involving alcohol, and alcohol-related traffic fatalities increased by > 10%.^13,23 These trends, along with a decrease in hospitalization for alcohol-related conditions, suggest missed opportunities for intervention with patients experiencing alcohol use disorder.

 

Limitations

In this study, hospitalizations under observation status were excluded, which could underestimate the total number of hospitalizations related to alcohol. We reasoned that this effect was likely to be small and not substantially different by year. ICD-10 codes were used to identify alcohol-related hospitalizations as any hospitalization with an included ICD-10 code listed as the primary discharge diagnosis code. This also likely underestimated the total number of alcohol-related hospitalizations. An ICD-10 code for COVID-19 was not in widespread use during our study period, which prohibited controlling explicitly for the volume of admissions due to COVID-19. The prelockdown period only contains data from the preceding 3 years, which might not be long enough for secular trends to become apparent. We assumed the population at risk remained constant when in reality, the net movement of patients into and out of VA care during the pandemic likely was more complex but not readily quantifiable. Nonetheless, the large drop in absolute number of alcohol-related hospitalizations is not likely to be sensitive to this change. In the absence of an objective measure of care-seeking behavior, we used the total daily number of hospitalizations as a surrogate for patient propensity to seek care. The total daily number of hospitalizations also reflects changes in physician admitting behavior over time. This allowed explicit modeling of care-seeking behavior as a covariate but does not capture other important determinants such as hospital capacity.

Conclusions

In this interrupted time-series analysis, the daily number of alcohol-related hospitalizations during the initial COVID-19 pandemic–associated lockdown period at VABHS decreased by 80% and remained 28% lower in the postlockdown period compared with the prepandemic baseline. In the context of evidence suggesting that alcohol-related mortality increased during the COVID-19 pandemic, alternate strategies to reach vulnerable individuals are needed. Because of high rates of relapse, hospitalization is an important opportunity to engage patients experiencing alcohol use disorder in treatment through referral to substance use treatment programs and medication-assisted therapy. Considering the reduction in alcohol-related hospitalizations during lockdown, other strategies are needed to ensure comprehensive and longitudinal care for this vulnerable population.

The United States’ initial public health response to the COVID-19 pandemic included containment measures that varied by state but generally required closing or suspending schools, nonessential businesses, and travel (commonly called lockdown).¹ During these periods, hospitalizations for serious and common conditions declined.^2,3 In Massachusetts, a state of emergency was declared on March 10, 2020, which remained in place until May 18, 2020, when a phased reopening of businesses began.

Although the evidence on the mental health impact of containment periods has been mixed, it has been suggested that these measures could lead to increases in alcohol-related hospitalizations.⁴ Social isolation and increased psychosocial and financial stressors raise the risk of relapse among patients with substance use disorders.^5-7 Marketing and survey data from the US and United Kingdom from the early months of the pandemic suggest that in-home alcohol consumption and sales of alcoholic beverages increased, while consumption of alcohol outside the home decreased.^8-10 Other research has shown an increase in the percentage—but not necessarily the absolute number—of emergency department (ED) visits and hospitalizations for alcohol-related diagnoses during periods of containment.^11,12 At least 1 study suggests that alcohol-related deaths increased beginning in the lockdown period and persisting into mid-2021.¹³

Because earlier studies suggest that lockdown periods are associated with increased alcohol consumption and relapse of alcohol use disorder, we hypothesized that the spring 2020 lockdown period in Massachusetts would be associated temporally with an increase in alcohol-related hospitalizations. To evaluate this hypothesis, we examined all hospitalizations in the US Department of Veterans Affairs (VA) Boston Healthcare System (VABHS) before, during, and after this lockdown period. VABHS includes a 160-bed acute care hospital and a 50-bed inpatient psychiatric facility.

 

Methods

We conducted an interrupted time-series analysis including all inpatient hospitalizations at VABHS from January 1, 2017, to December 31, 2020, to compare the daily number of alcohol-related hospitalizations across 3 exposure groups: prelockdown (the reference group, 1/1/2017-3/9/2020); lockdown (3/10/2020-5/18/2020); and postlockdown (5/19/2020-12/31/2020).

The VA Corporate Data Warehouse at VABHS was queried to identify all hospitalizations on the medical, psychiatry, and neurology services during the study period. Hospitalizations were considered alcohol-related if the International Statistical Classification of Diseases, Tenth Revision (ICD-10) primary diagnosis code (the main reason for hospitalization) was defined as an alcohol-related diagnosis by the VA Centralized Interactive Phenomics Resource (eAppendix 1, available online at doi:10.1278/fp.0404). This database, which has been previously used for COVID-19 research, is a catalog and knowledge-sharing platform of VA electronic health record–based phenotype algorithms, definitions, and metadata that builds on the Million Veteran Program and Cooperative Studies Program.^14,15 Hospitalizations under observation status were excluded.

To examine whether alcohol-related hospitalizations could have been categorized as COVID-19 when the conditions were co-occurring, we identified 244 hospitalizations coded with a primary ICD-10 code for COVID-19 during the lockdown and postlockdown periods. At the time of admission, each hospitalization carries an initial (free text) diagnosis, of which 3 had an initial diagnosis related to alcohol use. The population at risk for alcohol-related hospitalizations was estimated as the number of patients actively engaged in care at the VABHS. This was defined as the number of patients enrolled in VA care who have previously received any VA care; patients who are enrolled but have never received VA care were excluded from the population-at-risk denominator. Population-at-risk data were available for each fiscal year (FY) of the study period (9/30-10/1); the following population-at-risk sizes were used: 38,057 for FY 2017, 38,527 for FY 2018, 39,472 for FY 2019, and 37,893 for FY 2020.

The primary outcome was the daily number of alcohol-related hospitalizations in the prelockdown, lockdown, and postlockdown periods. A sensitivity analysis was performed using an alternate definition of the primary outcome using a broader set of alcohol-related ICD-10 codes (eAppendix 2, available online at doi:10.1278/fp.0404).

Statistical Analysis

To visually examine hospitalization trends during the study period, we generated a smoothed time-series plot of the 7-day moving average of the daily number of all-cause hospitalizations and the daily number of alcohol-related hospitalizations from January 1, 2017, to December 31, 2020. We used multivariable regression to model the daily number of alcohol-related hospitalizations over prelockdown (the reference group), lockdown, and postlockdown. In addition to the exposure, we included the following covariates in our model: day of the week, calendar date (to account for secular trends), and harmonic polynomials of the day of the year (to account for seasonal variation).¹⁶

We also examined models that included the daily total number of hospitalizations to account for the reduced likelihood of hospital admission for any reason during the pandemic. We used generalized linear models with a Poisson link to generate rate ratios and corresponding 95% CIs for estimates of the daily number of alcohol-related hospitalizations. We estimated the population incidence of alcohol-related hospitalizations per 100,000 patient-months for the exposure periods using the population denominators previously described. All analyses were performed in Stata 16.1.

 

Results

During the study period, 27,508 hospitalizations were available for analysis. The 7-day moving average of total daily hospitalizations and total daily alcohol-related hospitalizations over time for the period January 1, 2017, to December 31, 2020, are shown in the Figure.

Compared with the prelockdown period, the 7-day average of hospitalizations per day for all hospitalizations and alcohol-related hospitalizations decreased substantially during the lockdown and did not return to the prelockdown baseline during the postlockdown period.

The incidence of alcohol-related hospitalizations in the population dropped from 72 per 100,000 patient-months to 10 per 100,000 patient-months during the lockdown period and increased to 46 per 100,000 patient-months during the postlockdown period (Table).

Compared with the 3-year prelockdown period, the rate ratio for daily alcohol-related hospitalizations during the lockdown period decreased to 0.20 (95% CI, 0.10-0.39). In the postlockdown period, the rate ratio for daily alcohol-related hospitalizations increased, but to only 0.72 (95% CI, 0.57-0.92) compared with the prepandemic baseline.

Our results were not substantially different when we ran a sensitivity analysis that excluded the total daily number of admissions from our model. Compared with the prelockdown period, the rate ratio for the number of alcohol-related hospitalizations during the lockdown period was 0.16 (95% CI, 0.08-0.30), and the rate ratio for the postlockdown period was 0.65 (95% CI, 0.52-0.82). We conducted an additional sensitivity analysis using a broader definition of the primary outcome to include all alcohol-related diagnosis codes; however, the results were unchanged.

Discussion

During the spring 2020 COVID-19 lockdown period in Massachusetts, the daily number of VABHS alcohol-related hospitalizations decreased by nearly 80% compared with the prelockdown period. During the postlockdown period, the daily number of alcohol-related hospitalizations increased but only to 72% of the prelockdown baseline by the end of December 2020. A similar trend was observed for all-cause hospitalizations for the same exposure periods.

These results differ from 2 related studies on the effect of the COVID-19 pandemic on alcohol-related hospitalizations.^10,11 In a retrospective study of ED visits to 4 hospitals in New York City, Schimmel and colleagues reported that from March 1 to 31, 2020 (the initial COVID-19 peak), hospital visits for alcohol withdrawal increased while those for alcohol use decreased.¹⁰ However, these results are reported as a percentage of total ED visits rather than the total number of visits, which are vulnerable to spurious correlation because of concomitant changes in the total number of ED visits. In their study, the absolute number of alcohol-related ED visits did not increase during the initial 2020 COVID-19 peak, and the number of visits for alcohol withdrawal syndrome declined slightly (195 in 2019 and 180 in 2020). However, the percentage of visits increased from 7% to 10% because of a greater decline in total ED visits. This pattern of decline in the number of alcohol-related ED visits, accompanied by an increase in the percentage of alcohol-related ED visits, has been observed in at least 1 nationwide surveillance study.¹⁷ This apparent increase does not reflect an absolute increase in ED visits for alcohol withdrawal syndrome and represents a greater relative decline in visits for other causes during the study period.

Sharma and colleagues reported an increase in the percentage of patients who developed alcohol withdrawal syndrome while hospitalized in Delaware per 1000 hospitalizations during consecutive 2-week periods during the pandemic in 2020 compared with corresponding weeks in 2019.¹¹ The greatest increase occurred during the last 2 weeks of the Delaware stay-at-home order. The Clinical Institute Withdrawal Assessment of Alcohol Scale, revised (CIWA-Ar) score of > 8 was used to define alcohol withdrawal syndrome. The American Society of Addiction Medicine does not recommend using CIWA-Ar to diagnose alcohol withdrawal syndrome because the scale was developed to monitor response to treatment, not to establish a diagnosis.¹⁸

Although the true population incidence of alcohol-related hospitalizations is difficult to estimate because the size of the population at risk (ie, the denominator) often is not known, the total number of hospitalizations is not a reliable surrogate.¹⁹ Individuals hospitalized for nonalcohol causes are no longer at risk for alcohol-related hospitalization.

In our study, we assume the population at risk during the study period is constant and model changes in the absolute number—rather than percentage—of alcohol-related ED visits. These absolute estimates of alcohol-related hospitalizations better reflect the true burden on the health care system and avoid the confounding effect of declining total ED visits and hospitalizations that could lead to artificially increased percentages and spurious correlation.²⁰ The absolute percentage of alcohol-related hospitalizations also decreased during this period; therefore, our results are not sensitive to this approach.

Several factors could have contributed to the decrease in alcohol-related hospitalizations. Our findings suggest that patient likelihood to seek care and clinician threshold to admit patients for alcohol-related conditions are influenced by external factors, in this case, a public health lockdown. Although our data do not inform why hospitalizations did not return to prelockdown levels, our experience suggests that limited bed capacity and longer length of stay might have contributed. Other hypotheses include a shift to outpatient care, increased use of telehealth (a significant focus early in the pandemic), and avoiding care for less severe alcohol-related complications because of lingering concerns about exposure to COVID-19 in health care settings reported early in the pandemic. Massachusetts experienced a particularly deadly outbreak of COVID-19 in the Soldiers’ Home, a long-term care facility for veterans in Holyoke.²¹

Evidence suggests that in-home consumption of alcohol increased during lockdowns.^8-10 Our results show that during this period hospitalizations for alcohol-related conditions decreased at VABHS, a large urban VA medical system, while alcohol-related deaths increased nationally.¹³ Although this observation is not evidence of causality, these outcomes could be related.

In the 2 decades before the pandemic, alcohol-related deaths increased by about 2% per year.²² From 2019 to 2020, there was a 25% increase that continued through 2021.¹³ Death certificate data often are inaccurate, and it is difficult to determine whether COVID-19 had a substantial contributing role to these deaths, particularly during the initial period when testing was limited or unavailable. Nonetheless, deaths due to alcohol-associated liver disease, overdoses involving alcohol, and alcohol-related traffic fatalities increased by > 10%.^13,23 These trends, along with a decrease in hospitalization for alcohol-related conditions, suggest missed opportunities for intervention with patients experiencing alcohol use disorder.

 

Limitations

In this study, hospitalizations under observation status were excluded, which could underestimate the total number of hospitalizations related to alcohol. We reasoned that this effect was likely to be small and not substantially different by year. ICD-10 codes were used to identify alcohol-related hospitalizations as any hospitalization with an included ICD-10 code listed as the primary discharge diagnosis code. This also likely underestimated the total number of alcohol-related hospitalizations. An ICD-10 code for COVID-19 was not in widespread use during our study period, which prohibited controlling explicitly for the volume of admissions due to COVID-19. The prelockdown period only contains data from the preceding 3 years, which might not be long enough for secular trends to become apparent. We assumed the population at risk remained constant when in reality, the net movement of patients into and out of VA care during the pandemic likely was more complex but not readily quantifiable. Nonetheless, the large drop in absolute number of alcohol-related hospitalizations is not likely to be sensitive to this change. In the absence of an objective measure of care-seeking behavior, we used the total daily number of hospitalizations as a surrogate for patient propensity to seek care. The total daily number of hospitalizations also reflects changes in physician admitting behavior over time. This allowed explicit modeling of care-seeking behavior as a covariate but does not capture other important determinants such as hospital capacity.

Conclusions

In this interrupted time-series analysis, the daily number of alcohol-related hospitalizations during the initial COVID-19 pandemic–associated lockdown period at VABHS decreased by 80% and remained 28% lower in the postlockdown period compared with the prepandemic baseline. In the context of evidence suggesting that alcohol-related mortality increased during the COVID-19 pandemic, alternate strategies to reach vulnerable individuals are needed. Because of high rates of relapse, hospitalization is an important opportunity to engage patients experiencing alcohol use disorder in treatment through referral to substance use treatment programs and medication-assisted therapy. Considering the reduction in alcohol-related hospitalizations during lockdown, other strategies are needed to ensure comprehensive and longitudinal care for this vulnerable population.

The United States’ initial public health response to the COVID-19 pandemic included containment measures that varied by state but generally required closing or suspending schools, nonessential businesses, and travel (commonly called lockdown).¹ During these periods, hospitalizations for serious and common conditions declined.^2,3 In Massachusetts, a state of emergency was declared on March 10, 2020, which remained in place until May 18, 2020, when a phased reopening of businesses began.

Although the evidence on the mental health impact of containment periods has been mixed, it has been suggested that these measures could lead to increases in alcohol-related hospitalizations.⁴ Social isolation and increased psychosocial and financial stressors raise the risk of relapse among patients with substance use disorders.^5-7 Marketing and survey data from the US and United Kingdom from the early months of the pandemic suggest that in-home alcohol consumption and sales of alcoholic beverages increased, while consumption of alcohol outside the home decreased.^8-10 Other research has shown an increase in the percentage—but not necessarily the absolute number—of emergency department (ED) visits and hospitalizations for alcohol-related diagnoses during periods of containment.^11,12 At least 1 study suggests that alcohol-related deaths increased beginning in the lockdown period and persisting into mid-2021.¹³

Because earlier studies suggest that lockdown periods are associated with increased alcohol consumption and relapse of alcohol use disorder, we hypothesized that the spring 2020 lockdown period in Massachusetts would be associated temporally with an increase in alcohol-related hospitalizations. To evaluate this hypothesis, we examined all hospitalizations in the US Department of Veterans Affairs (VA) Boston Healthcare System (VABHS) before, during, and after this lockdown period. VABHS includes a 160-bed acute care hospital and a 50-bed inpatient psychiatric facility.

 

Methods

We conducted an interrupted time-series analysis including all inpatient hospitalizations at VABHS from January 1, 2017, to December 31, 2020, to compare the daily number of alcohol-related hospitalizations across 3 exposure groups: prelockdown (the reference group, 1/1/2017-3/9/2020); lockdown (3/10/2020-5/18/2020); and postlockdown (5/19/2020-12/31/2020).

The VA Corporate Data Warehouse at VABHS was queried to identify all hospitalizations on the medical, psychiatry, and neurology services during the study period. Hospitalizations were considered alcohol-related if the International Statistical Classification of Diseases, Tenth Revision (ICD-10) primary diagnosis code (the main reason for hospitalization) was defined as an alcohol-related diagnosis by the VA Centralized Interactive Phenomics Resource (eAppendix 1, available online at doi:10.1278/fp.0404). This database, which has been previously used for COVID-19 research, is a catalog and knowledge-sharing platform of VA electronic health record–based phenotype algorithms, definitions, and metadata that builds on the Million Veteran Program and Cooperative Studies Program.^14,15 Hospitalizations under observation status were excluded.

To examine whether alcohol-related hospitalizations could have been categorized as COVID-19 when the conditions were co-occurring, we identified 244 hospitalizations coded with a primary ICD-10 code for COVID-19 during the lockdown and postlockdown periods. At the time of admission, each hospitalization carries an initial (free text) diagnosis, of which 3 had an initial diagnosis related to alcohol use. The population at risk for alcohol-related hospitalizations was estimated as the number of patients actively engaged in care at the VABHS. This was defined as the number of patients enrolled in VA care who have previously received any VA care; patients who are enrolled but have never received VA care were excluded from the population-at-risk denominator. Population-at-risk data were available for each fiscal year (FY) of the study period (9/30-10/1); the following population-at-risk sizes were used: 38,057 for FY 2017, 38,527 for FY 2018, 39,472 for FY 2019, and 37,893 for FY 2020.

The primary outcome was the daily number of alcohol-related hospitalizations in the prelockdown, lockdown, and postlockdown periods. A sensitivity analysis was performed using an alternate definition of the primary outcome using a broader set of alcohol-related ICD-10 codes (eAppendix 2, available online at doi:10.1278/fp.0404).

Statistical Analysis

To visually examine hospitalization trends during the study period, we generated a smoothed time-series plot of the 7-day moving average of the daily number of all-cause hospitalizations and the daily number of alcohol-related hospitalizations from January 1, 2017, to December 31, 2020. We used multivariable regression to model the daily number of alcohol-related hospitalizations over prelockdown (the reference group), lockdown, and postlockdown. In addition to the exposure, we included the following covariates in our model: day of the week, calendar date (to account for secular trends), and harmonic polynomials of the day of the year (to account for seasonal variation).¹⁶

We also examined models that included the daily total number of hospitalizations to account for the reduced likelihood of hospital admission for any reason during the pandemic. We used generalized linear models with a Poisson link to generate rate ratios and corresponding 95% CIs for estimates of the daily number of alcohol-related hospitalizations. We estimated the population incidence of alcohol-related hospitalizations per 100,000 patient-months for the exposure periods using the population denominators previously described. All analyses were performed in Stata 16.1.

 

Results

During the study period, 27,508 hospitalizations were available for analysis. The 7-day moving average of total daily hospitalizations and total daily alcohol-related hospitalizations over time for the period January 1, 2017, to December 31, 2020, are shown in the Figure.

Compared with the prelockdown period, the 7-day average of hospitalizations per day for all hospitalizations and alcohol-related hospitalizations decreased substantially during the lockdown and did not return to the prelockdown baseline during the postlockdown period.

The incidence of alcohol-related hospitalizations in the population dropped from 72 per 100,000 patient-months to 10 per 100,000 patient-months during the lockdown period and increased to 46 per 100,000 patient-months during the postlockdown period (Table).

Compared with the 3-year prelockdown period, the rate ratio for daily alcohol-related hospitalizations during the lockdown period decreased to 0.20 (95% CI, 0.10-0.39). In the postlockdown period, the rate ratio for daily alcohol-related hospitalizations increased, but to only 0.72 (95% CI, 0.57-0.92) compared with the prepandemic baseline.

Our results were not substantially different when we ran a sensitivity analysis that excluded the total daily number of admissions from our model. Compared with the prelockdown period, the rate ratio for the number of alcohol-related hospitalizations during the lockdown period was 0.16 (95% CI, 0.08-0.30), and the rate ratio for the postlockdown period was 0.65 (95% CI, 0.52-0.82). We conducted an additional sensitivity analysis using a broader definition of the primary outcome to include all alcohol-related diagnosis codes; however, the results were unchanged.

Discussion

During the spring 2020 COVID-19 lockdown period in Massachusetts, the daily number of VABHS alcohol-related hospitalizations decreased by nearly 80% compared with the prelockdown period. During the postlockdown period, the daily number of alcohol-related hospitalizations increased but only to 72% of the prelockdown baseline by the end of December 2020. A similar trend was observed for all-cause hospitalizations for the same exposure periods.

These results differ from 2 related studies on the effect of the COVID-19 pandemic on alcohol-related hospitalizations.^10,11 In a retrospective study of ED visits to 4 hospitals in New York City, Schimmel and colleagues reported that from March 1 to 31, 2020 (the initial COVID-19 peak), hospital visits for alcohol withdrawal increased while those for alcohol use decreased.¹⁰ However, these results are reported as a percentage of total ED visits rather than the total number of visits, which are vulnerable to spurious correlation because of concomitant changes in the total number of ED visits. In their study, the absolute number of alcohol-related ED visits did not increase during the initial 2020 COVID-19 peak, and the number of visits for alcohol withdrawal syndrome declined slightly (195 in 2019 and 180 in 2020). However, the percentage of visits increased from 7% to 10% because of a greater decline in total ED visits. This pattern of decline in the number of alcohol-related ED visits, accompanied by an increase in the percentage of alcohol-related ED visits, has been observed in at least 1 nationwide surveillance study.¹⁷ This apparent increase does not reflect an absolute increase in ED visits for alcohol withdrawal syndrome and represents a greater relative decline in visits for other causes during the study period.

Sharma and colleagues reported an increase in the percentage of patients who developed alcohol withdrawal syndrome while hospitalized in Delaware per 1000 hospitalizations during consecutive 2-week periods during the pandemic in 2020 compared with corresponding weeks in 2019.¹¹ The greatest increase occurred during the last 2 weeks of the Delaware stay-at-home order. The Clinical Institute Withdrawal Assessment of Alcohol Scale, revised (CIWA-Ar) score of > 8 was used to define alcohol withdrawal syndrome. The American Society of Addiction Medicine does not recommend using CIWA-Ar to diagnose alcohol withdrawal syndrome because the scale was developed to monitor response to treatment, not to establish a diagnosis.¹⁸

Although the true population incidence of alcohol-related hospitalizations is difficult to estimate because the size of the population at risk (ie, the denominator) often is not known, the total number of hospitalizations is not a reliable surrogate.¹⁹ Individuals hospitalized for nonalcohol causes are no longer at risk for alcohol-related hospitalization.

In our study, we assume the population at risk during the study period is constant and model changes in the absolute number—rather than percentage—of alcohol-related ED visits. These absolute estimates of alcohol-related hospitalizations better reflect the true burden on the health care system and avoid the confounding effect of declining total ED visits and hospitalizations that could lead to artificially increased percentages and spurious correlation.²⁰ The absolute percentage of alcohol-related hospitalizations also decreased during this period; therefore, our results are not sensitive to this approach.

Several factors could have contributed to the decrease in alcohol-related hospitalizations. Our findings suggest that patient likelihood to seek care and clinician threshold to admit patients for alcohol-related conditions are influenced by external factors, in this case, a public health lockdown. Although our data do not inform why hospitalizations did not return to prelockdown levels, our experience suggests that limited bed capacity and longer length of stay might have contributed. Other hypotheses include a shift to outpatient care, increased use of telehealth (a significant focus early in the pandemic), and avoiding care for less severe alcohol-related complications because of lingering concerns about exposure to COVID-19 in health care settings reported early in the pandemic. Massachusetts experienced a particularly deadly outbreak of COVID-19 in the Soldiers’ Home, a long-term care facility for veterans in Holyoke.²¹

Evidence suggests that in-home consumption of alcohol increased during lockdowns.^8-10 Our results show that during this period hospitalizations for alcohol-related conditions decreased at VABHS, a large urban VA medical system, while alcohol-related deaths increased nationally.¹³ Although this observation is not evidence of causality, these outcomes could be related.

In the 2 decades before the pandemic, alcohol-related deaths increased by about 2% per year.²² From 2019 to 2020, there was a 25% increase that continued through 2021.¹³ Death certificate data often are inaccurate, and it is difficult to determine whether COVID-19 had a substantial contributing role to these deaths, particularly during the initial period when testing was limited or unavailable. Nonetheless, deaths due to alcohol-associated liver disease, overdoses involving alcohol, and alcohol-related traffic fatalities increased by > 10%.^13,23 These trends, along with a decrease in hospitalization for alcohol-related conditions, suggest missed opportunities for intervention with patients experiencing alcohol use disorder.

 

Limitations

In this study, hospitalizations under observation status were excluded, which could underestimate the total number of hospitalizations related to alcohol. We reasoned that this effect was likely to be small and not substantially different by year. ICD-10 codes were used to identify alcohol-related hospitalizations as any hospitalization with an included ICD-10 code listed as the primary discharge diagnosis code. This also likely underestimated the total number of alcohol-related hospitalizations. An ICD-10 code for COVID-19 was not in widespread use during our study period, which prohibited controlling explicitly for the volume of admissions due to COVID-19. The prelockdown period only contains data from the preceding 3 years, which might not be long enough for secular trends to become apparent. We assumed the population at risk remained constant when in reality, the net movement of patients into and out of VA care during the pandemic likely was more complex but not readily quantifiable. Nonetheless, the large drop in absolute number of alcohol-related hospitalizations is not likely to be sensitive to this change. In the absence of an objective measure of care-seeking behavior, we used the total daily number of hospitalizations as a surrogate for patient propensity to seek care. The total daily number of hospitalizations also reflects changes in physician admitting behavior over time. This allowed explicit modeling of care-seeking behavior as a covariate but does not capture other important determinants such as hospital capacity.

Conclusions

In this interrupted time-series analysis, the daily number of alcohol-related hospitalizations during the initial COVID-19 pandemic–associated lockdown period at VABHS decreased by 80% and remained 28% lower in the postlockdown period compared with the prepandemic baseline. In the context of evidence suggesting that alcohol-related mortality increased during the COVID-19 pandemic, alternate strategies to reach vulnerable individuals are needed. Because of high rates of relapse, hospitalization is an important opportunity to engage patients experiencing alcohol use disorder in treatment through referral to substance use treatment programs and medication-assisted therapy. Considering the reduction in alcohol-related hospitalizations during lockdown, other strategies are needed to ensure comprehensive and longitudinal care for this vulnerable population.

To the Editor:

Micrographic surgery and dermatologic oncology (MSDO) is a highly competitive subspecialty fellowship in dermatology. Prospective applicants often depend on the Internet to obtain pertinent information about fellowship programs to navigate the application process. An up-to-date and comprehensive fellowship website has the potential to be advantageous for both applicants and programs—applicants can more readily identify programs that align with their goals and values, and programs can effectively attract compatible applicants. These advantages are increasingly relevant with the virtual application process that has become essential considering the COVID-19 pandemic. At the height of the COVID-19 pandemic in 2020, we sought to evaluate the comprehensiveness of the content of Accreditation Council for Graduate Medical Education (ACGME) MSDO fellowship program websites to identify possible areas for improvement.

We obtained a list of all ACGME MSDO fellowships from the ACGME website (https://www.acgme.org/) and verified it against the list of MSDO programs in FREIDA, the American Medical Association residency and fellowship database (https://freida.ama-assn.org/). All programs without a website were excluded from further analysis. All data collection from currently accessible fellowship websites and evaluation occurred in April 2020.

The remaining MSDO fellowship program websites were evaluated using 25 criteria distributed among 5 domains: education/research, clinical training, program information, application process, and incentives. These criteria were determined based on earlier studies that similarly evaluated the website content of fellowship programs with inclusion of information that was considered valuable in the appraisal of fellowship programs.^1,2 Criteria were further refined by direct consideration of relevance and importance to MSDO fellowship applicants (eg, inclusion of case volume, exclusion of call schedule).

Each criterion was independently assessed by 2 investigators (J.Y.C. and S.J.E.S.). A third investigator (J.R.P.) then independently evaluated those 2 assessments for agreement. Where disagreement was discovered, the third evaluator (J.R.P.) provided a final appraisal. Cohen’s kappa (κ) was conducted to evaluate for concordance between the 2 primary website evaluators. We found there to be substantial agreement between the reviewers within the education/research (κ [SD]=0.772 [0.077]), clinical training (κ [SD]=0.740 [0.051]), application process (κ [SD]=0.726 [0.103]), and incentives domains (κ [SD]=0.730 [0.110]). There was moderate agreement (κ [SD]=0.603 [0.128]) between the reviewers within the program information domain.

We identified 77 active MSDO fellowship programs. Sixty of those 77 programs (77.9%) had a dedicated fellowship website that was readily accessible. Most programs that had a dedicated fellowship website had a core or affiliated residency program (49/60 [81.7%]).

Websites that we evaluated fulfilled a mean (SD) of 9.37 (4.17) of the 25 identified criteria. Only 13 of 60 (21.7%) websites fulfilled more than 50% of evaluated criteria.

There was no statistical difference in the number of criteria fulfilled based on whether the fellowship program had a core or affiliated residency program.

 

Upon reviewing website accessibility directly from FREIDA, only 5 of 60 programs (8.3%) provided applicants with a link directly to their fellowship page (Table). Most programs (41 [68.3%]) provided a link to the dermatology department website, not to the specific fellowship program page, thus requiring a multistep process to find the fellowship-specific page. The remaining programs had an inaccessible (4 [6.7%]) or absent (10 [16.7%]) link on FREIDA, though a fellowship website could be identified by an Internet search of the program name.

The domain most fulfilled was program information with an average of 51.1% of programs satisfying the criteria, whereas the incentives domain was least fulfilled with an average of only 20.8% of programs satisfying the criteria. Across the various criteria, websites more often included a description of the program (58 [96.6%]), mentioned accreditation (53 [88.3%]), and provided case descriptions (48 [80.0%]). They less often reported information regarding a fellow’s call responsibility (3 [5%]); evaluation criteria (5 [8.3%]); and rotation schedule or options (6 [10.0%]).

The highest number of criteria fulfilled by a single program was 19 (76%). The lowest number of criteria met was 2 (8%). These findings suggest a large variation in comprehensiveness across fellowship websites.

Our research suggests that many current MSDO fellowship programs have room to maximize the information provided to applicants through their websites, which is particularly relevant following the COVID-19 pandemic, as the value of providing comprehensive and transparent information through an online platform is greater than ever. Given the ongoing desire to limit travel, virtual methods for navigating the application process have been readily used, including online videoconferencing for interviews and virtual program visits. This scenario has placed applicants in a challenging situation—their ability to directly evaluate their compatibility with a given program has been limited.³

Earlier studies that analyzed rheumatology fellowship recruitment during the COVID-19 pandemic found that programs may have more difficulty highlighting the strengths of their institution (eg, clinical facilities, professional opportunities, educational environment).⁴ An updated and comprehensive fellowship website was recommended⁴ as a key part in facing these new challenges. On the other hand, given the large number of applicants each year for fellowship positions in any given program, we acknowledge the potential benefit programs may obtain from limiting electronic information that is readily accessible to all applicants, as doing so may encourage applicants to communicate directly with a program and allow programs to identify candidates who are more interested.

In light of the movement to a more virtual-friendly and technology-driven fellowship application process, we identified 25 content areas that fellowships may want to include on their websites so that potential applicants can be well informed about the program before submitting an application and scheduling an interview. Efforts to improve accessibility and maximize the content of these websites may help programs attract compatible candidates, improve transparency, and guide applicants throughout the application process.

To the Editor:

Micrographic surgery and dermatologic oncology (MSDO) is a highly competitive subspecialty fellowship in dermatology. Prospective applicants often depend on the Internet to obtain pertinent information about fellowship programs to navigate the application process. An up-to-date and comprehensive fellowship website has the potential to be advantageous for both applicants and programs—applicants can more readily identify programs that align with their goals and values, and programs can effectively attract compatible applicants. These advantages are increasingly relevant with the virtual application process that has become essential considering the COVID-19 pandemic. At the height of the COVID-19 pandemic in 2020, we sought to evaluate the comprehensiveness of the content of Accreditation Council for Graduate Medical Education (ACGME) MSDO fellowship program websites to identify possible areas for improvement.

We obtained a list of all ACGME MSDO fellowships from the ACGME website (https://www.acgme.org/) and verified it against the list of MSDO programs in FREIDA, the American Medical Association residency and fellowship database (https://freida.ama-assn.org/). All programs without a website were excluded from further analysis. All data collection from currently accessible fellowship websites and evaluation occurred in April 2020.

The remaining MSDO fellowship program websites were evaluated using 25 criteria distributed among 5 domains: education/research, clinical training, program information, application process, and incentives. These criteria were determined based on earlier studies that similarly evaluated the website content of fellowship programs with inclusion of information that was considered valuable in the appraisal of fellowship programs.^1,2 Criteria were further refined by direct consideration of relevance and importance to MSDO fellowship applicants (eg, inclusion of case volume, exclusion of call schedule).

Each criterion was independently assessed by 2 investigators (J.Y.C. and S.J.E.S.). A third investigator (J.R.P.) then independently evaluated those 2 assessments for agreement. Where disagreement was discovered, the third evaluator (J.R.P.) provided a final appraisal. Cohen’s kappa (κ) was conducted to evaluate for concordance between the 2 primary website evaluators. We found there to be substantial agreement between the reviewers within the education/research (κ [SD]=0.772 [0.077]), clinical training (κ [SD]=0.740 [0.051]), application process (κ [SD]=0.726 [0.103]), and incentives domains (κ [SD]=0.730 [0.110]). There was moderate agreement (κ [SD]=0.603 [0.128]) between the reviewers within the program information domain.

We identified 77 active MSDO fellowship programs. Sixty of those 77 programs (77.9%) had a dedicated fellowship website that was readily accessible. Most programs that had a dedicated fellowship website had a core or affiliated residency program (49/60 [81.7%]).

Websites that we evaluated fulfilled a mean (SD) of 9.37 (4.17) of the 25 identified criteria. Only 13 of 60 (21.7%) websites fulfilled more than 50% of evaluated criteria.

There was no statistical difference in the number of criteria fulfilled based on whether the fellowship program had a core or affiliated residency program.

 

Upon reviewing website accessibility directly from FREIDA, only 5 of 60 programs (8.3%) provided applicants with a link directly to their fellowship page (Table). Most programs (41 [68.3%]) provided a link to the dermatology department website, not to the specific fellowship program page, thus requiring a multistep process to find the fellowship-specific page. The remaining programs had an inaccessible (4 [6.7%]) or absent (10 [16.7%]) link on FREIDA, though a fellowship website could be identified by an Internet search of the program name.

The domain most fulfilled was program information with an average of 51.1% of programs satisfying the criteria, whereas the incentives domain was least fulfilled with an average of only 20.8% of programs satisfying the criteria. Across the various criteria, websites more often included a description of the program (58 [96.6%]), mentioned accreditation (53 [88.3%]), and provided case descriptions (48 [80.0%]). They less often reported information regarding a fellow’s call responsibility (3 [5%]); evaluation criteria (5 [8.3%]); and rotation schedule or options (6 [10.0%]).

The highest number of criteria fulfilled by a single program was 19 (76%). The lowest number of criteria met was 2 (8%). These findings suggest a large variation in comprehensiveness across fellowship websites.

Our research suggests that many current MSDO fellowship programs have room to maximize the information provided to applicants through their websites, which is particularly relevant following the COVID-19 pandemic, as the value of providing comprehensive and transparent information through an online platform is greater than ever. Given the ongoing desire to limit travel, virtual methods for navigating the application process have been readily used, including online videoconferencing for interviews and virtual program visits. This scenario has placed applicants in a challenging situation—their ability to directly evaluate their compatibility with a given program has been limited.³

Earlier studies that analyzed rheumatology fellowship recruitment during the COVID-19 pandemic found that programs may have more difficulty highlighting the strengths of their institution (eg, clinical facilities, professional opportunities, educational environment).⁴ An updated and comprehensive fellowship website was recommended⁴ as a key part in facing these new challenges. On the other hand, given the large number of applicants each year for fellowship positions in any given program, we acknowledge the potential benefit programs may obtain from limiting electronic information that is readily accessible to all applicants, as doing so may encourage applicants to communicate directly with a program and allow programs to identify candidates who are more interested.

In light of the movement to a more virtual-friendly and technology-driven fellowship application process, we identified 25 content areas that fellowships may want to include on their websites so that potential applicants can be well informed about the program before submitting an application and scheduling an interview. Efforts to improve accessibility and maximize the content of these websites may help programs attract compatible candidates, improve transparency, and guide applicants throughout the application process.

To the Editor:

Micrographic surgery and dermatologic oncology (MSDO) is a highly competitive subspecialty fellowship in dermatology. Prospective applicants often depend on the Internet to obtain pertinent information about fellowship programs to navigate the application process. An up-to-date and comprehensive fellowship website has the potential to be advantageous for both applicants and programs—applicants can more readily identify programs that align with their goals and values, and programs can effectively attract compatible applicants. These advantages are increasingly relevant with the virtual application process that has become essential considering the COVID-19 pandemic. At the height of the COVID-19 pandemic in 2020, we sought to evaluate the comprehensiveness of the content of Accreditation Council for Graduate Medical Education (ACGME) MSDO fellowship program websites to identify possible areas for improvement.

We obtained a list of all ACGME MSDO fellowships from the ACGME website (https://www.acgme.org/) and verified it against the list of MSDO programs in FREIDA, the American Medical Association residency and fellowship database (https://freida.ama-assn.org/). All programs without a website were excluded from further analysis. All data collection from currently accessible fellowship websites and evaluation occurred in April 2020.

The remaining MSDO fellowship program websites were evaluated using 25 criteria distributed among 5 domains: education/research, clinical training, program information, application process, and incentives. These criteria were determined based on earlier studies that similarly evaluated the website content of fellowship programs with inclusion of information that was considered valuable in the appraisal of fellowship programs.^1,2 Criteria were further refined by direct consideration of relevance and importance to MSDO fellowship applicants (eg, inclusion of case volume, exclusion of call schedule).

Each criterion was independently assessed by 2 investigators (J.Y.C. and S.J.E.S.). A third investigator (J.R.P.) then independently evaluated those 2 assessments for agreement. Where disagreement was discovered, the third evaluator (J.R.P.) provided a final appraisal. Cohen’s kappa (κ) was conducted to evaluate for concordance between the 2 primary website evaluators. We found there to be substantial agreement between the reviewers within the education/research (κ [SD]=0.772 [0.077]), clinical training (κ [SD]=0.740 [0.051]), application process (κ [SD]=0.726 [0.103]), and incentives domains (κ [SD]=0.730 [0.110]). There was moderate agreement (κ [SD]=0.603 [0.128]) between the reviewers within the program information domain.

We identified 77 active MSDO fellowship programs. Sixty of those 77 programs (77.9%) had a dedicated fellowship website that was readily accessible. Most programs that had a dedicated fellowship website had a core or affiliated residency program (49/60 [81.7%]).

Websites that we evaluated fulfilled a mean (SD) of 9.37 (4.17) of the 25 identified criteria. Only 13 of 60 (21.7%) websites fulfilled more than 50% of evaluated criteria.

There was no statistical difference in the number of criteria fulfilled based on whether the fellowship program had a core or affiliated residency program.

 

Upon reviewing website accessibility directly from FREIDA, only 5 of 60 programs (8.3%) provided applicants with a link directly to their fellowship page (Table). Most programs (41 [68.3%]) provided a link to the dermatology department website, not to the specific fellowship program page, thus requiring a multistep process to find the fellowship-specific page. The remaining programs had an inaccessible (4 [6.7%]) or absent (10 [16.7%]) link on FREIDA, though a fellowship website could be identified by an Internet search of the program name.

The domain most fulfilled was program information with an average of 51.1% of programs satisfying the criteria, whereas the incentives domain was least fulfilled with an average of only 20.8% of programs satisfying the criteria. Across the various criteria, websites more often included a description of the program (58 [96.6%]), mentioned accreditation (53 [88.3%]), and provided case descriptions (48 [80.0%]). They less often reported information regarding a fellow’s call responsibility (3 [5%]); evaluation criteria (5 [8.3%]); and rotation schedule or options (6 [10.0%]).

The highest number of criteria fulfilled by a single program was 19 (76%). The lowest number of criteria met was 2 (8%). These findings suggest a large variation in comprehensiveness across fellowship websites.

Our research suggests that many current MSDO fellowship programs have room to maximize the information provided to applicants through their websites, which is particularly relevant following the COVID-19 pandemic, as the value of providing comprehensive and transparent information through an online platform is greater than ever. Given the ongoing desire to limit travel, virtual methods for navigating the application process have been readily used, including online videoconferencing for interviews and virtual program visits. This scenario has placed applicants in a challenging situation—their ability to directly evaluate their compatibility with a given program has been limited.³

Earlier studies that analyzed rheumatology fellowship recruitment during the COVID-19 pandemic found that programs may have more difficulty highlighting the strengths of their institution (eg, clinical facilities, professional opportunities, educational environment).⁴ An updated and comprehensive fellowship website was recommended⁴ as a key part in facing these new challenges. On the other hand, given the large number of applicants each year for fellowship positions in any given program, we acknowledge the potential benefit programs may obtain from limiting electronic information that is readily accessible to all applicants, as doing so may encourage applicants to communicate directly with a program and allow programs to identify candidates who are more interested.

In light of the movement to a more virtual-friendly and technology-driven fellowship application process, we identified 25 content areas that fellowships may want to include on their websites so that potential applicants can be well informed about the program before submitting an application and scheduling an interview. Efforts to improve accessibility and maximize the content of these websites may help programs attract compatible candidates, improve transparency, and guide applicants throughout the application process.

Atopic dermatitis (AD) is a common inflammatory skin disease that may severely decrease quality of life (QOL) and lead to psychiatric comorbidities.^1-3 Prior studies have indicated that AD causes a substantial economic burden, and disease severity has been proportionally linked to medical costs.^4,5 Results of a multicenter cost-of-illness study from Germany estimated that a relapse of AD costs approximately €123 (US $136). The authors calculated the average annual cost of AD per patient to be €1425 (US $1580), whereas it is €956 (US $1060) in moderate disease and €2068 (US $2293) in severe disease (direct and indirect medical costs included).⁶ An observational cohort study from the Netherlands found that total direct cost per patient-year (PPY) was €4401 (US $4879) for patients with controlled AD vs €6993 (US $7756) for patients with uncontrolled AD.⁷

In a retrospective survey-based study, it was estimated that the annual cost of AD in Canada was approximately CAD $1.4 billion. The cost per patient varied from CAD $282 to CAD $1242 depending on disease severity.⁸ In another retrospective cohort study from the Netherlands, the average direct medical cost per patient with AD seeing a general practitioner was US $71 during follow-up in primary care. If the patient needed specialist consultation, the cost increased to an average of US $186.⁹

We aimed to assess the direct and indirect medical costs in adult patients with moderate to severe AD who attended a tertiary health care center in Finland. In addition, we evaluated the impact of AD on QOL in this patient cohort.

Methods

Study Design—Patients with AD who were treated at the Department of Dermatology and Allergology, Helsinki University Hospital, Finland, between February 2018 and December 2019 were randomly selected to participate in our survey study. All participants provided written informed consent. In Finland, patients with mild AD generally are treated in primary health care centers, and only patients with moderate to severe AD are referred to specialists and tertiary care centers. Patients were excluded if they were younger than 18 years, had AD confined to the hands, or reported the presence of other concomitant skin diseases that were being treated with topical or systemic therapies. The protocol for the study was approved by the local ethics committee of the University of Helsinki.

Questionnaire and Analysis of Disease Severity—The survey included the medical history, signs of atopy, former treatment(s) for AD, skin infections, visits to dermatologists or general practitioners, questions on mental health and hospitalization, and absence from work due to AD in the last 12 months. Disease severity was evaluated using the patient-oriented Rajka & Langeland eczema severity score and Patient Oriented Eczema Measure (POEM).^10,11 The impact on QOL was evaluated by the Dermatology Life Quality Index (DLQI).¹²

Medication Costs—The cost of prescription drugs was based on data from the Finnish national electronic prescription center. In Finland, all prescriptions are made electronically in the database. We analyzed all topical medications (eg, topical corticosteroids [TCSs], topical calcineurin inhibitors [TCIs], and emollients) and systemic medicaments (eg, antibiotics, antihistamines, cyclosporine, methotrexate, and corticosteroids) prescribed for the treatment of AD. In Finland, dupilumab was introduced for the treatment of severe AD in early 2019, and patients receiving dupilumab were excluded from the study. Over-the-counter medications were not included. The costs for laboratory testing were estimations based on the standard monitoring protocols of the Helsinki University Hospital. All costs were based on the Finnish price level standard for the year 2019.

Inpatient/Outpatient Visits and Sick Leave Due to AD—The number of inpatient and outpatient visits due to AD in the last 12 months was evaluated. Outpatient specialist consultations or nurse appointments at Helsinki University Hospital were verified from electronic patient records. In addition, inpatient treatment and phototherapy sessions were calculated from the database.

 

We assessed the number of sick leave days from work or educational activities during the last year. All costs of transportation for doctors’ appointments, laboratory monitoring, and phototherapy treatments were summed together to estimate the total transportation cost. Visits to nurse and inpatient visits were not included in the total transportation cost because patients often were hospitalized directly after consultation visits, and nurse appointments often were combined with inpatient and outpatient visits. To calculate the total transportation cost, we used a rate of €0.43 per kilometer measured from the patients’ home addresses, which was the official compensation rate of the Finnish Tax Administration for 2019.¹³

Statistical Analysis—Statistical analyses were performed using SPSS Statistics 25 (IBM). Descriptive analyses were used to describe baseline characteristics and to evaluate the mean costs of AD. The patients were divided into 2 groups according to POEM: (1) controlled AD (patients with clear skin or only mild AD; POEM score 0–7) and (2) uncontrolled AD (patients with moderate to very severe AD; POEM score 8–28). The Mann-Whitney U statistic was used to evaluate differences between the study groups.

Results

Patient Characteristics—One hundred sixty-seven patients answered the survey, of which 69 (41.3%) were males and 98 (58.7%) were females. There were 16 patients with controlled AD and 148 patients with uncontrolled AD. Three patients did not answer to POEM and were excluded. The baseline characteristics are presented in Table 1 and include self-reported symptoms related to atopy.

The most-used topical treatments were TCSs (n=155; 92.8%) and emollients (n=166; 99.4%). One hundred sixteen (69.5%) patients had used TCIs. The median amount of TCSs used was 300 g/y vs 30 g/y for TCIs (range, 0-5160 g/y) and 1200 g/y for emollients.

Fifteen (9.0%) patients had been hospitalized for AD in the last year. The mean (SD) length of hospitalization was 6.5 (2.8) days. Thirty-four (20.4%) patients received UVB phototherapy. Thirty-four (20.4%) patients were treated with at least 1 antibiotic course for secondary AD infection. Thirty-six (21.6%) patients needed at least 1 oral corticosteroid course for the treatment of an AD flare.

Fifteen (9.0%) patients reported a diagnosed psychiatric illness, and 17 (10.2%) patients were using prescription drugs for psychiatric illness. Forty-nine (29.3%) patients reported anxiety or depression often or very often, 54 (32.3%) patients reported sometimes, 33 (19.8%) patients reported rarely, and only 30 (18.0%) patients reported none.

Medication Costs—Mean medication cost PPY was €457.40 (US $507.34)(Figure 1 and Table 2). On average, one patient spent €87.50 (US $97.05) for TCSs, €121.90 (US $135.21) for emollients, and €225.10 (US $249.68) for TCIs. The average cost PPY for antibiotics was €6.10 (US $6.77). Other systemic treatments, including (US $18.65). Seventeen patients (10.2%) were on methotrexate therapy for AD in the last year, and 1 patient also used cyclosporine. The costs for laboratory monitoring in these patients were included in the direct cost calculations. The mean cost PPY of laboratory monitoring in the whole study cohort was €6.60 (US $7.32). In patients with systemic immunosuppressive therapy, the mean cost PPY for laboratory monitoring was €65.00 (US $72.09). Five patients had been tested for contact dermatitis; the costs of patch tests or other diagnostic tests were not included.

 

Visits to Health Care Providers—In the last year, patients had an average of 1.83 dermatologist consultations in the tertiary center (Table 2). In addition, the mean number of visits to private dermatologists was 0.61 and 1.42 visits to general practitioners. The mean cost of physician visits was €302.70 (US $335.75) in the tertiary center, €66.60 (US $73.87) in the private sector, and €141.90 (US $157.39) in primary health care. In total, the average cost of doctors’ appointments PPY was €506.30 (US $561.57). The mean estimated distance traveled per visit was 9.5 km.

The mean cost PPY of inpatient treatments was €329.90 (US $365.92) and €239.00 (US $265.09) for UV phototherapy. Only 4 patients had visited a nurse in the last year, with an average cost PPY of €2.50 (US $2.78).

In total, the cost PPY for health care provider visits was €1084.20, which included specialist consultations in a tertiary center and private sector, visits in primary health care, inpatient treatments, UV phototherapy sessions, nurse appointments in a tertiary center, and laboratory monitoring. The average transportation cost PPY was €34.00 (US $37.71). The mean number of visits to health care providers was 8.3 per year. Altogether, the direct cost PPY in the study cohort was €1580.60 (US $1752.39)(Table 2 and Figure 2).

Comparison of Medical Costs in Controlled vs Uncontrolled AD—In the controlled AD group (POEM score <8), the mean medication cost PPY was €567.15 (US $629.13), and the mean total direct cost PPY was €2040.46 (US $2263.24). In the uncontrolled AD group (POEM score ≥8), the mean medication cost PPY was €449.55 (US $498.63), and the mean total direct cost PPY was €1539.39 (US $1707.36)(Table 2). The comparisons of the study groups—controlled vs uncontrolled AD—showed no significant differences regarding medication costs PPY (P=.305, Mann-Whitney U statistic) and total direct costs PPY (P=.361, Mann-Whitney U statistic)(Figure 3). Thus, the distribution of medical costs was similar across all categories of the POEM score.

AD Severity and QOL—The mean (SD) POEM score in the study cohort was 17.9 (6.9). Sixteen (9.6%) patients had clear to almost clear skin or mild AD (POEM score 0–7). Forty-two (25.1%) patients had moderate AD (POEM score 8–16). Most of the patients (106; 63.5%) had severe or very severe AD (POEM score 17–28). According to the Rajka & Langeland score, 5 (3.0%) patients had mild disease (score 34), 81 (48.5%) patients had moderate disease (score 5–7), and 81 (48.5%) patients had severe disease (score 8–9). Eighty-one (48.5%) patients answered that AD affects their lives greatly, and 58 (34.7%) patients answered that it affects their lives extremely. Twenty-five (15.0%) patients answered that AD affects their everyday life to some extent, and only 2 (1.2%) patients answered that AD had little or no effect.

The mean (SD) DLQI was 13 (7.2). Based on the DLQI, 31 (18.6%) patients answered that AD had no effect or only a small effect on QOL (DLQI 0–5). In 36 (21.6%) patients, AD had a moderate effect on QOL (DLQI 6–10). The QOL impact was large (DLQI 11–20) and very large (DLQI 21–30) in 67 (40.1%) and 33 (19.8%) patients, respectively.

There was no significant difference in the impact of disease severity (POEM score) on the decrease of QOL (severe or very severe disease; P=.305, Mann-Whitney U statistic).

 

Absence From Work or Studies—At the study inclusion, 12 (7.2%) patients were not working or studying. Of the remaining 155 patients, 73 (47.1%) reported absence from work or educational activities due to AD in the last 12 months. The mean (SD) length of absence was 11.6 (10.2) days.

Comment

In this survey-based study of Finnish patients with moderate to severe AD, we observed that AD creates a substantial economic burden¹⁴ and negative impact on everyday life and QOL. According to DLQI, AD had a large or very large effect on most of the patients’ (59.9%) lives, and 90.2% of the included patients had self-reported moderate to very severe symptoms (POEM score 8–28). Our observations can partly be explained by characteristics of the Finnish health care system, in which patients with moderate to severe AD mainly are referred to specialist consultation. In the investigated cohort, many patients had used antibiotics (20.4%) and/or oral corticosteroids (21.6%) in the last year for the treatment of AD, which might indicate inadequate treatment of AD in the Finnish health care system.

Motivating patients to remain compliant is one of the main challenges in AD therapy.¹⁵ Fear of adverse effects from TCSs is common among patients and may cause poor treatment adherence.¹⁶ In a prospective study from the United Kingdom, the use of emollients in moderate to severe AD was considerably lower than AD guidelines recommend—approximately 10 g/d on average in adult patients. The median use of TCSs was between 35 and 38 g/mo.¹⁷ In our Finnish patient cohort, the amount of topical treatments was even lower, with a median use of emollients of 3.3 g/d and median use of TCSs of 25 g/mo. In another study from Denmark (N=322), 31% of patients with AD did not redeem their topical prescription medicaments, indicating poor adherence to topical treatment.¹⁸

It has been demonstrated that most of the patients’ habituation (tachyphylaxis) to TCSs is due to poor adherence instead of physiologic changes in tissue corticosteroid receptors.^19,20 Treatment adherence may be increased by scheduling early follow-up visits and providing adequate therapeutic patient education,²¹ which requires major efforts by the health care system and a financial investment.

Inadequate treatment will lead to more frequent disease flares and subsequently increase the medical costs for the patients and the health care system.²² In our Finnish patient cohort, a large part of direct treatment costs was due to inpatient treatment (Figure 2) even though only a small proportion of patients had been hospitalized. The patients were frequently young and otherwise in good general health, and they did not necessarily need continuous inpatient treatment and monitoring. In Finland, it will be necessary to develop more cost-effective treatment regimens for patients with AD with severe and frequent flares. Many patients would benefit from subsequent and regular sessions of topical treatment in an outpatient setting. In addition, the prevention of flares in moderate to severe AD will decrease medical costs.²³

The mean medication cost PPY was €457.40 (US $507.34), and mean total direct cost PPY was €1579.90 (US $1752.40), which indicates that AD causes a major economic burden to Finnish patients and to the Finnish health care system (Figures 1 and 2).²⁴ We did not observe significant differences between controlled and uncontrolled AD medical costs in our patient cohort (Figure 3), which may have been due to the relatively small sample size of only 16 patients in the controlled AD group. All patients attending the tertiary care hospital had moderate to severe AD, so it is likely that the patients with lower POEM scores had better-controlled disease. The POEM score estimates the grade of AD in the last 7 days, but based on the relapsing course of the disease, the grading score may differ substantially during the year in the same patient depending on the timing.^25,26

Topical calcineurin inhibitors comprised almost half of the medication costs (Figure 1), which may be caused by their higher prices compared with TCSs in Finland. In the beginning of 2019, a 50% less expensive biosimilar of tacrolimus ointment 0.1% was introduced to the Finnish market, which might decrease future treatment costs of TCIs. However, availability problems in both topical tacrolimus products were seen throughout 2019, which also may have affected the results in our study cohort. The median use of TCIs was unexpectedly low (only 30 g/y), which may be explained by different application habits. The use of large TCI amounts in some patients may have elevated mean costs.²⁷

 

In the Finnish public health care system, 40% of the cost for prescription medication and emollients is reimbursed after an initial deductible of €50. Emollients are reimbursed up to an amount of 1500 g/mo. Therefore, patients mostly acquired emollients as prescription medicine and not over-the-counter. Nonprescription medicaments were not included in our study, so the actual costs of topical treatment may have been higher.²⁸

In our cohort, 61.7% of the patients reported food allergies, and 70.1% reported allergic conjunctivitis. However, the study included only questionnaire-based data, and many of these patients probably had symptoms not associated with IgE-mediated allergies. The high prevalence indicates a substantial concomitant burden of more than skin symptoms in patients with AD.²⁹ Nine percent of patients reported a diagnosed psychiatric disorder, and 29.3% had self-reported anxiety or depression often or very often in the last year. Based on these findings, there may be high percentages of undiagnosed psychiatric comorbidities such as depression and anxiety disorders in patients with moderate to severe AD in Finland.³⁰ An important limitation of our study was that the patient data were based on a voluntary and anonymous survey and that depression and anxiety were addressed solely by a single question. In addition, the response rate cannot be analyzed correctly, and the demographics of the survey responders likely will differ substantially from all patients with AD at the university hospital.

Atopic dermatitis had a substantial effect on QOL in our patient cohort. Inadequate treatment of AD is known to negatively affect patient QOL and may lead to hospitalization or frequent oral corticosteroid courses.^31,32 In most cases, structured patient education and early follow-up visits may improve patient adherence to treatment and should be considered as an integral part of AD treatment.³³ In the investigated Finnish tertiary care hospital, a structured patient education system unfortunately was still lacking, though it has been proven effective elsewhere.³⁴ In addition, patient-centred educational programs are recommended in European guidelines for the treatment of AD.³⁵

Medical costs of AD may increase in the future as new treatments with higher direct costs, such as dupilumab, are introduced. Eichenfeld et al³⁶ analyzed electronic health plan claims in patients with AD with newly introduced systemic therapies and phototherapies after the availability of dupilumab in the United States (March 2017). Mean annualized total cost in all patients was $20,722; the highest in the dupilumab group with $36,505. Compared to our data, the total costs are much higher, but these are likely to rise in Finland in the future if a substantial amount (eg, 1%–5%) of patients will be on advanced therapies, including dupilumab. If advanced therapies will be introduced more broadly in Finland (eg, in the treatment of moderate AD [10%–20% of patients]), they will represent a major direct cost to the health care system. Zimmermann et al³⁷ showed in a cost-utility analysis that dupilumab improves health outcomes but with additional direct costs, and it is likely more cost-effective in patients with severe AD. Conversely, more efficient treatments may improve severe AD, reduce the need for hospitalization and recurrent doctors’ appointments as well as absence from work, and improve patient QOL,³⁸ consequently decreasing indirect medical costs and disease burden. Ariëns et al³⁹ showed in a recent registry-based study that dupilumab treatment induces a notable rise in work productivity and reduction of associated costs in patients with difficult-to-treat AD.

Conclusion

We aimed to analyze the economic burden of AD in Finland before the introduction of dupilumab. It will be interesting to see what the introduction of dupilumab and other novel systemic therapies have on total economic burden and medical costs. Most patients with AD in Finland can achieve disease control with topical treatments, but it is important to efficiently manage the patients who require additional supportive measures and specialist consultations, which may be challenging in the primary health care system because of the relapsing and remitting nature of the disease.

Atopic dermatitis (AD) is a common inflammatory skin disease that may severely decrease quality of life (QOL) and lead to psychiatric comorbidities.^1-3 Prior studies have indicated that AD causes a substantial economic burden, and disease severity has been proportionally linked to medical costs.^4,5 Results of a multicenter cost-of-illness study from Germany estimated that a relapse of AD costs approximately €123 (US $136). The authors calculated the average annual cost of AD per patient to be €1425 (US $1580), whereas it is €956 (US $1060) in moderate disease and €2068 (US $2293) in severe disease (direct and indirect medical costs included).⁶ An observational cohort study from the Netherlands found that total direct cost per patient-year (PPY) was €4401 (US $4879) for patients with controlled AD vs €6993 (US $7756) for patients with uncontrolled AD.⁷

In a retrospective survey-based study, it was estimated that the annual cost of AD in Canada was approximately CAD $1.4 billion. The cost per patient varied from CAD $282 to CAD $1242 depending on disease severity.⁸ In another retrospective cohort study from the Netherlands, the average direct medical cost per patient with AD seeing a general practitioner was US $71 during follow-up in primary care. If the patient needed specialist consultation, the cost increased to an average of US $186.⁹

We aimed to assess the direct and indirect medical costs in adult patients with moderate to severe AD who attended a tertiary health care center in Finland. In addition, we evaluated the impact of AD on QOL in this patient cohort.

Methods

Study Design—Patients with AD who were treated at the Department of Dermatology and Allergology, Helsinki University Hospital, Finland, between February 2018 and December 2019 were randomly selected to participate in our survey study. All participants provided written informed consent. In Finland, patients with mild AD generally are treated in primary health care centers, and only patients with moderate to severe AD are referred to specialists and tertiary care centers. Patients were excluded if they were younger than 18 years, had AD confined to the hands, or reported the presence of other concomitant skin diseases that were being treated with topical or systemic therapies. The protocol for the study was approved by the local ethics committee of the University of Helsinki.

Questionnaire and Analysis of Disease Severity—The survey included the medical history, signs of atopy, former treatment(s) for AD, skin infections, visits to dermatologists or general practitioners, questions on mental health and hospitalization, and absence from work due to AD in the last 12 months. Disease severity was evaluated using the patient-oriented Rajka & Langeland eczema severity score and Patient Oriented Eczema Measure (POEM).^10,11 The impact on QOL was evaluated by the Dermatology Life Quality Index (DLQI).¹²

Medication Costs—The cost of prescription drugs was based on data from the Finnish national electronic prescription center. In Finland, all prescriptions are made electronically in the database. We analyzed all topical medications (eg, topical corticosteroids [TCSs], topical calcineurin inhibitors [TCIs], and emollients) and systemic medicaments (eg, antibiotics, antihistamines, cyclosporine, methotrexate, and corticosteroids) prescribed for the treatment of AD. In Finland, dupilumab was introduced for the treatment of severe AD in early 2019, and patients receiving dupilumab were excluded from the study. Over-the-counter medications were not included. The costs for laboratory testing were estimations based on the standard monitoring protocols of the Helsinki University Hospital. All costs were based on the Finnish price level standard for the year 2019.

Inpatient/Outpatient Visits and Sick Leave Due to AD—The number of inpatient and outpatient visits due to AD in the last 12 months was evaluated. Outpatient specialist consultations or nurse appointments at Helsinki University Hospital were verified from electronic patient records. In addition, inpatient treatment and phototherapy sessions were calculated from the database.

 

We assessed the number of sick leave days from work or educational activities during the last year. All costs of transportation for doctors’ appointments, laboratory monitoring, and phototherapy treatments were summed together to estimate the total transportation cost. Visits to nurse and inpatient visits were not included in the total transportation cost because patients often were hospitalized directly after consultation visits, and nurse appointments often were combined with inpatient and outpatient visits. To calculate the total transportation cost, we used a rate of €0.43 per kilometer measured from the patients’ home addresses, which was the official compensation rate of the Finnish Tax Administration for 2019.¹³

Statistical Analysis—Statistical analyses were performed using SPSS Statistics 25 (IBM). Descriptive analyses were used to describe baseline characteristics and to evaluate the mean costs of AD. The patients were divided into 2 groups according to POEM: (1) controlled AD (patients with clear skin or only mild AD; POEM score 0–7) and (2) uncontrolled AD (patients with moderate to very severe AD; POEM score 8–28). The Mann-Whitney U statistic was used to evaluate differences between the study groups.

Results

Patient Characteristics—One hundred sixty-seven patients answered the survey, of which 69 (41.3%) were males and 98 (58.7%) were females. There were 16 patients with controlled AD and 148 patients with uncontrolled AD. Three patients did not answer to POEM and were excluded. The baseline characteristics are presented in Table 1 and include self-reported symptoms related to atopy.

The most-used topical treatments were TCSs (n=155; 92.8%) and emollients (n=166; 99.4%). One hundred sixteen (69.5%) patients had used TCIs. The median amount of TCSs used was 300 g/y vs 30 g/y for TCIs (range, 0-5160 g/y) and 1200 g/y for emollients.

Fifteen (9.0%) patients had been hospitalized for AD in the last year. The mean (SD) length of hospitalization was 6.5 (2.8) days. Thirty-four (20.4%) patients received UVB phototherapy. Thirty-four (20.4%) patients were treated with at least 1 antibiotic course for secondary AD infection. Thirty-six (21.6%) patients needed at least 1 oral corticosteroid course for the treatment of an AD flare.

Fifteen (9.0%) patients reported a diagnosed psychiatric illness, and 17 (10.2%) patients were using prescription drugs for psychiatric illness. Forty-nine (29.3%) patients reported anxiety or depression often or very often, 54 (32.3%) patients reported sometimes, 33 (19.8%) patients reported rarely, and only 30 (18.0%) patients reported none.

Medication Costs—Mean medication cost PPY was €457.40 (US $507.34)(Figure 1 and Table 2). On average, one patient spent €87.50 (US $97.05) for TCSs, €121.90 (US $135.21) for emollients, and €225.10 (US $249.68) for TCIs. The average cost PPY for antibiotics was €6.10 (US $6.77). Other systemic treatments, including (US $18.65). Seventeen patients (10.2%) were on methotrexate therapy for AD in the last year, and 1 patient also used cyclosporine. The costs for laboratory monitoring in these patients were included in the direct cost calculations. The mean cost PPY of laboratory monitoring in the whole study cohort was €6.60 (US $7.32). In patients with systemic immunosuppressive therapy, the mean cost PPY for laboratory monitoring was €65.00 (US $72.09). Five patients had been tested for contact dermatitis; the costs of patch tests or other diagnostic tests were not included.

 

Visits to Health Care Providers—In the last year, patients had an average of 1.83 dermatologist consultations in the tertiary center (Table 2). In addition, the mean number of visits to private dermatologists was 0.61 and 1.42 visits to general practitioners. The mean cost of physician visits was €302.70 (US $335.75) in the tertiary center, €66.60 (US $73.87) in the private sector, and €141.90 (US $157.39) in primary health care. In total, the average cost of doctors’ appointments PPY was €506.30 (US $561.57). The mean estimated distance traveled per visit was 9.5 km.

The mean cost PPY of inpatient treatments was €329.90 (US $365.92) and €239.00 (US $265.09) for UV phototherapy. Only 4 patients had visited a nurse in the last year, with an average cost PPY of €2.50 (US $2.78).

In total, the cost PPY for health care provider visits was €1084.20, which included specialist consultations in a tertiary center and private sector, visits in primary health care, inpatient treatments, UV phototherapy sessions, nurse appointments in a tertiary center, and laboratory monitoring. The average transportation cost PPY was €34.00 (US $37.71). The mean number of visits to health care providers was 8.3 per year. Altogether, the direct cost PPY in the study cohort was €1580.60 (US $1752.39)(Table 2 and Figure 2).

Comparison of Medical Costs in Controlled vs Uncontrolled AD—In the controlled AD group (POEM score <8), the mean medication cost PPY was €567.15 (US $629.13), and the mean total direct cost PPY was €2040.46 (US $2263.24). In the uncontrolled AD group (POEM score ≥8), the mean medication cost PPY was €449.55 (US $498.63), and the mean total direct cost PPY was €1539.39 (US $1707.36)(Table 2). The comparisons of the study groups—controlled vs uncontrolled AD—showed no significant differences regarding medication costs PPY (P=.305, Mann-Whitney U statistic) and total direct costs PPY (P=.361, Mann-Whitney U statistic)(Figure 3). Thus, the distribution of medical costs was similar across all categories of the POEM score.

AD Severity and QOL—The mean (SD) POEM score in the study cohort was 17.9 (6.9). Sixteen (9.6%) patients had clear to almost clear skin or mild AD (POEM score 0–7). Forty-two (25.1%) patients had moderate AD (POEM score 8–16). Most of the patients (106; 63.5%) had severe or very severe AD (POEM score 17–28). According to the Rajka & Langeland score, 5 (3.0%) patients had mild disease (score 34), 81 (48.5%) patients had moderate disease (score 5–7), and 81 (48.5%) patients had severe disease (score 8–9). Eighty-one (48.5%) patients answered that AD affects their lives greatly, and 58 (34.7%) patients answered that it affects their lives extremely. Twenty-five (15.0%) patients answered that AD affects their everyday life to some extent, and only 2 (1.2%) patients answered that AD had little or no effect.

The mean (SD) DLQI was 13 (7.2). Based on the DLQI, 31 (18.6%) patients answered that AD had no effect or only a small effect on QOL (DLQI 0–5). In 36 (21.6%) patients, AD had a moderate effect on QOL (DLQI 6–10). The QOL impact was large (DLQI 11–20) and very large (DLQI 21–30) in 67 (40.1%) and 33 (19.8%) patients, respectively.

There was no significant difference in the impact of disease severity (POEM score) on the decrease of QOL (severe or very severe disease; P=.305, Mann-Whitney U statistic).

 

Absence From Work or Studies—At the study inclusion, 12 (7.2%) patients were not working or studying. Of the remaining 155 patients, 73 (47.1%) reported absence from work or educational activities due to AD in the last 12 months. The mean (SD) length of absence was 11.6 (10.2) days.

Comment

In this survey-based study of Finnish patients with moderate to severe AD, we observed that AD creates a substantial economic burden¹⁴ and negative impact on everyday life and QOL. According to DLQI, AD had a large or very large effect on most of the patients’ (59.9%) lives, and 90.2% of the included patients had self-reported moderate to very severe symptoms (POEM score 8–28). Our observations can partly be explained by characteristics of the Finnish health care system, in which patients with moderate to severe AD mainly are referred to specialist consultation. In the investigated cohort, many patients had used antibiotics (20.4%) and/or oral corticosteroids (21.6%) in the last year for the treatment of AD, which might indicate inadequate treatment of AD in the Finnish health care system.

Motivating patients to remain compliant is one of the main challenges in AD therapy.¹⁵ Fear of adverse effects from TCSs is common among patients and may cause poor treatment adherence.¹⁶ In a prospective study from the United Kingdom, the use of emollients in moderate to severe AD was considerably lower than AD guidelines recommend—approximately 10 g/d on average in adult patients. The median use of TCSs was between 35 and 38 g/mo.¹⁷ In our Finnish patient cohort, the amount of topical treatments was even lower, with a median use of emollients of 3.3 g/d and median use of TCSs of 25 g/mo. In another study from Denmark (N=322), 31% of patients with AD did not redeem their topical prescription medicaments, indicating poor adherence to topical treatment.¹⁸

It has been demonstrated that most of the patients’ habituation (tachyphylaxis) to TCSs is due to poor adherence instead of physiologic changes in tissue corticosteroid receptors.^19,20 Treatment adherence may be increased by scheduling early follow-up visits and providing adequate therapeutic patient education,²¹ which requires major efforts by the health care system and a financial investment.

Inadequate treatment will lead to more frequent disease flares and subsequently increase the medical costs for the patients and the health care system.²² In our Finnish patient cohort, a large part of direct treatment costs was due to inpatient treatment (Figure 2) even though only a small proportion of patients had been hospitalized. The patients were frequently young and otherwise in good general health, and they did not necessarily need continuous inpatient treatment and monitoring. In Finland, it will be necessary to develop more cost-effective treatment regimens for patients with AD with severe and frequent flares. Many patients would benefit from subsequent and regular sessions of topical treatment in an outpatient setting. In addition, the prevention of flares in moderate to severe AD will decrease medical costs.²³

The mean medication cost PPY was €457.40 (US $507.34), and mean total direct cost PPY was €1579.90 (US $1752.40), which indicates that AD causes a major economic burden to Finnish patients and to the Finnish health care system (Figures 1 and 2).²⁴ We did not observe significant differences between controlled and uncontrolled AD medical costs in our patient cohort (Figure 3), which may have been due to the relatively small sample size of only 16 patients in the controlled AD group. All patients attending the tertiary care hospital had moderate to severe AD, so it is likely that the patients with lower POEM scores had better-controlled disease. The POEM score estimates the grade of AD in the last 7 days, but based on the relapsing course of the disease, the grading score may differ substantially during the year in the same patient depending on the timing.^25,26

Topical calcineurin inhibitors comprised almost half of the medication costs (Figure 1), which may be caused by their higher prices compared with TCSs in Finland. In the beginning of 2019, a 50% less expensive biosimilar of tacrolimus ointment 0.1% was introduced to the Finnish market, which might decrease future treatment costs of TCIs. However, availability problems in both topical tacrolimus products were seen throughout 2019, which also may have affected the results in our study cohort. The median use of TCIs was unexpectedly low (only 30 g/y), which may be explained by different application habits. The use of large TCI amounts in some patients may have elevated mean costs.²⁷

 

In the Finnish public health care system, 40% of the cost for prescription medication and emollients is reimbursed after an initial deductible of €50. Emollients are reimbursed up to an amount of 1500 g/mo. Therefore, patients mostly acquired emollients as prescription medicine and not over-the-counter. Nonprescription medicaments were not included in our study, so the actual costs of topical treatment may have been higher.²⁸

In our cohort, 61.7% of the patients reported food allergies, and 70.1% reported allergic conjunctivitis. However, the study included only questionnaire-based data, and many of these patients probably had symptoms not associated with IgE-mediated allergies. The high prevalence indicates a substantial concomitant burden of more than skin symptoms in patients with AD.²⁹ Nine percent of patients reported a diagnosed psychiatric disorder, and 29.3% had self-reported anxiety or depression often or very often in the last year. Based on these findings, there may be high percentages of undiagnosed psychiatric comorbidities such as depression and anxiety disorders in patients with moderate to severe AD in Finland.³⁰ An important limitation of our study was that the patient data were based on a voluntary and anonymous survey and that depression and anxiety were addressed solely by a single question. In addition, the response rate cannot be analyzed correctly, and the demographics of the survey responders likely will differ substantially from all patients with AD at the university hospital.

Atopic dermatitis had a substantial effect on QOL in our patient cohort. Inadequate treatment of AD is known to negatively affect patient QOL and may lead to hospitalization or frequent oral corticosteroid courses.^31,32 In most cases, structured patient education and early follow-up visits may improve patient adherence to treatment and should be considered as an integral part of AD treatment.³³ In the investigated Finnish tertiary care hospital, a structured patient education system unfortunately was still lacking, though it has been proven effective elsewhere.³⁴ In addition, patient-centred educational programs are recommended in European guidelines for the treatment of AD.³⁵

Medical costs of AD may increase in the future as new treatments with higher direct costs, such as dupilumab, are introduced. Eichenfeld et al³⁶ analyzed electronic health plan claims in patients with AD with newly introduced systemic therapies and phototherapies after the availability of dupilumab in the United States (March 2017). Mean annualized total cost in all patients was $20,722; the highest in the dupilumab group with $36,505. Compared to our data, the total costs are much higher, but these are likely to rise in Finland in the future if a substantial amount (eg, 1%–5%) of patients will be on advanced therapies, including dupilumab. If advanced therapies will be introduced more broadly in Finland (eg, in the treatment of moderate AD [10%–20% of patients]), they will represent a major direct cost to the health care system. Zimmermann et al³⁷ showed in a cost-utility analysis that dupilumab improves health outcomes but with additional direct costs, and it is likely more cost-effective in patients with severe AD. Conversely, more efficient treatments may improve severe AD, reduce the need for hospitalization and recurrent doctors’ appointments as well as absence from work, and improve patient QOL,³⁸ consequently decreasing indirect medical costs and disease burden. Ariëns et al³⁹ showed in a recent registry-based study that dupilumab treatment induces a notable rise in work productivity and reduction of associated costs in patients with difficult-to-treat AD.

Conclusion

We aimed to analyze the economic burden of AD in Finland before the introduction of dupilumab. It will be interesting to see what the introduction of dupilumab and other novel systemic therapies have on total economic burden and medical costs. Most patients with AD in Finland can achieve disease control with topical treatments, but it is important to efficiently manage the patients who require additional supportive measures and specialist consultations, which may be challenging in the primary health care system because of the relapsing and remitting nature of the disease.

Atopic dermatitis (AD) is a common inflammatory skin disease that may severely decrease quality of life (QOL) and lead to psychiatric comorbidities.^1-3 Prior studies have indicated that AD causes a substantial economic burden, and disease severity has been proportionally linked to medical costs.^4,5 Results of a multicenter cost-of-illness study from Germany estimated that a relapse of AD costs approximately €123 (US $136). The authors calculated the average annual cost of AD per patient to be €1425 (US $1580), whereas it is €956 (US $1060) in moderate disease and €2068 (US $2293) in severe disease (direct and indirect medical costs included).⁶ An observational cohort study from the Netherlands found that total direct cost per patient-year (PPY) was €4401 (US $4879) for patients with controlled AD vs €6993 (US $7756) for patients with uncontrolled AD.⁷

In a retrospective survey-based study, it was estimated that the annual cost of AD in Canada was approximately CAD $1.4 billion. The cost per patient varied from CAD $282 to CAD $1242 depending on disease severity.⁸ In another retrospective cohort study from the Netherlands, the average direct medical cost per patient with AD seeing a general practitioner was US $71 during follow-up in primary care. If the patient needed specialist consultation, the cost increased to an average of US $186.⁹

We aimed to assess the direct and indirect medical costs in adult patients with moderate to severe AD who attended a tertiary health care center in Finland. In addition, we evaluated the impact of AD on QOL in this patient cohort.

Methods

Study Design—Patients with AD who were treated at the Department of Dermatology and Allergology, Helsinki University Hospital, Finland, between February 2018 and December 2019 were randomly selected to participate in our survey study. All participants provided written informed consent. In Finland, patients with mild AD generally are treated in primary health care centers, and only patients with moderate to severe AD are referred to specialists and tertiary care centers. Patients were excluded if they were younger than 18 years, had AD confined to the hands, or reported the presence of other concomitant skin diseases that were being treated with topical or systemic therapies. The protocol for the study was approved by the local ethics committee of the University of Helsinki.

Questionnaire and Analysis of Disease Severity—The survey included the medical history, signs of atopy, former treatment(s) for AD, skin infections, visits to dermatologists or general practitioners, questions on mental health and hospitalization, and absence from work due to AD in the last 12 months. Disease severity was evaluated using the patient-oriented Rajka & Langeland eczema severity score and Patient Oriented Eczema Measure (POEM).^10,11 The impact on QOL was evaluated by the Dermatology Life Quality Index (DLQI).¹²

Medication Costs—The cost of prescription drugs was based on data from the Finnish national electronic prescription center. In Finland, all prescriptions are made electronically in the database. We analyzed all topical medications (eg, topical corticosteroids [TCSs], topical calcineurin inhibitors [TCIs], and emollients) and systemic medicaments (eg, antibiotics, antihistamines, cyclosporine, methotrexate, and corticosteroids) prescribed for the treatment of AD. In Finland, dupilumab was introduced for the treatment of severe AD in early 2019, and patients receiving dupilumab were excluded from the study. Over-the-counter medications were not included. The costs for laboratory testing were estimations based on the standard monitoring protocols of the Helsinki University Hospital. All costs were based on the Finnish price level standard for the year 2019.

Inpatient/Outpatient Visits and Sick Leave Due to AD—The number of inpatient and outpatient visits due to AD in the last 12 months was evaluated. Outpatient specialist consultations or nurse appointments at Helsinki University Hospital were verified from electronic patient records. In addition, inpatient treatment and phototherapy sessions were calculated from the database.

 

We assessed the number of sick leave days from work or educational activities during the last year. All costs of transportation for doctors’ appointments, laboratory monitoring, and phototherapy treatments were summed together to estimate the total transportation cost. Visits to nurse and inpatient visits were not included in the total transportation cost because patients often were hospitalized directly after consultation visits, and nurse appointments often were combined with inpatient and outpatient visits. To calculate the total transportation cost, we used a rate of €0.43 per kilometer measured from the patients’ home addresses, which was the official compensation rate of the Finnish Tax Administration for 2019.¹³

Statistical Analysis—Statistical analyses were performed using SPSS Statistics 25 (IBM). Descriptive analyses were used to describe baseline characteristics and to evaluate the mean costs of AD. The patients were divided into 2 groups according to POEM: (1) controlled AD (patients with clear skin or only mild AD; POEM score 0–7) and (2) uncontrolled AD (patients with moderate to very severe AD; POEM score 8–28). The Mann-Whitney U statistic was used to evaluate differences between the study groups.

Results

Patient Characteristics—One hundred sixty-seven patients answered the survey, of which 69 (41.3%) were males and 98 (58.7%) were females. There were 16 patients with controlled AD and 148 patients with uncontrolled AD. Three patients did not answer to POEM and were excluded. The baseline characteristics are presented in Table 1 and include self-reported symptoms related to atopy.

The most-used topical treatments were TCSs (n=155; 92.8%) and emollients (n=166; 99.4%). One hundred sixteen (69.5%) patients had used TCIs. The median amount of TCSs used was 300 g/y vs 30 g/y for TCIs (range, 0-5160 g/y) and 1200 g/y for emollients.

Fifteen (9.0%) patients had been hospitalized for AD in the last year. The mean (SD) length of hospitalization was 6.5 (2.8) days. Thirty-four (20.4%) patients received UVB phototherapy. Thirty-four (20.4%) patients were treated with at least 1 antibiotic course for secondary AD infection. Thirty-six (21.6%) patients needed at least 1 oral corticosteroid course for the treatment of an AD flare.

Fifteen (9.0%) patients reported a diagnosed psychiatric illness, and 17 (10.2%) patients were using prescription drugs for psychiatric illness. Forty-nine (29.3%) patients reported anxiety or depression often or very often, 54 (32.3%) patients reported sometimes, 33 (19.8%) patients reported rarely, and only 30 (18.0%) patients reported none.

Medication Costs—Mean medication cost PPY was €457.40 (US $507.34)(Figure 1 and Table 2). On average, one patient spent €87.50 (US $97.05) for TCSs, €121.90 (US $135.21) for emollients, and €225.10 (US $249.68) for TCIs. The average cost PPY for antibiotics was €6.10 (US $6.77). Other systemic treatments, including (US $18.65). Seventeen patients (10.2%) were on methotrexate therapy for AD in the last year, and 1 patient also used cyclosporine. The costs for laboratory monitoring in these patients were included in the direct cost calculations. The mean cost PPY of laboratory monitoring in the whole study cohort was €6.60 (US $7.32). In patients with systemic immunosuppressive therapy, the mean cost PPY for laboratory monitoring was €65.00 (US $72.09). Five patients had been tested for contact dermatitis; the costs of patch tests or other diagnostic tests were not included.

 

Visits to Health Care Providers—In the last year, patients had an average of 1.83 dermatologist consultations in the tertiary center (Table 2). In addition, the mean number of visits to private dermatologists was 0.61 and 1.42 visits to general practitioners. The mean cost of physician visits was €302.70 (US $335.75) in the tertiary center, €66.60 (US $73.87) in the private sector, and €141.90 (US $157.39) in primary health care. In total, the average cost of doctors’ appointments PPY was €506.30 (US $561.57). The mean estimated distance traveled per visit was 9.5 km.

The mean cost PPY of inpatient treatments was €329.90 (US $365.92) and €239.00 (US $265.09) for UV phototherapy. Only 4 patients had visited a nurse in the last year, with an average cost PPY of €2.50 (US $2.78).

In total, the cost PPY for health care provider visits was €1084.20, which included specialist consultations in a tertiary center and private sector, visits in primary health care, inpatient treatments, UV phototherapy sessions, nurse appointments in a tertiary center, and laboratory monitoring. The average transportation cost PPY was €34.00 (US $37.71). The mean number of visits to health care providers was 8.3 per year. Altogether, the direct cost PPY in the study cohort was €1580.60 (US $1752.39)(Table 2 and Figure 2).

Comparison of Medical Costs in Controlled vs Uncontrolled AD—In the controlled AD group (POEM score <8), the mean medication cost PPY was €567.15 (US $629.13), and the mean total direct cost PPY was €2040.46 (US $2263.24). In the uncontrolled AD group (POEM score ≥8), the mean medication cost PPY was €449.55 (US $498.63), and the mean total direct cost PPY was €1539.39 (US $1707.36)(Table 2). The comparisons of the study groups—controlled vs uncontrolled AD—showed no significant differences regarding medication costs PPY (P=.305, Mann-Whitney U statistic) and total direct costs PPY (P=.361, Mann-Whitney U statistic)(Figure 3). Thus, the distribution of medical costs was similar across all categories of the POEM score.

AD Severity and QOL—The mean (SD) POEM score in the study cohort was 17.9 (6.9). Sixteen (9.6%) patients had clear to almost clear skin or mild AD (POEM score 0–7). Forty-two (25.1%) patients had moderate AD (POEM score 8–16). Most of the patients (106; 63.5%) had severe or very severe AD (POEM score 17–28). According to the Rajka & Langeland score, 5 (3.0%) patients had mild disease (score 34), 81 (48.5%) patients had moderate disease (score 5–7), and 81 (48.5%) patients had severe disease (score 8–9). Eighty-one (48.5%) patients answered that AD affects their lives greatly, and 58 (34.7%) patients answered that it affects their lives extremely. Twenty-five (15.0%) patients answered that AD affects their everyday life to some extent, and only 2 (1.2%) patients answered that AD had little or no effect.

The mean (SD) DLQI was 13 (7.2). Based on the DLQI, 31 (18.6%) patients answered that AD had no effect or only a small effect on QOL (DLQI 0–5). In 36 (21.6%) patients, AD had a moderate effect on QOL (DLQI 6–10). The QOL impact was large (DLQI 11–20) and very large (DLQI 21–30) in 67 (40.1%) and 33 (19.8%) patients, respectively.

There was no significant difference in the impact of disease severity (POEM score) on the decrease of QOL (severe or very severe disease; P=.305, Mann-Whitney U statistic).

 

Absence From Work or Studies—At the study inclusion, 12 (7.2%) patients were not working or studying. Of the remaining 155 patients, 73 (47.1%) reported absence from work or educational activities due to AD in the last 12 months. The mean (SD) length of absence was 11.6 (10.2) days.

Comment

In this survey-based study of Finnish patients with moderate to severe AD, we observed that AD creates a substantial economic burden¹⁴ and negative impact on everyday life and QOL. According to DLQI, AD had a large or very large effect on most of the patients’ (59.9%) lives, and 90.2% of the included patients had self-reported moderate to very severe symptoms (POEM score 8–28). Our observations can partly be explained by characteristics of the Finnish health care system, in which patients with moderate to severe AD mainly are referred to specialist consultation. In the investigated cohort, many patients had used antibiotics (20.4%) and/or oral corticosteroids (21.6%) in the last year for the treatment of AD, which might indicate inadequate treatment of AD in the Finnish health care system.

Motivating patients to remain compliant is one of the main challenges in AD therapy.¹⁵ Fear of adverse effects from TCSs is common among patients and may cause poor treatment adherence.¹⁶ In a prospective study from the United Kingdom, the use of emollients in moderate to severe AD was considerably lower than AD guidelines recommend—approximately 10 g/d on average in adult patients. The median use of TCSs was between 35 and 38 g/mo.¹⁷ In our Finnish patient cohort, the amount of topical treatments was even lower, with a median use of emollients of 3.3 g/d and median use of TCSs of 25 g/mo. In another study from Denmark (N=322), 31% of patients with AD did not redeem their topical prescription medicaments, indicating poor adherence to topical treatment.¹⁸

It has been demonstrated that most of the patients’ habituation (tachyphylaxis) to TCSs is due to poor adherence instead of physiologic changes in tissue corticosteroid receptors.^19,20 Treatment adherence may be increased by scheduling early follow-up visits and providing adequate therapeutic patient education,²¹ which requires major efforts by the health care system and a financial investment.

Inadequate treatment will lead to more frequent disease flares and subsequently increase the medical costs for the patients and the health care system.²² In our Finnish patient cohort, a large part of direct treatment costs was due to inpatient treatment (Figure 2) even though only a small proportion of patients had been hospitalized. The patients were frequently young and otherwise in good general health, and they did not necessarily need continuous inpatient treatment and monitoring. In Finland, it will be necessary to develop more cost-effective treatment regimens for patients with AD with severe and frequent flares. Many patients would benefit from subsequent and regular sessions of topical treatment in an outpatient setting. In addition, the prevention of flares in moderate to severe AD will decrease medical costs.²³

The mean medication cost PPY was €457.40 (US $507.34), and mean total direct cost PPY was €1579.90 (US $1752.40), which indicates that AD causes a major economic burden to Finnish patients and to the Finnish health care system (Figures 1 and 2).²⁴ We did not observe significant differences between controlled and uncontrolled AD medical costs in our patient cohort (Figure 3), which may have been due to the relatively small sample size of only 16 patients in the controlled AD group. All patients attending the tertiary care hospital had moderate to severe AD, so it is likely that the patients with lower POEM scores had better-controlled disease. The POEM score estimates the grade of AD in the last 7 days, but based on the relapsing course of the disease, the grading score may differ substantially during the year in the same patient depending on the timing.^25,26

Topical calcineurin inhibitors comprised almost half of the medication costs (Figure 1), which may be caused by their higher prices compared with TCSs in Finland. In the beginning of 2019, a 50% less expensive biosimilar of tacrolimus ointment 0.1% was introduced to the Finnish market, which might decrease future treatment costs of TCIs. However, availability problems in both topical tacrolimus products were seen throughout 2019, which also may have affected the results in our study cohort. The median use of TCIs was unexpectedly low (only 30 g/y), which may be explained by different application habits. The use of large TCI amounts in some patients may have elevated mean costs.²⁷

 

In the Finnish public health care system, 40% of the cost for prescription medication and emollients is reimbursed after an initial deductible of €50. Emollients are reimbursed up to an amount of 1500 g/mo. Therefore, patients mostly acquired emollients as prescription medicine and not over-the-counter. Nonprescription medicaments were not included in our study, so the actual costs of topical treatment may have been higher.²⁸

In our cohort, 61.7% of the patients reported food allergies, and 70.1% reported allergic conjunctivitis. However, the study included only questionnaire-based data, and many of these patients probably had symptoms not associated with IgE-mediated allergies. The high prevalence indicates a substantial concomitant burden of more than skin symptoms in patients with AD.²⁹ Nine percent of patients reported a diagnosed psychiatric disorder, and 29.3% had self-reported anxiety or depression often or very often in the last year. Based on these findings, there may be high percentages of undiagnosed psychiatric comorbidities such as depression and anxiety disorders in patients with moderate to severe AD in Finland.³⁰ An important limitation of our study was that the patient data were based on a voluntary and anonymous survey and that depression and anxiety were addressed solely by a single question. In addition, the response rate cannot be analyzed correctly, and the demographics of the survey responders likely will differ substantially from all patients with AD at the university hospital.

Atopic dermatitis had a substantial effect on QOL in our patient cohort. Inadequate treatment of AD is known to negatively affect patient QOL and may lead to hospitalization or frequent oral corticosteroid courses.^31,32 In most cases, structured patient education and early follow-up visits may improve patient adherence to treatment and should be considered as an integral part of AD treatment.³³ In the investigated Finnish tertiary care hospital, a structured patient education system unfortunately was still lacking, though it has been proven effective elsewhere.³⁴ In addition, patient-centred educational programs are recommended in European guidelines for the treatment of AD.³⁵

Medical costs of AD may increase in the future as new treatments with higher direct costs, such as dupilumab, are introduced. Eichenfeld et al³⁶ analyzed electronic health plan claims in patients with AD with newly introduced systemic therapies and phototherapies after the availability of dupilumab in the United States (March 2017). Mean annualized total cost in all patients was $20,722; the highest in the dupilumab group with $36,505. Compared to our data, the total costs are much higher, but these are likely to rise in Finland in the future if a substantial amount (eg, 1%–5%) of patients will be on advanced therapies, including dupilumab. If advanced therapies will be introduced more broadly in Finland (eg, in the treatment of moderate AD [10%–20% of patients]), they will represent a major direct cost to the health care system. Zimmermann et al³⁷ showed in a cost-utility analysis that dupilumab improves health outcomes but with additional direct costs, and it is likely more cost-effective in patients with severe AD. Conversely, more efficient treatments may improve severe AD, reduce the need for hospitalization and recurrent doctors’ appointments as well as absence from work, and improve patient QOL,³⁸ consequently decreasing indirect medical costs and disease burden. Ariëns et al³⁹ showed in a recent registry-based study that dupilumab treatment induces a notable rise in work productivity and reduction of associated costs in patients with difficult-to-treat AD.

Conclusion

We aimed to analyze the economic burden of AD in Finland before the introduction of dupilumab. It will be interesting to see what the introduction of dupilumab and other novel systemic therapies have on total economic burden and medical costs. Most patients with AD in Finland can achieve disease control with topical treatments, but it is important to efficiently manage the patients who require additional supportive measures and specialist consultations, which may be challenging in the primary health care system because of the relapsing and remitting nature of the disease.

Dermatomyositis (DM) is an uncommon idiopathic inflammatory myopathy (IIM) characterized by muscle inflammation; proximal muscle weakness; and dermatologic findings, such as the heliotrope eruption and Gottron papules.^1-3 Dermatomyositis is associated with an increased malignancy risk compared to other IIMs, with a 13% to 42% lifetime risk for malignancy development.^4,5 The incidence for malignancy peaks during the first year following diagnosis and falls gradually over 5 years but remains increased compared to the general population.^6-11 Adenocarcinoma represents the majority of cancers associated with DM, particularly of the ovaries, lungs, breasts, gastrointestinal tract, pancreas, bladder, and prostate. The lymphatic system (non-Hodgkin lymphoma) also is overrepresented among cancers in DM.¹²

Because of the increased malignancy risk and cancer-related mortality in patients with DM, cancer screening generally is recommended following diagnosis.^13,14 However, consensus guidelines for screening modalities and frequency currently do not exist, resulting in widely varying practice patterns.¹⁵ Some experts advocate for a conventional cancer screening panel (CSP), as summarized in Table 1.^15-18 These tests may be repeated annually for 3 to 5 years following the diagnosis of DM. Although the use of myositis-specific antibodies (MSAs) recently has helped to risk-stratify DM patients, up to half of patients are MSA negative,¹⁹ and broad malignancy screening remains essential. Individualized discussions with patients about their risk factors, screening options, and risks and benefits of screening also are strongly encouraged.^19-22 Studies of the direct costs and effectiveness of streamlined screening with positron emission tomography/computed tomography (PET/CT) compared with a CSP have shown similar efficacy and lower out-of-pocket costs for patients receiving PET/CT imaging.^16-18

The goal of our study was to further characterize patients’ perspectives and experience of cancer screening in DM as well as indirect costs, both of which must be taken into consideration when developing consensus guidelines for DM malignancy screening. Inclusion of patient voice is essential given the similar efficacy of both screening methods. We assessed the indirect costs (eg, travel, lost work or wages, childcare) of a CSP in patients with DM. We theorized that the large quantity of tests involved in a CSP, which are performed at various locations on multiple days over the course of several years, may have substantial costs to patients beyond the co-pay and deductible. We also sought to measure patients’ perception of the burden associated with an annual CSP, which we defined to participants as the inconvenience or unpleasantness experienced by the patient, compared with an annual whole-body PET/CT. Finally, we examined the relative value of these screening methods to patients using a willingness-to-pay (WTP) analysis.

Materials and Methods

Patient Eligibility—Our study included Penn State Health (Hershey, Pennsylvania) patients 18 years or older with a recent diagnosis of DM—International Classification of Diseases, Ninth Revision code 710.3 or International Classification of Diseases, Tenth Revision codes M33.10 or M33.90—who were undergoing or had recently completed a CSP. Patients were excluded from the study if they had a concurrent or preceding diagnosis of malignancy (excluding nonmelanoma skin cancers) or had another IIM. The institutional review board at Penn State Health College of Medicine approved the study. Data for all patients were prospectively obtained.

Survey Design—A survey was generated to assess the burden and indirect costs associated with a CSP, which was modified from work done by Tchuenche et al²³ and Teni et al.²⁴ Focus groups were held in 2018 and 2019 with patients who met our inclusion criteria with the purpose of refining the survey instrument based on patient input. A summary explanation of research was provided to all participants, and informed consent was obtained. Patients were compensated for their time for focus groups. Audio of each focus group was then transcribed and analyzed for common themes. Following focus group feedback, a finalized survey was generated for assessing burden and indirect costs (survey instrument provided in the Supplementary Information). REDCap (Vanderbilt University), a secure web application, was used to construct the finalized survey and to collect and manage data.²⁵

Patients who fit our inclusion criteria were identified and recruited in multiple ways. Patients with appointments at the Penn State Milton S. Hershey Medical Center Department of Dermatology were presented with the opportunity to participate, Penn State Health records with the appropriate billing codes were collected and patients were contacted, and an advertisement for the study was posted on StudyFinder. Surveys constructed on REDCap were then sent electronically to patients who agreed to participate in the study. A second summary explanation of research was included on the first page of the survey to describe the process.

The survey had 3 main sections. The first section collected demographic information. In the second section, we surveyed patients regarding the various aspects of a CSP that focus groups identified as burdensome. In addition, patients were asked to compare their feelings regarding an annual CSP vs whole-body PET/CT for a 3-year period utilizing a rating scale of strongly disagree, somewhat disagree, somewhat agree, and strongly agree. This section also included a willingness-to-pay (WTP) analysis for each modality. We defined WTP as the maximum out-of-pocket cost that the patient would be willing to pay to receive testing, which was measured in a hypothetical scenario where neither whole-body PET/CT nor CSP was covered by insurance.²⁶ Although WTP may be influenced by external factors such as patient income, it can serve as a numerical measure of how much the patient values each service. Furthermore, these external factors become less relevant when comparing the relative value of 2 separate tests, as such factors apply equally in both scenarios. In the third section of the survey, patients were queried regarding various indirect costs associated with a CSP. Descriptions for a CSP and whole-body PET/CT, including risks and benefits, were provided to allow patients to make informed decisions.

 

Statistical Analysis—Because of the rarity of DM and the subsequently limited sample size, summary and descriptive statistics were utilized to characterize the sample and identify patterns in the results. Continuous variables are presented with means and standard deviations, and proportions are presented with frequencies and percentages. All analyses were done using SAS Version 9.4 (SAS Institute Inc).

Results

Patient Demographics—Fifty-four patients were identified using StudyFinder, physician referral, and search of the electronic health record. Nine patients agreed to take part in the focus groups, and 27 offered email addresses to be contacted for the survey. Of those 27 patients, 16 (59.3%) fit our inclusion criteria and completed the survey. Patient demographics are detailed in Table 2. The mean age was 55 years, and most patients were White (88% [14/16]), female (81% [13/16]), and had at least a bachelor’s degree (69% [11/16]). Most patients (69% [11/16]) had an annual income of less than $50,000, and half (50% [8/16]) were employed. All patients had been diagnosed with DM in or after 2013. Two patients were diagnosed with basal cell carcinoma during or after cancer screening.

Patient Preference for Screening and WTP—A majority (81% [13/16]) of patients desired some form of screening for occult malignancy following the diagnosis of DM, even in the hypothetical situation in which screening did not provide survival benefit (Figure 1). Twenty-five percent (4/16) of patients expressed that a CSP was burdensome, and 12.5% of patients (2/16) missed a CSP appointment; all of these patients rescheduled or were planning to reschedule. Assuming that both screening methods had similar predictive value in detecting malignancy, all 16 patients felt annual whole-body PET/CT for a 3-year period would be less burdensome than a CSP, and most (73% [11/15]) felt that it would decrease the likelihood of missed appointments. Overall, 93% (13/14) of patients preferred whole-body PET/CT over a CSP when given the choice between the 2 options (Figure 2). This preference was consistent with the patients’ WTP for these tests; patients reliably reported that they would pay more for annual whole-body PET/CT than for a CSP (Figure 3). Specifically, 75% (12/16) and 38% (6/16) of patients were willing to spend $250 or more and $1000 or more for annual whole-body PET/CT, respectively, compared with 56% (9/16) and 19% (3/16), respectively, for an annual CSP. Many patients (38% [6/16]) reported that they would not be willing to pay any out-of-pocket cost for a CSP compared with 13% (2/16) for PET/CT.Indirect Costs of Screening for Patients—Indirect costs incurred by patients undergoing a CSP are summarized in Table 3. Specifically, a large percentage of employed patients missed work (63% [5/8]) or had family miss work (38% [3/8]), necessitating the use of vacation and/or sick days to attend CSP appointments. A subset (25% [2/8]) lost income (average, $1500), and 1 patient reported that a family member lost income due to attending a CSP appointment. Most (75% [12/16]) patients also incurred substantial transportation costs (average, $243), with 1 patient spending $1000. No patients incurred child or elder care costs. One patient paid a small sum for lodging/meals while traveling to attend a CSP appointment.

Comment

Patients with DM have an increased incidence of malignancy, thus cancer screening serves a crucial role in the detection of occult disease.¹³ Up to half of DM patients are MSA negative, and most cancers in these patients are found with blind screening. Whole-body PET/CT has emerged as an alternative to a CSP. Evidence suggests that it has similar efficacy in detecting malignancy and may be particularly useful for identifying malignancies not routinely screened for in a CSP. In a prospective study of patients diagnosed with DM and polymyositis (N=55), whole-body PET/CT had a positive predictive value of 85.7% and negative predictive value for detecting occult malignancy of 93.8% compared with 77.8% and 95.7%, respectively, for a CSP.¹⁷

The results of our study showed that cancer screening is important to patients diagnosed with DM and that most of these patients desire some form of cancer screening. This finding held true even when patients were presented with a hypothetical situation in which screening was proven to have no survival benefit. Based on focus group data, this desire was likely driven by the fear generated by not knowing whether cancer is present, as reported by the following DM patients:

“I mean [cancer screening] is peace of mind. It is ultimately worth it. You know, better than . . . not doing the screenings and finding 3 years down the road that you have, you know, a serious problem . . . you had the cancer, and you didn’t have the screenings.” (DM patient 1)

“I would rather know than not know, even if it is bad news, just tell me. The sooner the better, and give me the whole spiel . . . maybe all the screenings don’t need to be done, done so much, so often afterwards if the initial ones are ok, but I think too, for peace of mind, I would rather know it all up front.” (DM patient 2)

 

Further, when presented with the hypothetical situation that insurance would not cover screenings, a few patients remarked they would relocate to obtain them:

“I would find a place where the screenings were done. I’d move.” (DM patient 4)

“If it was just sky high and [insurance companies] weren’t willing to negotiate, I would consider moving.” (DM patient 3).

Sentiments such as these emphasize the importance and value that DM patients place on being screened for cancer and also may explain why only 25% of patients felt a CSP was burdensome and only 13% reported missing appointments, all of whom planned on making them up at a later time.

When presented with the choice of a CSP or annual whole-body PET/CT for a 3-year period following the diagnosis of DM, all patients expressed that whole-body PET/CT would be less burdensome. Most preferred annual whole-body PET/CT despite the slightly increased radiation exposure associated and thought that it would limit missed appointments. Accordingly, more patients responded that they would pay more money out-of-pocket for annual whole-body PET/CT. Given that WTP can function as a numerical measure of value, our results showed that patients placed a higher value on whole-body PET/CT compared with a CSP. The indirect costs associated with a CSP also were substantial, particularly regarding missed work, use of vacation and/or sick days, and travel expenses, which is particularly important because most patients reported an annual income less than $50,000.

The direct costs of a CSP and whole-body PET/CT have been studied. Specifically, Kundrick et al¹⁸ found that whole-body PET/CT was less expensive for patients (by approximately $111) out-of-pocket compared with a CSP, though cost to insurance companies was slightly greater. The present study adds to these findings by better illustrating the burden and indirect costs that patients experience while undergoing a CSP and by characterizing the patient’s perception and preference of these 2 screening methods.

Limitations of our study include a small sample size willing to complete the survey. There also was a predominance of White and female participants, partially attributed to the greater number of female patients who develop DM compared to male patients. However, this still may limit applicability of this study to males and patients of other races. Another limitation includes recall bias on survey responses, particularly regarding indirect costs incurred with a CSP. A final limitation was that only patients with a recent diagnosis of DM who were actively undergoing screening or had recently completed malignancy screening were included in the study. Given that these patients were receiving (or had completed) exclusively a CSP, patients were comparing their personal experience with a described experience. In addition, only 2 patients were diagnosed with cancer—both with basal cell carcinoma diagnosed on physical examination—which may have influenced their perception of a CSP, given that nothing was found on an extensive number of tests. However, these patients still greatly valued their screening, as evidenced in the survey.

Conclusion

Our study contributes to a better understanding of the costs patients face while undergoing malignancy screening for DM and highlights the great value patients assign to undergoing screening regardless of impact on outcome. Our study also shows a preference for streamlined testing, which whole-body PET/CT may represent. Patients incurred substantial indirect costs with a CSP and perceived that a single test, such as whole-body PET/CT, would be less burdensome and result in better compliance with screening. As groups work to establish consensus guidelines for cancer screening in DM, it is important to include the patient’s perspective. Ultimately, prospective trials comparing these modalities are needed, at which time the efficacy, direct and indirect costs, and burden of each modality can be compared.

Dermatomyositis (DM) is an uncommon idiopathic inflammatory myopathy (IIM) characterized by muscle inflammation; proximal muscle weakness; and dermatologic findings, such as the heliotrope eruption and Gottron papules.^1-3 Dermatomyositis is associated with an increased malignancy risk compared to other IIMs, with a 13% to 42% lifetime risk for malignancy development.^4,5 The incidence for malignancy peaks during the first year following diagnosis and falls gradually over 5 years but remains increased compared to the general population.^6-11 Adenocarcinoma represents the majority of cancers associated with DM, particularly of the ovaries, lungs, breasts, gastrointestinal tract, pancreas, bladder, and prostate. The lymphatic system (non-Hodgkin lymphoma) also is overrepresented among cancers in DM.¹²

Because of the increased malignancy risk and cancer-related mortality in patients with DM, cancer screening generally is recommended following diagnosis.^13,14 However, consensus guidelines for screening modalities and frequency currently do not exist, resulting in widely varying practice patterns.¹⁵ Some experts advocate for a conventional cancer screening panel (CSP), as summarized in Table 1.^15-18 These tests may be repeated annually for 3 to 5 years following the diagnosis of DM. Although the use of myositis-specific antibodies (MSAs) recently has helped to risk-stratify DM patients, up to half of patients are MSA negative,¹⁹ and broad malignancy screening remains essential. Individualized discussions with patients about their risk factors, screening options, and risks and benefits of screening also are strongly encouraged.^19-22 Studies of the direct costs and effectiveness of streamlined screening with positron emission tomography/computed tomography (PET/CT) compared with a CSP have shown similar efficacy and lower out-of-pocket costs for patients receiving PET/CT imaging.^16-18

The goal of our study was to further characterize patients’ perspectives and experience of cancer screening in DM as well as indirect costs, both of which must be taken into consideration when developing consensus guidelines for DM malignancy screening. Inclusion of patient voice is essential given the similar efficacy of both screening methods. We assessed the indirect costs (eg, travel, lost work or wages, childcare) of a CSP in patients with DM. We theorized that the large quantity of tests involved in a CSP, which are performed at various locations on multiple days over the course of several years, may have substantial costs to patients beyond the co-pay and deductible. We also sought to measure patients’ perception of the burden associated with an annual CSP, which we defined to participants as the inconvenience or unpleasantness experienced by the patient, compared with an annual whole-body PET/CT. Finally, we examined the relative value of these screening methods to patients using a willingness-to-pay (WTP) analysis.

Materials and Methods

Patient Eligibility—Our study included Penn State Health (Hershey, Pennsylvania) patients 18 years or older with a recent diagnosis of DM—International Classification of Diseases, Ninth Revision code 710.3 or International Classification of Diseases, Tenth Revision codes M33.10 or M33.90—who were undergoing or had recently completed a CSP. Patients were excluded from the study if they had a concurrent or preceding diagnosis of malignancy (excluding nonmelanoma skin cancers) or had another IIM. The institutional review board at Penn State Health College of Medicine approved the study. Data for all patients were prospectively obtained.

Survey Design—A survey was generated to assess the burden and indirect costs associated with a CSP, which was modified from work done by Tchuenche et al²³ and Teni et al.²⁴ Focus groups were held in 2018 and 2019 with patients who met our inclusion criteria with the purpose of refining the survey instrument based on patient input. A summary explanation of research was provided to all participants, and informed consent was obtained. Patients were compensated for their time for focus groups. Audio of each focus group was then transcribed and analyzed for common themes. Following focus group feedback, a finalized survey was generated for assessing burden and indirect costs (survey instrument provided in the Supplementary Information). REDCap (Vanderbilt University), a secure web application, was used to construct the finalized survey and to collect and manage data.²⁵

Patients who fit our inclusion criteria were identified and recruited in multiple ways. Patients with appointments at the Penn State Milton S. Hershey Medical Center Department of Dermatology were presented with the opportunity to participate, Penn State Health records with the appropriate billing codes were collected and patients were contacted, and an advertisement for the study was posted on StudyFinder. Surveys constructed on REDCap were then sent electronically to patients who agreed to participate in the study. A second summary explanation of research was included on the first page of the survey to describe the process.

The survey had 3 main sections. The first section collected demographic information. In the second section, we surveyed patients regarding the various aspects of a CSP that focus groups identified as burdensome. In addition, patients were asked to compare their feelings regarding an annual CSP vs whole-body PET/CT for a 3-year period utilizing a rating scale of strongly disagree, somewhat disagree, somewhat agree, and strongly agree. This section also included a willingness-to-pay (WTP) analysis for each modality. We defined WTP as the maximum out-of-pocket cost that the patient would be willing to pay to receive testing, which was measured in a hypothetical scenario where neither whole-body PET/CT nor CSP was covered by insurance.²⁶ Although WTP may be influenced by external factors such as patient income, it can serve as a numerical measure of how much the patient values each service. Furthermore, these external factors become less relevant when comparing the relative value of 2 separate tests, as such factors apply equally in both scenarios. In the third section of the survey, patients were queried regarding various indirect costs associated with a CSP. Descriptions for a CSP and whole-body PET/CT, including risks and benefits, were provided to allow patients to make informed decisions.

 

Statistical Analysis—Because of the rarity of DM and the subsequently limited sample size, summary and descriptive statistics were utilized to characterize the sample and identify patterns in the results. Continuous variables are presented with means and standard deviations, and proportions are presented with frequencies and percentages. All analyses were done using SAS Version 9.4 (SAS Institute Inc).

Results

Patient Demographics—Fifty-four patients were identified using StudyFinder, physician referral, and search of the electronic health record. Nine patients agreed to take part in the focus groups, and 27 offered email addresses to be contacted for the survey. Of those 27 patients, 16 (59.3%) fit our inclusion criteria and completed the survey. Patient demographics are detailed in Table 2. The mean age was 55 years, and most patients were White (88% [14/16]), female (81% [13/16]), and had at least a bachelor’s degree (69% [11/16]). Most patients (69% [11/16]) had an annual income of less than $50,000, and half (50% [8/16]) were employed. All patients had been diagnosed with DM in or after 2013. Two patients were diagnosed with basal cell carcinoma during or after cancer screening.

Patient Preference for Screening and WTP—A majority (81% [13/16]) of patients desired some form of screening for occult malignancy following the diagnosis of DM, even in the hypothetical situation in which screening did not provide survival benefit (Figure 1). Twenty-five percent (4/16) of patients expressed that a CSP was burdensome, and 12.5% of patients (2/16) missed a CSP appointment; all of these patients rescheduled or were planning to reschedule. Assuming that both screening methods had similar predictive value in detecting malignancy, all 16 patients felt annual whole-body PET/CT for a 3-year period would be less burdensome than a CSP, and most (73% [11/15]) felt that it would decrease the likelihood of missed appointments. Overall, 93% (13/14) of patients preferred whole-body PET/CT over a CSP when given the choice between the 2 options (Figure 2). This preference was consistent with the patients’ WTP for these tests; patients reliably reported that they would pay more for annual whole-body PET/CT than for a CSP (Figure 3). Specifically, 75% (12/16) and 38% (6/16) of patients were willing to spend $250 or more and $1000 or more for annual whole-body PET/CT, respectively, compared with 56% (9/16) and 19% (3/16), respectively, for an annual CSP. Many patients (38% [6/16]) reported that they would not be willing to pay any out-of-pocket cost for a CSP compared with 13% (2/16) for PET/CT.Indirect Costs of Screening for Patients—Indirect costs incurred by patients undergoing a CSP are summarized in Table 3. Specifically, a large percentage of employed patients missed work (63% [5/8]) or had family miss work (38% [3/8]), necessitating the use of vacation and/or sick days to attend CSP appointments. A subset (25% [2/8]) lost income (average, $1500), and 1 patient reported that a family member lost income due to attending a CSP appointment. Most (75% [12/16]) patients also incurred substantial transportation costs (average, $243), with 1 patient spending $1000. No patients incurred child or elder care costs. One patient paid a small sum for lodging/meals while traveling to attend a CSP appointment.

Comment

Patients with DM have an increased incidence of malignancy, thus cancer screening serves a crucial role in the detection of occult disease.¹³ Up to half of DM patients are MSA negative, and most cancers in these patients are found with blind screening. Whole-body PET/CT has emerged as an alternative to a CSP. Evidence suggests that it has similar efficacy in detecting malignancy and may be particularly useful for identifying malignancies not routinely screened for in a CSP. In a prospective study of patients diagnosed with DM and polymyositis (N=55), whole-body PET/CT had a positive predictive value of 85.7% and negative predictive value for detecting occult malignancy of 93.8% compared with 77.8% and 95.7%, respectively, for a CSP.¹⁷

The results of our study showed that cancer screening is important to patients diagnosed with DM and that most of these patients desire some form of cancer screening. This finding held true even when patients were presented with a hypothetical situation in which screening was proven to have no survival benefit. Based on focus group data, this desire was likely driven by the fear generated by not knowing whether cancer is present, as reported by the following DM patients:

“I mean [cancer screening] is peace of mind. It is ultimately worth it. You know, better than . . . not doing the screenings and finding 3 years down the road that you have, you know, a serious problem . . . you had the cancer, and you didn’t have the screenings.” (DM patient 1)

“I would rather know than not know, even if it is bad news, just tell me. The sooner the better, and give me the whole spiel . . . maybe all the screenings don’t need to be done, done so much, so often afterwards if the initial ones are ok, but I think too, for peace of mind, I would rather know it all up front.” (DM patient 2)

 

Further, when presented with the hypothetical situation that insurance would not cover screenings, a few patients remarked they would relocate to obtain them:

“I would find a place where the screenings were done. I’d move.” (DM patient 4)

“If it was just sky high and [insurance companies] weren’t willing to negotiate, I would consider moving.” (DM patient 3).

Sentiments such as these emphasize the importance and value that DM patients place on being screened for cancer and also may explain why only 25% of patients felt a CSP was burdensome and only 13% reported missing appointments, all of whom planned on making them up at a later time.

When presented with the choice of a CSP or annual whole-body PET/CT for a 3-year period following the diagnosis of DM, all patients expressed that whole-body PET/CT would be less burdensome. Most preferred annual whole-body PET/CT despite the slightly increased radiation exposure associated and thought that it would limit missed appointments. Accordingly, more patients responded that they would pay more money out-of-pocket for annual whole-body PET/CT. Given that WTP can function as a numerical measure of value, our results showed that patients placed a higher value on whole-body PET/CT compared with a CSP. The indirect costs associated with a CSP also were substantial, particularly regarding missed work, use of vacation and/or sick days, and travel expenses, which is particularly important because most patients reported an annual income less than $50,000.

The direct costs of a CSP and whole-body PET/CT have been studied. Specifically, Kundrick et al¹⁸ found that whole-body PET/CT was less expensive for patients (by approximately $111) out-of-pocket compared with a CSP, though cost to insurance companies was slightly greater. The present study adds to these findings by better illustrating the burden and indirect costs that patients experience while undergoing a CSP and by characterizing the patient’s perception and preference of these 2 screening methods.

Limitations of our study include a small sample size willing to complete the survey. There also was a predominance of White and female participants, partially attributed to the greater number of female patients who develop DM compared to male patients. However, this still may limit applicability of this study to males and patients of other races. Another limitation includes recall bias on survey responses, particularly regarding indirect costs incurred with a CSP. A final limitation was that only patients with a recent diagnosis of DM who were actively undergoing screening or had recently completed malignancy screening were included in the study. Given that these patients were receiving (or had completed) exclusively a CSP, patients were comparing their personal experience with a described experience. In addition, only 2 patients were diagnosed with cancer—both with basal cell carcinoma diagnosed on physical examination—which may have influenced their perception of a CSP, given that nothing was found on an extensive number of tests. However, these patients still greatly valued their screening, as evidenced in the survey.

Conclusion

Our study contributes to a better understanding of the costs patients face while undergoing malignancy screening for DM and highlights the great value patients assign to undergoing screening regardless of impact on outcome. Our study also shows a preference for streamlined testing, which whole-body PET/CT may represent. Patients incurred substantial indirect costs with a CSP and perceived that a single test, such as whole-body PET/CT, would be less burdensome and result in better compliance with screening. As groups work to establish consensus guidelines for cancer screening in DM, it is important to include the patient’s perspective. Ultimately, prospective trials comparing these modalities are needed, at which time the efficacy, direct and indirect costs, and burden of each modality can be compared.

Dermatomyositis (DM) is an uncommon idiopathic inflammatory myopathy (IIM) characterized by muscle inflammation; proximal muscle weakness; and dermatologic findings, such as the heliotrope eruption and Gottron papules.^1-3 Dermatomyositis is associated with an increased malignancy risk compared to other IIMs, with a 13% to 42% lifetime risk for malignancy development.^4,5 The incidence for malignancy peaks during the first year following diagnosis and falls gradually over 5 years but remains increased compared to the general population.^6-11 Adenocarcinoma represents the majority of cancers associated with DM, particularly of the ovaries, lungs, breasts, gastrointestinal tract, pancreas, bladder, and prostate. The lymphatic system (non-Hodgkin lymphoma) also is overrepresented among cancers in DM.¹²

Because of the increased malignancy risk and cancer-related mortality in patients with DM, cancer screening generally is recommended following diagnosis.^13,14 However, consensus guidelines for screening modalities and frequency currently do not exist, resulting in widely varying practice patterns.¹⁵ Some experts advocate for a conventional cancer screening panel (CSP), as summarized in Table 1.^15-18 These tests may be repeated annually for 3 to 5 years following the diagnosis of DM. Although the use of myositis-specific antibodies (MSAs) recently has helped to risk-stratify DM patients, up to half of patients are MSA negative,¹⁹ and broad malignancy screening remains essential. Individualized discussions with patients about their risk factors, screening options, and risks and benefits of screening also are strongly encouraged.^19-22 Studies of the direct costs and effectiveness of streamlined screening with positron emission tomography/computed tomography (PET/CT) compared with a CSP have shown similar efficacy and lower out-of-pocket costs for patients receiving PET/CT imaging.^16-18

The goal of our study was to further characterize patients’ perspectives and experience of cancer screening in DM as well as indirect costs, both of which must be taken into consideration when developing consensus guidelines for DM malignancy screening. Inclusion of patient voice is essential given the similar efficacy of both screening methods. We assessed the indirect costs (eg, travel, lost work or wages, childcare) of a CSP in patients with DM. We theorized that the large quantity of tests involved in a CSP, which are performed at various locations on multiple days over the course of several years, may have substantial costs to patients beyond the co-pay and deductible. We also sought to measure patients’ perception of the burden associated with an annual CSP, which we defined to participants as the inconvenience or unpleasantness experienced by the patient, compared with an annual whole-body PET/CT. Finally, we examined the relative value of these screening methods to patients using a willingness-to-pay (WTP) analysis.

Materials and Methods

Patient Eligibility—Our study included Penn State Health (Hershey, Pennsylvania) patients 18 years or older with a recent diagnosis of DM—International Classification of Diseases, Ninth Revision code 710.3 or International Classification of Diseases, Tenth Revision codes M33.10 or M33.90—who were undergoing or had recently completed a CSP. Patients were excluded from the study if they had a concurrent or preceding diagnosis of malignancy (excluding nonmelanoma skin cancers) or had another IIM. The institutional review board at Penn State Health College of Medicine approved the study. Data for all patients were prospectively obtained.

Survey Design—A survey was generated to assess the burden and indirect costs associated with a CSP, which was modified from work done by Tchuenche et al²³ and Teni et al.²⁴ Focus groups were held in 2018 and 2019 with patients who met our inclusion criteria with the purpose of refining the survey instrument based on patient input. A summary explanation of research was provided to all participants, and informed consent was obtained. Patients were compensated for their time for focus groups. Audio of each focus group was then transcribed and analyzed for common themes. Following focus group feedback, a finalized survey was generated for assessing burden and indirect costs (survey instrument provided in the Supplementary Information). REDCap (Vanderbilt University), a secure web application, was used to construct the finalized survey and to collect and manage data.²⁵

Patients who fit our inclusion criteria were identified and recruited in multiple ways. Patients with appointments at the Penn State Milton S. Hershey Medical Center Department of Dermatology were presented with the opportunity to participate, Penn State Health records with the appropriate billing codes were collected and patients were contacted, and an advertisement for the study was posted on StudyFinder. Surveys constructed on REDCap were then sent electronically to patients who agreed to participate in the study. A second summary explanation of research was included on the first page of the survey to describe the process.

The survey had 3 main sections. The first section collected demographic information. In the second section, we surveyed patients regarding the various aspects of a CSP that focus groups identified as burdensome. In addition, patients were asked to compare their feelings regarding an annual CSP vs whole-body PET/CT for a 3-year period utilizing a rating scale of strongly disagree, somewhat disagree, somewhat agree, and strongly agree. This section also included a willingness-to-pay (WTP) analysis for each modality. We defined WTP as the maximum out-of-pocket cost that the patient would be willing to pay to receive testing, which was measured in a hypothetical scenario where neither whole-body PET/CT nor CSP was covered by insurance.²⁶ Although WTP may be influenced by external factors such as patient income, it can serve as a numerical measure of how much the patient values each service. Furthermore, these external factors become less relevant when comparing the relative value of 2 separate tests, as such factors apply equally in both scenarios. In the third section of the survey, patients were queried regarding various indirect costs associated with a CSP. Descriptions for a CSP and whole-body PET/CT, including risks and benefits, were provided to allow patients to make informed decisions.

 

Statistical Analysis—Because of the rarity of DM and the subsequently limited sample size, summary and descriptive statistics were utilized to characterize the sample and identify patterns in the results. Continuous variables are presented with means and standard deviations, and proportions are presented with frequencies and percentages. All analyses were done using SAS Version 9.4 (SAS Institute Inc).

Results

Patient Demographics—Fifty-four patients were identified using StudyFinder, physician referral, and search of the electronic health record. Nine patients agreed to take part in the focus groups, and 27 offered email addresses to be contacted for the survey. Of those 27 patients, 16 (59.3%) fit our inclusion criteria and completed the survey. Patient demographics are detailed in Table 2. The mean age was 55 years, and most patients were White (88% [14/16]), female (81% [13/16]), and had at least a bachelor’s degree (69% [11/16]). Most patients (69% [11/16]) had an annual income of less than $50,000, and half (50% [8/16]) were employed. All patients had been diagnosed with DM in or after 2013. Two patients were diagnosed with basal cell carcinoma during or after cancer screening.

Patient Preference for Screening and WTP—A majority (81% [13/16]) of patients desired some form of screening for occult malignancy following the diagnosis of DM, even in the hypothetical situation in which screening did not provide survival benefit (Figure 1). Twenty-five percent (4/16) of patients expressed that a CSP was burdensome, and 12.5% of patients (2/16) missed a CSP appointment; all of these patients rescheduled or were planning to reschedule. Assuming that both screening methods had similar predictive value in detecting malignancy, all 16 patients felt annual whole-body PET/CT for a 3-year period would be less burdensome than a CSP, and most (73% [11/15]) felt that it would decrease the likelihood of missed appointments. Overall, 93% (13/14) of patients preferred whole-body PET/CT over a CSP when given the choice between the 2 options (Figure 2). This preference was consistent with the patients’ WTP for these tests; patients reliably reported that they would pay more for annual whole-body PET/CT than for a CSP (Figure 3). Specifically, 75% (12/16) and 38% (6/16) of patients were willing to spend $250 or more and $1000 or more for annual whole-body PET/CT, respectively, compared with 56% (9/16) and 19% (3/16), respectively, for an annual CSP. Many patients (38% [6/16]) reported that they would not be willing to pay any out-of-pocket cost for a CSP compared with 13% (2/16) for PET/CT.Indirect Costs of Screening for Patients—Indirect costs incurred by patients undergoing a CSP are summarized in Table 3. Specifically, a large percentage of employed patients missed work (63% [5/8]) or had family miss work (38% [3/8]), necessitating the use of vacation and/or sick days to attend CSP appointments. A subset (25% [2/8]) lost income (average, $1500), and 1 patient reported that a family member lost income due to attending a CSP appointment. Most (75% [12/16]) patients also incurred substantial transportation costs (average, $243), with 1 patient spending $1000. No patients incurred child or elder care costs. One patient paid a small sum for lodging/meals while traveling to attend a CSP appointment.

Comment

Patients with DM have an increased incidence of malignancy, thus cancer screening serves a crucial role in the detection of occult disease.¹³ Up to half of DM patients are MSA negative, and most cancers in these patients are found with blind screening. Whole-body PET/CT has emerged as an alternative to a CSP. Evidence suggests that it has similar efficacy in detecting malignancy and may be particularly useful for identifying malignancies not routinely screened for in a CSP. In a prospective study of patients diagnosed with DM and polymyositis (N=55), whole-body PET/CT had a positive predictive value of 85.7% and negative predictive value for detecting occult malignancy of 93.8% compared with 77.8% and 95.7%, respectively, for a CSP.¹⁷

The results of our study showed that cancer screening is important to patients diagnosed with DM and that most of these patients desire some form of cancer screening. This finding held true even when patients were presented with a hypothetical situation in which screening was proven to have no survival benefit. Based on focus group data, this desire was likely driven by the fear generated by not knowing whether cancer is present, as reported by the following DM patients:

“I mean [cancer screening] is peace of mind. It is ultimately worth it. You know, better than . . . not doing the screenings and finding 3 years down the road that you have, you know, a serious problem . . . you had the cancer, and you didn’t have the screenings.” (DM patient 1)

“I would rather know than not know, even if it is bad news, just tell me. The sooner the better, and give me the whole spiel . . . maybe all the screenings don’t need to be done, done so much, so often afterwards if the initial ones are ok, but I think too, for peace of mind, I would rather know it all up front.” (DM patient 2)

 

Further, when presented with the hypothetical situation that insurance would not cover screenings, a few patients remarked they would relocate to obtain them:

“I would find a place where the screenings were done. I’d move.” (DM patient 4)

“If it was just sky high and [insurance companies] weren’t willing to negotiate, I would consider moving.” (DM patient 3).

Sentiments such as these emphasize the importance and value that DM patients place on being screened for cancer and also may explain why only 25% of patients felt a CSP was burdensome and only 13% reported missing appointments, all of whom planned on making them up at a later time.

When presented with the choice of a CSP or annual whole-body PET/CT for a 3-year period following the diagnosis of DM, all patients expressed that whole-body PET/CT would be less burdensome. Most preferred annual whole-body PET/CT despite the slightly increased radiation exposure associated and thought that it would limit missed appointments. Accordingly, more patients responded that they would pay more money out-of-pocket for annual whole-body PET/CT. Given that WTP can function as a numerical measure of value, our results showed that patients placed a higher value on whole-body PET/CT compared with a CSP. The indirect costs associated with a CSP also were substantial, particularly regarding missed work, use of vacation and/or sick days, and travel expenses, which is particularly important because most patients reported an annual income less than $50,000.

The direct costs of a CSP and whole-body PET/CT have been studied. Specifically, Kundrick et al¹⁸ found that whole-body PET/CT was less expensive for patients (by approximately $111) out-of-pocket compared with a CSP, though cost to insurance companies was slightly greater. The present study adds to these findings by better illustrating the burden and indirect costs that patients experience while undergoing a CSP and by characterizing the patient’s perception and preference of these 2 screening methods.

Limitations of our study include a small sample size willing to complete the survey. There also was a predominance of White and female participants, partially attributed to the greater number of female patients who develop DM compared to male patients. However, this still may limit applicability of this study to males and patients of other races. Another limitation includes recall bias on survey responses, particularly regarding indirect costs incurred with a CSP. A final limitation was that only patients with a recent diagnosis of DM who were actively undergoing screening or had recently completed malignancy screening were included in the study. Given that these patients were receiving (or had completed) exclusively a CSP, patients were comparing their personal experience with a described experience. In addition, only 2 patients were diagnosed with cancer—both with basal cell carcinoma diagnosed on physical examination—which may have influenced their perception of a CSP, given that nothing was found on an extensive number of tests. However, these patients still greatly valued their screening, as evidenced in the survey.

Conclusion

Our study contributes to a better understanding of the costs patients face while undergoing malignancy screening for DM and highlights the great value patients assign to undergoing screening regardless of impact on outcome. Our study also shows a preference for streamlined testing, which whole-body PET/CT may represent. Patients incurred substantial indirect costs with a CSP and perceived that a single test, such as whole-body PET/CT, would be less burdensome and result in better compliance with screening. As groups work to establish consensus guidelines for cancer screening in DM, it is important to include the patient’s perspective. Ultimately, prospective trials comparing these modalities are needed, at which time the efficacy, direct and indirect costs, and burden of each modality can be compared.

To the Editor:

With health care constituting one of the larger segments of the US economy, medical practice is increasingly subject to business considerations.¹ Patients, providers, and organizations are all required to make decisions that reflect choices beyond clinical needs alone. Given the impact of market forces, clinicians often are asked to navigate operational and business decisions. Accordingly, education about the policy and systems that shape care delivery can improve quality and help patients.²

The ability to understand the ecosystem of health care is of utmost importance for medical providers and can be achieved through resident education. Teaching fundamental business concepts enables residents to deliver care that is responsive to the constraints and opportunities encountered by patients and organizations, which ultimately will better prepare them to serve as advocates in alignment with their principal duties as physicians.

Despite the recognizable relationship between business and medicine, training has not yet been standardized to include topics in business education, and clinicians in dermatology are remarkably positioned to benefit because of the variety of practice settings and services they can provide. In dermatology, the diversity of services provided gives rise to complex coding and use of modifiers. Proper utilization of coding and billing is critical to create accurate documentation and receive appropriate reimbursement.³ Furthermore, clinicians in dermatology have to contend with the influence of insurance at many points of care, such as with coverage of pharmaceuticals. Formularies often have wide variability in coverage and are changing as new drugs come to market in the dermatologic space.⁴

The landscape of practice structure also has undergone change with increasing consolidation and mergers. The acquisition of practices by private equity firms has induced changes in practice infrastructure. The impact of changing organizational and managerial influences continues to be a topic of debate, with disparate opinions on how these developments shape standards of physician satisfaction and patient care.⁵

The convergence of these factors points to an important question that is gaining popularity: How will young dermatologists work within the context of all these parameters to best advocate and care for their patients? These questions are garnering more attention and were recently investigated through a survey of participants in a pilot program to evaluate the importance of business education in dermatology residency.

A survey of residency program directors was created by Patrinley and Dewan,⁶ which found that business education during residency was important and additional training should be implemented. Despite the perceived importance of business education, only half of the programs represented by survey respondents offered any structured educational opportunities, revealing a discrepancy between believed importance and practical implementation of business training, which suggests the need to develop a standardized, dermatology-specific curriculum that could be accessed by all residents in training.⁶

We performed a search of the medical literature to identify models of business education in residency programs. Only a few programs were identified, in which courses were predominantly instructed to trainees in primary care–based fields. According to course graduates, the programs were beneficial.^7,8 Programs that had descriptive information about curriculum structure and content were chosen for further investigation and included internal medicine programs at the University of California San Francisco (UCSF) and Columbia University Vagelos College of Physicians and Surgeons (New York, New York). UCSF implemented a Program in Residency Investigation Methods and Epidemiology (PRIME program) to deliver seven 90-minute sessions dedicated to introducing residents to medical economics. Sessions were constructed with the intent of being interactive seminars that took on a variety of forms, including reading-based discussions, case-based analysis, and simulation-based learning.⁷ Columbia University developed a pilot program of week-long didactic sessions that were delivered to third-year internal medicine residents. These seminars featured discussions on health policy and economics, health insurance, technology and cost assessment, legal medicine, public health, community-oriented primary care, and local health department initiatives.⁸ We drew on both courses to build a lecture series focused on the business of dermatology that was delivered to dermatology residents at UMass Chan Medical School (Worcester, Massachusetts). Topic selection also was informed by qualitative input collected via email from recent graduates of the UMass dermatology residency program, focusing on the following areas: the US medical economy and health care costs; billing, coding, and claims processing; quality, relative value units (RVUs), reimbursement, and the merit-based incentive payment system; coverage of pharmaceuticals and teledermatology; and management. Residents were not required to prepare for any of the sessions; they were provided with handouts and slideshow presentations for reference to review at their convenience if desired. Five seminars were virtually conducted by an MD/MBA candidate at the institution (E.H.). They were recorded over the course of an academic year at 1- to 2-month intervals. Each 45-minute session was conducted in a lecture-discussion format and included case examples to help illustrate key principles and stimulate conversation. For example, the lecture on reimbursement incorporated a fee schedule calculation for a shave biopsy, using RVU and geographic pricing cost index (GCPI) multipliers. This demonstrated the variation in Centers for Medicare & Medicaid Services reimbursement in relation to (1) constituents of the RVU calculation (ie, work, practice expense, and malpractice) and (2) practice in a particular location (ie, the GCPI). Following this example, a conversation ensued among participants regarding the factors that drive valuation, with particular interest in variation based on urban vs suburban locations across the United States. Participants also found it of interest to examine the percentage of the valuation dedicated to each constituent and how features such as lesion size informed the final assessment of the charge. Another stylistic choice in developing the model was to include prompts for further consideration prior to transitioning topics in the lectures. For example: when examining the burden of skin disease, the audience was prompted to consider: “What is driving cost escalations, and how will services of the clinical domain meet these evolving needs?” At another point in the introductory lecture, residents were asked: “How do different types of insurance plans impact the management of patients with dermatologic concerns?” These questions were intended to transition residents to the next topic of discussion and highlight take-home points of consideration for medical practice. The project was reviewed by the UMass institutional review board and met criteria for exemption.

 

Residents who participated in at least 1 lecture (N=10) were surveyed after attendance; there were 7 responses (70% response rate). Residents were asked to rate a series of statements on a scale of 1 (strongly disagree) to 5 (strongly agree) and to provide commentary via an online form. Respondents indicated that the course was enjoyable (average score, 4.00), provided an appropriate level of detail (average score, 4.00), would be beneficial to integrate into a dermatology residency curriculum (average score, 3.86), and informed how they would practice as a clinician (average score, 3.86)(Figure). The respondents agreed that the course met the main goals of this initiative: it helped them develop knowledge about the interface between business and dermatology (4.14) and exposed residents to topics they had not learned about previously (4.71).

Although the course generally was well received, areas for improvement were identified from respondents’ comments, relating to audience engagement and refining the level of detail in the lectures. Recommendations included “less technical jargon and more focus on ‘big picture’ concepts, given audience’s low baseline knowledge”; “more case examples in each module”; and “more diagrams or interactive activities (polls, quizzes, break-out rooms) because the lectures were a bit dense.” This input was taken into consideration when revising the lectures for future use; they were reconstructed to have more case-based examples and prompts to encourage participation.

Resident commentary also demonstrated appreciation for education in this subject material. Statements such as “this is an important topic for future dermatologists” and “thank you so much for taking the time to implement this course” reflected the perceived value of this material during critical academic time. Another resident remarked: “This was great, thanks for putting it together.”

Given the positive experience of the residents and successful implementation of the series, this course was made available to all dermatology trainees on a network server with accompanying written documents. It is planned to be offered on a 3-year cycle in the future and will be updated to reflect inevitable changes in health care.

Although the relationship between business and medicine is increasingly important, teaching business principles has not become standardized or required in medical training. Despite the perception that this content is of value, implementation of programming has lagged behind that recognition, likely due to challenges in designing the curriculum and diffusing content into an already-saturated schedule. A model course that can be replicated in other residency programs would be valuable. We introduced a dermatology-specific lecture series to help prepare trainees for dermatology practice in a variety of clinical settings and train them with the language of business and operations that will equip them to respond to the needs of their patients, their practice, and the medical environment. Findings of this pilot study may not be generalizable to all dermatology residency programs because the sample size was small; the study was conducted at a single institution; and the content was delivered entirely online.

To the Editor:

With health care constituting one of the larger segments of the US economy, medical practice is increasingly subject to business considerations.¹ Patients, providers, and organizations are all required to make decisions that reflect choices beyond clinical needs alone. Given the impact of market forces, clinicians often are asked to navigate operational and business decisions. Accordingly, education about the policy and systems that shape care delivery can improve quality and help patients.²

The ability to understand the ecosystem of health care is of utmost importance for medical providers and can be achieved through resident education. Teaching fundamental business concepts enables residents to deliver care that is responsive to the constraints and opportunities encountered by patients and organizations, which ultimately will better prepare them to serve as advocates in alignment with their principal duties as physicians.

Despite the recognizable relationship between business and medicine, training has not yet been standardized to include topics in business education, and clinicians in dermatology are remarkably positioned to benefit because of the variety of practice settings and services they can provide. In dermatology, the diversity of services provided gives rise to complex coding and use of modifiers. Proper utilization of coding and billing is critical to create accurate documentation and receive appropriate reimbursement.³ Furthermore, clinicians in dermatology have to contend with the influence of insurance at many points of care, such as with coverage of pharmaceuticals. Formularies often have wide variability in coverage and are changing as new drugs come to market in the dermatologic space.⁴

The landscape of practice structure also has undergone change with increasing consolidation and mergers. The acquisition of practices by private equity firms has induced changes in practice infrastructure. The impact of changing organizational and managerial influences continues to be a topic of debate, with disparate opinions on how these developments shape standards of physician satisfaction and patient care.⁵

The convergence of these factors points to an important question that is gaining popularity: How will young dermatologists work within the context of all these parameters to best advocate and care for their patients? These questions are garnering more attention and were recently investigated through a survey of participants in a pilot program to evaluate the importance of business education in dermatology residency.

A survey of residency program directors was created by Patrinley and Dewan,⁶ which found that business education during residency was important and additional training should be implemented. Despite the perceived importance of business education, only half of the programs represented by survey respondents offered any structured educational opportunities, revealing a discrepancy between believed importance and practical implementation of business training, which suggests the need to develop a standardized, dermatology-specific curriculum that could be accessed by all residents in training.⁶

We performed a search of the medical literature to identify models of business education in residency programs. Only a few programs were identified, in which courses were predominantly instructed to trainees in primary care–based fields. According to course graduates, the programs were beneficial.^7,8 Programs that had descriptive information about curriculum structure and content were chosen for further investigation and included internal medicine programs at the University of California San Francisco (UCSF) and Columbia University Vagelos College of Physicians and Surgeons (New York, New York). UCSF implemented a Program in Residency Investigation Methods and Epidemiology (PRIME program) to deliver seven 90-minute sessions dedicated to introducing residents to medical economics. Sessions were constructed with the intent of being interactive seminars that took on a variety of forms, including reading-based discussions, case-based analysis, and simulation-based learning.⁷ Columbia University developed a pilot program of week-long didactic sessions that were delivered to third-year internal medicine residents. These seminars featured discussions on health policy and economics, health insurance, technology and cost assessment, legal medicine, public health, community-oriented primary care, and local health department initiatives.⁸ We drew on both courses to build a lecture series focused on the business of dermatology that was delivered to dermatology residents at UMass Chan Medical School (Worcester, Massachusetts). Topic selection also was informed by qualitative input collected via email from recent graduates of the UMass dermatology residency program, focusing on the following areas: the US medical economy and health care costs; billing, coding, and claims processing; quality, relative value units (RVUs), reimbursement, and the merit-based incentive payment system; coverage of pharmaceuticals and teledermatology; and management. Residents were not required to prepare for any of the sessions; they were provided with handouts and slideshow presentations for reference to review at their convenience if desired. Five seminars were virtually conducted by an MD/MBA candidate at the institution (E.H.). They were recorded over the course of an academic year at 1- to 2-month intervals. Each 45-minute session was conducted in a lecture-discussion format and included case examples to help illustrate key principles and stimulate conversation. For example, the lecture on reimbursement incorporated a fee schedule calculation for a shave biopsy, using RVU and geographic pricing cost index (GCPI) multipliers. This demonstrated the variation in Centers for Medicare & Medicaid Services reimbursement in relation to (1) constituents of the RVU calculation (ie, work, practice expense, and malpractice) and (2) practice in a particular location (ie, the GCPI). Following this example, a conversation ensued among participants regarding the factors that drive valuation, with particular interest in variation based on urban vs suburban locations across the United States. Participants also found it of interest to examine the percentage of the valuation dedicated to each constituent and how features such as lesion size informed the final assessment of the charge. Another stylistic choice in developing the model was to include prompts for further consideration prior to transitioning topics in the lectures. For example: when examining the burden of skin disease, the audience was prompted to consider: “What is driving cost escalations, and how will services of the clinical domain meet these evolving needs?” At another point in the introductory lecture, residents were asked: “How do different types of insurance plans impact the management of patients with dermatologic concerns?” These questions were intended to transition residents to the next topic of discussion and highlight take-home points of consideration for medical practice. The project was reviewed by the UMass institutional review board and met criteria for exemption.

 

Residents who participated in at least 1 lecture (N=10) were surveyed after attendance; there were 7 responses (70% response rate). Residents were asked to rate a series of statements on a scale of 1 (strongly disagree) to 5 (strongly agree) and to provide commentary via an online form. Respondents indicated that the course was enjoyable (average score, 4.00), provided an appropriate level of detail (average score, 4.00), would be beneficial to integrate into a dermatology residency curriculum (average score, 3.86), and informed how they would practice as a clinician (average score, 3.86)(Figure). The respondents agreed that the course met the main goals of this initiative: it helped them develop knowledge about the interface between business and dermatology (4.14) and exposed residents to topics they had not learned about previously (4.71).

Although the course generally was well received, areas for improvement were identified from respondents’ comments, relating to audience engagement and refining the level of detail in the lectures. Recommendations included “less technical jargon and more focus on ‘big picture’ concepts, given audience’s low baseline knowledge”; “more case examples in each module”; and “more diagrams or interactive activities (polls, quizzes, break-out rooms) because the lectures were a bit dense.” This input was taken into consideration when revising the lectures for future use; they were reconstructed to have more case-based examples and prompts to encourage participation.

Resident commentary also demonstrated appreciation for education in this subject material. Statements such as “this is an important topic for future dermatologists” and “thank you so much for taking the time to implement this course” reflected the perceived value of this material during critical academic time. Another resident remarked: “This was great, thanks for putting it together.”

Given the positive experience of the residents and successful implementation of the series, this course was made available to all dermatology trainees on a network server with accompanying written documents. It is planned to be offered on a 3-year cycle in the future and will be updated to reflect inevitable changes in health care.

Although the relationship between business and medicine is increasingly important, teaching business principles has not become standardized or required in medical training. Despite the perception that this content is of value, implementation of programming has lagged behind that recognition, likely due to challenges in designing the curriculum and diffusing content into an already-saturated schedule. A model course that can be replicated in other residency programs would be valuable. We introduced a dermatology-specific lecture series to help prepare trainees for dermatology practice in a variety of clinical settings and train them with the language of business and operations that will equip them to respond to the needs of their patients, their practice, and the medical environment. Findings of this pilot study may not be generalizable to all dermatology residency programs because the sample size was small; the study was conducted at a single institution; and the content was delivered entirely online.

To the Editor:

With health care constituting one of the larger segments of the US economy, medical practice is increasingly subject to business considerations.¹ Patients, providers, and organizations are all required to make decisions that reflect choices beyond clinical needs alone. Given the impact of market forces, clinicians often are asked to navigate operational and business decisions. Accordingly, education about the policy and systems that shape care delivery can improve quality and help patients.²

The ability to understand the ecosystem of health care is of utmost importance for medical providers and can be achieved through resident education. Teaching fundamental business concepts enables residents to deliver care that is responsive to the constraints and opportunities encountered by patients and organizations, which ultimately will better prepare them to serve as advocates in alignment with their principal duties as physicians.

Despite the recognizable relationship between business and medicine, training has not yet been standardized to include topics in business education, and clinicians in dermatology are remarkably positioned to benefit because of the variety of practice settings and services they can provide. In dermatology, the diversity of services provided gives rise to complex coding and use of modifiers. Proper utilization of coding and billing is critical to create accurate documentation and receive appropriate reimbursement.³ Furthermore, clinicians in dermatology have to contend with the influence of insurance at many points of care, such as with coverage of pharmaceuticals. Formularies often have wide variability in coverage and are changing as new drugs come to market in the dermatologic space.⁴

The landscape of practice structure also has undergone change with increasing consolidation and mergers. The acquisition of practices by private equity firms has induced changes in practice infrastructure. The impact of changing organizational and managerial influences continues to be a topic of debate, with disparate opinions on how these developments shape standards of physician satisfaction and patient care.⁵

The convergence of these factors points to an important question that is gaining popularity: How will young dermatologists work within the context of all these parameters to best advocate and care for their patients? These questions are garnering more attention and were recently investigated through a survey of participants in a pilot program to evaluate the importance of business education in dermatology residency.

A survey of residency program directors was created by Patrinley and Dewan,⁶ which found that business education during residency was important and additional training should be implemented. Despite the perceived importance of business education, only half of the programs represented by survey respondents offered any structured educational opportunities, revealing a discrepancy between believed importance and practical implementation of business training, which suggests the need to develop a standardized, dermatology-specific curriculum that could be accessed by all residents in training.⁶

We performed a search of the medical literature to identify models of business education in residency programs. Only a few programs were identified, in which courses were predominantly instructed to trainees in primary care–based fields. According to course graduates, the programs were beneficial.^7,8 Programs that had descriptive information about curriculum structure and content were chosen for further investigation and included internal medicine programs at the University of California San Francisco (UCSF) and Columbia University Vagelos College of Physicians and Surgeons (New York, New York). UCSF implemented a Program in Residency Investigation Methods and Epidemiology (PRIME program) to deliver seven 90-minute sessions dedicated to introducing residents to medical economics. Sessions were constructed with the intent of being interactive seminars that took on a variety of forms, including reading-based discussions, case-based analysis, and simulation-based learning.⁷ Columbia University developed a pilot program of week-long didactic sessions that were delivered to third-year internal medicine residents. These seminars featured discussions on health policy and economics, health insurance, technology and cost assessment, legal medicine, public health, community-oriented primary care, and local health department initiatives.⁸ We drew on both courses to build a lecture series focused on the business of dermatology that was delivered to dermatology residents at UMass Chan Medical School (Worcester, Massachusetts). Topic selection also was informed by qualitative input collected via email from recent graduates of the UMass dermatology residency program, focusing on the following areas: the US medical economy and health care costs; billing, coding, and claims processing; quality, relative value units (RVUs), reimbursement, and the merit-based incentive payment system; coverage of pharmaceuticals and teledermatology; and management. Residents were not required to prepare for any of the sessions; they were provided with handouts and slideshow presentations for reference to review at their convenience if desired. Five seminars were virtually conducted by an MD/MBA candidate at the institution (E.H.). They were recorded over the course of an academic year at 1- to 2-month intervals. Each 45-minute session was conducted in a lecture-discussion format and included case examples to help illustrate key principles and stimulate conversation. For example, the lecture on reimbursement incorporated a fee schedule calculation for a shave biopsy, using RVU and geographic pricing cost index (GCPI) multipliers. This demonstrated the variation in Centers for Medicare & Medicaid Services reimbursement in relation to (1) constituents of the RVU calculation (ie, work, practice expense, and malpractice) and (2) practice in a particular location (ie, the GCPI). Following this example, a conversation ensued among participants regarding the factors that drive valuation, with particular interest in variation based on urban vs suburban locations across the United States. Participants also found it of interest to examine the percentage of the valuation dedicated to each constituent and how features such as lesion size informed the final assessment of the charge. Another stylistic choice in developing the model was to include prompts for further consideration prior to transitioning topics in the lectures. For example: when examining the burden of skin disease, the audience was prompted to consider: “What is driving cost escalations, and how will services of the clinical domain meet these evolving needs?” At another point in the introductory lecture, residents were asked: “How do different types of insurance plans impact the management of patients with dermatologic concerns?” These questions were intended to transition residents to the next topic of discussion and highlight take-home points of consideration for medical practice. The project was reviewed by the UMass institutional review board and met criteria for exemption.

 

Residents who participated in at least 1 lecture (N=10) were surveyed after attendance; there were 7 responses (70% response rate). Residents were asked to rate a series of statements on a scale of 1 (strongly disagree) to 5 (strongly agree) and to provide commentary via an online form. Respondents indicated that the course was enjoyable (average score, 4.00), provided an appropriate level of detail (average score, 4.00), would be beneficial to integrate into a dermatology residency curriculum (average score, 3.86), and informed how they would practice as a clinician (average score, 3.86)(Figure). The respondents agreed that the course met the main goals of this initiative: it helped them develop knowledge about the interface between business and dermatology (4.14) and exposed residents to topics they had not learned about previously (4.71).

Although the course generally was well received, areas for improvement were identified from respondents’ comments, relating to audience engagement and refining the level of detail in the lectures. Recommendations included “less technical jargon and more focus on ‘big picture’ concepts, given audience’s low baseline knowledge”; “more case examples in each module”; and “more diagrams or interactive activities (polls, quizzes, break-out rooms) because the lectures were a bit dense.” This input was taken into consideration when revising the lectures for future use; they were reconstructed to have more case-based examples and prompts to encourage participation.

Resident commentary also demonstrated appreciation for education in this subject material. Statements such as “this is an important topic for future dermatologists” and “thank you so much for taking the time to implement this course” reflected the perceived value of this material during critical academic time. Another resident remarked: “This was great, thanks for putting it together.”

Given the positive experience of the residents and successful implementation of the series, this course was made available to all dermatology trainees on a network server with accompanying written documents. It is planned to be offered on a 3-year cycle in the future and will be updated to reflect inevitable changes in health care.

Although the relationship between business and medicine is increasingly important, teaching business principles has not become standardized or required in medical training. Despite the perception that this content is of value, implementation of programming has lagged behind that recognition, likely due to challenges in designing the curriculum and diffusing content into an already-saturated schedule. A model course that can be replicated in other residency programs would be valuable. We introduced a dermatology-specific lecture series to help prepare trainees for dermatology practice in a variety of clinical settings and train them with the language of business and operations that will equip them to respond to the needs of their patients, their practice, and the medical environment. Findings of this pilot study may not be generalizable to all dermatology residency programs because the sample size was small; the study was conducted at a single institution; and the content was delivered entirely online.

To the Editor:

Citation number and Hirsch index (h-index) have long been employed as metrics of productivity for academic scholarship. The h-index is defined as the highest number of publications (the maximum h value) of an author who has published at least h papers, each cited by other authors at least h times.¹ In a bibliometric analysis of the most frequently cited authors in dermatology from 1974 to 2019 (N=378,276), females comprised 12% of first and 11% of senior authors of the most cited publications, and 6 of the most cited authors in dermatology were women.² In another study analyzing the most prolific dermatologic authors based on h-index, 0% from 1980 to 1989 and 19% from 2010 to 2019 were female (N=393,488).³ Because citation number and h-index favor longer-practicing dermatologists, we examined dermatology author productivity and gender trends by recent publication rates.

The Scopus database was searched for dermatology publications by using the field category “dermatology”from January 1, 2017, to October 7, 2022. Nondermatologists and authors with the same initials were excluded. Authors were ranked by number of publications, including original articles, case reports, letters, and reviews. Sex, degree, and years of experience were determined via a Google search of the author’s name. The h-index; number of citations; and percentages of first, middle, and last authorship were recorded.

Of the top 50 published dermatologists, 30% were female (n=15) and 56% (n=28) held both MD and PhD degrees (Table). The mean years of experience was 26.27 years (range, 6–44 years), with a mean of 29.23 years in females and 25.87 years in males. The mean h-index was 27.96 (range, 8–88), with 24.87 for females and 29.29 for males. The mean number of citations was 4032.64 (range, 235–36,908), with 2891.13 for females and 4521.86 for males. Thirty-one authors were most frequently middle authors, 18 were senior authors, and 1 was a first author. On average (SD), authors were senior or first author in 47.97% (20.08%) of their publications (range, 6.32%–94.93%).

Our study shows that females were more highly represented as top dermatology authors (30%) as measured by publication numbers from 2017 to 2022 than in studies measuring citation rate from 1974 to 2019 (12%)² or h-index from 2010 to 2019 (19%).³ Similarly, in a study of dermatology authorship from 2009 to 2019, on average, females represented 51.06% first and 38.18% last authors.⁴

The proportion of females in the dermatology workforce has increased, with 3964 of 10,385 (38.2%) active dermatologists in 2007⁵ being female vs 6372 of 12,505 (51.0%) in 2019.⁶ The lower proportion of practicing female dermatologists in earlier years likely accounts for the lower percentage of females in dermatology citations and h-index top lists during that time, given that citation and h-index metrics are biased to dermatologists with longer careers.

Although our data are encouraging, females still accounted for less than one-third of the top 50 authors by publication numbers. Gender inequalities persist, with only one-third of a total of 1292 National Institutes of Health dermatology grants and one-fourth of Research Project Grant Program (R01) grants being awarded to females in the years 2009 to 2014.⁷ Therefore, formal and informal mentorship, protected time for research, resources for childcare, and opportunities for funding will be critical in supporting female dermatologists to both publish highly impactful research and obtain research grants.

Limitations of our study include the omission of authors with identical initials and the inability to account for name changes. Furthermore, Scopus does not include all articles published by each author. Finally, publication number reflects quantity but may not reflect quality.

By quantitating dermatology author publication numbers, we found better representation of female authors compared with studies measuring citation number and h-index. With higher proportions of female dermatology trainees and efforts to increase mentorship and research support for female dermatologists, we expect improved equality in top lists of dermatology citations and h-index values.

To the Editor:

Citation number and Hirsch index (h-index) have long been employed as metrics of productivity for academic scholarship. The h-index is defined as the highest number of publications (the maximum h value) of an author who has published at least h papers, each cited by other authors at least h times.¹ In a bibliometric analysis of the most frequently cited authors in dermatology from 1974 to 2019 (N=378,276), females comprised 12% of first and 11% of senior authors of the most cited publications, and 6 of the most cited authors in dermatology were women.² In another study analyzing the most prolific dermatologic authors based on h-index, 0% from 1980 to 1989 and 19% from 2010 to 2019 were female (N=393,488).³ Because citation number and h-index favor longer-practicing dermatologists, we examined dermatology author productivity and gender trends by recent publication rates.

The Scopus database was searched for dermatology publications by using the field category “dermatology”from January 1, 2017, to October 7, 2022. Nondermatologists and authors with the same initials were excluded. Authors were ranked by number of publications, including original articles, case reports, letters, and reviews. Sex, degree, and years of experience were determined via a Google search of the author’s name. The h-index; number of citations; and percentages of first, middle, and last authorship were recorded.

Of the top 50 published dermatologists, 30% were female (n=15) and 56% (n=28) held both MD and PhD degrees (Table). The mean years of experience was 26.27 years (range, 6–44 years), with a mean of 29.23 years in females and 25.87 years in males. The mean h-index was 27.96 (range, 8–88), with 24.87 for females and 29.29 for males. The mean number of citations was 4032.64 (range, 235–36,908), with 2891.13 for females and 4521.86 for males. Thirty-one authors were most frequently middle authors, 18 were senior authors, and 1 was a first author. On average (SD), authors were senior or first author in 47.97% (20.08%) of their publications (range, 6.32%–94.93%).

Our study shows that females were more highly represented as top dermatology authors (30%) as measured by publication numbers from 2017 to 2022 than in studies measuring citation rate from 1974 to 2019 (12%)² or h-index from 2010 to 2019 (19%).³ Similarly, in a study of dermatology authorship from 2009 to 2019, on average, females represented 51.06% first and 38.18% last authors.⁴

The proportion of females in the dermatology workforce has increased, with 3964 of 10,385 (38.2%) active dermatologists in 2007⁵ being female vs 6372 of 12,505 (51.0%) in 2019.⁶ The lower proportion of practicing female dermatologists in earlier years likely accounts for the lower percentage of females in dermatology citations and h-index top lists during that time, given that citation and h-index metrics are biased to dermatologists with longer careers.

Although our data are encouraging, females still accounted for less than one-third of the top 50 authors by publication numbers. Gender inequalities persist, with only one-third of a total of 1292 National Institutes of Health dermatology grants and one-fourth of Research Project Grant Program (R01) grants being awarded to females in the years 2009 to 2014.⁷ Therefore, formal and informal mentorship, protected time for research, resources for childcare, and opportunities for funding will be critical in supporting female dermatologists to both publish highly impactful research and obtain research grants.

Limitations of our study include the omission of authors with identical initials and the inability to account for name changes. Furthermore, Scopus does not include all articles published by each author. Finally, publication number reflects quantity but may not reflect quality.

By quantitating dermatology author publication numbers, we found better representation of female authors compared with studies measuring citation number and h-index. With higher proportions of female dermatology trainees and efforts to increase mentorship and research support for female dermatologists, we expect improved equality in top lists of dermatology citations and h-index values.

To the Editor:

Citation number and Hirsch index (h-index) have long been employed as metrics of productivity for academic scholarship. The h-index is defined as the highest number of publications (the maximum h value) of an author who has published at least h papers, each cited by other authors at least h times.¹ In a bibliometric analysis of the most frequently cited authors in dermatology from 1974 to 2019 (N=378,276), females comprised 12% of first and 11% of senior authors of the most cited publications, and 6 of the most cited authors in dermatology were women.² In another study analyzing the most prolific dermatologic authors based on h-index, 0% from 1980 to 1989 and 19% from 2010 to 2019 were female (N=393,488).³ Because citation number and h-index favor longer-practicing dermatologists, we examined dermatology author productivity and gender trends by recent publication rates.

The Scopus database was searched for dermatology publications by using the field category “dermatology”from January 1, 2017, to October 7, 2022. Nondermatologists and authors with the same initials were excluded. Authors were ranked by number of publications, including original articles, case reports, letters, and reviews. Sex, degree, and years of experience were determined via a Google search of the author’s name. The h-index; number of citations; and percentages of first, middle, and last authorship were recorded.

Of the top 50 published dermatologists, 30% were female (n=15) and 56% (n=28) held both MD and PhD degrees (Table). The mean years of experience was 26.27 years (range, 6–44 years), with a mean of 29.23 years in females and 25.87 years in males. The mean h-index was 27.96 (range, 8–88), with 24.87 for females and 29.29 for males. The mean number of citations was 4032.64 (range, 235–36,908), with 2891.13 for females and 4521.86 for males. Thirty-one authors were most frequently middle authors, 18 were senior authors, and 1 was a first author. On average (SD), authors were senior or first author in 47.97% (20.08%) of their publications (range, 6.32%–94.93%).

Our study shows that females were more highly represented as top dermatology authors (30%) as measured by publication numbers from 2017 to 2022 than in studies measuring citation rate from 1974 to 2019 (12%)² or h-index from 2010 to 2019 (19%).³ Similarly, in a study of dermatology authorship from 2009 to 2019, on average, females represented 51.06% first and 38.18% last authors.⁴

The proportion of females in the dermatology workforce has increased, with 3964 of 10,385 (38.2%) active dermatologists in 2007⁵ being female vs 6372 of 12,505 (51.0%) in 2019.⁶ The lower proportion of practicing female dermatologists in earlier years likely accounts for the lower percentage of females in dermatology citations and h-index top lists during that time, given that citation and h-index metrics are biased to dermatologists with longer careers.

Although our data are encouraging, females still accounted for less than one-third of the top 50 authors by publication numbers. Gender inequalities persist, with only one-third of a total of 1292 National Institutes of Health dermatology grants and one-fourth of Research Project Grant Program (R01) grants being awarded to females in the years 2009 to 2014.⁷ Therefore, formal and informal mentorship, protected time for research, resources for childcare, and opportunities for funding will be critical in supporting female dermatologists to both publish highly impactful research and obtain research grants.

Limitations of our study include the omission of authors with identical initials and the inability to account for name changes. Furthermore, Scopus does not include all articles published by each author. Finally, publication number reflects quantity but may not reflect quality.

By quantitating dermatology author publication numbers, we found better representation of female authors compared with studies measuring citation number and h-index. With higher proportions of female dermatology trainees and efforts to increase mentorship and research support for female dermatologists, we expect improved equality in top lists of dermatology citations and h-index values.

Hidradenitis suppurativa (HS), a chronic inflammatory disease that is characterized by tender inflamed nodules of the skin and subcutaneous tissue, disproportionately affects postpubertal females as well as Black/African American individuals. The nodules can rupture, form sinus tracts, and scar. ¹ Hidradenitis suppurativa has been associated with cardiovascular disease, type 2 diabetes mellitus, polycystic ovary syndrome, depression, suicide, and substance use disorders. Because of the symptom burden and associated conditions, HS can be a painful and distressing disease that substantially impairs the quality of life for individuals with this condition. ²

Pain is a commonly reported symptom in HS that often goes untreated. Furthermore, HS-related pain is complex due to the involvement of different pain types that require various treatment modalities.³ According to Savage et al,⁴ recognizing whether HS-related pain is acute, chronic, neuropathic, or nociceptive is vital in establishing a framework for an effective pain management scheme. Currently, such established multimodal pain management strategies in dermatology do not exist. In 2021, dermatology-specific pain management strategies proposed the use of a multimodal regimen to address the multifaceted nature of HS-related pain.⁴ However, these strategies failed to recognize the systemic racial and ethnic biases in the US health care system that undermine pain management care for minority groups.^5,6 One approach to combatting racial disparities in pain management is determining average pain levels across racial groups.⁷ This study sought to compare HS-related pain scores by racial groups. Furthermore, we assessed differences in perception of patients’ respective pain management regimens by race. We hypothesized that the average HS-related pain intensities and pain management would differ between self-reported racial groups.

Methods  

This cross-sectional study took place over 5 months (August through December 2021). A survey was emailed to 2198 adult patients with HS in the University of Alabama Health System. The survey consisted of demographic and general questions about a patient’s HS. Pain scores were captured using the numeric rating scale (NRS), a measurement tool for pain intensity on a scale from 0 to 10. ⁸ Age at diagnosis, gender, education level, household income, total body areas affected by HS, disease severity (categorized as mild, moderate, and severe), comorbidities including mood disorders, tobacco use, and HS and HS-related pain medication regimens also were collected. Additionally, participants were asked about their level of agreement with the following statements: “I am satisfied with how my pain related to HS is being managed by my doctors” and “My pain related to HS is under control.” The level of agreement was measured using a 5-point Likert scale, with responses ranging from strongly disagree to strongly agree. All data included in the analysis were self-reported. The study received institutional review board approval for the University of Alabama at Birmingham.

Statistical Analysis—Descriptive statistics were used to assess statistical differences in patient characteristics of Black/African American participants compared to other participants, including White, Asian, and Hispanic/Latino participants. Thirteen participants were excluded from the final analysis: 2 participants were missing data, and 11 biracial participants were excluded due to overlapping White and Black/African American races that may have confounded the analysis. Categorical variables were reported as frequencies and percentages, and χ² and Fisher exact tests, when necessary, were used to test for statistically significant differences. Continuous variables were summarized with means and standard deviations, and a t test was used for statistically significant differences.

Logistic regression was performed to assess the relationship between race and pain after adjusting for confounding variables such as obesity, current tobacco use, self-reported HS severity, and the presence of comorbidities. A total of 204 patient records were included in the analysis, of which 70 (34.3%) had a pain score of 8 or higher, which indicates very severe pain intensity levels on the NRS,⁸ and were selected as a cut point based on the distribution of responses. For this cross-sectional cohort, our approach was to compare characteristics of those classified with a top score of 8 or higher (n=70) vs a top score of 0 to 7 (n=134)(cases vs noncases). Statistical analyses were performed using JMP Pro 16 (JMP Statistical Discovery LLC) at an α=.05 significance level; logistic regression was performed using SPSS Statistics (IBM). For the logistic regression, we grouped patient race into 2 categories: Black/African American and Other, which included White, Asian, and Hispanic/Latino participants.

Crude and adjusted multivariable logistic regression analyses were used to calculate prevalence odds ratios with 95% confidence intervals. Covariate inclusion in the multivariable logistic regression was based on a priori hypothesis/knowledge and was meant to estimate the independent effect of race after adjustment for income, HS severity, and history of prescription pain medication use. Other variables, including tobacco use, obesity, mood disorders, and current HS treatments, were all individually tested in the multivariate analysis and did not significantly impact the odds ratio for high pain. Statistical adjustment slightly decreased (19%) the magnitude between crude and adjusted prevalence odds ratios for the association between Black/African American race and high pain score.

Results  

Survey Demographics —The final analysis included 204 survey respondents. Most respondents were Black/African American (58.82%), and nearly all were female (89.71%)(Table 1). The mean age (SD) of respondents was 37.37 (11.29) years (range, 19-70 years). Many respondents reported having completed some college (36.27%) or receiving a bachelor’s degree (19.12%). Of patients who were not Black/African American, 10.71% had higher than a master’s degree, whereas no Black/African American patients held a degree higher than a master’s ( P = .0052). Additionally, more Black/African American respondents (35.83%) reported an annual household income level of less than $25,000 compared with respondents who were not Black/African American (19.05%, P = .0001). Most respondents rated the severity of their HS as moderate or severe (46.57% and 41.67%, respectively), and there was no significant difference in reported severity of HS between racial groups ( P = .5395).

 

Pain Scores—As documented in the Methods, respondents were asked to rate their HS-related pain intensity from 0 to 10 using the NRS. The average pain score (SD)—the level of pain intensity over the prior month—was 6.39 (2.56)(range, 0–10). The mean pain score (SD) at the time of the survey was 3.61 (2.98)(range, 0–10)(Table 1). These data revealed that Black/African American patients had a significantly higher average pain score (SD) than patients with HS who were not Black/African American (7.08 [2.49] and 5.40 [2.35], respectively; P<.0001). After adjustment with multivariable logistical regression, Black/African American patients had 4-fold increased odds for very severe levels of pain (score of ≥8) compared with patients who were not Black/African American.

Pain Management—Although pain scores were higher for Black/African American patients with HS, there was no significant difference in the perception of pain control between racial groups (P=.0761). Additionally, we found low income (adjusted prevalence odds ratio [POR], 0.22; 95% CI, 0.05-0.91), a history of prescription pain medication use (adjusted POR, 2.25; 95% CI, 1.13-4.51), and HS severity (adjusted POR, 4.40; 95% CI, 1.11-17.36) all to be independent risk factors contributing to higher pain scores in patients with HS (Table 2). Lastly, we noted current or reported history of pain medication use was significantly correlated with higher pain scores (P=.0280 and P=.0213, respectively).

Satisfaction With Pain Management—The level of satisfaction with physician management of HS-related pain was significantly different between Black/African American patients and those who were not Black/African American (P=.0129). Of those who identified as Black/African American, 26.7% (n=32) strongly disagreed with the statement, “I am satisfied with how my pain related to HS is being managed by my doctors,” whereas only 15.5% (n=13) of patients who were not Black/African American strongly disagreed. 

Comment

There is no cure for HS, and a large focus of treatment is pain management. Because racial disparities in the treatment of chronic pain will affect those with HS, we conducted a cross-sectional analysis of pain and pain management among HS patients. We found that Black/African American patients with HS have higher average pain scores than those who are not Black/African American and were 4 times more likely to experience very severe pain. Prior studies have established that patients with HS often report higher pain levels than patients with other chronic inflammatory skin conditions, ^7,8 and our study identified racial disparities in HS-related pain management.

Measuring pain is challenging because of its multidimensional and subjective nature, making it essential to consider underlying causes and patients’ emotional responses to pain.⁹ By adjusting for confounding factors that may influence pain, such as mood disorders, disease severity, comorbidities, and medication use, we were able to gain better insight into fundamental differences in average pain intensity levels among racial groups and assess what factors may be contributing to a patient’s pain perception. Our study determined that lower income levels, higher HS disease severity, and a history of prescription pain medication use were all independent risk factors for high pain. Of note, obesity, tobacco use, and mood disorders such as anxiety and depression did not significantly differ between racial groups or increase the odds of high pain between racial groups identified.

With low income being an independent risk factor for high pain, we must consider the social determinants of health and how they may influence the pain experience in HS. We speculate that low income may be associated with other social determinants of health for the patients assessed in this study, such as lack of social and community support or limited health care access that contribute to worse health outcomes.^10,11 In addition, low income contributes to limited access to medical care or treatments¹²; without access to effective HS management, lower-income patients may be at risk for higher disease severity and thus higher pain levels. However, economic stability is only a part of the whole picture; therefore, assessing the other social determinants of health in patients with HS may lead to better health outcomes and quality of life.

Another identified risk factor for high pain was a reported history of prescription pain medication use. This finding suggests that patients with moderate to severe pain likely have required stronger analgesic medications in the past. We further speculate that high pain levels in patients who have received prescription pain medications indicate either undertreatment, mistreatment, or recalcitrant pain. More research is needed to assess the relationship between HS-related pain intensity, analgesic medications, and providers who manage HS-related pain.

We also found that Black/African American patients with HS had a significantly higher dissatisfaction with their physician’s management of their pain, which could be attributable to several factors, including biological differences in medication metabolism (in which the patient has medication-resistant HS), undertreatment of pain, and/or poor doctor-patient relations. These reasons coincide with other diseases where health disparities are found.^13-15 Recognizing these factors will be key to dismantling the disparities in HS that are noted within this study. The limitations of this work include the cross-sectional study design and its inability to evaluate causal factors of high pain levels across racial groups, the NRS lack of insight on pain chronicity or pain experience,⁷ the lack of provider or institution perspectives, and self-reported data. Additionally, only patients with email access were included, which may have excluded vulnerable populations with more pain associated with their HS.

Our findings highlight an area for further investigation to assess why these racial differences exist in HS-related pain. The results also emphasize the need for research evaluating whether systemic or health care provider biases contribute to racial differences in HS-related pain management.

Acknowledgment — Dr. Weir was supported by the Predoctoral Clinical/Translational Research Program (TL1), a National Institutes of Health Ruth L. Kirschstein National Research Service Award (NRSA), through the University of Alabama at Birmingham (UAB) Center for Clinical and Translational Science (CCTS).  

Hidradenitis suppurativa (HS), a chronic inflammatory disease that is characterized by tender inflamed nodules of the skin and subcutaneous tissue, disproportionately affects postpubertal females as well as Black/African American individuals. The nodules can rupture, form sinus tracts, and scar. ¹ Hidradenitis suppurativa has been associated with cardiovascular disease, type 2 diabetes mellitus, polycystic ovary syndrome, depression, suicide, and substance use disorders. Because of the symptom burden and associated conditions, HS can be a painful and distressing disease that substantially impairs the quality of life for individuals with this condition. ²

Pain is a commonly reported symptom in HS that often goes untreated. Furthermore, HS-related pain is complex due to the involvement of different pain types that require various treatment modalities.³ According to Savage et al,⁴ recognizing whether HS-related pain is acute, chronic, neuropathic, or nociceptive is vital in establishing a framework for an effective pain management scheme. Currently, such established multimodal pain management strategies in dermatology do not exist. In 2021, dermatology-specific pain management strategies proposed the use of a multimodal regimen to address the multifaceted nature of HS-related pain.⁴ However, these strategies failed to recognize the systemic racial and ethnic biases in the US health care system that undermine pain management care for minority groups.^5,6 One approach to combatting racial disparities in pain management is determining average pain levels across racial groups.⁷ This study sought to compare HS-related pain scores by racial groups. Furthermore, we assessed differences in perception of patients’ respective pain management regimens by race. We hypothesized that the average HS-related pain intensities and pain management would differ between self-reported racial groups.

Methods  

This cross-sectional study took place over 5 months (August through December 2021). A survey was emailed to 2198 adult patients with HS in the University of Alabama Health System. The survey consisted of demographic and general questions about a patient’s HS. Pain scores were captured using the numeric rating scale (NRS), a measurement tool for pain intensity on a scale from 0 to 10. ⁸ Age at diagnosis, gender, education level, household income, total body areas affected by HS, disease severity (categorized as mild, moderate, and severe), comorbidities including mood disorders, tobacco use, and HS and HS-related pain medication regimens also were collected. Additionally, participants were asked about their level of agreement with the following statements: “I am satisfied with how my pain related to HS is being managed by my doctors” and “My pain related to HS is under control.” The level of agreement was measured using a 5-point Likert scale, with responses ranging from strongly disagree to strongly agree. All data included in the analysis were self-reported. The study received institutional review board approval for the University of Alabama at Birmingham.

Statistical Analysis—Descriptive statistics were used to assess statistical differences in patient characteristics of Black/African American participants compared to other participants, including White, Asian, and Hispanic/Latino participants. Thirteen participants were excluded from the final analysis: 2 participants were missing data, and 11 biracial participants were excluded due to overlapping White and Black/African American races that may have confounded the analysis. Categorical variables were reported as frequencies and percentages, and χ² and Fisher exact tests, when necessary, were used to test for statistically significant differences. Continuous variables were summarized with means and standard deviations, and a t test was used for statistically significant differences.

Logistic regression was performed to assess the relationship between race and pain after adjusting for confounding variables such as obesity, current tobacco use, self-reported HS severity, and the presence of comorbidities. A total of 204 patient records were included in the analysis, of which 70 (34.3%) had a pain score of 8 or higher, which indicates very severe pain intensity levels on the NRS,⁸ and were selected as a cut point based on the distribution of responses. For this cross-sectional cohort, our approach was to compare characteristics of those classified with a top score of 8 or higher (n=70) vs a top score of 0 to 7 (n=134)(cases vs noncases). Statistical analyses were performed using JMP Pro 16 (JMP Statistical Discovery LLC) at an α=.05 significance level; logistic regression was performed using SPSS Statistics (IBM). For the logistic regression, we grouped patient race into 2 categories: Black/African American and Other, which included White, Asian, and Hispanic/Latino participants.

Crude and adjusted multivariable logistic regression analyses were used to calculate prevalence odds ratios with 95% confidence intervals. Covariate inclusion in the multivariable logistic regression was based on a priori hypothesis/knowledge and was meant to estimate the independent effect of race after adjustment for income, HS severity, and history of prescription pain medication use. Other variables, including tobacco use, obesity, mood disorders, and current HS treatments, were all individually tested in the multivariate analysis and did not significantly impact the odds ratio for high pain. Statistical adjustment slightly decreased (19%) the magnitude between crude and adjusted prevalence odds ratios for the association between Black/African American race and high pain score.

Results  

Survey Demographics —The final analysis included 204 survey respondents. Most respondents were Black/African American (58.82%), and nearly all were female (89.71%)(Table 1). The mean age (SD) of respondents was 37.37 (11.29) years (range, 19-70 years). Many respondents reported having completed some college (36.27%) or receiving a bachelor’s degree (19.12%). Of patients who were not Black/African American, 10.71% had higher than a master’s degree, whereas no Black/African American patients held a degree higher than a master’s ( P = .0052). Additionally, more Black/African American respondents (35.83%) reported an annual household income level of less than $25,000 compared with respondents who were not Black/African American (19.05%, P = .0001). Most respondents rated the severity of their HS as moderate or severe (46.57% and 41.67%, respectively), and there was no significant difference in reported severity of HS between racial groups ( P = .5395).

 

Pain Scores—As documented in the Methods, respondents were asked to rate their HS-related pain intensity from 0 to 10 using the NRS. The average pain score (SD)—the level of pain intensity over the prior month—was 6.39 (2.56)(range, 0–10). The mean pain score (SD) at the time of the survey was 3.61 (2.98)(range, 0–10)(Table 1). These data revealed that Black/African American patients had a significantly higher average pain score (SD) than patients with HS who were not Black/African American (7.08 [2.49] and 5.40 [2.35], respectively; P<.0001). After adjustment with multivariable logistical regression, Black/African American patients had 4-fold increased odds for very severe levels of pain (score of ≥8) compared with patients who were not Black/African American.

Pain Management—Although pain scores were higher for Black/African American patients with HS, there was no significant difference in the perception of pain control between racial groups (P=.0761). Additionally, we found low income (adjusted prevalence odds ratio [POR], 0.22; 95% CI, 0.05-0.91), a history of prescription pain medication use (adjusted POR, 2.25; 95% CI, 1.13-4.51), and HS severity (adjusted POR, 4.40; 95% CI, 1.11-17.36) all to be independent risk factors contributing to higher pain scores in patients with HS (Table 2). Lastly, we noted current or reported history of pain medication use was significantly correlated with higher pain scores (P=.0280 and P=.0213, respectively).

Satisfaction With Pain Management—The level of satisfaction with physician management of HS-related pain was significantly different between Black/African American patients and those who were not Black/African American (P=.0129). Of those who identified as Black/African American, 26.7% (n=32) strongly disagreed with the statement, “I am satisfied with how my pain related to HS is being managed by my doctors,” whereas only 15.5% (n=13) of patients who were not Black/African American strongly disagreed. 

Comment

There is no cure for HS, and a large focus of treatment is pain management. Because racial disparities in the treatment of chronic pain will affect those with HS, we conducted a cross-sectional analysis of pain and pain management among HS patients. We found that Black/African American patients with HS have higher average pain scores than those who are not Black/African American and were 4 times more likely to experience very severe pain. Prior studies have established that patients with HS often report higher pain levels than patients with other chronic inflammatory skin conditions, ^7,8 and our study identified racial disparities in HS-related pain management.

Measuring pain is challenging because of its multidimensional and subjective nature, making it essential to consider underlying causes and patients’ emotional responses to pain.⁹ By adjusting for confounding factors that may influence pain, such as mood disorders, disease severity, comorbidities, and medication use, we were able to gain better insight into fundamental differences in average pain intensity levels among racial groups and assess what factors may be contributing to a patient’s pain perception. Our study determined that lower income levels, higher HS disease severity, and a history of prescription pain medication use were all independent risk factors for high pain. Of note, obesity, tobacco use, and mood disorders such as anxiety and depression did not significantly differ between racial groups or increase the odds of high pain between racial groups identified.

With low income being an independent risk factor for high pain, we must consider the social determinants of health and how they may influence the pain experience in HS. We speculate that low income may be associated with other social determinants of health for the patients assessed in this study, such as lack of social and community support or limited health care access that contribute to worse health outcomes.^10,11 In addition, low income contributes to limited access to medical care or treatments¹²; without access to effective HS management, lower-income patients may be at risk for higher disease severity and thus higher pain levels. However, economic stability is only a part of the whole picture; therefore, assessing the other social determinants of health in patients with HS may lead to better health outcomes and quality of life.

Another identified risk factor for high pain was a reported history of prescription pain medication use. This finding suggests that patients with moderate to severe pain likely have required stronger analgesic medications in the past. We further speculate that high pain levels in patients who have received prescription pain medications indicate either undertreatment, mistreatment, or recalcitrant pain. More research is needed to assess the relationship between HS-related pain intensity, analgesic medications, and providers who manage HS-related pain.

We also found that Black/African American patients with HS had a significantly higher dissatisfaction with their physician’s management of their pain, which could be attributable to several factors, including biological differences in medication metabolism (in which the patient has medication-resistant HS), undertreatment of pain, and/or poor doctor-patient relations. These reasons coincide with other diseases where health disparities are found.^13-15 Recognizing these factors will be key to dismantling the disparities in HS that are noted within this study. The limitations of this work include the cross-sectional study design and its inability to evaluate causal factors of high pain levels across racial groups, the NRS lack of insight on pain chronicity or pain experience,⁷ the lack of provider or institution perspectives, and self-reported data. Additionally, only patients with email access were included, which may have excluded vulnerable populations with more pain associated with their HS.

Our findings highlight an area for further investigation to assess why these racial differences exist in HS-related pain. The results also emphasize the need for research evaluating whether systemic or health care provider biases contribute to racial differences in HS-related pain management.

Acknowledgment — Dr. Weir was supported by the Predoctoral Clinical/Translational Research Program (TL1), a National Institutes of Health Ruth L. Kirschstein National Research Service Award (NRSA), through the University of Alabama at Birmingham (UAB) Center for Clinical and Translational Science (CCTS).  

Hidradenitis suppurativa (HS), a chronic inflammatory disease that is characterized by tender inflamed nodules of the skin and subcutaneous tissue, disproportionately affects postpubertal females as well as Black/African American individuals. The nodules can rupture, form sinus tracts, and scar. ¹ Hidradenitis suppurativa has been associated with cardiovascular disease, type 2 diabetes mellitus, polycystic ovary syndrome, depression, suicide, and substance use disorders. Because of the symptom burden and associated conditions, HS can be a painful and distressing disease that substantially impairs the quality of life for individuals with this condition. ²

Pain is a commonly reported symptom in HS that often goes untreated. Furthermore, HS-related pain is complex due to the involvement of different pain types that require various treatment modalities.³ According to Savage et al,⁴ recognizing whether HS-related pain is acute, chronic, neuropathic, or nociceptive is vital in establishing a framework for an effective pain management scheme. Currently, such established multimodal pain management strategies in dermatology do not exist. In 2021, dermatology-specific pain management strategies proposed the use of a multimodal regimen to address the multifaceted nature of HS-related pain.⁴ However, these strategies failed to recognize the systemic racial and ethnic biases in the US health care system that undermine pain management care for minority groups.^5,6 One approach to combatting racial disparities in pain management is determining average pain levels across racial groups.⁷ This study sought to compare HS-related pain scores by racial groups. Furthermore, we assessed differences in perception of patients’ respective pain management regimens by race. We hypothesized that the average HS-related pain intensities and pain management would differ between self-reported racial groups.

Methods  

This cross-sectional study took place over 5 months (August through December 2021). A survey was emailed to 2198 adult patients with HS in the University of Alabama Health System. The survey consisted of demographic and general questions about a patient’s HS. Pain scores were captured using the numeric rating scale (NRS), a measurement tool for pain intensity on a scale from 0 to 10. ⁸ Age at diagnosis, gender, education level, household income, total body areas affected by HS, disease severity (categorized as mild, moderate, and severe), comorbidities including mood disorders, tobacco use, and HS and HS-related pain medication regimens also were collected. Additionally, participants were asked about their level of agreement with the following statements: “I am satisfied with how my pain related to HS is being managed by my doctors” and “My pain related to HS is under control.” The level of agreement was measured using a 5-point Likert scale, with responses ranging from strongly disagree to strongly agree. All data included in the analysis were self-reported. The study received institutional review board approval for the University of Alabama at Birmingham.

Statistical Analysis—Descriptive statistics were used to assess statistical differences in patient characteristics of Black/African American participants compared to other participants, including White, Asian, and Hispanic/Latino participants. Thirteen participants were excluded from the final analysis: 2 participants were missing data, and 11 biracial participants were excluded due to overlapping White and Black/African American races that may have confounded the analysis. Categorical variables were reported as frequencies and percentages, and χ² and Fisher exact tests, when necessary, were used to test for statistically significant differences. Continuous variables were summarized with means and standard deviations, and a t test was used for statistically significant differences.

Logistic regression was performed to assess the relationship between race and pain after adjusting for confounding variables such as obesity, current tobacco use, self-reported HS severity, and the presence of comorbidities. A total of 204 patient records were included in the analysis, of which 70 (34.3%) had a pain score of 8 or higher, which indicates very severe pain intensity levels on the NRS,⁸ and were selected as a cut point based on the distribution of responses. For this cross-sectional cohort, our approach was to compare characteristics of those classified with a top score of 8 or higher (n=70) vs a top score of 0 to 7 (n=134)(cases vs noncases). Statistical analyses were performed using JMP Pro 16 (JMP Statistical Discovery LLC) at an α=.05 significance level; logistic regression was performed using SPSS Statistics (IBM). For the logistic regression, we grouped patient race into 2 categories: Black/African American and Other, which included White, Asian, and Hispanic/Latino participants.

Crude and adjusted multivariable logistic regression analyses were used to calculate prevalence odds ratios with 95% confidence intervals. Covariate inclusion in the multivariable logistic regression was based on a priori hypothesis/knowledge and was meant to estimate the independent effect of race after adjustment for income, HS severity, and history of prescription pain medication use. Other variables, including tobacco use, obesity, mood disorders, and current HS treatments, were all individually tested in the multivariate analysis and did not significantly impact the odds ratio for high pain. Statistical adjustment slightly decreased (19%) the magnitude between crude and adjusted prevalence odds ratios for the association between Black/African American race and high pain score.

Results  

Survey Demographics —The final analysis included 204 survey respondents. Most respondents were Black/African American (58.82%), and nearly all were female (89.71%)(Table 1). The mean age (SD) of respondents was 37.37 (11.29) years (range, 19-70 years). Many respondents reported having completed some college (36.27%) or receiving a bachelor’s degree (19.12%). Of patients who were not Black/African American, 10.71% had higher than a master’s degree, whereas no Black/African American patients held a degree higher than a master’s ( P = .0052). Additionally, more Black/African American respondents (35.83%) reported an annual household income level of less than $25,000 compared with respondents who were not Black/African American (19.05%, P = .0001). Most respondents rated the severity of their HS as moderate or severe (46.57% and 41.67%, respectively), and there was no significant difference in reported severity of HS between racial groups ( P = .5395).

 

Pain Scores—As documented in the Methods, respondents were asked to rate their HS-related pain intensity from 0 to 10 using the NRS. The average pain score (SD)—the level of pain intensity over the prior month—was 6.39 (2.56)(range, 0–10). The mean pain score (SD) at the time of the survey was 3.61 (2.98)(range, 0–10)(Table 1). These data revealed that Black/African American patients had a significantly higher average pain score (SD) than patients with HS who were not Black/African American (7.08 [2.49] and 5.40 [2.35], respectively; P<.0001). After adjustment with multivariable logistical regression, Black/African American patients had 4-fold increased odds for very severe levels of pain (score of ≥8) compared with patients who were not Black/African American.

Pain Management—Although pain scores were higher for Black/African American patients with HS, there was no significant difference in the perception of pain control between racial groups (P=.0761). Additionally, we found low income (adjusted prevalence odds ratio [POR], 0.22; 95% CI, 0.05-0.91), a history of prescription pain medication use (adjusted POR, 2.25; 95% CI, 1.13-4.51), and HS severity (adjusted POR, 4.40; 95% CI, 1.11-17.36) all to be independent risk factors contributing to higher pain scores in patients with HS (Table 2). Lastly, we noted current or reported history of pain medication use was significantly correlated with higher pain scores (P=.0280 and P=.0213, respectively).

Satisfaction With Pain Management—The level of satisfaction with physician management of HS-related pain was significantly different between Black/African American patients and those who were not Black/African American (P=.0129). Of those who identified as Black/African American, 26.7% (n=32) strongly disagreed with the statement, “I am satisfied with how my pain related to HS is being managed by my doctors,” whereas only 15.5% (n=13) of patients who were not Black/African American strongly disagreed. 

Comment

There is no cure for HS, and a large focus of treatment is pain management. Because racial disparities in the treatment of chronic pain will affect those with HS, we conducted a cross-sectional analysis of pain and pain management among HS patients. We found that Black/African American patients with HS have higher average pain scores than those who are not Black/African American and were 4 times more likely to experience very severe pain. Prior studies have established that patients with HS often report higher pain levels than patients with other chronic inflammatory skin conditions, ^7,8 and our study identified racial disparities in HS-related pain management.

Measuring pain is challenging because of its multidimensional and subjective nature, making it essential to consider underlying causes and patients’ emotional responses to pain.⁹ By adjusting for confounding factors that may influence pain, such as mood disorders, disease severity, comorbidities, and medication use, we were able to gain better insight into fundamental differences in average pain intensity levels among racial groups and assess what factors may be contributing to a patient’s pain perception. Our study determined that lower income levels, higher HS disease severity, and a history of prescription pain medication use were all independent risk factors for high pain. Of note, obesity, tobacco use, and mood disorders such as anxiety and depression did not significantly differ between racial groups or increase the odds of high pain between racial groups identified.

With low income being an independent risk factor for high pain, we must consider the social determinants of health and how they may influence the pain experience in HS. We speculate that low income may be associated with other social determinants of health for the patients assessed in this study, such as lack of social and community support or limited health care access that contribute to worse health outcomes.^10,11 In addition, low income contributes to limited access to medical care or treatments¹²; without access to effective HS management, lower-income patients may be at risk for higher disease severity and thus higher pain levels. However, economic stability is only a part of the whole picture; therefore, assessing the other social determinants of health in patients with HS may lead to better health outcomes and quality of life.

Another identified risk factor for high pain was a reported history of prescription pain medication use. This finding suggests that patients with moderate to severe pain likely have required stronger analgesic medications in the past. We further speculate that high pain levels in patients who have received prescription pain medications indicate either undertreatment, mistreatment, or recalcitrant pain. More research is needed to assess the relationship between HS-related pain intensity, analgesic medications, and providers who manage HS-related pain.

We also found that Black/African American patients with HS had a significantly higher dissatisfaction with their physician’s management of their pain, which could be attributable to several factors, including biological differences in medication metabolism (in which the patient has medication-resistant HS), undertreatment of pain, and/or poor doctor-patient relations. These reasons coincide with other diseases where health disparities are found.^13-15 Recognizing these factors will be key to dismantling the disparities in HS that are noted within this study. The limitations of this work include the cross-sectional study design and its inability to evaluate causal factors of high pain levels across racial groups, the NRS lack of insight on pain chronicity or pain experience,⁷ the lack of provider or institution perspectives, and self-reported data. Additionally, only patients with email access were included, which may have excluded vulnerable populations with more pain associated with their HS.

Our findings highlight an area for further investigation to assess why these racial differences exist in HS-related pain. The results also emphasize the need for research evaluating whether systemic or health care provider biases contribute to racial differences in HS-related pain management.

Acknowledgment — Dr. Weir was supported by the Predoctoral Clinical/Translational Research Program (TL1), a National Institutes of Health Ruth L. Kirschstein National Research Service Award (NRSA), through the University of Alabama at Birmingham (UAB) Center for Clinical and Translational Science (CCTS).  

To the Editor:

Psoriasis is an immune-mediated dermatologic condition that is associated with various comorbidities, including obesity.¹ The underlying pathophysiology of psoriasis has been extensively studied, and recent research has discussed the role of obesity in IL-17 secretion.² The relationship between being overweight/obese and having psoriasis has been documented in the literature.^1,2 However, this association in a recent population is lacking. We sought to investigate the association between psoriasis and obesity utilizing a representative US population of adults—the 2009-2014 National Health and Nutrition Examination Survey (NHANES) data,³ which contains the most recent psoriasis data.

We conducted a population-based, cross-sectional study focused on patients 20 years and older with psoriasis from the 2009-2014 NHANES database. Three 2-year cycles of NHANES data were combined to create our 2009 to 2014 dataset. In the Table, numerous variables including age, sex, household income, race/ethnicity, education, diabetes status, tobacco use, body mass index (BMI), waist circumference, and being called overweight by a health care provider were analyzed using χ² or t test analyses to evaluate for differences among those with and without psoriasis. Diabetes status was assessed by the question “Other than during pregnancy, have you ever been told by a doctor or health professional that you have diabetes or sugar diabetes?” Tobacco use was assessed by the question “Have you smoked at least 100 cigarettes in your entire life?” Psoriasis status was assessed by a self-reported response to the question “Have you ever been told by a doctor or other health care professional that you had psoriasis?” Three different outcome variables were used to determine if patients were overweight or obese: BMI, waist circumference, and response to the question “Has a doctor or other health professional ever told you that you were overweight?” Obesity was defined as having a BMI of 30 or higher or waist circumference of 102 cm or more in males and 88 cm or more in females.⁴ Being overweight was defined as having a BMI of 25 to 29.99 or response of Yes to “Has a doctor or other health professional ever told you that you were overweight?”

Initially, there were 17,547 participants 20 years and older from 2009 to 2014, but 1654 participants were excluded because of missing data for obesity or psoriasis; therefore, 15,893 patients were included in our analysis. Multivariable logistic regressions were utilized to examine the association between psoriasis and being overweight/obese (eTable). Additionally, the models were adjusted based on age, sex, household income, race/ethnicity, diabetes status, and tobacco use. All data processing and analysis were performed in Stata/MP 17 (StataCorp LLC). P<.05 was considered statistically significant.

The Table shows characteristics of US adults with and without psoriasis in NHANES 2009-2014. We found that the variables of interest evaluating body weight that were significantly different on analysis between patients with and without psoriasis included waist circumference—patients with psoriasis had a significantly higher waist circumference (P=.009)—and being told by a health care provider that they are overweight (P<.0001), which supports the findings by Love et al,⁵ who reported abdominal obesity was the most common feature of metabolic syndrome exhibited among patients with psoriasis.

Multivariable logistic regression analysis (eTable) revealed that there was a significant association between psoriasis and BMI of 25 to 29.99 (adjusted odds ratio [AOR], 1.34; 95% CI, 1.02-1.76; P=.04) and being told by a health care provider that they are overweight (AOR, 1.91; 95% CI, 1.44-2.52; P<.001). After adjusting for confounding variables, there was no significant association between psoriasis and a BMI of 30 or higher (AOR, 1.00; 95% CI, 0.73-1.38; P=.99) or a waist circumference of 102 cm or more in males and 88 cm or more in females (AOR, 1.15; 95% CI, 0.86-1.53; P=.3).

Our findings suggest that a few variables indicative of being overweight or obese are associated with psoriasis. This relationship most likely is due to increased adipokine, including resistin, levels in overweight individuals, resulting in a proinflammatory state.⁶ It has been suggested that BMI alone is not a definitive marker for measuring fat storage levels in individuals. People can have a normal or slightly elevated BMI but possess excessive adiposity, resulting in chronic inflammation.⁶ Therefore, our findings of a significant association between psoriasis and being told by a health care provider that they are overweight might be a stronger measurement for possessing excessive fat, as this is likely due to clinical judgment rather than BMI measurement.

Moreover, it should be noted that the potential reason for the lack of association between BMI of 30 or higher and psoriasis in our analysis may be a result of BMI serving as a poor measurement for adiposity. Additionally, Armstrong and colleagues⁷ discussed that the association between BMI and psoriasis was stronger for patients with moderate to severe psoriasis. Our study consisted of NHANES data for self-reported psoriasis diagnoses without a psoriasis severity index, making it difficult to extrapolate which individuals had mild or moderate to severe psoriasis, which may have contributed to our finding of no association between BMI of 30 or higher and psoriasis.

The self-reported nature of the survey questions and lack of questions regarding psoriasis severity serve as limitations to the study. Both obesity and psoriasis can have various systemic consequences, such as cardiovascular disease, due to the development of an inflammatory state.⁸ Future studies may explore other body measurements that indicate being overweight or obese and the potential synergistic relationship of obesity and psoriasis severity, optimizing the development of effective treatment plans.

To the Editor:

Psoriasis is an immune-mediated dermatologic condition that is associated with various comorbidities, including obesity.¹ The underlying pathophysiology of psoriasis has been extensively studied, and recent research has discussed the role of obesity in IL-17 secretion.² The relationship between being overweight/obese and having psoriasis has been documented in the literature.^1,2 However, this association in a recent population is lacking. We sought to investigate the association between psoriasis and obesity utilizing a representative US population of adults—the 2009-2014 National Health and Nutrition Examination Survey (NHANES) data,³ which contains the most recent psoriasis data.

We conducted a population-based, cross-sectional study focused on patients 20 years and older with psoriasis from the 2009-2014 NHANES database. Three 2-year cycles of NHANES data were combined to create our 2009 to 2014 dataset. In the Table, numerous variables including age, sex, household income, race/ethnicity, education, diabetes status, tobacco use, body mass index (BMI), waist circumference, and being called overweight by a health care provider were analyzed using χ² or t test analyses to evaluate for differences among those with and without psoriasis. Diabetes status was assessed by the question “Other than during pregnancy, have you ever been told by a doctor or health professional that you have diabetes or sugar diabetes?” Tobacco use was assessed by the question “Have you smoked at least 100 cigarettes in your entire life?” Psoriasis status was assessed by a self-reported response to the question “Have you ever been told by a doctor or other health care professional that you had psoriasis?” Three different outcome variables were used to determine if patients were overweight or obese: BMI, waist circumference, and response to the question “Has a doctor or other health professional ever told you that you were overweight?” Obesity was defined as having a BMI of 30 or higher or waist circumference of 102 cm or more in males and 88 cm or more in females.⁴ Being overweight was defined as having a BMI of 25 to 29.99 or response of Yes to “Has a doctor or other health professional ever told you that you were overweight?”

Initially, there were 17,547 participants 20 years and older from 2009 to 2014, but 1654 participants were excluded because of missing data for obesity or psoriasis; therefore, 15,893 patients were included in our analysis. Multivariable logistic regressions were utilized to examine the association between psoriasis and being overweight/obese (eTable). Additionally, the models were adjusted based on age, sex, household income, race/ethnicity, diabetes status, and tobacco use. All data processing and analysis were performed in Stata/MP 17 (StataCorp LLC). P<.05 was considered statistically significant.

The Table shows characteristics of US adults with and without psoriasis in NHANES 2009-2014. We found that the variables of interest evaluating body weight that were significantly different on analysis between patients with and without psoriasis included waist circumference—patients with psoriasis had a significantly higher waist circumference (P=.009)—and being told by a health care provider that they are overweight (P<.0001), which supports the findings by Love et al,⁵ who reported abdominal obesity was the most common feature of metabolic syndrome exhibited among patients with psoriasis.

Multivariable logistic regression analysis (eTable) revealed that there was a significant association between psoriasis and BMI of 25 to 29.99 (adjusted odds ratio [AOR], 1.34; 95% CI, 1.02-1.76; P=.04) and being told by a health care provider that they are overweight (AOR, 1.91; 95% CI, 1.44-2.52; P<.001). After adjusting for confounding variables, there was no significant association between psoriasis and a BMI of 30 or higher (AOR, 1.00; 95% CI, 0.73-1.38; P=.99) or a waist circumference of 102 cm or more in males and 88 cm or more in females (AOR, 1.15; 95% CI, 0.86-1.53; P=.3).

Our findings suggest that a few variables indicative of being overweight or obese are associated with psoriasis. This relationship most likely is due to increased adipokine, including resistin, levels in overweight individuals, resulting in a proinflammatory state.⁶ It has been suggested that BMI alone is not a definitive marker for measuring fat storage levels in individuals. People can have a normal or slightly elevated BMI but possess excessive adiposity, resulting in chronic inflammation.⁶ Therefore, our findings of a significant association between psoriasis and being told by a health care provider that they are overweight might be a stronger measurement for possessing excessive fat, as this is likely due to clinical judgment rather than BMI measurement.

Moreover, it should be noted that the potential reason for the lack of association between BMI of 30 or higher and psoriasis in our analysis may be a result of BMI serving as a poor measurement for adiposity. Additionally, Armstrong and colleagues⁷ discussed that the association between BMI and psoriasis was stronger for patients with moderate to severe psoriasis. Our study consisted of NHANES data for self-reported psoriasis diagnoses without a psoriasis severity index, making it difficult to extrapolate which individuals had mild or moderate to severe psoriasis, which may have contributed to our finding of no association between BMI of 30 or higher and psoriasis.

The self-reported nature of the survey questions and lack of questions regarding psoriasis severity serve as limitations to the study. Both obesity and psoriasis can have various systemic consequences, such as cardiovascular disease, due to the development of an inflammatory state.⁸ Future studies may explore other body measurements that indicate being overweight or obese and the potential synergistic relationship of obesity and psoriasis severity, optimizing the development of effective treatment plans.

To the Editor:

Psoriasis is an immune-mediated dermatologic condition that is associated with various comorbidities, including obesity.¹ The underlying pathophysiology of psoriasis has been extensively studied, and recent research has discussed the role of obesity in IL-17 secretion.² The relationship between being overweight/obese and having psoriasis has been documented in the literature.^1,2 However, this association in a recent population is lacking. We sought to investigate the association between psoriasis and obesity utilizing a representative US population of adults—the 2009-2014 National Health and Nutrition Examination Survey (NHANES) data,³ which contains the most recent psoriasis data.

We conducted a population-based, cross-sectional study focused on patients 20 years and older with psoriasis from the 2009-2014 NHANES database. Three 2-year cycles of NHANES data were combined to create our 2009 to 2014 dataset. In the Table, numerous variables including age, sex, household income, race/ethnicity, education, diabetes status, tobacco use, body mass index (BMI), waist circumference, and being called overweight by a health care provider were analyzed using χ² or t test analyses to evaluate for differences among those with and without psoriasis. Diabetes status was assessed by the question “Other than during pregnancy, have you ever been told by a doctor or health professional that you have diabetes or sugar diabetes?” Tobacco use was assessed by the question “Have you smoked at least 100 cigarettes in your entire life?” Psoriasis status was assessed by a self-reported response to the question “Have you ever been told by a doctor or other health care professional that you had psoriasis?” Three different outcome variables were used to determine if patients were overweight or obese: BMI, waist circumference, and response to the question “Has a doctor or other health professional ever told you that you were overweight?” Obesity was defined as having a BMI of 30 or higher or waist circumference of 102 cm or more in males and 88 cm or more in females.⁴ Being overweight was defined as having a BMI of 25 to 29.99 or response of Yes to “Has a doctor or other health professional ever told you that you were overweight?”

Initially, there were 17,547 participants 20 years and older from 2009 to 2014, but 1654 participants were excluded because of missing data for obesity or psoriasis; therefore, 15,893 patients were included in our analysis. Multivariable logistic regressions were utilized to examine the association between psoriasis and being overweight/obese (eTable). Additionally, the models were adjusted based on age, sex, household income, race/ethnicity, diabetes status, and tobacco use. All data processing and analysis were performed in Stata/MP 17 (StataCorp LLC). P<.05 was considered statistically significant.

The Table shows characteristics of US adults with and without psoriasis in NHANES 2009-2014. We found that the variables of interest evaluating body weight that were significantly different on analysis between patients with and without psoriasis included waist circumference—patients with psoriasis had a significantly higher waist circumference (P=.009)—and being told by a health care provider that they are overweight (P<.0001), which supports the findings by Love et al,⁵ who reported abdominal obesity was the most common feature of metabolic syndrome exhibited among patients with psoriasis.

Multivariable logistic regression analysis (eTable) revealed that there was a significant association between psoriasis and BMI of 25 to 29.99 (adjusted odds ratio [AOR], 1.34; 95% CI, 1.02-1.76; P=.04) and being told by a health care provider that they are overweight (AOR, 1.91; 95% CI, 1.44-2.52; P<.001). After adjusting for confounding variables, there was no significant association between psoriasis and a BMI of 30 or higher (AOR, 1.00; 95% CI, 0.73-1.38; P=.99) or a waist circumference of 102 cm or more in males and 88 cm or more in females (AOR, 1.15; 95% CI, 0.86-1.53; P=.3).

Our findings suggest that a few variables indicative of being overweight or obese are associated with psoriasis. This relationship most likely is due to increased adipokine, including resistin, levels in overweight individuals, resulting in a proinflammatory state.⁶ It has been suggested that BMI alone is not a definitive marker for measuring fat storage levels in individuals. People can have a normal or slightly elevated BMI but possess excessive adiposity, resulting in chronic inflammation.⁶ Therefore, our findings of a significant association between psoriasis and being told by a health care provider that they are overweight might be a stronger measurement for possessing excessive fat, as this is likely due to clinical judgment rather than BMI measurement.

Moreover, it should be noted that the potential reason for the lack of association between BMI of 30 or higher and psoriasis in our analysis may be a result of BMI serving as a poor measurement for adiposity. Additionally, Armstrong and colleagues⁷ discussed that the association between BMI and psoriasis was stronger for patients with moderate to severe psoriasis. Our study consisted of NHANES data for self-reported psoriasis diagnoses without a psoriasis severity index, making it difficult to extrapolate which individuals had mild or moderate to severe psoriasis, which may have contributed to our finding of no association between BMI of 30 or higher and psoriasis.

The self-reported nature of the survey questions and lack of questions regarding psoriasis severity serve as limitations to the study. Both obesity and psoriasis can have various systemic consequences, such as cardiovascular disease, due to the development of an inflammatory state.⁸ Future studies may explore other body measurements that indicate being overweight or obese and the potential synergistic relationship of obesity and psoriasis severity, optimizing the development of effective treatment plans.