Melanoma

Patients with stage III melanoma who have circulating tumor cells (CTCs) at baseline may benefit from adjuvant therapy, according to investigators.

A prospective study showed that patients with at least one CTC upon first presentation had increased risks of both short-term and long-term recurrence, reported lead author Anthony Lucci, MD, of the University of Texas MD Anderson Cancer Center, Houston, and colleagues.

While previous studies have suggested that CTCs hold prognostic value for melanoma patients, no trials had evaluated the CellSearch CTC Test – a standardized technique approved by the Food and Drug Administration – in patients with stage III disease, the investigators wrote. Their report is in Clinical Cancer Research.

In the present study, the investigators tested the CellSearch system in 243 patients with stage III cutaneous melanoma who were treated at MD Anderson Cancer Center. Patients with uveal or mucosal melanoma, or distant metastatic disease, were excluded.

Baseline blood samples were drawn within 3 months of regional lymph node metastasis, determined by either lymphadenectomy or sentinel lymph node biopsy. CTC assay positivity required that at least one CTC was detected within a single 7.5 mL tube of blood.

Out of 243 patients, 90 (37%) had a positive test. Of these 90 patients, almost one-quarter (23%) relapsed within 6 months, compared with 8% of patients who had a negative CTC assay. Within the full follow-up period, which was as long as 64 months, 48% of patients with CTCs at baseline relapsed, compared with 37% of patients without CTCs.

Multivariable regression analysis, which was adjusted for age, sex, pathological nodal stage, Breslow thickness, ulceration, and lymphovascular invasion, showed that baseline CTC positivity was an independent risk factor for melanoma recurrence, both in the short term and the long term. Compared with patients who lacked CTCs, those who tested positive were three times as likely to have disease recurrence within 6 months (hazard ratio, 3.13; P = .018). For relapse-free survival within 54 months, this hazard ratio decreased to 2.25 (P = .006).

Although a Cochran-Armitage test suggested that recurrence risks increased with CTC count, the investigators noted that a minority of patients (17%) had two or more CTCs, and just 5% had three or more CTCs.

According to the investigators, CTCs at baseline could become the first reliable blood-based biomarker for this patient population.

“[CTCs] clearly identified a group of stage III patients at high risk for relapse,” the investigators wrote. “This would be clinically very significant as an independent risk factor to help identify the stage III patients who would benefit most from adjuvant systemic therapy.”

This study was funded by the Kiefer family, Sheila Prenowitz, the Simon and Linda Eyles Foundation, the Sam and Janna Moore family, and the Wintermann Foundation. The investigators reported no conflicts of interest.

SOURCE: Lucci et al. Clin Cancer Res. doi: 10.1158/1078-0432.CCR-19-2670.

Patients with stage III melanoma who have circulating tumor cells (CTCs) at baseline may benefit from adjuvant therapy, according to investigators.

A prospective study showed that patients with at least one CTC upon first presentation had increased risks of both short-term and long-term recurrence, reported lead author Anthony Lucci, MD, of the University of Texas MD Anderson Cancer Center, Houston, and colleagues.

While previous studies have suggested that CTCs hold prognostic value for melanoma patients, no trials had evaluated the CellSearch CTC Test – a standardized technique approved by the Food and Drug Administration – in patients with stage III disease, the investigators wrote. Their report is in Clinical Cancer Research.

In the present study, the investigators tested the CellSearch system in 243 patients with stage III cutaneous melanoma who were treated at MD Anderson Cancer Center. Patients with uveal or mucosal melanoma, or distant metastatic disease, were excluded.

Baseline blood samples were drawn within 3 months of regional lymph node metastasis, determined by either lymphadenectomy or sentinel lymph node biopsy. CTC assay positivity required that at least one CTC was detected within a single 7.5 mL tube of blood.

Out of 243 patients, 90 (37%) had a positive test. Of these 90 patients, almost one-quarter (23%) relapsed within 6 months, compared with 8% of patients who had a negative CTC assay. Within the full follow-up period, which was as long as 64 months, 48% of patients with CTCs at baseline relapsed, compared with 37% of patients without CTCs.

Multivariable regression analysis, which was adjusted for age, sex, pathological nodal stage, Breslow thickness, ulceration, and lymphovascular invasion, showed that baseline CTC positivity was an independent risk factor for melanoma recurrence, both in the short term and the long term. Compared with patients who lacked CTCs, those who tested positive were three times as likely to have disease recurrence within 6 months (hazard ratio, 3.13; P = .018). For relapse-free survival within 54 months, this hazard ratio decreased to 2.25 (P = .006).

Although a Cochran-Armitage test suggested that recurrence risks increased with CTC count, the investigators noted that a minority of patients (17%) had two or more CTCs, and just 5% had three or more CTCs.

According to the investigators, CTCs at baseline could become the first reliable blood-based biomarker for this patient population.

“[CTCs] clearly identified a group of stage III patients at high risk for relapse,” the investigators wrote. “This would be clinically very significant as an independent risk factor to help identify the stage III patients who would benefit most from adjuvant systemic therapy.”

This study was funded by the Kiefer family, Sheila Prenowitz, the Simon and Linda Eyles Foundation, the Sam and Janna Moore family, and the Wintermann Foundation. The investigators reported no conflicts of interest.

SOURCE: Lucci et al. Clin Cancer Res. doi: 10.1158/1078-0432.CCR-19-2670.

Patients with stage III melanoma who have circulating tumor cells (CTCs) at baseline may benefit from adjuvant therapy, according to investigators.

A prospective study showed that patients with at least one CTC upon first presentation had increased risks of both short-term and long-term recurrence, reported lead author Anthony Lucci, MD, of the University of Texas MD Anderson Cancer Center, Houston, and colleagues.

While previous studies have suggested that CTCs hold prognostic value for melanoma patients, no trials had evaluated the CellSearch CTC Test – a standardized technique approved by the Food and Drug Administration – in patients with stage III disease, the investigators wrote. Their report is in Clinical Cancer Research.

In the present study, the investigators tested the CellSearch system in 243 patients with stage III cutaneous melanoma who were treated at MD Anderson Cancer Center. Patients with uveal or mucosal melanoma, or distant metastatic disease, were excluded.

Baseline blood samples were drawn within 3 months of regional lymph node metastasis, determined by either lymphadenectomy or sentinel lymph node biopsy. CTC assay positivity required that at least one CTC was detected within a single 7.5 mL tube of blood.

Out of 243 patients, 90 (37%) had a positive test. Of these 90 patients, almost one-quarter (23%) relapsed within 6 months, compared with 8% of patients who had a negative CTC assay. Within the full follow-up period, which was as long as 64 months, 48% of patients with CTCs at baseline relapsed, compared with 37% of patients without CTCs.

Multivariable regression analysis, which was adjusted for age, sex, pathological nodal stage, Breslow thickness, ulceration, and lymphovascular invasion, showed that baseline CTC positivity was an independent risk factor for melanoma recurrence, both in the short term and the long term. Compared with patients who lacked CTCs, those who tested positive were three times as likely to have disease recurrence within 6 months (hazard ratio, 3.13; P = .018). For relapse-free survival within 54 months, this hazard ratio decreased to 2.25 (P = .006).

Although a Cochran-Armitage test suggested that recurrence risks increased with CTC count, the investigators noted that a minority of patients (17%) had two or more CTCs, and just 5% had three or more CTCs.

According to the investigators, CTCs at baseline could become the first reliable blood-based biomarker for this patient population.

“[CTCs] clearly identified a group of stage III patients at high risk for relapse,” the investigators wrote. “This would be clinically very significant as an independent risk factor to help identify the stage III patients who would benefit most from adjuvant systemic therapy.”

This study was funded by the Kiefer family, Sheila Prenowitz, the Simon and Linda Eyles Foundation, the Sam and Janna Moore family, and the Wintermann Foundation. The investigators reported no conflicts of interest.

SOURCE: Lucci et al. Clin Cancer Res. doi: 10.1158/1078-0432.CCR-19-2670.

To the Editor:

Pembrolizumab is an anti–programmed death receptor 1 humanized monoclonal antibody used for treating advanced or metastatic melanoma.¹ It is associated with several immune-related adverse events because it blocks a T-cell receptor checkpoint.² The most common dermatologic immune-related adverse event seen with anti–programmed death receptor 1 medications is a nonspecific morbilliform rash, usually seen after the second treatment cycle; however, pruritus, vitiligo, bullous disorders, and lichenoid reactions also have been reported.³ We report a case of pembrolizumab-induced, self-limited lobular panniculitis in a patient with metastatic melanoma.

A 37-year-old woman with malignant melanoma presented with tender, erythematous, subcutaneous nodules on the hips and legs of 2 weeks’ duration (Figure 1). Twelve years prior to the current presentation, she was diagnosed with metastases to the cecum, lung, and brain. A review of systems was otherwise negative. She had been receiving pembrolizumab infusions (2 mg/kg every 3 weeks) for the last 2.7 years as second-line therapy after previously undergoing chemotherapy, radiation, and resection. She was not taking oral contraceptives or other hormone-based medications and did not report any new medications.

Chest computed tomography was normal in our patient, and she reported no systemic symptoms. Additional laboratory studies to evaluate for sarcoidosis were not obtained. Furthermore, the lesions quickly resolved despite continued use of pembrolizumab. We report this case to highlight that pembrolizumab may induce an isolated, self-limited lobular panniculitis years after medication initiation.

To the Editor:

Pembrolizumab is an anti–programmed death receptor 1 humanized monoclonal antibody used for treating advanced or metastatic melanoma.¹ It is associated with several immune-related adverse events because it blocks a T-cell receptor checkpoint.² The most common dermatologic immune-related adverse event seen with anti–programmed death receptor 1 medications is a nonspecific morbilliform rash, usually seen after the second treatment cycle; however, pruritus, vitiligo, bullous disorders, and lichenoid reactions also have been reported.³ We report a case of pembrolizumab-induced, self-limited lobular panniculitis in a patient with metastatic melanoma.

A 37-year-old woman with malignant melanoma presented with tender, erythematous, subcutaneous nodules on the hips and legs of 2 weeks’ duration (Figure 1). Twelve years prior to the current presentation, she was diagnosed with metastases to the cecum, lung, and brain. A review of systems was otherwise negative. She had been receiving pembrolizumab infusions (2 mg/kg every 3 weeks) for the last 2.7 years as second-line therapy after previously undergoing chemotherapy, radiation, and resection. She was not taking oral contraceptives or other hormone-based medications and did not report any new medications.

Chest computed tomography was normal in our patient, and she reported no systemic symptoms. Additional laboratory studies to evaluate for sarcoidosis were not obtained. Furthermore, the lesions quickly resolved despite continued use of pembrolizumab. We report this case to highlight that pembrolizumab may induce an isolated, self-limited lobular panniculitis years after medication initiation.

To the Editor:

Pembrolizumab is an anti–programmed death receptor 1 humanized monoclonal antibody used for treating advanced or metastatic melanoma.¹ It is associated with several immune-related adverse events because it blocks a T-cell receptor checkpoint.² The most common dermatologic immune-related adverse event seen with anti–programmed death receptor 1 medications is a nonspecific morbilliform rash, usually seen after the second treatment cycle; however, pruritus, vitiligo, bullous disorders, and lichenoid reactions also have been reported.³ We report a case of pembrolizumab-induced, self-limited lobular panniculitis in a patient with metastatic melanoma.

A 37-year-old woman with malignant melanoma presented with tender, erythematous, subcutaneous nodules on the hips and legs of 2 weeks’ duration (Figure 1). Twelve years prior to the current presentation, she was diagnosed with metastases to the cecum, lung, and brain. A review of systems was otherwise negative. She had been receiving pembrolizumab infusions (2 mg/kg every 3 weeks) for the last 2.7 years as second-line therapy after previously undergoing chemotherapy, radiation, and resection. She was not taking oral contraceptives or other hormone-based medications and did not report any new medications.

Chest computed tomography was normal in our patient, and she reported no systemic symptoms. Additional laboratory studies to evaluate for sarcoidosis were not obtained. Furthermore, the lesions quickly resolved despite continued use of pembrolizumab. We report this case to highlight that pembrolizumab may induce an isolated, self-limited lobular panniculitis years after medication initiation.

To the Editor:

Melanoma is the third most common skin cancer. It is estimated that 18% of melanoma patients will develop skin metastases, with skin being the first site of involvement in 56% of cases.¹ Of all cancers, it is estimated that 5% will develop skin metastases. It is the presenting sign in nearly 1% of visceral cancers.² Melanoma and nonmelanoma metastases can have sundry presentations. We present a case of metastatic melanoma with multiple keratoacanthoma (KA)–like skin lesions in a patient with a known history of nonmelanoma skin cancer (NMSC) as well as melanoma.

A 76-year-old man with a history of pT2aNXMX melanoma on the left upper back presented for a routine 3-month follow-up and reported several new asymptomatic bumps on the chest, back, and right upper extremity within the last 2 weeks. The melanoma was removed via wide local excision 2 years prior at an outside facility with a Breslow depth of 1.05 mm and a negative sentinel lymph node biopsy. The mitotic rate or ulceration status was unknown. He also had a history of several NMSCs, as well as a medical history of coronary artery disease, myocardial infarction, and ventricular tachycardia with cardiac defibrillator placement. Physical examination revealed 5 pink, volcano-shaped nodules with central keratotic plugs on the upper back (Figure 1), chest, and right upper extremity, in addition to 1 pink pearly nodule on the right side of the chest. The history and appearance of the lesions were suspicious for eruptive KAs. There was no evidence of cancer recurrence at the prior melanoma and NMSC sites.

Although melanoma is well known to metastasize years and even decades later, KA-like lesions have not been reported as manifestations of metastatic melanoma.^4,5 Our patient likely had a primary amelanotic melanoma, as the medical records from the outside facility stated that basal cell carcinoma was ruled out via biopsy. The amelanotic nature of the primary melanoma may have influenced the amelanotic appearance of the metastases. Our patient had no signs of immunosuppression that could have contributed to the sudden skin metastases.

Depending on the subtype of cutaneous metastases (eg, satellitosis, in-transit disease, distant cutaneous metastases), the location prevalence of the primary melanoma varies. In a study of 4865 melanoma patients who were diagnosed and followed prospectively over a 30-year period, skin metastases were mostly locoregional and presentation on the leg and foot were disproportionate.¹ In contrast, the trunk was overrepresented for distant metastases. Distant metastases also were more associated with concurrent metastases to the viscera.¹ Accordingly, a patient’s prognosis and management will differ depending on the subtype of cutaneous metastases.

Eruptive or multiple KAs classically have been associated with the Grzybowski variant, the Ferguson-Smith familial variant, and Muir-Torre syndrome. It was reported as a paraneoplastic syndrome associated with colon cancer, ovarian cancer, and once with myelodysplastic syndrome.³ Keratoacanthomas are being classified as well-differentiated squamous cell carcinomas and have metastatic potential. A biopsy is recommended to diagnose KAs as opposed to historically being monitored for resolution. A skin biopsy is the standard of care in management of KAs.

In addition to being associated with Muir-Torre syndrome and classified as a paraneoplastic syndrome,³ eruptive KAs can occur following skin resurfacing for actinic damage, fractional photothermolysis, cryotherapy, Jessner peels, and trichloroacetic acid peels.⁶ A couple other uncommon settings include a case report of an arc welder with job-associated radiation and multiple reports of tattoo-induced KAs.^7,8 There is the new increasingly common association of squamous cell carcinomas with BRAF inhibitors, such as vemurafenib, for metastatic melanoma.⁹

In a 2012 review article on cutaneous metastases, Riahi and Cohen¹⁰ found 8 cases of cutaneous metastases presenting as KA-like lesions; none were metastatic melanoma. All were solitary lesions, not multiple lesions, as in our patient. The sources were lung (3 cases), breast, esophagus, chondrosarcoma, bronchial, and mesothelioma. The most common location was the upper lip. Additionally, similar to our patient, they behaved clinically as KAs with rapid growth and keratotic plugs and were asymptomatic.¹⁰

Metastatic melanoma may mimic many other cutaneous processes that may make the diagnosis more difficult. Our case indicates that cutaneous metastases may mimic KAs. Although multiple KA-like lesions can spontaneously occur, a paraneoplastic syndrome and other underlying etiologies should be considered.

To the Editor:

Melanoma is the third most common skin cancer. It is estimated that 18% of melanoma patients will develop skin metastases, with skin being the first site of involvement in 56% of cases.¹ Of all cancers, it is estimated that 5% will develop skin metastases. It is the presenting sign in nearly 1% of visceral cancers.² Melanoma and nonmelanoma metastases can have sundry presentations. We present a case of metastatic melanoma with multiple keratoacanthoma (KA)–like skin lesions in a patient with a known history of nonmelanoma skin cancer (NMSC) as well as melanoma.

A 76-year-old man with a history of pT2aNXMX melanoma on the left upper back presented for a routine 3-month follow-up and reported several new asymptomatic bumps on the chest, back, and right upper extremity within the last 2 weeks. The melanoma was removed via wide local excision 2 years prior at an outside facility with a Breslow depth of 1.05 mm and a negative sentinel lymph node biopsy. The mitotic rate or ulceration status was unknown. He also had a history of several NMSCs, as well as a medical history of coronary artery disease, myocardial infarction, and ventricular tachycardia with cardiac defibrillator placement. Physical examination revealed 5 pink, volcano-shaped nodules with central keratotic plugs on the upper back (Figure 1), chest, and right upper extremity, in addition to 1 pink pearly nodule on the right side of the chest. The history and appearance of the lesions were suspicious for eruptive KAs. There was no evidence of cancer recurrence at the prior melanoma and NMSC sites.

Although melanoma is well known to metastasize years and even decades later, KA-like lesions have not been reported as manifestations of metastatic melanoma.^4,5 Our patient likely had a primary amelanotic melanoma, as the medical records from the outside facility stated that basal cell carcinoma was ruled out via biopsy. The amelanotic nature of the primary melanoma may have influenced the amelanotic appearance of the metastases. Our patient had no signs of immunosuppression that could have contributed to the sudden skin metastases.

Depending on the subtype of cutaneous metastases (eg, satellitosis, in-transit disease, distant cutaneous metastases), the location prevalence of the primary melanoma varies. In a study of 4865 melanoma patients who were diagnosed and followed prospectively over a 30-year period, skin metastases were mostly locoregional and presentation on the leg and foot were disproportionate.¹ In contrast, the trunk was overrepresented for distant metastases. Distant metastases also were more associated with concurrent metastases to the viscera.¹ Accordingly, a patient’s prognosis and management will differ depending on the subtype of cutaneous metastases.

Eruptive or multiple KAs classically have been associated with the Grzybowski variant, the Ferguson-Smith familial variant, and Muir-Torre syndrome. It was reported as a paraneoplastic syndrome associated with colon cancer, ovarian cancer, and once with myelodysplastic syndrome.³ Keratoacanthomas are being classified as well-differentiated squamous cell carcinomas and have metastatic potential. A biopsy is recommended to diagnose KAs as opposed to historically being monitored for resolution. A skin biopsy is the standard of care in management of KAs.

In addition to being associated with Muir-Torre syndrome and classified as a paraneoplastic syndrome,³ eruptive KAs can occur following skin resurfacing for actinic damage, fractional photothermolysis, cryotherapy, Jessner peels, and trichloroacetic acid peels.⁶ A couple other uncommon settings include a case report of an arc welder with job-associated radiation and multiple reports of tattoo-induced KAs.^7,8 There is the new increasingly common association of squamous cell carcinomas with BRAF inhibitors, such as vemurafenib, for metastatic melanoma.⁹

In a 2012 review article on cutaneous metastases, Riahi and Cohen¹⁰ found 8 cases of cutaneous metastases presenting as KA-like lesions; none were metastatic melanoma. All were solitary lesions, not multiple lesions, as in our patient. The sources were lung (3 cases), breast, esophagus, chondrosarcoma, bronchial, and mesothelioma. The most common location was the upper lip. Additionally, similar to our patient, they behaved clinically as KAs with rapid growth and keratotic plugs and were asymptomatic.¹⁰

Metastatic melanoma may mimic many other cutaneous processes that may make the diagnosis more difficult. Our case indicates that cutaneous metastases may mimic KAs. Although multiple KA-like lesions can spontaneously occur, a paraneoplastic syndrome and other underlying etiologies should be considered.

To the Editor:

Melanoma is the third most common skin cancer. It is estimated that 18% of melanoma patients will develop skin metastases, with skin being the first site of involvement in 56% of cases.¹ Of all cancers, it is estimated that 5% will develop skin metastases. It is the presenting sign in nearly 1% of visceral cancers.² Melanoma and nonmelanoma metastases can have sundry presentations. We present a case of metastatic melanoma with multiple keratoacanthoma (KA)–like skin lesions in a patient with a known history of nonmelanoma skin cancer (NMSC) as well as melanoma.

A 76-year-old man with a history of pT2aNXMX melanoma on the left upper back presented for a routine 3-month follow-up and reported several new asymptomatic bumps on the chest, back, and right upper extremity within the last 2 weeks. The melanoma was removed via wide local excision 2 years prior at an outside facility with a Breslow depth of 1.05 mm and a negative sentinel lymph node biopsy. The mitotic rate or ulceration status was unknown. He also had a history of several NMSCs, as well as a medical history of coronary artery disease, myocardial infarction, and ventricular tachycardia with cardiac defibrillator placement. Physical examination revealed 5 pink, volcano-shaped nodules with central keratotic plugs on the upper back (Figure 1), chest, and right upper extremity, in addition to 1 pink pearly nodule on the right side of the chest. The history and appearance of the lesions were suspicious for eruptive KAs. There was no evidence of cancer recurrence at the prior melanoma and NMSC sites.

Although melanoma is well known to metastasize years and even decades later, KA-like lesions have not been reported as manifestations of metastatic melanoma.^4,5 Our patient likely had a primary amelanotic melanoma, as the medical records from the outside facility stated that basal cell carcinoma was ruled out via biopsy. The amelanotic nature of the primary melanoma may have influenced the amelanotic appearance of the metastases. Our patient had no signs of immunosuppression that could have contributed to the sudden skin metastases.

Depending on the subtype of cutaneous metastases (eg, satellitosis, in-transit disease, distant cutaneous metastases), the location prevalence of the primary melanoma varies. In a study of 4865 melanoma patients who were diagnosed and followed prospectively over a 30-year period, skin metastases were mostly locoregional and presentation on the leg and foot were disproportionate.¹ In contrast, the trunk was overrepresented for distant metastases. Distant metastases also were more associated with concurrent metastases to the viscera.¹ Accordingly, a patient’s prognosis and management will differ depending on the subtype of cutaneous metastases.

Eruptive or multiple KAs classically have been associated with the Grzybowski variant, the Ferguson-Smith familial variant, and Muir-Torre syndrome. It was reported as a paraneoplastic syndrome associated with colon cancer, ovarian cancer, and once with myelodysplastic syndrome.³ Keratoacanthomas are being classified as well-differentiated squamous cell carcinomas and have metastatic potential. A biopsy is recommended to diagnose KAs as opposed to historically being monitored for resolution. A skin biopsy is the standard of care in management of KAs.

In addition to being associated with Muir-Torre syndrome and classified as a paraneoplastic syndrome,³ eruptive KAs can occur following skin resurfacing for actinic damage, fractional photothermolysis, cryotherapy, Jessner peels, and trichloroacetic acid peels.⁶ A couple other uncommon settings include a case report of an arc welder with job-associated radiation and multiple reports of tattoo-induced KAs.^7,8 There is the new increasingly common association of squamous cell carcinomas with BRAF inhibitors, such as vemurafenib, for metastatic melanoma.⁹

In a 2012 review article on cutaneous metastases, Riahi and Cohen¹⁰ found 8 cases of cutaneous metastases presenting as KA-like lesions; none were metastatic melanoma. All were solitary lesions, not multiple lesions, as in our patient. The sources were lung (3 cases), breast, esophagus, chondrosarcoma, bronchial, and mesothelioma. The most common location was the upper lip. Additionally, similar to our patient, they behaved clinically as KAs with rapid growth and keratotic plugs and were asymptomatic.¹⁰

Metastatic melanoma may mimic many other cutaneous processes that may make the diagnosis more difficult. Our case indicates that cutaneous metastases may mimic KAs. Although multiple KA-like lesions can spontaneously occur, a paraneoplastic syndrome and other underlying etiologies should be considered.

A combination of the tyrosine kinase inhibitor lenvatinib (Lenvima) and the immune checkpoint inhibitor pembrolizumab (Keytruda) was safe and showed promising activity against advanced renal cell carcinoma and other solid tumors in a phase 1b/2 study.

Overall response rates (ORR) at 24 weeks ranged from 63% for patients with advanced renal cell carcinomas (RCC) to 25% for patients with urothelial cancers, reported Matthew H. Taylor, MD, of Knight Cancer Institute at Oregon Health & Science University in Portland, and colleagues.

The findings from this study sparked additional clinical trials for patients with gastric cancer, gastroesophageal cancer, and differentiated thyroid cancer, and set the stage for larger phase 3 trials in patients with advanced RCC, endometrial cancer, malignant melanoma, and non–small cell lung cancer (NSCLC).

“In the future, we also plan to study lenvatinib plus pembrolizumab in patients with RCC who have had disease progression after treatment with immune checkpoint inhibitors,” they wrote. The report was published in Journal of Clinical Oncology.

Lenvatinib is a multitargeted tyrosine kinase inhibitor (TKI) with action against vascular endothelial growth factor (VEGF) receptors 1-3, fibroblast growth factor (FGF) receptors 1-4, platelet-derived growth factor receptors alpha and the RET and KIT kinases.

“Preclinical and clinical studies suggest that modulation of VEGF-mediated immune suppression via angiogenesis inhibition could potentially augment the immunotherapeutic activity of immune checkpoint inhibitors,” the investigators wrote.

They reported results from the dose finding (1b) phase including 13 patients and initial phase 2 expansion cohorts with a total of 124 patients.

The maximum tolerated dose of lenvatinib in combination with pembrolizumab was established as 20 mg/day.

At 24 weeks of follow-up, the ORR for 30 patients with RCC was 63%; two additional patients had responses after week 24, for a total ORR at study cutoff in this cohort of 70%. The median duration of response for these patients was 20 months, and the median progression-free survival (PFS) was 19.8 months. At the time of data cutoff for this analysis, 9 of the 30 patients with RCC were still on treatment.

For 23 patients with endometrial cancer, the 24-week and overall ORR were 52%, with a median duration of response not reached, and a median PFS of 9.7 months. Seven patients were still on treatment at data cutoff.

For 21 patients with melanoma, the 24-week and overall ORR were 48%, median duration of response was 12.5 months, and median PFS was 5.5 months. Two of the patients were still on treatment at data cutoff.

For the 22 patients with squamous cell cancer of the head and neck, the 24-week ORR was 36%, with two patients having a response after week 24 for a total ORR at data cutoff of 46%. The median duration of response was 8.2 months and the median PFS was 4.7 months. Three patients remained on treatment at data cutoff.

For 21 patients with NSCLC, the 24-week and overall ORR were 33%, the median duration of response was 10.9 months, and median PFS was 5.9 months. Six of the patients were still receiving treatment at data cutoff.

For 20 patients with urothelial cancer, the 24-week and overall ORR were 25%, with a median duration of response not reached, and a median PFS of 5.4 months. Three patients were still receiving the combination at the time of data cutoff.

Treatment related adverse events (TRAEs) occurred in 133 of all 137 patients enrolled in the two study phases. The adverse events were similar across all cohorts, with any grade of events including fatigue in 58%, diarrhea in 52%, hypertension in 47%, hypothyroidism in 42%, and decreased appetite in 39%.

The most frequent grade 3 or 4 TRAEs were hypertension in 20%, fatigue in 12%, diarrhea in 9%, proteinuria in 8%, and increased lipase levels in 7%.

In all, 85% of patients had a TRAE leading to lenvatinib dose reduction and/or interruption, and 13% required lenvatinib discontinuation.

Events leading to pembrolizumab dose interruption occurred in 45% of patients, and pembrolizumab discontinuation in 15%.

The study was sponsored by Eisai with collaboration from Merck Sharp & Dohme. Dr. Taylor disclosed a consulting or advisory role for Bristol-Myers Squibb, Eisai, Array BioPharma, Loxo, Bayer, ArQule, Blueprint Medicines, Novartis, and Sanofi/Genzyme, and speakers bureau activities for BMS and Eisai.

SOURCE: Taylor MH et al. J Clin Oncol. 2020 Jan. 21 doi: 10.1200/JCO.19.01598.

A combination of the tyrosine kinase inhibitor lenvatinib (Lenvima) and the immune checkpoint inhibitor pembrolizumab (Keytruda) was safe and showed promising activity against advanced renal cell carcinoma and other solid tumors in a phase 1b/2 study.

Overall response rates (ORR) at 24 weeks ranged from 63% for patients with advanced renal cell carcinomas (RCC) to 25% for patients with urothelial cancers, reported Matthew H. Taylor, MD, of Knight Cancer Institute at Oregon Health & Science University in Portland, and colleagues.

The findings from this study sparked additional clinical trials for patients with gastric cancer, gastroesophageal cancer, and differentiated thyroid cancer, and set the stage for larger phase 3 trials in patients with advanced RCC, endometrial cancer, malignant melanoma, and non–small cell lung cancer (NSCLC).

“In the future, we also plan to study lenvatinib plus pembrolizumab in patients with RCC who have had disease progression after treatment with immune checkpoint inhibitors,” they wrote. The report was published in Journal of Clinical Oncology.

Lenvatinib is a multitargeted tyrosine kinase inhibitor (TKI) with action against vascular endothelial growth factor (VEGF) receptors 1-3, fibroblast growth factor (FGF) receptors 1-4, platelet-derived growth factor receptors alpha and the RET and KIT kinases.

“Preclinical and clinical studies suggest that modulation of VEGF-mediated immune suppression via angiogenesis inhibition could potentially augment the immunotherapeutic activity of immune checkpoint inhibitors,” the investigators wrote.

They reported results from the dose finding (1b) phase including 13 patients and initial phase 2 expansion cohorts with a total of 124 patients.

The maximum tolerated dose of lenvatinib in combination with pembrolizumab was established as 20 mg/day.

At 24 weeks of follow-up, the ORR for 30 patients with RCC was 63%; two additional patients had responses after week 24, for a total ORR at study cutoff in this cohort of 70%. The median duration of response for these patients was 20 months, and the median progression-free survival (PFS) was 19.8 months. At the time of data cutoff for this analysis, 9 of the 30 patients with RCC were still on treatment.

For 23 patients with endometrial cancer, the 24-week and overall ORR were 52%, with a median duration of response not reached, and a median PFS of 9.7 months. Seven patients were still on treatment at data cutoff.

For 21 patients with melanoma, the 24-week and overall ORR were 48%, median duration of response was 12.5 months, and median PFS was 5.5 months. Two of the patients were still on treatment at data cutoff.

For the 22 patients with squamous cell cancer of the head and neck, the 24-week ORR was 36%, with two patients having a response after week 24 for a total ORR at data cutoff of 46%. The median duration of response was 8.2 months and the median PFS was 4.7 months. Three patients remained on treatment at data cutoff.

For 21 patients with NSCLC, the 24-week and overall ORR were 33%, the median duration of response was 10.9 months, and median PFS was 5.9 months. Six of the patients were still receiving treatment at data cutoff.

For 20 patients with urothelial cancer, the 24-week and overall ORR were 25%, with a median duration of response not reached, and a median PFS of 5.4 months. Three patients were still receiving the combination at the time of data cutoff.

Treatment related adverse events (TRAEs) occurred in 133 of all 137 patients enrolled in the two study phases. The adverse events were similar across all cohorts, with any grade of events including fatigue in 58%, diarrhea in 52%, hypertension in 47%, hypothyroidism in 42%, and decreased appetite in 39%.

The most frequent grade 3 or 4 TRAEs were hypertension in 20%, fatigue in 12%, diarrhea in 9%, proteinuria in 8%, and increased lipase levels in 7%.

In all, 85% of patients had a TRAE leading to lenvatinib dose reduction and/or interruption, and 13% required lenvatinib discontinuation.

Events leading to pembrolizumab dose interruption occurred in 45% of patients, and pembrolizumab discontinuation in 15%.

The study was sponsored by Eisai with collaboration from Merck Sharp & Dohme. Dr. Taylor disclosed a consulting or advisory role for Bristol-Myers Squibb, Eisai, Array BioPharma, Loxo, Bayer, ArQule, Blueprint Medicines, Novartis, and Sanofi/Genzyme, and speakers bureau activities for BMS and Eisai.

SOURCE: Taylor MH et al. J Clin Oncol. 2020 Jan. 21 doi: 10.1200/JCO.19.01598.

A combination of the tyrosine kinase inhibitor lenvatinib (Lenvima) and the immune checkpoint inhibitor pembrolizumab (Keytruda) was safe and showed promising activity against advanced renal cell carcinoma and other solid tumors in a phase 1b/2 study.

Overall response rates (ORR) at 24 weeks ranged from 63% for patients with advanced renal cell carcinomas (RCC) to 25% for patients with urothelial cancers, reported Matthew H. Taylor, MD, of Knight Cancer Institute at Oregon Health & Science University in Portland, and colleagues.

The findings from this study sparked additional clinical trials for patients with gastric cancer, gastroesophageal cancer, and differentiated thyroid cancer, and set the stage for larger phase 3 trials in patients with advanced RCC, endometrial cancer, malignant melanoma, and non–small cell lung cancer (NSCLC).

“In the future, we also plan to study lenvatinib plus pembrolizumab in patients with RCC who have had disease progression after treatment with immune checkpoint inhibitors,” they wrote. The report was published in Journal of Clinical Oncology.

Lenvatinib is a multitargeted tyrosine kinase inhibitor (TKI) with action against vascular endothelial growth factor (VEGF) receptors 1-3, fibroblast growth factor (FGF) receptors 1-4, platelet-derived growth factor receptors alpha and the RET and KIT kinases.

“Preclinical and clinical studies suggest that modulation of VEGF-mediated immune suppression via angiogenesis inhibition could potentially augment the immunotherapeutic activity of immune checkpoint inhibitors,” the investigators wrote.

They reported results from the dose finding (1b) phase including 13 patients and initial phase 2 expansion cohorts with a total of 124 patients.

The maximum tolerated dose of lenvatinib in combination with pembrolizumab was established as 20 mg/day.

At 24 weeks of follow-up, the ORR for 30 patients with RCC was 63%; two additional patients had responses after week 24, for a total ORR at study cutoff in this cohort of 70%. The median duration of response for these patients was 20 months, and the median progression-free survival (PFS) was 19.8 months. At the time of data cutoff for this analysis, 9 of the 30 patients with RCC were still on treatment.

For 23 patients with endometrial cancer, the 24-week and overall ORR were 52%, with a median duration of response not reached, and a median PFS of 9.7 months. Seven patients were still on treatment at data cutoff.

For 21 patients with melanoma, the 24-week and overall ORR were 48%, median duration of response was 12.5 months, and median PFS was 5.5 months. Two of the patients were still on treatment at data cutoff.

For the 22 patients with squamous cell cancer of the head and neck, the 24-week ORR was 36%, with two patients having a response after week 24 for a total ORR at data cutoff of 46%. The median duration of response was 8.2 months and the median PFS was 4.7 months. Three patients remained on treatment at data cutoff.

For 21 patients with NSCLC, the 24-week and overall ORR were 33%, the median duration of response was 10.9 months, and median PFS was 5.9 months. Six of the patients were still receiving treatment at data cutoff.

For 20 patients with urothelial cancer, the 24-week and overall ORR were 25%, with a median duration of response not reached, and a median PFS of 5.4 months. Three patients were still receiving the combination at the time of data cutoff.

Treatment related adverse events (TRAEs) occurred in 133 of all 137 patients enrolled in the two study phases. The adverse events were similar across all cohorts, with any grade of events including fatigue in 58%, diarrhea in 52%, hypertension in 47%, hypothyroidism in 42%, and decreased appetite in 39%.

The most frequent grade 3 or 4 TRAEs were hypertension in 20%, fatigue in 12%, diarrhea in 9%, proteinuria in 8%, and increased lipase levels in 7%.

In all, 85% of patients had a TRAE leading to lenvatinib dose reduction and/or interruption, and 13% required lenvatinib discontinuation.

Events leading to pembrolizumab dose interruption occurred in 45% of patients, and pembrolizumab discontinuation in 15%.

The study was sponsored by Eisai with collaboration from Merck Sharp & Dohme. Dr. Taylor disclosed a consulting or advisory role for Bristol-Myers Squibb, Eisai, Array BioPharma, Loxo, Bayer, ArQule, Blueprint Medicines, Novartis, and Sanofi/Genzyme, and speakers bureau activities for BMS and Eisai.

SOURCE: Taylor MH et al. J Clin Oncol. 2020 Jan. 21 doi: 10.1200/JCO.19.01598.

This infographic is available in the PDF above.

This infographic is available in the PDF above.

This infographic is available in the PDF above.

Declines in death rates for lung cancer and melanoma have gained momentum in recent years, fueling a record drop in cancer mortality, the American Cancer Society says.

Lung cancer death rates, which were falling by 3% in men and 2% in women annually in 2008 through 2013, dropped by 5% in men and nearly 4% per year in women annually from 2013 to 2017, according to the society’s 2020 statistical report.

Those accelerating reductions in death rates helped fuel the biggest-ever single year decline in overall cancer mortality, of 2.2%, from 2016 to 2017, their report shows.

According to the investigators, the decline in melanoma death rates escalated to 6.9% per year among 20- to 49-year-olds over 2013-2017, compared with a decline of just 2.9% per year during 2006-2010. Likewise, the melanoma death rate decline was 7.2% annually for the more recent time period, compared with just 1.3% annually in the earlier time period. The finding was even more remarkable for those 65 years of age and older, according to investigators, since the declines in melanoma death rates reached 6.2% annually, compared with a 0.9% annual increase in the years before immunotherapy.

Smoking cessation has been the main driver of progress in cutting lung cancer death rates, according to the report, while in melanoma, death rates have dropped after the introduction of immune checkpoint inhibitors and targeted therapies.

By contrast, reductions in death rates have slowed for colorectal cancers and female breast cancers, and have stabilized for prostate cancer, Ms. Siegel and coauthors stated, adding that racial and geographic disparities persist in preventable cancers, including those of the lung and cervix.

“Increased investment in both the equitable application of existing cancer control interventions and basic and clinical research to further advance treatment options would undoubtedly accelerate progress against cancer,” said the investigators. The report appears in CA: A Cancer Journal for Clinicians.

While the decline in lung cancer death rates is good news, the disease remains a major killer, responsible for more deaths than breast, colorectal, and ovarian cancer combined, said Jacques P. Fontaine, MD, a thoracic surgeon at Moffitt Cancer Center in Tampa, Fla.

“Five-year survival rates are still around the 18%-20% range, which is much lower than breast and prostate cancer,” Dr. Fontaine said in an interview. “Nonetheless, we’ve made a little dent in that, and we’re improving.”

Two other factors that have helped spur that improvement, according to Dr. Fontaine, are the reduced incidence of squamous cell carcinomas, which are linked to smoking, and the increased use of lung cancer screening with low-dose computed tomography.

Squamous cell carcinomas tend to be a central rather than peripheral, which makes the tumors harder to resect: “Surgery is sometimes not an option, and even to this day in 2020, the single most effective treatment for lung cancer remains surgical resection,” said Dr. Fontaine.

Likewise, centrally located tumors may preclude giving high-dose radiation and may result in more “collateral damage” to healthy tissue, he added.

Landmark studies show that low-dose CT scans reduce lung cancer deaths by 20% or more; however, screening can have false-positive results that lead to unnecessary biopsies and other harms, suggesting that the procedures should be done in centers of excellence that provide high-quality, responsible screening for early lung cancer, Dr. Fontaine said.

While the drop in melanoma death rates is encouraging and, not surprising in light of new cutting-edge therapies, an ongoing unmet treatment need still exists, according to Vishal Anil Patel, MD, director of cutaneous oncology at the George Washington Cancer Center in Washington.

“We still have a lot to learn, and a way to go, because we’ve really just made the first breakthrough,” Dr. Patel said in an interview.

SOURCE: Siegel RL et al. CA Cancer J Clin. 2020;70(1):7-30. doi: 10.3322/caac.21590.

Declines in death rates for lung cancer and melanoma have gained momentum in recent years, fueling a record drop in cancer mortality, the American Cancer Society says.

Lung cancer death rates, which were falling by 3% in men and 2% in women annually in 2008 through 2013, dropped by 5% in men and nearly 4% per year in women annually from 2013 to 2017, according to the society’s 2020 statistical report.

Those accelerating reductions in death rates helped fuel the biggest-ever single year decline in overall cancer mortality, of 2.2%, from 2016 to 2017, their report shows.

According to the investigators, the decline in melanoma death rates escalated to 6.9% per year among 20- to 49-year-olds over 2013-2017, compared with a decline of just 2.9% per year during 2006-2010. Likewise, the melanoma death rate decline was 7.2% annually for the more recent time period, compared with just 1.3% annually in the earlier time period. The finding was even more remarkable for those 65 years of age and older, according to investigators, since the declines in melanoma death rates reached 6.2% annually, compared with a 0.9% annual increase in the years before immunotherapy.

Smoking cessation has been the main driver of progress in cutting lung cancer death rates, according to the report, while in melanoma, death rates have dropped after the introduction of immune checkpoint inhibitors and targeted therapies.

By contrast, reductions in death rates have slowed for colorectal cancers and female breast cancers, and have stabilized for prostate cancer, Ms. Siegel and coauthors stated, adding that racial and geographic disparities persist in preventable cancers, including those of the lung and cervix.

“Increased investment in both the equitable application of existing cancer control interventions and basic and clinical research to further advance treatment options would undoubtedly accelerate progress against cancer,” said the investigators. The report appears in CA: A Cancer Journal for Clinicians.

While the decline in lung cancer death rates is good news, the disease remains a major killer, responsible for more deaths than breast, colorectal, and ovarian cancer combined, said Jacques P. Fontaine, MD, a thoracic surgeon at Moffitt Cancer Center in Tampa, Fla.

“Five-year survival rates are still around the 18%-20% range, which is much lower than breast and prostate cancer,” Dr. Fontaine said in an interview. “Nonetheless, we’ve made a little dent in that, and we’re improving.”

Two other factors that have helped spur that improvement, according to Dr. Fontaine, are the reduced incidence of squamous cell carcinomas, which are linked to smoking, and the increased use of lung cancer screening with low-dose computed tomography.

Squamous cell carcinomas tend to be a central rather than peripheral, which makes the tumors harder to resect: “Surgery is sometimes not an option, and even to this day in 2020, the single most effective treatment for lung cancer remains surgical resection,” said Dr. Fontaine.

Likewise, centrally located tumors may preclude giving high-dose radiation and may result in more “collateral damage” to healthy tissue, he added.

Landmark studies show that low-dose CT scans reduce lung cancer deaths by 20% or more; however, screening can have false-positive results that lead to unnecessary biopsies and other harms, suggesting that the procedures should be done in centers of excellence that provide high-quality, responsible screening for early lung cancer, Dr. Fontaine said.

While the drop in melanoma death rates is encouraging and, not surprising in light of new cutting-edge therapies, an ongoing unmet treatment need still exists, according to Vishal Anil Patel, MD, director of cutaneous oncology at the George Washington Cancer Center in Washington.

“We still have a lot to learn, and a way to go, because we’ve really just made the first breakthrough,” Dr. Patel said in an interview.

SOURCE: Siegel RL et al. CA Cancer J Clin. 2020;70(1):7-30. doi: 10.3322/caac.21590.

Declines in death rates for lung cancer and melanoma have gained momentum in recent years, fueling a record drop in cancer mortality, the American Cancer Society says.

Lung cancer death rates, which were falling by 3% in men and 2% in women annually in 2008 through 2013, dropped by 5% in men and nearly 4% per year in women annually from 2013 to 2017, according to the society’s 2020 statistical report.

Those accelerating reductions in death rates helped fuel the biggest-ever single year decline in overall cancer mortality, of 2.2%, from 2016 to 2017, their report shows.

According to the investigators, the decline in melanoma death rates escalated to 6.9% per year among 20- to 49-year-olds over 2013-2017, compared with a decline of just 2.9% per year during 2006-2010. Likewise, the melanoma death rate decline was 7.2% annually for the more recent time period, compared with just 1.3% annually in the earlier time period. The finding was even more remarkable for those 65 years of age and older, according to investigators, since the declines in melanoma death rates reached 6.2% annually, compared with a 0.9% annual increase in the years before immunotherapy.

Smoking cessation has been the main driver of progress in cutting lung cancer death rates, according to the report, while in melanoma, death rates have dropped after the introduction of immune checkpoint inhibitors and targeted therapies.

By contrast, reductions in death rates have slowed for colorectal cancers and female breast cancers, and have stabilized for prostate cancer, Ms. Siegel and coauthors stated, adding that racial and geographic disparities persist in preventable cancers, including those of the lung and cervix.

“Increased investment in both the equitable application of existing cancer control interventions and basic and clinical research to further advance treatment options would undoubtedly accelerate progress against cancer,” said the investigators. The report appears in CA: A Cancer Journal for Clinicians.

While the decline in lung cancer death rates is good news, the disease remains a major killer, responsible for more deaths than breast, colorectal, and ovarian cancer combined, said Jacques P. Fontaine, MD, a thoracic surgeon at Moffitt Cancer Center in Tampa, Fla.

“Five-year survival rates are still around the 18%-20% range, which is much lower than breast and prostate cancer,” Dr. Fontaine said in an interview. “Nonetheless, we’ve made a little dent in that, and we’re improving.”

Two other factors that have helped spur that improvement, according to Dr. Fontaine, are the reduced incidence of squamous cell carcinomas, which are linked to smoking, and the increased use of lung cancer screening with low-dose computed tomography.

Squamous cell carcinomas tend to be a central rather than peripheral, which makes the tumors harder to resect: “Surgery is sometimes not an option, and even to this day in 2020, the single most effective treatment for lung cancer remains surgical resection,” said Dr. Fontaine.

Likewise, centrally located tumors may preclude giving high-dose radiation and may result in more “collateral damage” to healthy tissue, he added.

Landmark studies show that low-dose CT scans reduce lung cancer deaths by 20% or more; however, screening can have false-positive results that lead to unnecessary biopsies and other harms, suggesting that the procedures should be done in centers of excellence that provide high-quality, responsible screening for early lung cancer, Dr. Fontaine said.

While the drop in melanoma death rates is encouraging and, not surprising in light of new cutting-edge therapies, an ongoing unmet treatment need still exists, according to Vishal Anil Patel, MD, director of cutaneous oncology at the George Washington Cancer Center in Washington.

“We still have a lot to learn, and a way to go, because we’ve really just made the first breakthrough,” Dr. Patel said in an interview.

SOURCE: Siegel RL et al. CA Cancer J Clin. 2020;70(1):7-30. doi: 10.3322/caac.21590.

A new tool specifically designed to evaluate immune checkpoint inhibitor–related toxicity, developed based on direct patient involvement, picks up on cutaneous and other side effects that would be missed using traditional quality of life questionnaires, investigators say.

The 25-item list represents the first-ever health-related quality of life (HRQOL) toxicity subscale developed for patients receiving checkpoint inhibitors, according to the investigators, led by Aaron R. Hansen, MBBS, of the division of medical oncology and hematology in the department of medicine at the University of Toronto.

The toxicity subscale is combined with the Functional Assessment of Cancer Therapy–General (FACT-G), which measures physical, emotional, family and social, and functional domains, to form the Functional Assessment of Cancer Therapy–Immune Checkpoint Modulator (FACT-ICM), Dr. Hansen stated in a recent report that describes initial development and early validation efforts.

The FACT-ICM could become an important tool for measuring HRQOL in patients receiving checkpoint inhibitors, depending on results of further investigations including more patients, the authors wrote in that report.

“Currently, we would recommend that our toxicity subscale be validated first before use in clinical care, or in trials with QOL as a primary or secondary endpoint,” wrote Dr. Hansen and colleagues in the report, which appears in Cancer.

The toxicity subscale asks patients to rate items such as “I am bothered by dry skin,” “I feel pain, soreness or aches in some of my muscles,” and “My fatigue keeps me from doing the things I want do” on a scale of 0 (not at all) to 4 (very much).

Development of the toxicity subscale was based on focus groups and interviews with 37 patients with a variety of cancer types who were being treated with a PD-1, PD-L1, and CTLA-4 immune checkpoint inhibitors. Sixteen physicians were surveyed to evaluate the patient input, while 11 of them also participated in follow-up interviews.

“At every step in this process, the patients were central,” the investigators wrote in their report.

According to the investigators, that approach is in line with guidance from the Food and Drug Administration, which has said that meaningful patient input should be used in the upfront development of patient-reported outcome (PRO) measures, rather than obtaining patient endorsement after the fact.

By contrast, an electronic PRO immune-oncology module recently developed, based on 19 immune-related adverse events from drug labels and clinical trial reports, had “no evaluation” of effects on HRQOL, according to Dr. Hansen and coauthors, who added that the tool “did not adhere” to the FDA call for meaningful patient input.

Some previous studies of quality of life in immune checkpoint inhibitor–treated patients have used tumor-specific PROs and the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire–Core 30 Items (EORTC-QLQ-C30).

The new, immune checkpoint inhibitor–specific toxicity subscale has “broader coverage” of side effects that reportedly affect HRQOL in patients treated with these agents, including taste disturbance, cough, and fever or chills, according to the investigators.

Moreover, the EORTC-QLQ-C30 and the EORTC head and neck cancer–specific-35 module (EORTC QLQ-H&N35), do not include items related to cutaneous adverse events such as itch, rash, and dry skin that have been seen in some checkpoint inhibitor clinical trials, they noted.

“This represents a clear limitation of such preexisting PRO instruments, which should be addressed with our immune checkpoint moduator–specific tool,” they wrote.

The study was supported by a grant from the University of Toronto. Authors of the study provided disclosures related to Merck, Bristol-Myers Squibb, Karyopharm, Boston Biomedical, Novartis, Genentech, Hoffmann La Roche, GlaxoSmithKline, and others.

SOURCE: Hansen AR et al. Cancer. 2020 Jan 8. doi: 10.1002/cncr.32692.

A new tool specifically designed to evaluate immune checkpoint inhibitor–related toxicity, developed based on direct patient involvement, picks up on cutaneous and other side effects that would be missed using traditional quality of life questionnaires, investigators say.

The 25-item list represents the first-ever health-related quality of life (HRQOL) toxicity subscale developed for patients receiving checkpoint inhibitors, according to the investigators, led by Aaron R. Hansen, MBBS, of the division of medical oncology and hematology in the department of medicine at the University of Toronto.

The toxicity subscale is combined with the Functional Assessment of Cancer Therapy–General (FACT-G), which measures physical, emotional, family and social, and functional domains, to form the Functional Assessment of Cancer Therapy–Immune Checkpoint Modulator (FACT-ICM), Dr. Hansen stated in a recent report that describes initial development and early validation efforts.

The FACT-ICM could become an important tool for measuring HRQOL in patients receiving checkpoint inhibitors, depending on results of further investigations including more patients, the authors wrote in that report.

“Currently, we would recommend that our toxicity subscale be validated first before use in clinical care, or in trials with QOL as a primary or secondary endpoint,” wrote Dr. Hansen and colleagues in the report, which appears in Cancer.

The toxicity subscale asks patients to rate items such as “I am bothered by dry skin,” “I feel pain, soreness or aches in some of my muscles,” and “My fatigue keeps me from doing the things I want do” on a scale of 0 (not at all) to 4 (very much).

Development of the toxicity subscale was based on focus groups and interviews with 37 patients with a variety of cancer types who were being treated with a PD-1, PD-L1, and CTLA-4 immune checkpoint inhibitors. Sixteen physicians were surveyed to evaluate the patient input, while 11 of them also participated in follow-up interviews.

“At every step in this process, the patients were central,” the investigators wrote in their report.

According to the investigators, that approach is in line with guidance from the Food and Drug Administration, which has said that meaningful patient input should be used in the upfront development of patient-reported outcome (PRO) measures, rather than obtaining patient endorsement after the fact.

By contrast, an electronic PRO immune-oncology module recently developed, based on 19 immune-related adverse events from drug labels and clinical trial reports, had “no evaluation” of effects on HRQOL, according to Dr. Hansen and coauthors, who added that the tool “did not adhere” to the FDA call for meaningful patient input.

Some previous studies of quality of life in immune checkpoint inhibitor–treated patients have used tumor-specific PROs and the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire–Core 30 Items (EORTC-QLQ-C30).

The new, immune checkpoint inhibitor–specific toxicity subscale has “broader coverage” of side effects that reportedly affect HRQOL in patients treated with these agents, including taste disturbance, cough, and fever or chills, according to the investigators.

Moreover, the EORTC-QLQ-C30 and the EORTC head and neck cancer–specific-35 module (EORTC QLQ-H&N35), do not include items related to cutaneous adverse events such as itch, rash, and dry skin that have been seen in some checkpoint inhibitor clinical trials, they noted.

“This represents a clear limitation of such preexisting PRO instruments, which should be addressed with our immune checkpoint moduator–specific tool,” they wrote.

The study was supported by a grant from the University of Toronto. Authors of the study provided disclosures related to Merck, Bristol-Myers Squibb, Karyopharm, Boston Biomedical, Novartis, Genentech, Hoffmann La Roche, GlaxoSmithKline, and others.

SOURCE: Hansen AR et al. Cancer. 2020 Jan 8. doi: 10.1002/cncr.32692.

A new tool specifically designed to evaluate immune checkpoint inhibitor–related toxicity, developed based on direct patient involvement, picks up on cutaneous and other side effects that would be missed using traditional quality of life questionnaires, investigators say.

The 25-item list represents the first-ever health-related quality of life (HRQOL) toxicity subscale developed for patients receiving checkpoint inhibitors, according to the investigators, led by Aaron R. Hansen, MBBS, of the division of medical oncology and hematology in the department of medicine at the University of Toronto.

The toxicity subscale is combined with the Functional Assessment of Cancer Therapy–General (FACT-G), which measures physical, emotional, family and social, and functional domains, to form the Functional Assessment of Cancer Therapy–Immune Checkpoint Modulator (FACT-ICM), Dr. Hansen stated in a recent report that describes initial development and early validation efforts.

The FACT-ICM could become an important tool for measuring HRQOL in patients receiving checkpoint inhibitors, depending on results of further investigations including more patients, the authors wrote in that report.

“Currently, we would recommend that our toxicity subscale be validated first before use in clinical care, or in trials with QOL as a primary or secondary endpoint,” wrote Dr. Hansen and colleagues in the report, which appears in Cancer.

The toxicity subscale asks patients to rate items such as “I am bothered by dry skin,” “I feel pain, soreness or aches in some of my muscles,” and “My fatigue keeps me from doing the things I want do” on a scale of 0 (not at all) to 4 (very much).

Development of the toxicity subscale was based on focus groups and interviews with 37 patients with a variety of cancer types who were being treated with a PD-1, PD-L1, and CTLA-4 immune checkpoint inhibitors. Sixteen physicians were surveyed to evaluate the patient input, while 11 of them also participated in follow-up interviews.

“At every step in this process, the patients were central,” the investigators wrote in their report.

According to the investigators, that approach is in line with guidance from the Food and Drug Administration, which has said that meaningful patient input should be used in the upfront development of patient-reported outcome (PRO) measures, rather than obtaining patient endorsement after the fact.

By contrast, an electronic PRO immune-oncology module recently developed, based on 19 immune-related adverse events from drug labels and clinical trial reports, had “no evaluation” of effects on HRQOL, according to Dr. Hansen and coauthors, who added that the tool “did not adhere” to the FDA call for meaningful patient input.

Some previous studies of quality of life in immune checkpoint inhibitor–treated patients have used tumor-specific PROs and the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire–Core 30 Items (EORTC-QLQ-C30).

The new, immune checkpoint inhibitor–specific toxicity subscale has “broader coverage” of side effects that reportedly affect HRQOL in patients treated with these agents, including taste disturbance, cough, and fever or chills, according to the investigators.

Moreover, the EORTC-QLQ-C30 and the EORTC head and neck cancer–specific-35 module (EORTC QLQ-H&N35), do not include items related to cutaneous adverse events such as itch, rash, and dry skin that have been seen in some checkpoint inhibitor clinical trials, they noted.

“This represents a clear limitation of such preexisting PRO instruments, which should be addressed with our immune checkpoint moduator–specific tool,” they wrote.

The study was supported by a grant from the University of Toronto. Authors of the study provided disclosures related to Merck, Bristol-Myers Squibb, Karyopharm, Boston Biomedical, Novartis, Genentech, Hoffmann La Roche, GlaxoSmithKline, and others.

SOURCE: Hansen AR et al. Cancer. 2020 Jan 8. doi: 10.1002/cncr.32692.

Practice Gap

A 69-year-old man underwent staged excision for an invasive melanoma (0.4-mm Breslow depth; stage Ia) of the right dorsal nose. Two stages were required to achieve clear margins, leaving a 3.0×2.5-cm defect involving the nasal dorsum, right nasal sidewall, and nasal supratip (Figure 1). He declined any multistage repair and preferred a full-thickness skin graft (FTSG) over any interpolation flap.

Given the size of our patient’s defect, primary repair was not possible and second intention healing may have resulted in a suboptimal cosmetic outcome, potential alar distortion, and prolonged healing. No single local flap, such as the dorsal nasal rotation flap, crescentic advancement flap, bilobed flap, and Rintala flap, would have provided adequate coverage. A FTSG of the entire defect would not have been an ideal tissue match, and given the limited surrounding laxity, a Burow FTSG would have required the linear repair to extend well into the forehead with a questionable cosmetic outcome.

The Technique

We opted to repair the defect using a combination of local flaps for a single-stage repair. Using the right cheek reservoir, a crescentic advancement flap was performed to restore the right nasal sidewall as best as possible with a standing cone taken superiorly. To execute this flap, an incision was made extending from the alar sulcus into the nasolabial fold while preserving the apical triangle of the upper cutaneous lip. The flap was elevated submuscularly on the nose, and broad undermining was performed in the subcutaneous plane of the medial cheek. A crescentic redundancy above the alar sulcus was excised, and periosteal tacking sutures were placed to both help advance the flap and to recreate the nasofacial sulcus.¹

Next, a nasal tip spiral/rotation flap was designed to restore the remaining nasal defect.² An incision was made at the right inferiormost aspect of the defect and extended along the inferior border of the nasal tip as it crossed the midline to the left side of the nose. After incising and elevating the flap in the submuscular plane, there was not enough of a tissue reservoir to cover the entire remaining nasal defect.

To resolve this intraoperative conundrum, simple interrupted sutures were placed into the nasal cartilage at midline to narrow the structure of the nose (Figure 2). Three 4-0 polyglactin 910 sutures were placed beginning with the upper lateral cartilages and extending inferiorly to the lower lateral cartilages. Narrowing the nasal cartilages allowed for a smaller residual defect. The nasal tip rotation flap was then spiraled into place with adequate coverage. Some of the flap tip was trimmed after the superior aspect of the rotation flap was sutured to the inferior edge of the crescentic advancement flap. The immediate postoperative appearance is shown in Figure 3.

Practice Implications

Cartilage sutures highlight an underutilized technique in nasal reconstruction, with few cases reported in the dermatologic surgery literature.^3,4 The interdomal suture is placed through the left and right lower lateral cartilages to help narrow and redefine the nasal tip.⁵ Reported techniques include simple interrupted suture or horizontal mattress suture. Suture material for nasal cartilage may be permanent (nylon or polypropylene) or long-lasting (polydioxanone or polyglactin 910).⁵ The use of interdomal sutures has been reported to narrow and decrease the volume of nasal tip defects prior to repair with local flaps and FTSG.^3,4 In addition to the interdomal suture of the lower lateral nasal cartilage, simple interrupted sutures were placed in the upper lateral cartilages that created an even smaller residual defect. Sutures of the nasal cartilage may be a good option for select patients in dermatologic reconstruction, allowing for a simple repair with the added benefit of improved cosmetic result.

A combination of local flaps may be used to repair large nasal defects involving multiple subunits, especially in patients who decline multistage reconstruction. A nasal tip rotation/spiral flap can be considered for the appropriate nasal tip defect. Suturing the nasal cartilage with either permanent or long-lasting suture can narrow the cartilage and facilitate flap coverage for nasal defects while also improving the appearance of patients with wide prominent lower noses.

Practice Gap

A 69-year-old man underwent staged excision for an invasive melanoma (0.4-mm Breslow depth; stage Ia) of the right dorsal nose. Two stages were required to achieve clear margins, leaving a 3.0×2.5-cm defect involving the nasal dorsum, right nasal sidewall, and nasal supratip (Figure 1). He declined any multistage repair and preferred a full-thickness skin graft (FTSG) over any interpolation flap.

Given the size of our patient’s defect, primary repair was not possible and second intention healing may have resulted in a suboptimal cosmetic outcome, potential alar distortion, and prolonged healing. No single local flap, such as the dorsal nasal rotation flap, crescentic advancement flap, bilobed flap, and Rintala flap, would have provided adequate coverage. A FTSG of the entire defect would not have been an ideal tissue match, and given the limited surrounding laxity, a Burow FTSG would have required the linear repair to extend well into the forehead with a questionable cosmetic outcome.

The Technique

We opted to repair the defect using a combination of local flaps for a single-stage repair. Using the right cheek reservoir, a crescentic advancement flap was performed to restore the right nasal sidewall as best as possible with a standing cone taken superiorly. To execute this flap, an incision was made extending from the alar sulcus into the nasolabial fold while preserving the apical triangle of the upper cutaneous lip. The flap was elevated submuscularly on the nose, and broad undermining was performed in the subcutaneous plane of the medial cheek. A crescentic redundancy above the alar sulcus was excised, and periosteal tacking sutures were placed to both help advance the flap and to recreate the nasofacial sulcus.¹

Next, a nasal tip spiral/rotation flap was designed to restore the remaining nasal defect.² An incision was made at the right inferiormost aspect of the defect and extended along the inferior border of the nasal tip as it crossed the midline to the left side of the nose. After incising and elevating the flap in the submuscular plane, there was not enough of a tissue reservoir to cover the entire remaining nasal defect.

To resolve this intraoperative conundrum, simple interrupted sutures were placed into the nasal cartilage at midline to narrow the structure of the nose (Figure 2). Three 4-0 polyglactin 910 sutures were placed beginning with the upper lateral cartilages and extending inferiorly to the lower lateral cartilages. Narrowing the nasal cartilages allowed for a smaller residual defect. The nasal tip rotation flap was then spiraled into place with adequate coverage. Some of the flap tip was trimmed after the superior aspect of the rotation flap was sutured to the inferior edge of the crescentic advancement flap. The immediate postoperative appearance is shown in Figure 3.

Practice Implications

Cartilage sutures highlight an underutilized technique in nasal reconstruction, with few cases reported in the dermatologic surgery literature.^3,4 The interdomal suture is placed through the left and right lower lateral cartilages to help narrow and redefine the nasal tip.⁵ Reported techniques include simple interrupted suture or horizontal mattress suture. Suture material for nasal cartilage may be permanent (nylon or polypropylene) or long-lasting (polydioxanone or polyglactin 910).⁵ The use of interdomal sutures has been reported to narrow and decrease the volume of nasal tip defects prior to repair with local flaps and FTSG.^3,4 In addition to the interdomal suture of the lower lateral nasal cartilage, simple interrupted sutures were placed in the upper lateral cartilages that created an even smaller residual defect. Sutures of the nasal cartilage may be a good option for select patients in dermatologic reconstruction, allowing for a simple repair with the added benefit of improved cosmetic result.

A combination of local flaps may be used to repair large nasal defects involving multiple subunits, especially in patients who decline multistage reconstruction. A nasal tip rotation/spiral flap can be considered for the appropriate nasal tip defect. Suturing the nasal cartilage with either permanent or long-lasting suture can narrow the cartilage and facilitate flap coverage for nasal defects while also improving the appearance of patients with wide prominent lower noses.

Practice Gap

A 69-year-old man underwent staged excision for an invasive melanoma (0.4-mm Breslow depth; stage Ia) of the right dorsal nose. Two stages were required to achieve clear margins, leaving a 3.0×2.5-cm defect involving the nasal dorsum, right nasal sidewall, and nasal supratip (Figure 1). He declined any multistage repair and preferred a full-thickness skin graft (FTSG) over any interpolation flap.

Given the size of our patient’s defect, primary repair was not possible and second intention healing may have resulted in a suboptimal cosmetic outcome, potential alar distortion, and prolonged healing. No single local flap, such as the dorsal nasal rotation flap, crescentic advancement flap, bilobed flap, and Rintala flap, would have provided adequate coverage. A FTSG of the entire defect would not have been an ideal tissue match, and given the limited surrounding laxity, a Burow FTSG would have required the linear repair to extend well into the forehead with a questionable cosmetic outcome.

The Technique

We opted to repair the defect using a combination of local flaps for a single-stage repair. Using the right cheek reservoir, a crescentic advancement flap was performed to restore the right nasal sidewall as best as possible with a standing cone taken superiorly. To execute this flap, an incision was made extending from the alar sulcus into the nasolabial fold while preserving the apical triangle of the upper cutaneous lip. The flap was elevated submuscularly on the nose, and broad undermining was performed in the subcutaneous plane of the medial cheek. A crescentic redundancy above the alar sulcus was excised, and periosteal tacking sutures were placed to both help advance the flap and to recreate the nasofacial sulcus.¹

Next, a nasal tip spiral/rotation flap was designed to restore the remaining nasal defect.² An incision was made at the right inferiormost aspect of the defect and extended along the inferior border of the nasal tip as it crossed the midline to the left side of the nose. After incising and elevating the flap in the submuscular plane, there was not enough of a tissue reservoir to cover the entire remaining nasal defect.

To resolve this intraoperative conundrum, simple interrupted sutures were placed into the nasal cartilage at midline to narrow the structure of the nose (Figure 2). Three 4-0 polyglactin 910 sutures were placed beginning with the upper lateral cartilages and extending inferiorly to the lower lateral cartilages. Narrowing the nasal cartilages allowed for a smaller residual defect. The nasal tip rotation flap was then spiraled into place with adequate coverage. Some of the flap tip was trimmed after the superior aspect of the rotation flap was sutured to the inferior edge of the crescentic advancement flap. The immediate postoperative appearance is shown in Figure 3.

Practice Implications

Cartilage sutures highlight an underutilized technique in nasal reconstruction, with few cases reported in the dermatologic surgery literature.^3,4 The interdomal suture is placed through the left and right lower lateral cartilages to help narrow and redefine the nasal tip.⁵ Reported techniques include simple interrupted suture or horizontal mattress suture. Suture material for nasal cartilage may be permanent (nylon or polypropylene) or long-lasting (polydioxanone or polyglactin 910).⁵ The use of interdomal sutures has been reported to narrow and decrease the volume of nasal tip defects prior to repair with local flaps and FTSG.^3,4 In addition to the interdomal suture of the lower lateral nasal cartilage, simple interrupted sutures were placed in the upper lateral cartilages that created an even smaller residual defect. Sutures of the nasal cartilage may be a good option for select patients in dermatologic reconstruction, allowing for a simple repair with the added benefit of improved cosmetic result.

A combination of local flaps may be used to repair large nasal defects involving multiple subunits, especially in patients who decline multistage reconstruction. A nasal tip rotation/spiral flap can be considered for the appropriate nasal tip defect. Suturing the nasal cartilage with either permanent or long-lasting suture can narrow the cartilage and facilitate flap coverage for nasal defects while also improving the appearance of patients with wide prominent lower noses.

Artificial intelligence (AI) is a loosely defined term that refers to machines (ie, algorithms) simulating facets of human intelligence. Some examples of AI are seen in natural language-processing algorithms, including autocorrect and search engine autocomplete functions; voice recognition in virtual assistants; autopilot systems in airplanes and self-driving cars; and computer vision in image and object recognition. Since the dawn of the century, various forms of AI have been tested and introduced in health care. However, a gap exists between clinician viewpoints on AI and the engineering world’s assumptions of what can be automated in medicine.

In this article, we review the history and evolution of AI in medicine, focusing on radiology and dermatology; current capabilities of AI; challenges to clinical integration; and future directions. Our aim is to provide realistic expectations of current technologies in solving complex problems and to empower dermatologists in planning for a future that likely includes various forms of AI.

Early Stages of AI in Medical Decision-making

Some of the earliest forms of clinical decision-support software in medicine were computer-aided detection and computer-aided diagnosis (CAD) used in screening for breast and lung cancer on mammography and computed tomography.^1-3 Early research on the use of CAD systems in radiology date to the 1960s (Figure), with the first US Food and Drug Administration–approved CAD system in mammography in 1998 and for Centers for Medicare & Medicaid Services reimbursement in 2002.^1,2

Early CAD systems relied on rule-based classifiers, which use predefined features to classify images into desired categories. For example, to classify an image as a high-risk or benign mass, features such as contour and texture had to be explicitly defined. Although these systems showed on par with, or higher, accuracy vs a radiologist in validation studies, early CAD systems never achieved wide adoption because of an increased rate of false positives as well as added work burden on a radiologist, who had to silence overcalling by the software.^1,2,4,5

Computer-aided diagnosis–based melanoma diagnosis was introduced in early 2000 in dermatology (Figure) using the same feature-based classifiers. These systems claimed expert-level accuracy in proof-of-concept studies and prospective uncontrolled trials on proprietary devices using these classifiers.^6,7 Similar to radiology, however, real-world adoption did not happen; in fact, the last of these devices was taken off the market in 2017. A recent meta-analysis of studies using CAD-based melanoma diagnosis point to study bias; data overfitting; and lack of large controlled, prospective trials as possible reasons why results could not be replicated in a clinical setting.⁸

Beyond 2010: Deep Learning

New techniques in machine learning (ML), called deep learning, began to emerge after 2010 (Figure). In deep learning, instead of directing the computer to look for certain discriminative features, the machine learns those features from the large amount of data without being explicitly programed to do so. In other words, compared to predecessor forms of computing, there is less human supervision in the learning process (Table). The concept of ML has existed since the 1980s. The field saw exponential growth in the last decade with the improvement of algorithms; an increase in computing power; and emergence of large training data sets, such as open-source platforms on the Web.^9,10

Most ML methods today incorporate artificial neural networks (ANN), computer programs that imitate the architecture of biological neural networks and form dynamically changing systems that improve with continuous data exposure. The performance of an ANN is dependent on the number and architecture of its neural layers and (similar to CAD systems) the size, quality, and generalizability of the training data set.^9-12

In medicine, images (eg, clinical or dermoscopic images and imaging scans) are the most commonly used form of data for AI development. Convolutional neural networks (CNN), a subtype of ANN, are frequently used for this purpose. These networks use a hierarchical neural network architecture, similar to the visual cortex, that allows for composition of complex features (eg, shapes) from simpler features (eg, image intensities), which leads to more efficient data processing.^10-12

In recent years, CNNs have been applied in a number of image-based medical fields, including radiology, dermatology, and pathology. Initially, studies were largely led by computer scientists trying to match clinician performance in detection of disease categories. However, there has been a shift toward more physicians getting involved, which has motivated development of large curated (ie, expert-labeled) and standardized clinical data sets in training the CNN. Although training on quality-controlled data is a work in progress across medical disciplines, it has led to improved machine performance.^11,12

Recent Advances in AI

In recent years, the number of studies covering CNN in diagnosis has increased exponentially in several medical specialties. The goal is to improve software to close the gap between experts and the machine in live clinical settings. The current literature focuses on a comparison of experts with the machine in simulated settings; prospective clinical trials are still lagging in the real world.^9,11,13

We look at radiology to explore recent advances in AI diagnosis for 3 reasons: (1) radiology has the largest repository of digital data (using a picture archiving and communication system) among medical specialties; (2) radiology has well-defined, image-acquisition protocols in its clinical workflow¹⁴; and (3) gray-scale images are easier to standardize because they are impervious to environmental variables that are difficult to control (eg, recent sun exposure, rosacea flare, lighting, sweating). These are some of the reasons we think radiology is, and will be, ahead in training AI algorithms and integrating them into clinical practice. However, even radiology AI studies have limitations, including a lack of prospective, real-world clinical setting, generalizable studies, and a lack of large standardized available databases for training algorithms.

Narrowing our discussion to studies of mammography—given the repetitive nature and binary output of this modality, which has made it one of the first targets of automation in diagnostic imaging^1,2,5,13—AI-based CAD in mammography, much like its predecessor feature-based CAD, has shown promising results in artificial settings. Five key mammography CNN studies have reported a wide range of diagnostic accuracy (area under the curve, 69.2 to 97.8 [mean, 88.2]) compared to radiologists.^15-19

In the most recent study (2019), Rodriguez-Ruiz et al¹⁵ compared machines and a cohort of 101 radiologists, in which AI showed performance comparability. However, results in this artificial setting were not followed up with prospective analysis of the technology in a clinical setting. First-generation, feature-based CADs in mammography also showed expert-level performance in artificial settings, but the technology became extinct because these results were not generalizable to real-world in prospective trials. To our knowledge, a limitation of radiology AI is that all current CNNs have not yet been tested in a live clinical setting.^13-19

The second limitation of radiology AI is lack of standardization, which also applies to mammography, despite this subset having the largest and oldest publicly available data set. In a recent review of 23 studies on AI-based algorithms in mammography (2010-2019), clinicians point to one of the biggest flaws: the use of small, nonstandardized, and skewed public databases (often enriched for malignancy) as training algorithms.¹³

Standardization refers to quality-control measures in acquisition, processing, and image labeling that need to be met for images to be included in the training data set. At present, large stores of radiologic data that are standardized within each institution are not publicly accessible through a unified reference platform. Lack of large standardized training data sets leads to selection bias and increases the risk for overfitting, which occurs when algorithm models incorporate background noise in the data into its prediction scheme. Overfitting has been noted in several AI-based studies in mammography,¹³ which limits the generalizability of algorithm performance in the real-world setting.

To overcome this limitation, the American College of Radiology Data Science Institute recently took the lead on creating a reference platform for quality control and standardized data generation for AI integration in radiology. The goal of the institute is for radiologists to work collaboratively with industry to ensure that algorithms are trained on quality data that produces clinically useable output for the clinician and patient.^11,20

Similar to initial radiology studies utilizing AI mainly as a screening tool, AI-driven studies in dermatology are focused on classification of melanocytic lesions; the goal is to aid in melanoma screening. Two of the most-recent, most-cited articles on this topic are by Esteva et al²¹ and Tschandl et al.²² Esteva et al²¹ matched the performance of 21 dermatologists in binary classification (malignant or nonmalignant) of clinical and dermoscopic images in pigmented and nonpigmented categories. A CNN developed by Google was trained on 130,000 clinical images encompassing more than 2000 dermatologist-labeled diagnoses from 18 sites. Despite promising results, the question remains whether these findings are transferrable to the clinical setting. In addition to the limitation on generalizability, the authors do not elaborate on standardization of training image data sets. For example, it is unclear what percentage of the training data set’s image labels were based on biopsy results vs clinical diagnosis.²¹

The second study was the largest Web-based study to compare the performance of more than 500 dermatologists worldwide.²² The top 3–performing algorithms (among a pool of 139) were at least as good as the performance of 27 expert dermatologists (defined as having more than 10 years’ experience) in the classification of pigmented lesions into 7 predefined categories.²² However, images came from nonstandardized sources gathered from a 20-year period at one European academic center and a private practice in Australia. Tschandl et al²² looked at external validation with an independent data set, outside the training data set. Although not generalizable to a real-world setting, looking at external data sets helps correct for overfitting and is a good first step in understanding transferability of results. However, the external data set was chosen by the authors and therefore might be tainted by selection bias. Although only a 10% drop in algorithmic accuracy was noted using the external data set chosen by the authors, this drop does not apply to other data sets or more importantly to a real-world setting.²²

Current limitations and future goals of radiology also will most likely apply to dermatology AI research. In medicine and radiology, the goal of AI is to first help users by prioritizing what they should focus on. The concept of comparing AI to a radiologist or dermatologist is potentially shortsighted. Shortcomings of the current supervised or semisupervised algorithms used in medicine underscore the points that, first, to make their outputs clinically usable, it should be clinicians who procure and standardize training data sets and, second, it appears logical that the performance of these category of algorithms requires constant monitoring for bias. Therefore, these algorithms cannot operate as stand-alone diagnostic machines but as an aid to the clinician—if the performance of the algorithms is proved in large trials.

Near-Future Directions and Projections

Almost all recent state-of-the-art AI systems tested in medical disciplines fall under the engineering terminology of narrow or weak AI, meaning any given algorithm is trained to do only one specific task.⁹ An example of a task is classification of images into multiple categories (ie, benign or malignant). However, task classification only works with preselected images that will need substantial improvements in standardization.

Although it has been demonstrated that AI systems can excel at one task at a time, such as classification, better than a human cohort in simulated settings, these literal machines lack the ability to incorporate context; integrate various forms of sensory input such as visual, voice, or text; or make associations the way humans do.⁹ Multiple tasks and clinical context integration are required for predictive diagnosis or clinical decision-making, even in a simulated environment. In this sense, CNN is still similar to its antiquated linear CAD predecessor: It cannot make a diagnosis or a clinical decision but might be appropriate for triaging cases that are referred for evaluation by a dermatologist.

Medical AI also may use electronic health records or patient-gathered data (eg, apps). However, clinical images are more structured and less noisy and are more easily incorporated in AI training. Therefore, as we are already witnessing, earlier validation and adoption of AI will occur in image-based disciplines, beginning with radiology; then pathology; and eventually dermatology, which will be the most challenging of the 3 medical specialties to standardize.

Final Thoughts

Artificial intelligence in health care is in its infancy; specific task-driven algorithms are only beginning to be introduced. We project that in the next 5 to 10 years, clinicians will become increasingly involved in training and testing large-scale validation as well as monitoring narrow AI in clinical trials. Radiology has served as the pioneering area in medicine and is just beginning to utilize narrow AI to help specialists with very specific tasks. For example, a task would be to triage which scans to look at first for a radiologist or which pigmented lesion might need prompt evaluation by a dermatologist. Artificial intelligence in medicine is not replacing specialists or placing decision-making in the hands of a nonexpert. At this point, CNNs have not proven that they make us better at diagnosing because real-world clinical data are lacking, which may change in the future with large standardized training data sets and validation with prospective clinical trials. The near future for dermatology and pathology will follow what is already happening in radiology, with AI substantially increasing workflow efficiency by prioritizing tasks.

Artificial intelligence (AI) is a loosely defined term that refers to machines (ie, algorithms) simulating facets of human intelligence. Some examples of AI are seen in natural language-processing algorithms, including autocorrect and search engine autocomplete functions; voice recognition in virtual assistants; autopilot systems in airplanes and self-driving cars; and computer vision in image and object recognition. Since the dawn of the century, various forms of AI have been tested and introduced in health care. However, a gap exists between clinician viewpoints on AI and the engineering world’s assumptions of what can be automated in medicine.

In this article, we review the history and evolution of AI in medicine, focusing on radiology and dermatology; current capabilities of AI; challenges to clinical integration; and future directions. Our aim is to provide realistic expectations of current technologies in solving complex problems and to empower dermatologists in planning for a future that likely includes various forms of AI.

Early Stages of AI in Medical Decision-making

Some of the earliest forms of clinical decision-support software in medicine were computer-aided detection and computer-aided diagnosis (CAD) used in screening for breast and lung cancer on mammography and computed tomography.^1-3 Early research on the use of CAD systems in radiology date to the 1960s (Figure), with the first US Food and Drug Administration–approved CAD system in mammography in 1998 and for Centers for Medicare & Medicaid Services reimbursement in 2002.^1,2

Early CAD systems relied on rule-based classifiers, which use predefined features to classify images into desired categories. For example, to classify an image as a high-risk or benign mass, features such as contour and texture had to be explicitly defined. Although these systems showed on par with, or higher, accuracy vs a radiologist in validation studies, early CAD systems never achieved wide adoption because of an increased rate of false positives as well as added work burden on a radiologist, who had to silence overcalling by the software.^1,2,4,5

Computer-aided diagnosis–based melanoma diagnosis was introduced in early 2000 in dermatology (Figure) using the same feature-based classifiers. These systems claimed expert-level accuracy in proof-of-concept studies and prospective uncontrolled trials on proprietary devices using these classifiers.^6,7 Similar to radiology, however, real-world adoption did not happen; in fact, the last of these devices was taken off the market in 2017. A recent meta-analysis of studies using CAD-based melanoma diagnosis point to study bias; data overfitting; and lack of large controlled, prospective trials as possible reasons why results could not be replicated in a clinical setting.⁸

Beyond 2010: Deep Learning

New techniques in machine learning (ML), called deep learning, began to emerge after 2010 (Figure). In deep learning, instead of directing the computer to look for certain discriminative features, the machine learns those features from the large amount of data without being explicitly programed to do so. In other words, compared to predecessor forms of computing, there is less human supervision in the learning process (Table). The concept of ML has existed since the 1980s. The field saw exponential growth in the last decade with the improvement of algorithms; an increase in computing power; and emergence of large training data sets, such as open-source platforms on the Web.^9,10

Most ML methods today incorporate artificial neural networks (ANN), computer programs that imitate the architecture of biological neural networks and form dynamically changing systems that improve with continuous data exposure. The performance of an ANN is dependent on the number and architecture of its neural layers and (similar to CAD systems) the size, quality, and generalizability of the training data set.^9-12

In medicine, images (eg, clinical or dermoscopic images and imaging scans) are the most commonly used form of data for AI development. Convolutional neural networks (CNN), a subtype of ANN, are frequently used for this purpose. These networks use a hierarchical neural network architecture, similar to the visual cortex, that allows for composition of complex features (eg, shapes) from simpler features (eg, image intensities), which leads to more efficient data processing.^10-12

In recent years, CNNs have been applied in a number of image-based medical fields, including radiology, dermatology, and pathology. Initially, studies were largely led by computer scientists trying to match clinician performance in detection of disease categories. However, there has been a shift toward more physicians getting involved, which has motivated development of large curated (ie, expert-labeled) and standardized clinical data sets in training the CNN. Although training on quality-controlled data is a work in progress across medical disciplines, it has led to improved machine performance.^11,12

Recent Advances in AI

In recent years, the number of studies covering CNN in diagnosis has increased exponentially in several medical specialties. The goal is to improve software to close the gap between experts and the machine in live clinical settings. The current literature focuses on a comparison of experts with the machine in simulated settings; prospective clinical trials are still lagging in the real world.^9,11,13

We look at radiology to explore recent advances in AI diagnosis for 3 reasons: (1) radiology has the largest repository of digital data (using a picture archiving and communication system) among medical specialties; (2) radiology has well-defined, image-acquisition protocols in its clinical workflow¹⁴; and (3) gray-scale images are easier to standardize because they are impervious to environmental variables that are difficult to control (eg, recent sun exposure, rosacea flare, lighting, sweating). These are some of the reasons we think radiology is, and will be, ahead in training AI algorithms and integrating them into clinical practice. However, even radiology AI studies have limitations, including a lack of prospective, real-world clinical setting, generalizable studies, and a lack of large standardized available databases for training algorithms.

Narrowing our discussion to studies of mammography—given the repetitive nature and binary output of this modality, which has made it one of the first targets of automation in diagnostic imaging^1,2,5,13—AI-based CAD in mammography, much like its predecessor feature-based CAD, has shown promising results in artificial settings. Five key mammography CNN studies have reported a wide range of diagnostic accuracy (area under the curve, 69.2 to 97.8 [mean, 88.2]) compared to radiologists.^15-19

In the most recent study (2019), Rodriguez-Ruiz et al¹⁵ compared machines and a cohort of 101 radiologists, in which AI showed performance comparability. However, results in this artificial setting were not followed up with prospective analysis of the technology in a clinical setting. First-generation, feature-based CADs in mammography also showed expert-level performance in artificial settings, but the technology became extinct because these results were not generalizable to real-world in prospective trials. To our knowledge, a limitation of radiology AI is that all current CNNs have not yet been tested in a live clinical setting.^13-19

The second limitation of radiology AI is lack of standardization, which also applies to mammography, despite this subset having the largest and oldest publicly available data set. In a recent review of 23 studies on AI-based algorithms in mammography (2010-2019), clinicians point to one of the biggest flaws: the use of small, nonstandardized, and skewed public databases (often enriched for malignancy) as training algorithms.¹³

Standardization refers to quality-control measures in acquisition, processing, and image labeling that need to be met for images to be included in the training data set. At present, large stores of radiologic data that are standardized within each institution are not publicly accessible through a unified reference platform. Lack of large standardized training data sets leads to selection bias and increases the risk for overfitting, which occurs when algorithm models incorporate background noise in the data into its prediction scheme. Overfitting has been noted in several AI-based studies in mammography,¹³ which limits the generalizability of algorithm performance in the real-world setting.

To overcome this limitation, the American College of Radiology Data Science Institute recently took the lead on creating a reference platform for quality control and standardized data generation for AI integration in radiology. The goal of the institute is for radiologists to work collaboratively with industry to ensure that algorithms are trained on quality data that produces clinically useable output for the clinician and patient.^11,20

Similar to initial radiology studies utilizing AI mainly as a screening tool, AI-driven studies in dermatology are focused on classification of melanocytic lesions; the goal is to aid in melanoma screening. Two of the most-recent, most-cited articles on this topic are by Esteva et al²¹ and Tschandl et al.²² Esteva et al²¹ matched the performance of 21 dermatologists in binary classification (malignant or nonmalignant) of clinical and dermoscopic images in pigmented and nonpigmented categories. A CNN developed by Google was trained on 130,000 clinical images encompassing more than 2000 dermatologist-labeled diagnoses from 18 sites. Despite promising results, the question remains whether these findings are transferrable to the clinical setting. In addition to the limitation on generalizability, the authors do not elaborate on standardization of training image data sets. For example, it is unclear what percentage of the training data set’s image labels were based on biopsy results vs clinical diagnosis.²¹

The second study was the largest Web-based study to compare the performance of more than 500 dermatologists worldwide.²² The top 3–performing algorithms (among a pool of 139) were at least as good as the performance of 27 expert dermatologists (defined as having more than 10 years’ experience) in the classification of pigmented lesions into 7 predefined categories.²² However, images came from nonstandardized sources gathered from a 20-year period at one European academic center and a private practice in Australia. Tschandl et al²² looked at external validation with an independent data set, outside the training data set. Although not generalizable to a real-world setting, looking at external data sets helps correct for overfitting and is a good first step in understanding transferability of results. However, the external data set was chosen by the authors and therefore might be tainted by selection bias. Although only a 10% drop in algorithmic accuracy was noted using the external data set chosen by the authors, this drop does not apply to other data sets or more importantly to a real-world setting.²²

Current limitations and future goals of radiology also will most likely apply to dermatology AI research. In medicine and radiology, the goal of AI is to first help users by prioritizing what they should focus on. The concept of comparing AI to a radiologist or dermatologist is potentially shortsighted. Shortcomings of the current supervised or semisupervised algorithms used in medicine underscore the points that, first, to make their outputs clinically usable, it should be clinicians who procure and standardize training data sets and, second, it appears logical that the performance of these category of algorithms requires constant monitoring for bias. Therefore, these algorithms cannot operate as stand-alone diagnostic machines but as an aid to the clinician—if the performance of the algorithms is proved in large trials.

Near-Future Directions and Projections

Almost all recent state-of-the-art AI systems tested in medical disciplines fall under the engineering terminology of narrow or weak AI, meaning any given algorithm is trained to do only one specific task.⁹ An example of a task is classification of images into multiple categories (ie, benign or malignant). However, task classification only works with preselected images that will need substantial improvements in standardization.

Although it has been demonstrated that AI systems can excel at one task at a time, such as classification, better than a human cohort in simulated settings, these literal machines lack the ability to incorporate context; integrate various forms of sensory input such as visual, voice, or text; or make associations the way humans do.⁹ Multiple tasks and clinical context integration are required for predictive diagnosis or clinical decision-making, even in a simulated environment. In this sense, CNN is still similar to its antiquated linear CAD predecessor: It cannot make a diagnosis or a clinical decision but might be appropriate for triaging cases that are referred for evaluation by a dermatologist.

Medical AI also may use electronic health records or patient-gathered data (eg, apps). However, clinical images are more structured and less noisy and are more easily incorporated in AI training. Therefore, as we are already witnessing, earlier validation and adoption of AI will occur in image-based disciplines, beginning with radiology; then pathology; and eventually dermatology, which will be the most challenging of the 3 medical specialties to standardize.

Final Thoughts

Artificial intelligence in health care is in its infancy; specific task-driven algorithms are only beginning to be introduced. We project that in the next 5 to 10 years, clinicians will become increasingly involved in training and testing large-scale validation as well as monitoring narrow AI in clinical trials. Radiology has served as the pioneering area in medicine and is just beginning to utilize narrow AI to help specialists with very specific tasks. For example, a task would be to triage which scans to look at first for a radiologist or which pigmented lesion might need prompt evaluation by a dermatologist. Artificial intelligence in medicine is not replacing specialists or placing decision-making in the hands of a nonexpert. At this point, CNNs have not proven that they make us better at diagnosing because real-world clinical data are lacking, which may change in the future with large standardized training data sets and validation with prospective clinical trials. The near future for dermatology and pathology will follow what is already happening in radiology, with AI substantially increasing workflow efficiency by prioritizing tasks.

Artificial intelligence (AI) is a loosely defined term that refers to machines (ie, algorithms) simulating facets of human intelligence. Some examples of AI are seen in natural language-processing algorithms, including autocorrect and search engine autocomplete functions; voice recognition in virtual assistants; autopilot systems in airplanes and self-driving cars; and computer vision in image and object recognition. Since the dawn of the century, various forms of AI have been tested and introduced in health care. However, a gap exists between clinician viewpoints on AI and the engineering world’s assumptions of what can be automated in medicine.

In this article, we review the history and evolution of AI in medicine, focusing on radiology and dermatology; current capabilities of AI; challenges to clinical integration; and future directions. Our aim is to provide realistic expectations of current technologies in solving complex problems and to empower dermatologists in planning for a future that likely includes various forms of AI.

Early Stages of AI in Medical Decision-making

Some of the earliest forms of clinical decision-support software in medicine were computer-aided detection and computer-aided diagnosis (CAD) used in screening for breast and lung cancer on mammography and computed tomography.^1-3 Early research on the use of CAD systems in radiology date to the 1960s (Figure), with the first US Food and Drug Administration–approved CAD system in mammography in 1998 and for Centers for Medicare & Medicaid Services reimbursement in 2002.^1,2

Early CAD systems relied on rule-based classifiers, which use predefined features to classify images into desired categories. For example, to classify an image as a high-risk or benign mass, features such as contour and texture had to be explicitly defined. Although these systems showed on par with, or higher, accuracy vs a radiologist in validation studies, early CAD systems never achieved wide adoption because of an increased rate of false positives as well as added work burden on a radiologist, who had to silence overcalling by the software.^1,2,4,5

Computer-aided diagnosis–based melanoma diagnosis was introduced in early 2000 in dermatology (Figure) using the same feature-based classifiers. These systems claimed expert-level accuracy in proof-of-concept studies and prospective uncontrolled trials on proprietary devices using these classifiers.^6,7 Similar to radiology, however, real-world adoption did not happen; in fact, the last of these devices was taken off the market in 2017. A recent meta-analysis of studies using CAD-based melanoma diagnosis point to study bias; data overfitting; and lack of large controlled, prospective trials as possible reasons why results could not be replicated in a clinical setting.⁸

Beyond 2010: Deep Learning

New techniques in machine learning (ML), called deep learning, began to emerge after 2010 (Figure). In deep learning, instead of directing the computer to look for certain discriminative features, the machine learns those features from the large amount of data without being explicitly programed to do so. In other words, compared to predecessor forms of computing, there is less human supervision in the learning process (Table). The concept of ML has existed since the 1980s. The field saw exponential growth in the last decade with the improvement of algorithms; an increase in computing power; and emergence of large training data sets, such as open-source platforms on the Web.^9,10

Most ML methods today incorporate artificial neural networks (ANN), computer programs that imitate the architecture of biological neural networks and form dynamically changing systems that improve with continuous data exposure. The performance of an ANN is dependent on the number and architecture of its neural layers and (similar to CAD systems) the size, quality, and generalizability of the training data set.^9-12

In medicine, images (eg, clinical or dermoscopic images and imaging scans) are the most commonly used form of data for AI development. Convolutional neural networks (CNN), a subtype of ANN, are frequently used for this purpose. These networks use a hierarchical neural network architecture, similar to the visual cortex, that allows for composition of complex features (eg, shapes) from simpler features (eg, image intensities), which leads to more efficient data processing.^10-12

In recent years, CNNs have been applied in a number of image-based medical fields, including radiology, dermatology, and pathology. Initially, studies were largely led by computer scientists trying to match clinician performance in detection of disease categories. However, there has been a shift toward more physicians getting involved, which has motivated development of large curated (ie, expert-labeled) and standardized clinical data sets in training the CNN. Although training on quality-controlled data is a work in progress across medical disciplines, it has led to improved machine performance.^11,12

Recent Advances in AI

In recent years, the number of studies covering CNN in diagnosis has increased exponentially in several medical specialties. The goal is to improve software to close the gap between experts and the machine in live clinical settings. The current literature focuses on a comparison of experts with the machine in simulated settings; prospective clinical trials are still lagging in the real world.^9,11,13

We look at radiology to explore recent advances in AI diagnosis for 3 reasons: (1) radiology has the largest repository of digital data (using a picture archiving and communication system) among medical specialties; (2) radiology has well-defined, image-acquisition protocols in its clinical workflow¹⁴; and (3) gray-scale images are easier to standardize because they are impervious to environmental variables that are difficult to control (eg, recent sun exposure, rosacea flare, lighting, sweating). These are some of the reasons we think radiology is, and will be, ahead in training AI algorithms and integrating them into clinical practice. However, even radiology AI studies have limitations, including a lack of prospective, real-world clinical setting, generalizable studies, and a lack of large standardized available databases for training algorithms.

Narrowing our discussion to studies of mammography—given the repetitive nature and binary output of this modality, which has made it one of the first targets of automation in diagnostic imaging^1,2,5,13—AI-based CAD in mammography, much like its predecessor feature-based CAD, has shown promising results in artificial settings. Five key mammography CNN studies have reported a wide range of diagnostic accuracy (area under the curve, 69.2 to 97.8 [mean, 88.2]) compared to radiologists.^15-19

In the most recent study (2019), Rodriguez-Ruiz et al¹⁵ compared machines and a cohort of 101 radiologists, in which AI showed performance comparability. However, results in this artificial setting were not followed up with prospective analysis of the technology in a clinical setting. First-generation, feature-based CADs in mammography also showed expert-level performance in artificial settings, but the technology became extinct because these results were not generalizable to real-world in prospective trials. To our knowledge, a limitation of radiology AI is that all current CNNs have not yet been tested in a live clinical setting.^13-19

The second limitation of radiology AI is lack of standardization, which also applies to mammography, despite this subset having the largest and oldest publicly available data set. In a recent review of 23 studies on AI-based algorithms in mammography (2010-2019), clinicians point to one of the biggest flaws: the use of small, nonstandardized, and skewed public databases (often enriched for malignancy) as training algorithms.¹³

Standardization refers to quality-control measures in acquisition, processing, and image labeling that need to be met for images to be included in the training data set. At present, large stores of radiologic data that are standardized within each institution are not publicly accessible through a unified reference platform. Lack of large standardized training data sets leads to selection bias and increases the risk for overfitting, which occurs when algorithm models incorporate background noise in the data into its prediction scheme. Overfitting has been noted in several AI-based studies in mammography,¹³ which limits the generalizability of algorithm performance in the real-world setting.

To overcome this limitation, the American College of Radiology Data Science Institute recently took the lead on creating a reference platform for quality control and standardized data generation for AI integration in radiology. The goal of the institute is for radiologists to work collaboratively with industry to ensure that algorithms are trained on quality data that produces clinically useable output for the clinician and patient.^11,20

Similar to initial radiology studies utilizing AI mainly as a screening tool, AI-driven studies in dermatology are focused on classification of melanocytic lesions; the goal is to aid in melanoma screening. Two of the most-recent, most-cited articles on this topic are by Esteva et al²¹ and Tschandl et al.²² Esteva et al²¹ matched the performance of 21 dermatologists in binary classification (malignant or nonmalignant) of clinical and dermoscopic images in pigmented and nonpigmented categories. A CNN developed by Google was trained on 130,000 clinical images encompassing more than 2000 dermatologist-labeled diagnoses from 18 sites. Despite promising results, the question remains whether these findings are transferrable to the clinical setting. In addition to the limitation on generalizability, the authors do not elaborate on standardization of training image data sets. For example, it is unclear what percentage of the training data set’s image labels were based on biopsy results vs clinical diagnosis.²¹

The second study was the largest Web-based study to compare the performance of more than 500 dermatologists worldwide.²² The top 3–performing algorithms (among a pool of 139) were at least as good as the performance of 27 expert dermatologists (defined as having more than 10 years’ experience) in the classification of pigmented lesions into 7 predefined categories.²² However, images came from nonstandardized sources gathered from a 20-year period at one European academic center and a private practice in Australia. Tschandl et al²² looked at external validation with an independent data set, outside the training data set. Although not generalizable to a real-world setting, looking at external data sets helps correct for overfitting and is a good first step in understanding transferability of results. However, the external data set was chosen by the authors and therefore might be tainted by selection bias. Although only a 10% drop in algorithmic accuracy was noted using the external data set chosen by the authors, this drop does not apply to other data sets or more importantly to a real-world setting.²²

Current limitations and future goals of radiology also will most likely apply to dermatology AI research. In medicine and radiology, the goal of AI is to first help users by prioritizing what they should focus on. The concept of comparing AI to a radiologist or dermatologist is potentially shortsighted. Shortcomings of the current supervised or semisupervised algorithms used in medicine underscore the points that, first, to make their outputs clinically usable, it should be clinicians who procure and standardize training data sets and, second, it appears logical that the performance of these category of algorithms requires constant monitoring for bias. Therefore, these algorithms cannot operate as stand-alone diagnostic machines but as an aid to the clinician—if the performance of the algorithms is proved in large trials.

Near-Future Directions and Projections

Almost all recent state-of-the-art AI systems tested in medical disciplines fall under the engineering terminology of narrow or weak AI, meaning any given algorithm is trained to do only one specific task.⁹ An example of a task is classification of images into multiple categories (ie, benign or malignant). However, task classification only works with preselected images that will need substantial improvements in standardization.

Although it has been demonstrated that AI systems can excel at one task at a time, such as classification, better than a human cohort in simulated settings, these literal machines lack the ability to incorporate context; integrate various forms of sensory input such as visual, voice, or text; or make associations the way humans do.⁹ Multiple tasks and clinical context integration are required for predictive diagnosis or clinical decision-making, even in a simulated environment. In this sense, CNN is still similar to its antiquated linear CAD predecessor: It cannot make a diagnosis or a clinical decision but might be appropriate for triaging cases that are referred for evaluation by a dermatologist.

Medical AI also may use electronic health records or patient-gathered data (eg, apps). However, clinical images are more structured and less noisy and are more easily incorporated in AI training. Therefore, as we are already witnessing, earlier validation and adoption of AI will occur in image-based disciplines, beginning with radiology; then pathology; and eventually dermatology, which will be the most challenging of the 3 medical specialties to standardize.

Final Thoughts

Artificial intelligence in health care is in its infancy; specific task-driven algorithms are only beginning to be introduced. We project that in the next 5 to 10 years, clinicians will become increasingly involved in training and testing large-scale validation as well as monitoring narrow AI in clinical trials. Radiology has served as the pioneering area in medicine and is just beginning to utilize narrow AI to help specialists with very specific tasks. For example, a task would be to triage which scans to look at first for a radiologist or which pigmented lesion might need prompt evaluation by a dermatologist. Artificial intelligence in medicine is not replacing specialists or placing decision-making in the hands of a nonexpert. At this point, CNNs have not proven that they make us better at diagnosing because real-world clinical data are lacking, which may change in the future with large standardized training data sets and validation with prospective clinical trials. The near future for dermatology and pathology will follow what is already happening in radiology, with AI substantially increasing workflow efficiency by prioritizing tasks.

To the Editor:

Port-wine stains (PWSs) are the most common type of vascular malformations. Patients rarely develop cancers in the overlying skin. However, we describe a case of melanoma in situ occurring within a long-standing facial PWS.

A 60-year-old white man with a history of a large unilateral facial PWS covering the right ear, lateral cheek, jaw, and neck presented to clinic with a new dark lesion on the right ear that had been growing for a few weeks or more. His PWS had been previously treated intermittently with a pulsed dye laser (PDL) for decades with variable improvement. He had not undergone any laser procedures in the last 8 months but wanted to restart treatment with the PDL. Upon further discussion, he reported a new darker area on the right earlobe that was growing. He had no personal or family history of skin cancer and was otherwise healthy. Physical examination revealed a large red vascular patch encompassing the ear, cheek, chin, and lateral neck. Within the PWS there was a black and dark brown patch with irregular borders on the right earlobe (Figure 1A). A shave biopsy was performed for histopathologic examination. The biopsy showed a confluent proliferation of atypical melanocytes along the dermoepidermal junction extending down adnexal structures (Figure 2A) that stained positive for MART-1/Melan-A (Figure 2B). In the dermis, solar elastosis and prominent dilated and thin-walled vessels were present. These findings were consistent with a melanoma in situ, lentigo maligna type, overlying a capillary malformation.

Port-wine stains are congenital vascular malformations that affect approximately 0.3% of individuals.¹ Most are located on the head and neck along the distribution of the trigeminal nerve. Cases are thought to occur sporadically, with recent evidence for somatic GNAQ mutations in both nonsyndromic cases and in Sturge-Weber syndrome.² These lesions become progressively larger with time due to dilation of the capillary proliferation.³ Melanoma in situ, lentigo maligna type, usually affects white men in the sixth and seventh decades of life. It commonly arises on skin with chronic sun damage, particularly on the head and neck.⁴

Although uncommon, skin cancers have been known to arise in PWSs. Reports of basal cell carcinomas (BCCs) and squamous cell carcinomas (SCCs) have been published, but to date, there are no reports of melanoma or melanoma in situ arising in a PWS. According to a PubMed search of articles indexed for MEDLINE using the terms melanoma and port wine stain, squamous cell carcinoma and port wine stain, and basal cell carcinoma and port wine stain, fewer than 30 cases of BCCs in a PWS and only 4 cases of SCCs in a PWS have been documented, with 1 patient developing multiple BCCs and SCCs.^1,5 Most BCCs (approximately 75%) and SCCs have been associated with historical treatments used to treat PWS before the development of laser therapy, such as grenz rays, topical thorium X, and other radiotherapy techniques.^5,6 Interestingly, our patient’s PWS had only been treated with a PDL. Other risk factors for skin cancer in a PWS include sun exposure and smoking.⁵ There is no evidence that a PDL contributes to the development of skin cancer, but radiotherapy is a major factor.⁷

Treatment of these skin cancers is no different, with both Mohs micrographic surgery and standard excision used when appropriate. Despite the vascular nature of the lesion, there is only a minimal increase in bleeding risk.³ Most reports indicate no increase in perioperative bleeding.^5,7 One case documented a hematoma developing postoperatively.⁶

This case of melanoma in situ arising in a PWS expands the range of skin cancer types known to arise in these malformations. Because of the potential for skin cancer to develop in a PWS, it is important to routinely examine these vascular proliferations.

To the Editor:

Port-wine stains (PWSs) are the most common type of vascular malformations. Patients rarely develop cancers in the overlying skin. However, we describe a case of melanoma in situ occurring within a long-standing facial PWS.

A 60-year-old white man with a history of a large unilateral facial PWS covering the right ear, lateral cheek, jaw, and neck presented to clinic with a new dark lesion on the right ear that had been growing for a few weeks or more. His PWS had been previously treated intermittently with a pulsed dye laser (PDL) for decades with variable improvement. He had not undergone any laser procedures in the last 8 months but wanted to restart treatment with the PDL. Upon further discussion, he reported a new darker area on the right earlobe that was growing. He had no personal or family history of skin cancer and was otherwise healthy. Physical examination revealed a large red vascular patch encompassing the ear, cheek, chin, and lateral neck. Within the PWS there was a black and dark brown patch with irregular borders on the right earlobe (Figure 1A). A shave biopsy was performed for histopathologic examination. The biopsy showed a confluent proliferation of atypical melanocytes along the dermoepidermal junction extending down adnexal structures (Figure 2A) that stained positive for MART-1/Melan-A (Figure 2B). In the dermis, solar elastosis and prominent dilated and thin-walled vessels were present. These findings were consistent with a melanoma in situ, lentigo maligna type, overlying a capillary malformation.

Port-wine stains are congenital vascular malformations that affect approximately 0.3% of individuals.¹ Most are located on the head and neck along the distribution of the trigeminal nerve. Cases are thought to occur sporadically, with recent evidence for somatic GNAQ mutations in both nonsyndromic cases and in Sturge-Weber syndrome.² These lesions become progressively larger with time due to dilation of the capillary proliferation.³ Melanoma in situ, lentigo maligna type, usually affects white men in the sixth and seventh decades of life. It commonly arises on skin with chronic sun damage, particularly on the head and neck.⁴

Although uncommon, skin cancers have been known to arise in PWSs. Reports of basal cell carcinomas (BCCs) and squamous cell carcinomas (SCCs) have been published, but to date, there are no reports of melanoma or melanoma in situ arising in a PWS. According to a PubMed search of articles indexed for MEDLINE using the terms melanoma and port wine stain, squamous cell carcinoma and port wine stain, and basal cell carcinoma and port wine stain, fewer than 30 cases of BCCs in a PWS and only 4 cases of SCCs in a PWS have been documented, with 1 patient developing multiple BCCs and SCCs.^1,5 Most BCCs (approximately 75%) and SCCs have been associated with historical treatments used to treat PWS before the development of laser therapy, such as grenz rays, topical thorium X, and other radiotherapy techniques.^5,6 Interestingly, our patient’s PWS had only been treated with a PDL. Other risk factors for skin cancer in a PWS include sun exposure and smoking.⁵ There is no evidence that a PDL contributes to the development of skin cancer, but radiotherapy is a major factor.⁷

Treatment of these skin cancers is no different, with both Mohs micrographic surgery and standard excision used when appropriate. Despite the vascular nature of the lesion, there is only a minimal increase in bleeding risk.³ Most reports indicate no increase in perioperative bleeding.^5,7 One case documented a hematoma developing postoperatively.⁶

This case of melanoma in situ arising in a PWS expands the range of skin cancer types known to arise in these malformations. Because of the potential for skin cancer to develop in a PWS, it is important to routinely examine these vascular proliferations.

To the Editor:

Port-wine stains (PWSs) are the most common type of vascular malformations. Patients rarely develop cancers in the overlying skin. However, we describe a case of melanoma in situ occurring within a long-standing facial PWS.

A 60-year-old white man with a history of a large unilateral facial PWS covering the right ear, lateral cheek, jaw, and neck presented to clinic with a new dark lesion on the right ear that had been growing for a few weeks or more. His PWS had been previously treated intermittently with a pulsed dye laser (PDL) for decades with variable improvement. He had not undergone any laser procedures in the last 8 months but wanted to restart treatment with the PDL. Upon further discussion, he reported a new darker area on the right earlobe that was growing. He had no personal or family history of skin cancer and was otherwise healthy. Physical examination revealed a large red vascular patch encompassing the ear, cheek, chin, and lateral neck. Within the PWS there was a black and dark brown patch with irregular borders on the right earlobe (Figure 1A). A shave biopsy was performed for histopathologic examination. The biopsy showed a confluent proliferation of atypical melanocytes along the dermoepidermal junction extending down adnexal structures (Figure 2A) that stained positive for MART-1/Melan-A (Figure 2B). In the dermis, solar elastosis and prominent dilated and thin-walled vessels were present. These findings were consistent with a melanoma in situ, lentigo maligna type, overlying a capillary malformation.

Port-wine stains are congenital vascular malformations that affect approximately 0.3% of individuals.¹ Most are located on the head and neck along the distribution of the trigeminal nerve. Cases are thought to occur sporadically, with recent evidence for somatic GNAQ mutations in both nonsyndromic cases and in Sturge-Weber syndrome.² These lesions become progressively larger with time due to dilation of the capillary proliferation.³ Melanoma in situ, lentigo maligna type, usually affects white men in the sixth and seventh decades of life. It commonly arises on skin with chronic sun damage, particularly on the head and neck.⁴

Although uncommon, skin cancers have been known to arise in PWSs. Reports of basal cell carcinomas (BCCs) and squamous cell carcinomas (SCCs) have been published, but to date, there are no reports of melanoma or melanoma in situ arising in a PWS. According to a PubMed search of articles indexed for MEDLINE using the terms melanoma and port wine stain, squamous cell carcinoma and port wine stain, and basal cell carcinoma and port wine stain, fewer than 30 cases of BCCs in a PWS and only 4 cases of SCCs in a PWS have been documented, with 1 patient developing multiple BCCs and SCCs.^1,5 Most BCCs (approximately 75%) and SCCs have been associated with historical treatments used to treat PWS before the development of laser therapy, such as grenz rays, topical thorium X, and other radiotherapy techniques.^5,6 Interestingly, our patient’s PWS had only been treated with a PDL. Other risk factors for skin cancer in a PWS include sun exposure and smoking.⁵ There is no evidence that a PDL contributes to the development of skin cancer, but radiotherapy is a major factor.⁷

Treatment of these skin cancers is no different, with both Mohs micrographic surgery and standard excision used when appropriate. Despite the vascular nature of the lesion, there is only a minimal increase in bleeding risk.³ Most reports indicate no increase in perioperative bleeding.^5,7 One case documented a hematoma developing postoperatively.⁶

This case of melanoma in situ arising in a PWS expands the range of skin cancer types known to arise in these malformations. Because of the potential for skin cancer to develop in a PWS, it is important to routinely examine these vascular proliferations.