Howard P. Levy, MD, PhD

Family history has always been one of the most powerful and inexpensive tools for assessing a patient’s genetic risks.

In the case of rare single-gene (Mendelian) syndromes , such as Marfan syndrome, cystic fibrosis, or a cancer syndrome, knowledge of the family history can alert the patient and his/her physicians of the potential need for testing, surveillance and/or preventive measures. For common multifactorial disorders, such as diabetes, asthma, or non-Mendelian cancer predisposition, this is even more important, because a single genetic test result is less likely to drastically change an individual’s overall risk assessment.

Obstacles to widespread utilization of family history among primary care physicians have included the time required to gather the data; storage and retrieval of the data; and uncertainty about how to interpret it, especially in the era of evidence-based medicine.

Genetic testing provides another source of data, which shares many of the same obstacles to widespread adoption as family history: choosing and ordering appropriate tests; data storage and retrieval; evidence-based application of these data.

These challenges are not too difficult to solve when dealing with a single mutation or just one or two genes. But we are now in an era of easily accessible genome-wide mapping, with 1 million or more pieces of data in each test result. And clinical-grade sequencing of entire genomes, containing several billion pieces of data, is coming soon.

With specialty training, a human can learn to interpret moderately complicated family histories with a high degree of scientific accuracy. But global assessment of an entire genome, either for identification of the cause of a Mendelian syndrome or for recognition of factors contributing to common disease, is beyond reasonable human capacity.

Realistically, neither family history nor genetic testing is sufficient for robust risk assessment. Combining both sources of genetic information, together with environmental and other personal factors, is the ultimate goal, and is the essence of individualized medicine.

Much work is yet to be done to fully develop the evidence and refine the algorithms that will drive this type of medical care. But for many conditions, there is already sufficient qualitative and quantitative evidence to justify clinical application today. Consider, for example, the inclusion of family history in assessment of coronary artery disease risk or likelihood of carrying a mutation in one of the BRCA genes. So, the third obstacle, evidence-based interpretation, is not really the rate-limiting step today.

The biggest current barriers to integration of family history and genetic (or genomic) data in actual clinical care are data acquisition, storage, and retrieval.

The solution to this problem is obvious: computers.

More specifically, despite all of the grumbling and complaints they generate, electronic health records (EHRs) hold great promise for simplifying our lives while simultaneously improving patient care.

To be maximally useful, a family history should include data on three to four generations: children, siblings, parents/aunts/uncles, and grandparents. Depending upon the specific question at hand, children or grandparents may be more or less relevant to the risk assessment; cousins, nieces, and nephews may sometimes be important, too. For each relative, useful data include age, age of death, cause of death, and the diagnosis and approximate age of onset of each health problem. In many cases, presence or absence of common environmental risk factors is also important (for example, lung cancer without exposure to tobacco, asbestos, or radon suggests a greater likelihood of genetic predisposition). The ancestral origin on each side of the family is also sometimes informative.

Add to that the increasing availability of single-gene and whole-genome testing results, and the task of assembling all of this data becomes overwhelming.

EHRs have the ability to simplify all of this.

Just as is done today when collecting medical information about an individual patient, the family history and genetic testing elements discussed above can be stored as discrete pieces of data for each of that patient’s relatives. Studies have already shown that patients’ recall of their family history is generally pretty reliable. And if they’re encouraged to discuss that history with family members, the quality of the data improves further.

Patient portals, an integral component of meaningful use of EHRs, empower patients to collect and input this information from home, solving the problem of inadequate time during a clinical encounter.

With appropriate informed consent, it is also technically feasible to link EHR data of related individuals to increase the quantity and accuracy of family history data. Genomic sequencing data, from the patient and his/her relatives, can also be electronically linked to the family history.

Once gathered and stored, those discrete elements still need to be displayed in a logical way. People in general and physicians in particular, value flexibility and individuality in how they view data. Some prefer a graphical display of family history in a genogram or pedigree; others prefer tabular representation. Sometimes, it is helpful to see which relatives have a specific condition or group of conditions. Other times, it is useful to view the medical history of each relative individually. One may wish to view the entire family history all at once, but often a filtered view of only certain conditions or relatives of interest is more relevant.

For the most part, EHR vendors have not yet developed the software to support management of family history and genetic data as discussed here. They are not completely to blame. Like any other business, they focus their resources on tools and innovations that their customers need or want. We, the consumers of EHR products, have not yet made this a high enough priority.

Ultimately, detailed and discrete family history, genetic, and environmental data can be fed into evidence-based algorithms, some already in existence, others yet to be developed, to drive clinical decision support that helps physicians at the point of care to deliver safer, more efficient, and more effective medical care. What stands in our way is the ability to gather and organize this data.

EHRs, despite the frustration of implementing and learning to use them, are the key to unlocking this potential.

Please consider asking your EHR vendor to develop the family history and genomic data tools necessary to enable truly individualized medicine.

Dr. Levy is with the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine at Johns Hopkins University, Baltimore.

Family history has always been one of the most powerful and inexpensive tools for assessing a patient’s genetic risks.

In the case of rare single-gene (Mendelian) syndromes , such as Marfan syndrome, cystic fibrosis, or a cancer syndrome, knowledge of the family history can alert the patient and his/her physicians of the potential need for testing, surveillance and/or preventive measures. For common multifactorial disorders, such as diabetes, asthma, or non-Mendelian cancer predisposition, this is even more important, because a single genetic test result is less likely to drastically change an individual’s overall risk assessment.

Obstacles to widespread utilization of family history among primary care physicians have included the time required to gather the data; storage and retrieval of the data; and uncertainty about how to interpret it, especially in the era of evidence-based medicine.

Genetic testing provides another source of data, which shares many of the same obstacles to widespread adoption as family history: choosing and ordering appropriate tests; data storage and retrieval; evidence-based application of these data.

These challenges are not too difficult to solve when dealing with a single mutation or just one or two genes. But we are now in an era of easily accessible genome-wide mapping, with 1 million or more pieces of data in each test result. And clinical-grade sequencing of entire genomes, containing several billion pieces of data, is coming soon.

With specialty training, a human can learn to interpret moderately complicated family histories with a high degree of scientific accuracy. But global assessment of an entire genome, either for identification of the cause of a Mendelian syndrome or for recognition of factors contributing to common disease, is beyond reasonable human capacity.

Realistically, neither family history nor genetic testing is sufficient for robust risk assessment. Combining both sources of genetic information, together with environmental and other personal factors, is the ultimate goal, and is the essence of individualized medicine.

Much work is yet to be done to fully develop the evidence and refine the algorithms that will drive this type of medical care. But for many conditions, there is already sufficient qualitative and quantitative evidence to justify clinical application today. Consider, for example, the inclusion of family history in assessment of coronary artery disease risk or likelihood of carrying a mutation in one of the BRCA genes. So, the third obstacle, evidence-based interpretation, is not really the rate-limiting step today.

The biggest current barriers to integration of family history and genetic (or genomic) data in actual clinical care are data acquisition, storage, and retrieval.

The solution to this problem is obvious: computers.

More specifically, despite all of the grumbling and complaints they generate, electronic health records (EHRs) hold great promise for simplifying our lives while simultaneously improving patient care.

To be maximally useful, a family history should include data on three to four generations: children, siblings, parents/aunts/uncles, and grandparents. Depending upon the specific question at hand, children or grandparents may be more or less relevant to the risk assessment; cousins, nieces, and nephews may sometimes be important, too. For each relative, useful data include age, age of death, cause of death, and the diagnosis and approximate age of onset of each health problem. In many cases, presence or absence of common environmental risk factors is also important (for example, lung cancer without exposure to tobacco, asbestos, or radon suggests a greater likelihood of genetic predisposition). The ancestral origin on each side of the family is also sometimes informative.

Add to that the increasing availability of single-gene and whole-genome testing results, and the task of assembling all of this data becomes overwhelming.

EHRs have the ability to simplify all of this.

Just as is done today when collecting medical information about an individual patient, the family history and genetic testing elements discussed above can be stored as discrete pieces of data for each of that patient’s relatives. Studies have already shown that patients’ recall of their family history is generally pretty reliable. And if they’re encouraged to discuss that history with family members, the quality of the data improves further.

Patient portals, an integral component of meaningful use of EHRs, empower patients to collect and input this information from home, solving the problem of inadequate time during a clinical encounter.

With appropriate informed consent, it is also technically feasible to link EHR data of related individuals to increase the quantity and accuracy of family history data. Genomic sequencing data, from the patient and his/her relatives, can also be electronically linked to the family history.

Once gathered and stored, those discrete elements still need to be displayed in a logical way. People in general and physicians in particular, value flexibility and individuality in how they view data. Some prefer a graphical display of family history in a genogram or pedigree; others prefer tabular representation. Sometimes, it is helpful to see which relatives have a specific condition or group of conditions. Other times, it is useful to view the medical history of each relative individually. One may wish to view the entire family history all at once, but often a filtered view of only certain conditions or relatives of interest is more relevant.

For the most part, EHR vendors have not yet developed the software to support management of family history and genetic data as discussed here. They are not completely to blame. Like any other business, they focus their resources on tools and innovations that their customers need or want. We, the consumers of EHR products, have not yet made this a high enough priority.

Ultimately, detailed and discrete family history, genetic, and environmental data can be fed into evidence-based algorithms, some already in existence, others yet to be developed, to drive clinical decision support that helps physicians at the point of care to deliver safer, more efficient, and more effective medical care. What stands in our way is the ability to gather and organize this data.

EHRs, despite the frustration of implementing and learning to use them, are the key to unlocking this potential.

Please consider asking your EHR vendor to develop the family history and genomic data tools necessary to enable truly individualized medicine.

Dr. Levy is with the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine at Johns Hopkins University, Baltimore.

Family history has always been one of the most powerful and inexpensive tools for assessing a patient’s genetic risks.

In the case of rare single-gene (Mendelian) syndromes , such as Marfan syndrome, cystic fibrosis, or a cancer syndrome, knowledge of the family history can alert the patient and his/her physicians of the potential need for testing, surveillance and/or preventive measures. For common multifactorial disorders, such as diabetes, asthma, or non-Mendelian cancer predisposition, this is even more important, because a single genetic test result is less likely to drastically change an individual’s overall risk assessment.

Obstacles to widespread utilization of family history among primary care physicians have included the time required to gather the data; storage and retrieval of the data; and uncertainty about how to interpret it, especially in the era of evidence-based medicine.

Genetic testing provides another source of data, which shares many of the same obstacles to widespread adoption as family history: choosing and ordering appropriate tests; data storage and retrieval; evidence-based application of these data.

These challenges are not too difficult to solve when dealing with a single mutation or just one or two genes. But we are now in an era of easily accessible genome-wide mapping, with 1 million or more pieces of data in each test result. And clinical-grade sequencing of entire genomes, containing several billion pieces of data, is coming soon.

With specialty training, a human can learn to interpret moderately complicated family histories with a high degree of scientific accuracy. But global assessment of an entire genome, either for identification of the cause of a Mendelian syndrome or for recognition of factors contributing to common disease, is beyond reasonable human capacity.

Realistically, neither family history nor genetic testing is sufficient for robust risk assessment. Combining both sources of genetic information, together with environmental and other personal factors, is the ultimate goal, and is the essence of individualized medicine.

Much work is yet to be done to fully develop the evidence and refine the algorithms that will drive this type of medical care. But for many conditions, there is already sufficient qualitative and quantitative evidence to justify clinical application today. Consider, for example, the inclusion of family history in assessment of coronary artery disease risk or likelihood of carrying a mutation in one of the BRCA genes. So, the third obstacle, evidence-based interpretation, is not really the rate-limiting step today.

The biggest current barriers to integration of family history and genetic (or genomic) data in actual clinical care are data acquisition, storage, and retrieval.

The solution to this problem is obvious: computers.

More specifically, despite all of the grumbling and complaints they generate, electronic health records (EHRs) hold great promise for simplifying our lives while simultaneously improving patient care.

To be maximally useful, a family history should include data on three to four generations: children, siblings, parents/aunts/uncles, and grandparents. Depending upon the specific question at hand, children or grandparents may be more or less relevant to the risk assessment; cousins, nieces, and nephews may sometimes be important, too. For each relative, useful data include age, age of death, cause of death, and the diagnosis and approximate age of onset of each health problem. In many cases, presence or absence of common environmental risk factors is also important (for example, lung cancer without exposure to tobacco, asbestos, or radon suggests a greater likelihood of genetic predisposition). The ancestral origin on each side of the family is also sometimes informative.

Add to that the increasing availability of single-gene and whole-genome testing results, and the task of assembling all of this data becomes overwhelming.

EHRs have the ability to simplify all of this.

Just as is done today when collecting medical information about an individual patient, the family history and genetic testing elements discussed above can be stored as discrete pieces of data for each of that patient’s relatives. Studies have already shown that patients’ recall of their family history is generally pretty reliable. And if they’re encouraged to discuss that history with family members, the quality of the data improves further.

Patient portals, an integral component of meaningful use of EHRs, empower patients to collect and input this information from home, solving the problem of inadequate time during a clinical encounter.

With appropriate informed consent, it is also technically feasible to link EHR data of related individuals to increase the quantity and accuracy of family history data. Genomic sequencing data, from the patient and his/her relatives, can also be electronically linked to the family history.

Once gathered and stored, those discrete elements still need to be displayed in a logical way. People in general and physicians in particular, value flexibility and individuality in how they view data. Some prefer a graphical display of family history in a genogram or pedigree; others prefer tabular representation. Sometimes, it is helpful to see which relatives have a specific condition or group of conditions. Other times, it is useful to view the medical history of each relative individually. One may wish to view the entire family history all at once, but often a filtered view of only certain conditions or relatives of interest is more relevant.

For the most part, EHR vendors have not yet developed the software to support management of family history and genetic data as discussed here. They are not completely to blame. Like any other business, they focus their resources on tools and innovations that their customers need or want. We, the consumers of EHR products, have not yet made this a high enough priority.

Ultimately, detailed and discrete family history, genetic, and environmental data can be fed into evidence-based algorithms, some already in existence, others yet to be developed, to drive clinical decision support that helps physicians at the point of care to deliver safer, more efficient, and more effective medical care. What stands in our way is the ability to gather and organize this data.

EHRs, despite the frustration of implementing and learning to use them, are the key to unlocking this potential.

Please consider asking your EHR vendor to develop the family history and genomic data tools necessary to enable truly individualized medicine.

Dr. Levy is with the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine at Johns Hopkins University, Baltimore.

Direct-to-consumer genetic testing, including personal genome analysis, has been available for several years and raises challenging questions of risk and benefit.

Among the potential harms are that clients may not fully understand their results, that abnormal results may lead to undue anxiety and/or inappropriate responses, and that normal results may lead to false reassurance ("Direct-to-consumer genetic analysis," Internal Medicine News, Oct. 1, 2008).

This is particularly concerning in the context of single-gene disorders such as hereditary breast and ovarian cancer syndrome, in which a mutation in either the BRCA1 or BRCA2 gene confers up to 50%-60% lifetime risk of female breast cancer and up to 40% lifetime risk of ovarian cancer – as well as increasing the risk of female primary papillary serous peritoneal carcinoma, male breast cancer, prostate cancer, and pancreatic cancer ("BRCA1 and BRCA2 Hereditary Breast and Ovarian Cancer," GeneReviews 1998 [Updated 2011 Jan. 20]).

The three mutations in these two genes that are most common among Ashkenazi Jews are now included in commercial direct-to-consumer (DTC) personal genome analysis, making it possible for consumers to screen themselves for these mutations.

In February, a DTC genetic testing company published a report suggesting that the potential harms of such testing may be overstated (Peer J. 2013;1:e8). They invited all 136 adult clients who had one of the three common Ashkenazi BRCA gene mutations and had elected to view their BRCA results to participate in an interview about their experience. A total of 32 agreed: 16 men and 16 women.

For 25 of them, this was a newly discovered mutation, while 7 previously knew that they carried a mutation. Thirty-one mutation-negative clients were interviewed as controls. The groups were identical for most demographics, but differed as expected with respect to cancer history.

Four of the 16 women with mutations had a personal history of breast and/or ovarian cancer, but only 1 of 18 mutation-negative women had breast cancer. There was a positive family history of breast/ovarian cancer for 72% of those with mutations and 48% of those without mutations.

Some of those with mutations were surprised at the news, because they had no significant family history of breast/ovarian cancer. But most were not surprised, because they either knew their mutation status or suspected it, based on family history or Ashkenazi ancestry.

However, no one reported feeling extremely upset, and only four people (12.5%) reported moderate upset feelings. Nine (28%) had self-limited disappointment or anxiety, and 17 (53%) had neutral feelings about their result. Even among the 25 for whom this was a new finding, 11 (44%) felt neutral. Only 1 of the 32 participants expressed regret about learning his result, citing the emotional cost of knowing he has a mutation and that he might pass it on to his children.

Overall, these anxiety results are mostly reassuring and are in line with other published studies of patients’ emotional response to receiving adverse test results (both genetic and nongenetic).

Half of the mutation-negative clients felt neutral about their result, and the other half expressed relief. Almost all of those who felt relieved had a positive family history of breast/ovarian cancer.

Fears of false assurance

The question of false reassurance for these individuals remains a concern. The DTC company required all clients to read a brief written passage about the meaning and limitations of these results prior to viewing them, including a statement that this test does not identify all genetic causes of breast and ovarian cancer.

The interview responses are summarized to indicate that the majority of mutation-negative participants understood that their risk of developing breast cancer was unchanged as a result of their negative testing.

However, because the testing did not involve full sequencing of all known genes associated with breast and ovarian cancer, it is quite possible that some of those who were negative in this study might in fact truly have a genetic cancer predisposition syndrome and be at substantially increased (but unrecognized) risk of breast, ovarian, and possibly other cancers.

Actions taken by clients after learning of a BRCA mutation are generally reassuring: 60% discussed their results with a physician (28% with their primary care physician). Most of the women obtained appropriate medical counseling and took steps to reduce their risks (with prophylactic surgery and/or increased cancer screening). Many people shared the information with family members, resulting in identification of additional relatives with the same mutations, who then sought appropriate screening and prevention.

Several of the mutation carriers, both those in the original study and their relatives, do not meet current criteria for BRCA gene testing or intensified screening and prophylaxis. Assuming such identification reduces future morbidity and mortality, these results argue in favor of more widespread screening for such mutations. But many more data need to be considered before implementing such a program.

Interestingly, a few clients who initially felt neutral about their abnormal test result reported anxiety only after learning more about the implications and management options, including mastectomy and oophorectomy.

This difference in emotional response could be related to the method of receiving information: reading it online in a comfortable environment when the client chooses to, vs. receiving it in a medical setting when the clinician is able to. Or it might reflect the physical and temporal separation of receiving an abnormal test result from the counseling and discussion of its management. More research is needed to better understand these possibilities.

Only one of the mutation-positive men sought medical advice, but most reported intent to pursue regular breast and prostate screening. However, it is unclear how they would act on that intent without involving a physician – leaving open the possibility that they may not fully understand the clinical significance of their positive result.

Also of concern, only seven (23%) of the mutation-negative clients discussed their result with a health care provider. Among those seven, most discussed it with their primary care physician, who generally did not know what to do with the information.

This represents another missed opportunity to identify patients with a positive family history who might benefit from additional genetic counseling and/or testing.

There are multiple limitations to this study. It is based on a very small and highly selected sample that is biased toward the desire to participate in research, to pursue genetic testing, to specifically know one’s BRCA mutation status, and to share one’s experience learning that status.

There is also the risk of observer bias (the authors are all employees of the company that performed the testing) and participant bias (the subjects are all clients of the company and knew that they were being interviewed by a representative of the company).

Autonomy vs. beneficence

Overall, the results suggest that, at least for a segment of the population, direct access to genetic data of potentially great medical significance may not be associated with as high a risk of emotional distress and inappropriate action as previously thought. But some individuals likely are at risk for unnecessary anxiety and/or misinterpretation of the results.

Central to the discussion is tension over how best to balance autonomy, in the form of a patient’s right to access personal medical information however she or he desires, against beneficence, in the form of minimizing harm.

There is some basic information that, ideally, should be available to all consumers (patients) for every test, whether genetic or not. This includes the meaning of a positive, negative, or inconclusive result; the limitations of conclusions that can be drawn from the result; the possibility of strong emotional reactions to both negative and positive results; and the possible courses of action to consider in response to the result.

Physician-ordered testing is probably the best way to accomplish those goals, but direct-to-consumer testing is here to stay and may be appropriate for some individuals. Hopefully, the companies offering such testing are providing accurate and understandable information to help their clients make the best and safest use of their results.

As primary care physicians, we should encourage best practices among ourselves and DTC companies, and remain available to provide advice, interpretation, and specialty referral to patients who may have started down the path of DTC testing and later decide to seek our professional assistance.

Dr. Levy is with the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine at Johns Hopkins University, Baltimore. He reports having no conflicts of interest.

Direct-to-consumer genetic testing, including personal genome analysis, has been available for several years and raises challenging questions of risk and benefit.

Among the potential harms are that clients may not fully understand their results, that abnormal results may lead to undue anxiety and/or inappropriate responses, and that normal results may lead to false reassurance ("Direct-to-consumer genetic analysis," Internal Medicine News, Oct. 1, 2008).

This is particularly concerning in the context of single-gene disorders such as hereditary breast and ovarian cancer syndrome, in which a mutation in either the BRCA1 or BRCA2 gene confers up to 50%-60% lifetime risk of female breast cancer and up to 40% lifetime risk of ovarian cancer – as well as increasing the risk of female primary papillary serous peritoneal carcinoma, male breast cancer, prostate cancer, and pancreatic cancer ("BRCA1 and BRCA2 Hereditary Breast and Ovarian Cancer," GeneReviews 1998 [Updated 2011 Jan. 20]).

The three mutations in these two genes that are most common among Ashkenazi Jews are now included in commercial direct-to-consumer (DTC) personal genome analysis, making it possible for consumers to screen themselves for these mutations.

In February, a DTC genetic testing company published a report suggesting that the potential harms of such testing may be overstated (Peer J. 2013;1:e8). They invited all 136 adult clients who had one of the three common Ashkenazi BRCA gene mutations and had elected to view their BRCA results to participate in an interview about their experience. A total of 32 agreed: 16 men and 16 women.

For 25 of them, this was a newly discovered mutation, while 7 previously knew that they carried a mutation. Thirty-one mutation-negative clients were interviewed as controls. The groups were identical for most demographics, but differed as expected with respect to cancer history.

Four of the 16 women with mutations had a personal history of breast and/or ovarian cancer, but only 1 of 18 mutation-negative women had breast cancer. There was a positive family history of breast/ovarian cancer for 72% of those with mutations and 48% of those without mutations.

Some of those with mutations were surprised at the news, because they had no significant family history of breast/ovarian cancer. But most were not surprised, because they either knew their mutation status or suspected it, based on family history or Ashkenazi ancestry.

However, no one reported feeling extremely upset, and only four people (12.5%) reported moderate upset feelings. Nine (28%) had self-limited disappointment or anxiety, and 17 (53%) had neutral feelings about their result. Even among the 25 for whom this was a new finding, 11 (44%) felt neutral. Only 1 of the 32 participants expressed regret about learning his result, citing the emotional cost of knowing he has a mutation and that he might pass it on to his children.

Overall, these anxiety results are mostly reassuring and are in line with other published studies of patients’ emotional response to receiving adverse test results (both genetic and nongenetic).

Half of the mutation-negative clients felt neutral about their result, and the other half expressed relief. Almost all of those who felt relieved had a positive family history of breast/ovarian cancer.

Fears of false assurance

The question of false reassurance for these individuals remains a concern. The DTC company required all clients to read a brief written passage about the meaning and limitations of these results prior to viewing them, including a statement that this test does not identify all genetic causes of breast and ovarian cancer.

The interview responses are summarized to indicate that the majority of mutation-negative participants understood that their risk of developing breast cancer was unchanged as a result of their negative testing.

However, because the testing did not involve full sequencing of all known genes associated with breast and ovarian cancer, it is quite possible that some of those who were negative in this study might in fact truly have a genetic cancer predisposition syndrome and be at substantially increased (but unrecognized) risk of breast, ovarian, and possibly other cancers.

Actions taken by clients after learning of a BRCA mutation are generally reassuring: 60% discussed their results with a physician (28% with their primary care physician). Most of the women obtained appropriate medical counseling and took steps to reduce their risks (with prophylactic surgery and/or increased cancer screening). Many people shared the information with family members, resulting in identification of additional relatives with the same mutations, who then sought appropriate screening and prevention.

Several of the mutation carriers, both those in the original study and their relatives, do not meet current criteria for BRCA gene testing or intensified screening and prophylaxis. Assuming such identification reduces future morbidity and mortality, these results argue in favor of more widespread screening for such mutations. But many more data need to be considered before implementing such a program.

Interestingly, a few clients who initially felt neutral about their abnormal test result reported anxiety only after learning more about the implications and management options, including mastectomy and oophorectomy.

This difference in emotional response could be related to the method of receiving information: reading it online in a comfortable environment when the client chooses to, vs. receiving it in a medical setting when the clinician is able to. Or it might reflect the physical and temporal separation of receiving an abnormal test result from the counseling and discussion of its management. More research is needed to better understand these possibilities.

Only one of the mutation-positive men sought medical advice, but most reported intent to pursue regular breast and prostate screening. However, it is unclear how they would act on that intent without involving a physician – leaving open the possibility that they may not fully understand the clinical significance of their positive result.

Also of concern, only seven (23%) of the mutation-negative clients discussed their result with a health care provider. Among those seven, most discussed it with their primary care physician, who generally did not know what to do with the information.

This represents another missed opportunity to identify patients with a positive family history who might benefit from additional genetic counseling and/or testing.

There are multiple limitations to this study. It is based on a very small and highly selected sample that is biased toward the desire to participate in research, to pursue genetic testing, to specifically know one’s BRCA mutation status, and to share one’s experience learning that status.

There is also the risk of observer bias (the authors are all employees of the company that performed the testing) and participant bias (the subjects are all clients of the company and knew that they were being interviewed by a representative of the company).

Autonomy vs. beneficence

Overall, the results suggest that, at least for a segment of the population, direct access to genetic data of potentially great medical significance may not be associated with as high a risk of emotional distress and inappropriate action as previously thought. But some individuals likely are at risk for unnecessary anxiety and/or misinterpretation of the results.

Central to the discussion is tension over how best to balance autonomy, in the form of a patient’s right to access personal medical information however she or he desires, against beneficence, in the form of minimizing harm.

There is some basic information that, ideally, should be available to all consumers (patients) for every test, whether genetic or not. This includes the meaning of a positive, negative, or inconclusive result; the limitations of conclusions that can be drawn from the result; the possibility of strong emotional reactions to both negative and positive results; and the possible courses of action to consider in response to the result.

Physician-ordered testing is probably the best way to accomplish those goals, but direct-to-consumer testing is here to stay and may be appropriate for some individuals. Hopefully, the companies offering such testing are providing accurate and understandable information to help their clients make the best and safest use of their results.

As primary care physicians, we should encourage best practices among ourselves and DTC companies, and remain available to provide advice, interpretation, and specialty referral to patients who may have started down the path of DTC testing and later decide to seek our professional assistance.

Dr. Levy is with the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine at Johns Hopkins University, Baltimore. He reports having no conflicts of interest.

Direct-to-consumer genetic testing, including personal genome analysis, has been available for several years and raises challenging questions of risk and benefit.

Among the potential harms are that clients may not fully understand their results, that abnormal results may lead to undue anxiety and/or inappropriate responses, and that normal results may lead to false reassurance ("Direct-to-consumer genetic analysis," Internal Medicine News, Oct. 1, 2008).

This is particularly concerning in the context of single-gene disorders such as hereditary breast and ovarian cancer syndrome, in which a mutation in either the BRCA1 or BRCA2 gene confers up to 50%-60% lifetime risk of female breast cancer and up to 40% lifetime risk of ovarian cancer – as well as increasing the risk of female primary papillary serous peritoneal carcinoma, male breast cancer, prostate cancer, and pancreatic cancer ("BRCA1 and BRCA2 Hereditary Breast and Ovarian Cancer," GeneReviews 1998 [Updated 2011 Jan. 20]).

The three mutations in these two genes that are most common among Ashkenazi Jews are now included in commercial direct-to-consumer (DTC) personal genome analysis, making it possible for consumers to screen themselves for these mutations.

In February, a DTC genetic testing company published a report suggesting that the potential harms of such testing may be overstated (Peer J. 2013;1:e8). They invited all 136 adult clients who had one of the three common Ashkenazi BRCA gene mutations and had elected to view their BRCA results to participate in an interview about their experience. A total of 32 agreed: 16 men and 16 women.

For 25 of them, this was a newly discovered mutation, while 7 previously knew that they carried a mutation. Thirty-one mutation-negative clients were interviewed as controls. The groups were identical for most demographics, but differed as expected with respect to cancer history.

Four of the 16 women with mutations had a personal history of breast and/or ovarian cancer, but only 1 of 18 mutation-negative women had breast cancer. There was a positive family history of breast/ovarian cancer for 72% of those with mutations and 48% of those without mutations.

Some of those with mutations were surprised at the news, because they had no significant family history of breast/ovarian cancer. But most were not surprised, because they either knew their mutation status or suspected it, based on family history or Ashkenazi ancestry.

However, no one reported feeling extremely upset, and only four people (12.5%) reported moderate upset feelings. Nine (28%) had self-limited disappointment or anxiety, and 17 (53%) had neutral feelings about their result. Even among the 25 for whom this was a new finding, 11 (44%) felt neutral. Only 1 of the 32 participants expressed regret about learning his result, citing the emotional cost of knowing he has a mutation and that he might pass it on to his children.

Overall, these anxiety results are mostly reassuring and are in line with other published studies of patients’ emotional response to receiving adverse test results (both genetic and nongenetic).

Half of the mutation-negative clients felt neutral about their result, and the other half expressed relief. Almost all of those who felt relieved had a positive family history of breast/ovarian cancer.

Fears of false assurance

The question of false reassurance for these individuals remains a concern. The DTC company required all clients to read a brief written passage about the meaning and limitations of these results prior to viewing them, including a statement that this test does not identify all genetic causes of breast and ovarian cancer.

The interview responses are summarized to indicate that the majority of mutation-negative participants understood that their risk of developing breast cancer was unchanged as a result of their negative testing.

However, because the testing did not involve full sequencing of all known genes associated with breast and ovarian cancer, it is quite possible that some of those who were negative in this study might in fact truly have a genetic cancer predisposition syndrome and be at substantially increased (but unrecognized) risk of breast, ovarian, and possibly other cancers.

Actions taken by clients after learning of a BRCA mutation are generally reassuring: 60% discussed their results with a physician (28% with their primary care physician). Most of the women obtained appropriate medical counseling and took steps to reduce their risks (with prophylactic surgery and/or increased cancer screening). Many people shared the information with family members, resulting in identification of additional relatives with the same mutations, who then sought appropriate screening and prevention.

Several of the mutation carriers, both those in the original study and their relatives, do not meet current criteria for BRCA gene testing or intensified screening and prophylaxis. Assuming such identification reduces future morbidity and mortality, these results argue in favor of more widespread screening for such mutations. But many more data need to be considered before implementing such a program.

Interestingly, a few clients who initially felt neutral about their abnormal test result reported anxiety only after learning more about the implications and management options, including mastectomy and oophorectomy.

This difference in emotional response could be related to the method of receiving information: reading it online in a comfortable environment when the client chooses to, vs. receiving it in a medical setting when the clinician is able to. Or it might reflect the physical and temporal separation of receiving an abnormal test result from the counseling and discussion of its management. More research is needed to better understand these possibilities.

Only one of the mutation-positive men sought medical advice, but most reported intent to pursue regular breast and prostate screening. However, it is unclear how they would act on that intent without involving a physician – leaving open the possibility that they may not fully understand the clinical significance of their positive result.

Also of concern, only seven (23%) of the mutation-negative clients discussed their result with a health care provider. Among those seven, most discussed it with their primary care physician, who generally did not know what to do with the information.

This represents another missed opportunity to identify patients with a positive family history who might benefit from additional genetic counseling and/or testing.

There are multiple limitations to this study. It is based on a very small and highly selected sample that is biased toward the desire to participate in research, to pursue genetic testing, to specifically know one’s BRCA mutation status, and to share one’s experience learning that status.

There is also the risk of observer bias (the authors are all employees of the company that performed the testing) and participant bias (the subjects are all clients of the company and knew that they were being interviewed by a representative of the company).

Autonomy vs. beneficence

Overall, the results suggest that, at least for a segment of the population, direct access to genetic data of potentially great medical significance may not be associated with as high a risk of emotional distress and inappropriate action as previously thought. But some individuals likely are at risk for unnecessary anxiety and/or misinterpretation of the results.

Central to the discussion is tension over how best to balance autonomy, in the form of a patient’s right to access personal medical information however she or he desires, against beneficence, in the form of minimizing harm.

There is some basic information that, ideally, should be available to all consumers (patients) for every test, whether genetic or not. This includes the meaning of a positive, negative, or inconclusive result; the limitations of conclusions that can be drawn from the result; the possibility of strong emotional reactions to both negative and positive results; and the possible courses of action to consider in response to the result.

Physician-ordered testing is probably the best way to accomplish those goals, but direct-to-consumer testing is here to stay and may be appropriate for some individuals. Hopefully, the companies offering such testing are providing accurate and understandable information to help their clients make the best and safest use of their results.

As primary care physicians, we should encourage best practices among ourselves and DTC companies, and remain available to provide advice, interpretation, and specialty referral to patients who may have started down the path of DTC testing and later decide to seek our professional assistance.

Dr. Levy is with the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine at Johns Hopkins University, Baltimore. He reports having no conflicts of interest.

Genome-wide association studies continue to identify variants that affect the likelihood of developing one or more of several hundred common health conditions, but the risk associated with any one variant is generally quite small. To make this information clinically relevant, the cumulative effects of multiple variants need to be analyzed within the context of known environmental factors. As the cost of genotyping and sequencing falls, this problem is being addressed by including genetic variants in large epidemiologic trials.

One example is the recent study by Harvard investigators on the interaction between genetic predisposition to obesity and intake of sugar-sweetened colas, non-cola soft drinks, and fruit drinks (N. Engl. J. Med. 2012 [doi:10.1056/NEJMoa1203039).

The authors analyzed data from three prospective cohort studies: the Nurses’ Health Study (NHS) of female registered nurses, the Health Professionals Follow-Up Study (HPFS) of male health professionals, and the Women’s Genome Health Study (WGHS) of female health professionals. A total of 11,357 initially healthy NHS and HPFS participants had genotype data available and were used for primary analysis. From WGHS, 21,740 initially healthy women were used for a replication set.

Intake of sugar-sweetened beverages (SSBs) was measured by periodic questionnaire and divided into categories of less than 1 serving per month, 1-4 servings per month, 2-6 servings per week, and 1 or more serving per day. Height, weight, physical activity, and other dietary data also were obtained by periodic questionnaire.

All 32 single nucleotide polymorphisms (SNPs) currently known to be associated with obesity were included. A combined risk score was calculated for each participant, using a weighted value for each SNP according to its relative effect size. Since there are two copies of each SNP, the potential range of scores was 0-64; actual scores varied from 13-43, with a mean of 29. Effects on BMI were determined by 10-point increments in risk score.

As expected, there was a significant correlation between greater SSB intake and higher BMI, but this effect was much more pronounced among participants with higher genetic risk scores. In those with the lowest SSB intake, a 10-point increase in genetic risk corresponded with 1.00 kg/m² increase in BMI in the pooled NHS/HPFS studies and a 1.46 kg/m² increase in BMI in the WGHS.

At the other end of the spectrum, among those who consumed 1 or more SSBs daily, a 10-point increase in genetic risk corresponded to a 1.85 kg/m² increase in BMI in the NHS/HPFS studies and a 2.43 kg/m² increase in WGHS.

By contrast, intake of artificially sweetened beverages had no effect on the association between genetic risk score and BMI. In addition, while those with the greatest intake of SSB also had higher total calorie intake and lower physical activity, alcohol intake, and overall diet quality, statistical adjustment for these factors had no effect on the observed association between SSB intake and genetic predisposition to obesity. Furthermore, while the total genetic risk score showed strong statistical significance, very few of the individual SNPs demonstrated a significant effect by themselves, and exclusion of variation in the single strongest marker (in the FTO gene) did not change the overall findings.

The authors also looked at the incidence of obesity according to genetic risk score and SSB intake. Pooling data from the three prospective studies, the relative risk of new-onset obesity per increment of 10 genetic risk points was 1.35 for SSB intake less than 1 per month, 1.59 and 1.56 for the intermediate levels of SSB intake, and a striking 3.35 among participants consuming 1 or more SSB per day.

There are still plenty of limitations in genetic and obesity research. The genetic risk score calculated in this study accounted for different effect sizes of each variant, but we don’t yet know if there are interactions between individual variants that might require further adjustment. Most of the genetic variability in BMI is not accounted for by the currently known SNPs, so additional genetic factors will need to be considered in the future. SSB intake may or may not be the primary cause of the observed associations; other unmeasured lifestyle or environmental factors are likely also important. And understanding actual pathophysiologic mechanisms of interaction between genetic variants and dietary factors is still in its infancy. Nonetheless, this study provides compelling evidence that such interactions do exist and are clinically relevant.

For the majority of patients, these findings are unlikely to alter the general weight management advice to maximize exercise and limit total calories, especially sugars, simple starches and fats. The most immediate clinical utility may lie in counseling nonobese patients about their future risk. Family history remains an inexpensive test for overall genetic predisposition, and for those with a strong family history of obesity it may be especially important to limit SSB intake (and perhaps other sources of sugar and simple starch, too). Selective testing of a panel of obesity-linked genetic variants might be of value for a subset of patients who need extra motivation to make the appropriate lifestyle changes.

And perhaps elucidation of the actual mechanism of this gene-environment interaction will lead to innovative and more effective therapies for obesity in the future.

Dr. Levy is with the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine at Johns Hopkins University, Baltimore. He reports having no conflicts of interest. E-mail him at imnews@elsevier.com.

Genome-wide association studies continue to identify variants that affect the likelihood of developing one or more of several hundred common health conditions, but the risk associated with any one variant is generally quite small. To make this information clinically relevant, the cumulative effects of multiple variants need to be analyzed within the context of known environmental factors. As the cost of genotyping and sequencing falls, this problem is being addressed by including genetic variants in large epidemiologic trials.

One example is the recent study by Harvard investigators on the interaction between genetic predisposition to obesity and intake of sugar-sweetened colas, non-cola soft drinks, and fruit drinks (N. Engl. J. Med. 2012 [doi:10.1056/NEJMoa1203039).

The authors analyzed data from three prospective cohort studies: the Nurses’ Health Study (NHS) of female registered nurses, the Health Professionals Follow-Up Study (HPFS) of male health professionals, and the Women’s Genome Health Study (WGHS) of female health professionals. A total of 11,357 initially healthy NHS and HPFS participants had genotype data available and were used for primary analysis. From WGHS, 21,740 initially healthy women were used for a replication set.

Intake of sugar-sweetened beverages (SSBs) was measured by periodic questionnaire and divided into categories of less than 1 serving per month, 1-4 servings per month, 2-6 servings per week, and 1 or more serving per day. Height, weight, physical activity, and other dietary data also were obtained by periodic questionnaire.

All 32 single nucleotide polymorphisms (SNPs) currently known to be associated with obesity were included. A combined risk score was calculated for each participant, using a weighted value for each SNP according to its relative effect size. Since there are two copies of each SNP, the potential range of scores was 0-64; actual scores varied from 13-43, with a mean of 29. Effects on BMI were determined by 10-point increments in risk score.

As expected, there was a significant correlation between greater SSB intake and higher BMI, but this effect was much more pronounced among participants with higher genetic risk scores. In those with the lowest SSB intake, a 10-point increase in genetic risk corresponded with 1.00 kg/m² increase in BMI in the pooled NHS/HPFS studies and a 1.46 kg/m² increase in BMI in the WGHS.

At the other end of the spectrum, among those who consumed 1 or more SSBs daily, a 10-point increase in genetic risk corresponded to a 1.85 kg/m² increase in BMI in the NHS/HPFS studies and a 2.43 kg/m² increase in WGHS.

By contrast, intake of artificially sweetened beverages had no effect on the association between genetic risk score and BMI. In addition, while those with the greatest intake of SSB also had higher total calorie intake and lower physical activity, alcohol intake, and overall diet quality, statistical adjustment for these factors had no effect on the observed association between SSB intake and genetic predisposition to obesity. Furthermore, while the total genetic risk score showed strong statistical significance, very few of the individual SNPs demonstrated a significant effect by themselves, and exclusion of variation in the single strongest marker (in the FTO gene) did not change the overall findings.

The authors also looked at the incidence of obesity according to genetic risk score and SSB intake. Pooling data from the three prospective studies, the relative risk of new-onset obesity per increment of 10 genetic risk points was 1.35 for SSB intake less than 1 per month, 1.59 and 1.56 for the intermediate levels of SSB intake, and a striking 3.35 among participants consuming 1 or more SSB per day.

There are still plenty of limitations in genetic and obesity research. The genetic risk score calculated in this study accounted for different effect sizes of each variant, but we don’t yet know if there are interactions between individual variants that might require further adjustment. Most of the genetic variability in BMI is not accounted for by the currently known SNPs, so additional genetic factors will need to be considered in the future. SSB intake may or may not be the primary cause of the observed associations; other unmeasured lifestyle or environmental factors are likely also important. And understanding actual pathophysiologic mechanisms of interaction between genetic variants and dietary factors is still in its infancy. Nonetheless, this study provides compelling evidence that such interactions do exist and are clinically relevant.

For the majority of patients, these findings are unlikely to alter the general weight management advice to maximize exercise and limit total calories, especially sugars, simple starches and fats. The most immediate clinical utility may lie in counseling nonobese patients about their future risk. Family history remains an inexpensive test for overall genetic predisposition, and for those with a strong family history of obesity it may be especially important to limit SSB intake (and perhaps other sources of sugar and simple starch, too). Selective testing of a panel of obesity-linked genetic variants might be of value for a subset of patients who need extra motivation to make the appropriate lifestyle changes.

And perhaps elucidation of the actual mechanism of this gene-environment interaction will lead to innovative and more effective therapies for obesity in the future.

Dr. Levy is with the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine at Johns Hopkins University, Baltimore. He reports having no conflicts of interest. E-mail him at imnews@elsevier.com.

Genome-wide association studies continue to identify variants that affect the likelihood of developing one or more of several hundred common health conditions, but the risk associated with any one variant is generally quite small. To make this information clinically relevant, the cumulative effects of multiple variants need to be analyzed within the context of known environmental factors. As the cost of genotyping and sequencing falls, this problem is being addressed by including genetic variants in large epidemiologic trials.

One example is the recent study by Harvard investigators on the interaction between genetic predisposition to obesity and intake of sugar-sweetened colas, non-cola soft drinks, and fruit drinks (N. Engl. J. Med. 2012 [doi:10.1056/NEJMoa1203039).

The authors analyzed data from three prospective cohort studies: the Nurses’ Health Study (NHS) of female registered nurses, the Health Professionals Follow-Up Study (HPFS) of male health professionals, and the Women’s Genome Health Study (WGHS) of female health professionals. A total of 11,357 initially healthy NHS and HPFS participants had genotype data available and were used for primary analysis. From WGHS, 21,740 initially healthy women were used for a replication set.

Intake of sugar-sweetened beverages (SSBs) was measured by periodic questionnaire and divided into categories of less than 1 serving per month, 1-4 servings per month, 2-6 servings per week, and 1 or more serving per day. Height, weight, physical activity, and other dietary data also were obtained by periodic questionnaire.

All 32 single nucleotide polymorphisms (SNPs) currently known to be associated with obesity were included. A combined risk score was calculated for each participant, using a weighted value for each SNP according to its relative effect size. Since there are two copies of each SNP, the potential range of scores was 0-64; actual scores varied from 13-43, with a mean of 29. Effects on BMI were determined by 10-point increments in risk score.

As expected, there was a significant correlation between greater SSB intake and higher BMI, but this effect was much more pronounced among participants with higher genetic risk scores. In those with the lowest SSB intake, a 10-point increase in genetic risk corresponded with 1.00 kg/m² increase in BMI in the pooled NHS/HPFS studies and a 1.46 kg/m² increase in BMI in the WGHS.

At the other end of the spectrum, among those who consumed 1 or more SSBs daily, a 10-point increase in genetic risk corresponded to a 1.85 kg/m² increase in BMI in the NHS/HPFS studies and a 2.43 kg/m² increase in WGHS.

By contrast, intake of artificially sweetened beverages had no effect on the association between genetic risk score and BMI. In addition, while those with the greatest intake of SSB also had higher total calorie intake and lower physical activity, alcohol intake, and overall diet quality, statistical adjustment for these factors had no effect on the observed association between SSB intake and genetic predisposition to obesity. Furthermore, while the total genetic risk score showed strong statistical significance, very few of the individual SNPs demonstrated a significant effect by themselves, and exclusion of variation in the single strongest marker (in the FTO gene) did not change the overall findings.

The authors also looked at the incidence of obesity according to genetic risk score and SSB intake. Pooling data from the three prospective studies, the relative risk of new-onset obesity per increment of 10 genetic risk points was 1.35 for SSB intake less than 1 per month, 1.59 and 1.56 for the intermediate levels of SSB intake, and a striking 3.35 among participants consuming 1 or more SSB per day.

There are still plenty of limitations in genetic and obesity research. The genetic risk score calculated in this study accounted for different effect sizes of each variant, but we don’t yet know if there are interactions between individual variants that might require further adjustment. Most of the genetic variability in BMI is not accounted for by the currently known SNPs, so additional genetic factors will need to be considered in the future. SSB intake may or may not be the primary cause of the observed associations; other unmeasured lifestyle or environmental factors are likely also important. And understanding actual pathophysiologic mechanisms of interaction between genetic variants and dietary factors is still in its infancy. Nonetheless, this study provides compelling evidence that such interactions do exist and are clinically relevant.

For the majority of patients, these findings are unlikely to alter the general weight management advice to maximize exercise and limit total calories, especially sugars, simple starches and fats. The most immediate clinical utility may lie in counseling nonobese patients about their future risk. Family history remains an inexpensive test for overall genetic predisposition, and for those with a strong family history of obesity it may be especially important to limit SSB intake (and perhaps other sources of sugar and simple starch, too). Selective testing of a panel of obesity-linked genetic variants might be of value for a subset of patients who need extra motivation to make the appropriate lifestyle changes.

And perhaps elucidation of the actual mechanism of this gene-environment interaction will lead to innovative and more effective therapies for obesity in the future.

Dr. Levy is with the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine at Johns Hopkins University, Baltimore. He reports having no conflicts of interest. E-mail him at imnews@elsevier.com.

Tamoxifen is a well-established adjuvant therapy for estrogen and/or progesterone receptor (HR)–positive breast cancer, and is the only hormonal treatment available to premenopausal patients. Tamoxifen itself is considered a relatively clinically inactive prodrug. Its most abundant metabolite, endoxifen, has much greater affinity for the estrogen receptor than tamoxifen, and is thought to be the primary compound responsible for clinical benefit.

Bioactivation of tamoxifen to endoxifen depends in large part upon the function of the P450 enzyme CYP2D6. This enzyme is known to have significant genetic variability, and can also be inhibited by several medications, especially fluoxetine, paroxetine and bupropion. There is ample evidence that diminished activity of CYP2D6 (whether caused by pharmacogenetic variation or use of an inhibiting drug) correlates with reduced circulating levels of endoxifen in humans treated with tamoxifen.

In theory, patients with normal CYP2D6 activity (extensive metabolizers, EM) should have a better clinical response to tamoxifen than those with moderately reduced (intermediate metabolizers, IM) or severely reduced/absent CYP2D6 activity (poor metabolizers, PM). In addition, strong inhibitors of CYP2D6 would also reduce the efficacy of tamoxifen.

However, clinical outcome studies have been mixed, with some confirming a reduced recurrence-free interval in IM and PM patients, while others showed no correlation with CYP2D6 phenotype. Most of these are small, retrospective studies that suffer from multiple limitations. The result is uncertainty as to whether pharmacogenetic testing of CYP2D6 prior to prescribing tamoxifen, and/or drug or dose selection based upon CYP2D6 phenotype, is clinically appropriate.

A review of 17 independent retrospective studies of CYP2D6 and tamoxifen response evaluated several possible factors that might explain some of this discrepancy (Oncologist 2012;17:620-30). Some of the more significant cofounders were misclassification of CYP2D6 activity and failure to account for tamoxifen adherence or use of concurrent treatments.

• Hormone receptor classification. Some of the studies included patients whose cancers were HR negative. Since tamoxifen is only effective in HR-positive cancers, inclusion of HR-negative patients could easily confound the data. However, there was no obvious correlation between inclusion of HR-negative cases and finding or not finding a CYP2D6 effect on outcome.

• Menopausal status. In postmenopausal women, tamoxifen and one of its many other non-endoxifen metabolites saturate more than 99.9% of estrogen receptors, as opposed to only 90%–95% saturation by these two molecules in premenopausal women. This leads to speculation that endoxifen availability (and thus CYP2D6 activity) is only relevant in premenopausal women. Again, however, this review found no obvious influence of menopausal status on the results of these studies of CYP2D6 and tamoxifen response.

• Tamoxifen combination therapy. Coadministration of chemotherapy or an aromatase inhibitor with tamoxifen could mask the variability in tamoxifen response attributable to CYP2D6 variation. Indeed, there was a clear trend toward not finding a CYP2D6 effect in studies with fewer patients receiving tamoxifen monotherapy and positive CYP2D6 effects in studies with more monotherapy patients.

• Genotyping comprehensiveness. The normal (EM) version of the CYP2D6 gene cannot actually be positively identified. Rather, it is assumed to be present when a known variant conferring IM or PM status is not found. There are several dozen known genetic variants in CYP2D6, but most of the published studies only looked at one or a few of them. The result can be misclassification of IM or PM patients as EM, thus masking possible pharmacogenetic effects. This review showed a trend toward positive CYP2D6 effects in studies that used more extensive genotyping.

• CYP2D6 inhibitor coadministration. Studies that accounted for inhibitor use were more likely to indicate CYP2D6 effects than those that did not. Furthermore, two studies that did not support a CYP2D6 association despite accounting for inhibitor use appeared to overestimate the effects of some relatively weak or noninhibitory drugs, which might have confounded the results.

• Tamoxifen adherence. Tamoxifen can be a difficult drug to take, especially because of hot flashes, and discontinuation is more common in CYP2D6 EM patients (who have higher circulating endoxifen levels). On average, just under 80% of patients are adherent at 1 year, with further declines to as low as 50% by the fourth year. Only one of the studies adjusted their analysis for estimated tamoxifen adherence, which resulted in a stronger correlation between CYP2D6 activity and tamoxifen benefit.

Two recent reports from large randomized trials – ATAC (Arimidex, Tamoxifen, Alone or in Combination) and BIG (Breast International Group 1-98) – failed to identify any significant difference in clinical outcome among EM, IM, and PM patients treated with tamoxifen (J. Natl. Cancer Inst. 2012;104:452-60; J. Natl. Cancer Inst. 2012;104:441-51). The accompanying editorial asserts that the question of CYP2D6 clinical utility for tamoxifen therapy "has likely been laid to rest" (J. Natl. Cancer Inst. 2012;104:427-8).

However, a more thorough consideration of the details of these two studies still leaves open the possibility that CYP2D6 genotyping is clinically relevant.

Both studies involved the retrospective analysis of a subgroup from a larger prospective randomized trial. Their strengths include large sample sizes and analysis restricted to tamoxifen monotherapy. But there are multiple limiting factors.

In the ATAC study, only 92.5% of the tumors were HR positive. All tumors in BIG were HR positive.

In both studies, the genetic analyses were performed on a convenience sample of tumor blocks from which adequate DNA could be extracted, and both had statistical anomalies that might suggest sampling bias. In ATAC, there were significant differences between the genotyped subgroup and the overall study population with respect to frequency of surgical, radiation, and chemotherapy treatment prior to adjuvant tamoxifen therapy. In BIG, the frequencies of genotyped subjects with 0, 1, or 2 copies of one of the most common variants (CYP2D6*4) were inconsistent with the expected proportions in a random population sample.

The genetic analyses looked at only six or nine known CYP2D6 variants.

Only ATAC adjusted for potent and intermediate CYP2D6 inhibitors, although those drugs did not appear to affect the results.

Neither study included significant numbers of premenopausal patients, so the results may apply only to postmenopausal women.

Despite the relatively large populations studied, modest effects of CYP2D6 variation may have been missed. ATAC was estimated to be able to identify a 35% or greater variation associated with PM status; BIG was powered to detect only a 55% or greater effect size.

Neither study measured actual endoxifen levels or had any other way of assessing adherence to therapy.

In summary, it is certainly possible that CYP2D6 variation is not valuable in predicting clinical response to tamoxifen therapy. Perhaps endoxifen is not the most biologically relevant metabolite, or perhaps the effect of reduced CYP2D6 activity is small enough that it does not account for a significant proportion of tamoxifen failures. But it is also possible that CYP2D6 activity is clinically important for at least a subset of tamoxifen-treated patients. A more definitive answer awaits results from truly prospective trials that control for multiple confounding factors and measure actual endoxifen levels.

For now, in the absence of better data to address this issue, it does not seem appropriate to order CYP2D6 genotyping prior to starting tamoxifen therapy. Likewise, even if the CYP2D6 genotype is known, it should not influence the choice between tamoxifen and an aromatase inhibitor, or the dosing of tamoxifen. However, if a patient receiving tamoxifen requires medication for depression and/or hot flashes, it certainly seems reasonable to start with something other than fluoxetine, paroxetine, or bupropion, just in case those potent CYP2D6 inhibitors do turn out to reduce clinical efficacy.

Dr. Levy is at the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine of Johns Hopkins University, Baltimore. He reports having no conflicts of interest.

Tamoxifen is a well-established adjuvant therapy for estrogen and/or progesterone receptor (HR)–positive breast cancer, and is the only hormonal treatment available to premenopausal patients. Tamoxifen itself is considered a relatively clinically inactive prodrug. Its most abundant metabolite, endoxifen, has much greater affinity for the estrogen receptor than tamoxifen, and is thought to be the primary compound responsible for clinical benefit.

Bioactivation of tamoxifen to endoxifen depends in large part upon the function of the P450 enzyme CYP2D6. This enzyme is known to have significant genetic variability, and can also be inhibited by several medications, especially fluoxetine, paroxetine and bupropion. There is ample evidence that diminished activity of CYP2D6 (whether caused by pharmacogenetic variation or use of an inhibiting drug) correlates with reduced circulating levels of endoxifen in humans treated with tamoxifen.

In theory, patients with normal CYP2D6 activity (extensive metabolizers, EM) should have a better clinical response to tamoxifen than those with moderately reduced (intermediate metabolizers, IM) or severely reduced/absent CYP2D6 activity (poor metabolizers, PM). In addition, strong inhibitors of CYP2D6 would also reduce the efficacy of tamoxifen.

However, clinical outcome studies have been mixed, with some confirming a reduced recurrence-free interval in IM and PM patients, while others showed no correlation with CYP2D6 phenotype. Most of these are small, retrospective studies that suffer from multiple limitations. The result is uncertainty as to whether pharmacogenetic testing of CYP2D6 prior to prescribing tamoxifen, and/or drug or dose selection based upon CYP2D6 phenotype, is clinically appropriate.

A review of 17 independent retrospective studies of CYP2D6 and tamoxifen response evaluated several possible factors that might explain some of this discrepancy (Oncologist 2012;17:620-30). Some of the more significant cofounders were misclassification of CYP2D6 activity and failure to account for tamoxifen adherence or use of concurrent treatments.

• Hormone receptor classification. Some of the studies included patients whose cancers were HR negative. Since tamoxifen is only effective in HR-positive cancers, inclusion of HR-negative patients could easily confound the data. However, there was no obvious correlation between inclusion of HR-negative cases and finding or not finding a CYP2D6 effect on outcome.

• Menopausal status. In postmenopausal women, tamoxifen and one of its many other non-endoxifen metabolites saturate more than 99.9% of estrogen receptors, as opposed to only 90%–95% saturation by these two molecules in premenopausal women. This leads to speculation that endoxifen availability (and thus CYP2D6 activity) is only relevant in premenopausal women. Again, however, this review found no obvious influence of menopausal status on the results of these studies of CYP2D6 and tamoxifen response.

• Tamoxifen combination therapy. Coadministration of chemotherapy or an aromatase inhibitor with tamoxifen could mask the variability in tamoxifen response attributable to CYP2D6 variation. Indeed, there was a clear trend toward not finding a CYP2D6 effect in studies with fewer patients receiving tamoxifen monotherapy and positive CYP2D6 effects in studies with more monotherapy patients.

• Genotyping comprehensiveness. The normal (EM) version of the CYP2D6 gene cannot actually be positively identified. Rather, it is assumed to be present when a known variant conferring IM or PM status is not found. There are several dozen known genetic variants in CYP2D6, but most of the published studies only looked at one or a few of them. The result can be misclassification of IM or PM patients as EM, thus masking possible pharmacogenetic effects. This review showed a trend toward positive CYP2D6 effects in studies that used more extensive genotyping.

• CYP2D6 inhibitor coadministration. Studies that accounted for inhibitor use were more likely to indicate CYP2D6 effects than those that did not. Furthermore, two studies that did not support a CYP2D6 association despite accounting for inhibitor use appeared to overestimate the effects of some relatively weak or noninhibitory drugs, which might have confounded the results.

• Tamoxifen adherence. Tamoxifen can be a difficult drug to take, especially because of hot flashes, and discontinuation is more common in CYP2D6 EM patients (who have higher circulating endoxifen levels). On average, just under 80% of patients are adherent at 1 year, with further declines to as low as 50% by the fourth year. Only one of the studies adjusted their analysis for estimated tamoxifen adherence, which resulted in a stronger correlation between CYP2D6 activity and tamoxifen benefit.

Two recent reports from large randomized trials – ATAC (Arimidex, Tamoxifen, Alone or in Combination) and BIG (Breast International Group 1-98) – failed to identify any significant difference in clinical outcome among EM, IM, and PM patients treated with tamoxifen (J. Natl. Cancer Inst. 2012;104:452-60; J. Natl. Cancer Inst. 2012;104:441-51). The accompanying editorial asserts that the question of CYP2D6 clinical utility for tamoxifen therapy "has likely been laid to rest" (J. Natl. Cancer Inst. 2012;104:427-8).

However, a more thorough consideration of the details of these two studies still leaves open the possibility that CYP2D6 genotyping is clinically relevant.

Both studies involved the retrospective analysis of a subgroup from a larger prospective randomized trial. Their strengths include large sample sizes and analysis restricted to tamoxifen monotherapy. But there are multiple limiting factors.

In the ATAC study, only 92.5% of the tumors were HR positive. All tumors in BIG were HR positive.

In both studies, the genetic analyses were performed on a convenience sample of tumor blocks from which adequate DNA could be extracted, and both had statistical anomalies that might suggest sampling bias. In ATAC, there were significant differences between the genotyped subgroup and the overall study population with respect to frequency of surgical, radiation, and chemotherapy treatment prior to adjuvant tamoxifen therapy. In BIG, the frequencies of genotyped subjects with 0, 1, or 2 copies of one of the most common variants (CYP2D6*4) were inconsistent with the expected proportions in a random population sample.

The genetic analyses looked at only six or nine known CYP2D6 variants.

Only ATAC adjusted for potent and intermediate CYP2D6 inhibitors, although those drugs did not appear to affect the results.

Neither study included significant numbers of premenopausal patients, so the results may apply only to postmenopausal women.

Despite the relatively large populations studied, modest effects of CYP2D6 variation may have been missed. ATAC was estimated to be able to identify a 35% or greater variation associated with PM status; BIG was powered to detect only a 55% or greater effect size.

Neither study measured actual endoxifen levels or had any other way of assessing adherence to therapy.

In summary, it is certainly possible that CYP2D6 variation is not valuable in predicting clinical response to tamoxifen therapy. Perhaps endoxifen is not the most biologically relevant metabolite, or perhaps the effect of reduced CYP2D6 activity is small enough that it does not account for a significant proportion of tamoxifen failures. But it is also possible that CYP2D6 activity is clinically important for at least a subset of tamoxifen-treated patients. A more definitive answer awaits results from truly prospective trials that control for multiple confounding factors and measure actual endoxifen levels.

For now, in the absence of better data to address this issue, it does not seem appropriate to order CYP2D6 genotyping prior to starting tamoxifen therapy. Likewise, even if the CYP2D6 genotype is known, it should not influence the choice between tamoxifen and an aromatase inhibitor, or the dosing of tamoxifen. However, if a patient receiving tamoxifen requires medication for depression and/or hot flashes, it certainly seems reasonable to start with something other than fluoxetine, paroxetine, or bupropion, just in case those potent CYP2D6 inhibitors do turn out to reduce clinical efficacy.

Dr. Levy is at the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine of Johns Hopkins University, Baltimore. He reports having no conflicts of interest.

Tamoxifen is a well-established adjuvant therapy for estrogen and/or progesterone receptor (HR)–positive breast cancer, and is the only hormonal treatment available to premenopausal patients. Tamoxifen itself is considered a relatively clinically inactive prodrug. Its most abundant metabolite, endoxifen, has much greater affinity for the estrogen receptor than tamoxifen, and is thought to be the primary compound responsible for clinical benefit.

Bioactivation of tamoxifen to endoxifen depends in large part upon the function of the P450 enzyme CYP2D6. This enzyme is known to have significant genetic variability, and can also be inhibited by several medications, especially fluoxetine, paroxetine and bupropion. There is ample evidence that diminished activity of CYP2D6 (whether caused by pharmacogenetic variation or use of an inhibiting drug) correlates with reduced circulating levels of endoxifen in humans treated with tamoxifen.

In theory, patients with normal CYP2D6 activity (extensive metabolizers, EM) should have a better clinical response to tamoxifen than those with moderately reduced (intermediate metabolizers, IM) or severely reduced/absent CYP2D6 activity (poor metabolizers, PM). In addition, strong inhibitors of CYP2D6 would also reduce the efficacy of tamoxifen.

However, clinical outcome studies have been mixed, with some confirming a reduced recurrence-free interval in IM and PM patients, while others showed no correlation with CYP2D6 phenotype. Most of these are small, retrospective studies that suffer from multiple limitations. The result is uncertainty as to whether pharmacogenetic testing of CYP2D6 prior to prescribing tamoxifen, and/or drug or dose selection based upon CYP2D6 phenotype, is clinically appropriate.

A review of 17 independent retrospective studies of CYP2D6 and tamoxifen response evaluated several possible factors that might explain some of this discrepancy (Oncologist 2012;17:620-30). Some of the more significant cofounders were misclassification of CYP2D6 activity and failure to account for tamoxifen adherence or use of concurrent treatments.

• Hormone receptor classification. Some of the studies included patients whose cancers were HR negative. Since tamoxifen is only effective in HR-positive cancers, inclusion of HR-negative patients could easily confound the data. However, there was no obvious correlation between inclusion of HR-negative cases and finding or not finding a CYP2D6 effect on outcome.

• Menopausal status. In postmenopausal women, tamoxifen and one of its many other non-endoxifen metabolites saturate more than 99.9% of estrogen receptors, as opposed to only 90%–95% saturation by these two molecules in premenopausal women. This leads to speculation that endoxifen availability (and thus CYP2D6 activity) is only relevant in premenopausal women. Again, however, this review found no obvious influence of menopausal status on the results of these studies of CYP2D6 and tamoxifen response.

• Tamoxifen combination therapy. Coadministration of chemotherapy or an aromatase inhibitor with tamoxifen could mask the variability in tamoxifen response attributable to CYP2D6 variation. Indeed, there was a clear trend toward not finding a CYP2D6 effect in studies with fewer patients receiving tamoxifen monotherapy and positive CYP2D6 effects in studies with more monotherapy patients.

• Genotyping comprehensiveness. The normal (EM) version of the CYP2D6 gene cannot actually be positively identified. Rather, it is assumed to be present when a known variant conferring IM or PM status is not found. There are several dozen known genetic variants in CYP2D6, but most of the published studies only looked at one or a few of them. The result can be misclassification of IM or PM patients as EM, thus masking possible pharmacogenetic effects. This review showed a trend toward positive CYP2D6 effects in studies that used more extensive genotyping.

• CYP2D6 inhibitor coadministration. Studies that accounted for inhibitor use were more likely to indicate CYP2D6 effects than those that did not. Furthermore, two studies that did not support a CYP2D6 association despite accounting for inhibitor use appeared to overestimate the effects of some relatively weak or noninhibitory drugs, which might have confounded the results.

• Tamoxifen adherence. Tamoxifen can be a difficult drug to take, especially because of hot flashes, and discontinuation is more common in CYP2D6 EM patients (who have higher circulating endoxifen levels). On average, just under 80% of patients are adherent at 1 year, with further declines to as low as 50% by the fourth year. Only one of the studies adjusted their analysis for estimated tamoxifen adherence, which resulted in a stronger correlation between CYP2D6 activity and tamoxifen benefit.

Two recent reports from large randomized trials – ATAC (Arimidex, Tamoxifen, Alone or in Combination) and BIG (Breast International Group 1-98) – failed to identify any significant difference in clinical outcome among EM, IM, and PM patients treated with tamoxifen (J. Natl. Cancer Inst. 2012;104:452-60; J. Natl. Cancer Inst. 2012;104:441-51). The accompanying editorial asserts that the question of CYP2D6 clinical utility for tamoxifen therapy "has likely been laid to rest" (J. Natl. Cancer Inst. 2012;104:427-8).

However, a more thorough consideration of the details of these two studies still leaves open the possibility that CYP2D6 genotyping is clinically relevant.

Both studies involved the retrospective analysis of a subgroup from a larger prospective randomized trial. Their strengths include large sample sizes and analysis restricted to tamoxifen monotherapy. But there are multiple limiting factors.

In the ATAC study, only 92.5% of the tumors were HR positive. All tumors in BIG were HR positive.

In both studies, the genetic analyses were performed on a convenience sample of tumor blocks from which adequate DNA could be extracted, and both had statistical anomalies that might suggest sampling bias. In ATAC, there were significant differences between the genotyped subgroup and the overall study population with respect to frequency of surgical, radiation, and chemotherapy treatment prior to adjuvant tamoxifen therapy. In BIG, the frequencies of genotyped subjects with 0, 1, or 2 copies of one of the most common variants (CYP2D6*4) were inconsistent with the expected proportions in a random population sample.

The genetic analyses looked at only six or nine known CYP2D6 variants.

Only ATAC adjusted for potent and intermediate CYP2D6 inhibitors, although those drugs did not appear to affect the results.

Neither study included significant numbers of premenopausal patients, so the results may apply only to postmenopausal women.

Despite the relatively large populations studied, modest effects of CYP2D6 variation may have been missed. ATAC was estimated to be able to identify a 35% or greater variation associated with PM status; BIG was powered to detect only a 55% or greater effect size.

Neither study measured actual endoxifen levels or had any other way of assessing adherence to therapy.

In summary, it is certainly possible that CYP2D6 variation is not valuable in predicting clinical response to tamoxifen therapy. Perhaps endoxifen is not the most biologically relevant metabolite, or perhaps the effect of reduced CYP2D6 activity is small enough that it does not account for a significant proportion of tamoxifen failures. But it is also possible that CYP2D6 activity is clinically important for at least a subset of tamoxifen-treated patients. A more definitive answer awaits results from truly prospective trials that control for multiple confounding factors and measure actual endoxifen levels.

For now, in the absence of better data to address this issue, it does not seem appropriate to order CYP2D6 genotyping prior to starting tamoxifen therapy. Likewise, even if the CYP2D6 genotype is known, it should not influence the choice between tamoxifen and an aromatase inhibitor, or the dosing of tamoxifen. However, if a patient receiving tamoxifen requires medication for depression and/or hot flashes, it certainly seems reasonable to start with something other than fluoxetine, paroxetine, or bupropion, just in case those potent CYP2D6 inhibitors do turn out to reduce clinical efficacy.

Dr. Levy is at the division of general internal medicine and the McKusick-Nathans Institute of Genetic Medicine of Johns Hopkins University, Baltimore. He reports having no conflicts of interest.

In the recent column, (“CAD: One Step Forward, One Step Back,” Internal Medicine News digital network Oct. 28, 2010), Dr. Matthew Taylor discussed the “step backward” that occurred in genetic risk prediction of coronary artery disease (CAD) with the discovery that variations in the KIF6 gene might not be clinically associated with CAD after all. More research and vigorous discussion about the role of this gene in this disease – and about early clinical adoption of newly discovered risk factors in general – are sure to follow.

Of course, KIF6 is only one of many genetic markers that have been linked to CAD risk. Many are already in the public domain and available clinically through health care providers and via direct-to-consumer marketing. What are we to do with these?

One approach to answering such questions is to turn to an evidence review. The Evaluation of Genomic Applications in Practice and Prevention (EGAPP) initiative seeks to evaluate genetic tests and innovations with respect to implementation in clinical practice and public health. Much like the U.S. Preventive Services Task Force, EGAPP commissions evidence reviews and then issues recommendations based upon the evidence. More information, as well as evidence reviews and recommendations reports, is available at www.egappreviews.org.

In December, EGAPP published a report on genomic profiling to assess cardiovascular risk (Genet. Med. 2010;12:839-43). The overarching question was whether genomic profiling to identify undiagnosed individuals who are at increased risk for cardiovascular disease (CVD) leads to improved cardiovascular outcomes. Overall, the assessment was that there is insufficient evidence to support such testing in the general population. The net health benefit was deemed to be low, and clinical use was discouraged unless improved clinical outcomes are demonstrated in future research.

It is worth exploring the report further to understand the current state of knowledge and anticipate potential new developments. The evidence review looked at all eight genetic testing panels marketed for CVD risk prediction that were commercially available in February 2008. Collectively, 58 genetic variants were identified among 29 genes. Of these, 38 were reported to have some association with risk for coronary heart disease (CHD), consisting of coronary artery disease, ischemic heart disease, or myocardial infarction. Data for association with stroke were weaker.

Only two test manufacturers (deCODE Genetics and Interleukin Genetics) indicated that they are Clinical Laboratory Improvement Amendments–certified and provided detailed test method and validation information. For the other tests, there was frequently insufficient information to identify even the specific genetic variants being tested, much less the analytic validity of the approach. Nonetheless, the existing scientific technology was considered adequate to allow satisfactory accuracy and reliability of the tests for detecting genetic variation.

Regarding clinical utility, there are no published data on the long-term outcomes associated with genetic testing for CHD risk prediction. The anticipated benefit is improved identification of those at higher risk of disease, which could lead to improved clinical outcomes by virtue of increased efforts at risk reduction (through behavior change and pharmacologic treatment) and more aggressive screening for and management of manifest disease.

Among all the markers, 24 were deemed to have some degree of credibility and/or statistical significance. However, combining them in a statistical model did not provide a clinically useful stand-alone predictive test. Even the 9p21 genetic marker, when combined with traditional cardiac risk factors, provided only 0%-5% improvement in risk assessment for CHD.

Regarding clinical utility, there are no published data on the long-term outcomes associated with such testing. In theory, such tests would improve identification of those at higher risk of disease, which could lead to improved clinical outcomes by virtue of increased efforts at risk reduction (through behavior change and pharmacologic treatment) and more aggressive screening for and management of manifest disease. Potential harms must be considered, too. Among false positives – those identified incorrectly as being at increased risk for CHD – there may be unnecessarily increased anxiety and treatment-associated adverse events, without any reduction in morbidity or mortality. There also are financial costs associated with false positives. An additional risk is false reassurance for those who are at increased risk, but who are not identified by genetic testing. There is room for optimism, as early data suggest at least short-term cardiac risk reduction without clinical harms, but additional and longer-term studies are still needed.

There are notable limitations to this report. Most of the data so far were obtained in whites of European ancestry. Gene-disease associations and effect sizes may be quite different in other populations. In addition, many of the studies are relatively small and underpowered. Larger and newer studies are still identifying additional candidate genes and markers. Even more challenging is that we still do not know how to combine multiple genetic factors to develop a composite risk assessment, nor do we know how to combine a genetic risk assessment with traditional cardiac risk predications. Most models assume complete independence and simply multiply the odds ratio of each identified variant. However, this approach has not been validated and could yield falsely elevated or diminished risk scores.

The true power of genetic association studies to identify risk factors for common diseases may in fact not lie in better identification of those at increased risk. Rather, elucidation of the biological basis of such associations holds the promise of improving our understanding and eventual treatment of the underlying disease process.

So how should a physician deal with the availability of cardiac risk assessment tests? Clearly, these are best suited for people who consider themselves early adopters. The 9p21 marker appears to be the most clinically relevant marker at this point. Indeed, the risk associated with this genetic variant appears highest in younger individuals, so perhaps it might be appropriate to consider obtaining such information in a person under age 55 years who needs further encouragement to modify his or her cardiac risks.

Ideally, patients and their providers will engage in a discussion of the potential risks and benefits prior to any testing. Then, in the truest and oldest model of personalized medicine, they can decide together if pursuing such testing is appropriate on a case-by-case basis.

In the recent column, (“CAD: One Step Forward, One Step Back,” Internal Medicine News digital network Oct. 28, 2010), Dr. Matthew Taylor discussed the “step backward” that occurred in genetic risk prediction of coronary artery disease (CAD) with the discovery that variations in the KIF6 gene might not be clinically associated with CAD after all. More research and vigorous discussion about the role of this gene in this disease – and about early clinical adoption of newly discovered risk factors in general – are sure to follow.

Of course, KIF6 is only one of many genetic markers that have been linked to CAD risk. Many are already in the public domain and available clinically through health care providers and via direct-to-consumer marketing. What are we to do with these?

One approach to answering such questions is to turn to an evidence review. The Evaluation of Genomic Applications in Practice and Prevention (EGAPP) initiative seeks to evaluate genetic tests and innovations with respect to implementation in clinical practice and public health. Much like the U.S. Preventive Services Task Force, EGAPP commissions evidence reviews and then issues recommendations based upon the evidence. More information, as well as evidence reviews and recommendations reports, is available at www.egappreviews.org.

In December, EGAPP published a report on genomic profiling to assess cardiovascular risk (Genet. Med. 2010;12:839-43). The overarching question was whether genomic profiling to identify undiagnosed individuals who are at increased risk for cardiovascular disease (CVD) leads to improved cardiovascular outcomes. Overall, the assessment was that there is insufficient evidence to support such testing in the general population. The net health benefit was deemed to be low, and clinical use was discouraged unless improved clinical outcomes are demonstrated in future research.

It is worth exploring the report further to understand the current state of knowledge and anticipate potential new developments. The evidence review looked at all eight genetic testing panels marketed for CVD risk prediction that were commercially available in February 2008. Collectively, 58 genetic variants were identified among 29 genes. Of these, 38 were reported to have some association with risk for coronary heart disease (CHD), consisting of coronary artery disease, ischemic heart disease, or myocardial infarction. Data for association with stroke were weaker.

Only two test manufacturers (deCODE Genetics and Interleukin Genetics) indicated that they are Clinical Laboratory Improvement Amendments–certified and provided detailed test method and validation information. For the other tests, there was frequently insufficient information to identify even the specific genetic variants being tested, much less the analytic validity of the approach. Nonetheless, the existing scientific technology was considered adequate to allow satisfactory accuracy and reliability of the tests for detecting genetic variation.

Regarding clinical utility, there are no published data on the long-term outcomes associated with genetic testing for CHD risk prediction. The anticipated benefit is improved identification of those at higher risk of disease, which could lead to improved clinical outcomes by virtue of increased efforts at risk reduction (through behavior change and pharmacologic treatment) and more aggressive screening for and management of manifest disease.

Among all the markers, 24 were deemed to have some degree of credibility and/or statistical significance. However, combining them in a statistical model did not provide a clinically useful stand-alone predictive test. Even the 9p21 genetic marker, when combined with traditional cardiac risk factors, provided only 0%-5% improvement in risk assessment for CHD.

Regarding clinical utility, there are no published data on the long-term outcomes associated with such testing. In theory, such tests would improve identification of those at higher risk of disease, which could lead to improved clinical outcomes by virtue of increased efforts at risk reduction (through behavior change and pharmacologic treatment) and more aggressive screening for and management of manifest disease. Potential harms must be considered, too. Among false positives – those identified incorrectly as being at increased risk for CHD – there may be unnecessarily increased anxiety and treatment-associated adverse events, without any reduction in morbidity or mortality. There also are financial costs associated with false positives. An additional risk is false reassurance for those who are at increased risk, but who are not identified by genetic testing. There is room for optimism, as early data suggest at least short-term cardiac risk reduction without clinical harms, but additional and longer-term studies are still needed.

There are notable limitations to this report. Most of the data so far were obtained in whites of European ancestry. Gene-disease associations and effect sizes may be quite different in other populations. In addition, many of the studies are relatively small and underpowered. Larger and newer studies are still identifying additional candidate genes and markers. Even more challenging is that we still do not know how to combine multiple genetic factors to develop a composite risk assessment, nor do we know how to combine a genetic risk assessment with traditional cardiac risk predications. Most models assume complete independence and simply multiply the odds ratio of each identified variant. However, this approach has not been validated and could yield falsely elevated or diminished risk scores.

The true power of genetic association studies to identify risk factors for common diseases may in fact not lie in better identification of those at increased risk. Rather, elucidation of the biological basis of such associations holds the promise of improving our understanding and eventual treatment of the underlying disease process.

So how should a physician deal with the availability of cardiac risk assessment tests? Clearly, these are best suited for people who consider themselves early adopters. The 9p21 marker appears to be the most clinically relevant marker at this point. Indeed, the risk associated with this genetic variant appears highest in younger individuals, so perhaps it might be appropriate to consider obtaining such information in a person under age 55 years who needs further encouragement to modify his or her cardiac risks.

Ideally, patients and their providers will engage in a discussion of the potential risks and benefits prior to any testing. Then, in the truest and oldest model of personalized medicine, they can decide together if pursuing such testing is appropriate on a case-by-case basis.

In the recent column, (“CAD: One Step Forward, One Step Back,” Internal Medicine News digital network Oct. 28, 2010), Dr. Matthew Taylor discussed the “step backward” that occurred in genetic risk prediction of coronary artery disease (CAD) with the discovery that variations in the KIF6 gene might not be clinically associated with CAD after all. More research and vigorous discussion about the role of this gene in this disease – and about early clinical adoption of newly discovered risk factors in general – are sure to follow.

Of course, KIF6 is only one of many genetic markers that have been linked to CAD risk. Many are already in the public domain and available clinically through health care providers and via direct-to-consumer marketing. What are we to do with these?

One approach to answering such questions is to turn to an evidence review. The Evaluation of Genomic Applications in Practice and Prevention (EGAPP) initiative seeks to evaluate genetic tests and innovations with respect to implementation in clinical practice and public health. Much like the U.S. Preventive Services Task Force, EGAPP commissions evidence reviews and then issues recommendations based upon the evidence. More information, as well as evidence reviews and recommendations reports, is available at www.egappreviews.org.

In December, EGAPP published a report on genomic profiling to assess cardiovascular risk (Genet. Med. 2010;12:839-43). The overarching question was whether genomic profiling to identify undiagnosed individuals who are at increased risk for cardiovascular disease (CVD) leads to improved cardiovascular outcomes. Overall, the assessment was that there is insufficient evidence to support such testing in the general population. The net health benefit was deemed to be low, and clinical use was discouraged unless improved clinical outcomes are demonstrated in future research.

It is worth exploring the report further to understand the current state of knowledge and anticipate potential new developments. The evidence review looked at all eight genetic testing panels marketed for CVD risk prediction that were commercially available in February 2008. Collectively, 58 genetic variants were identified among 29 genes. Of these, 38 were reported to have some association with risk for coronary heart disease (CHD), consisting of coronary artery disease, ischemic heart disease, or myocardial infarction. Data for association with stroke were weaker.

Only two test manufacturers (deCODE Genetics and Interleukin Genetics) indicated that they are Clinical Laboratory Improvement Amendments–certified and provided detailed test method and validation information. For the other tests, there was frequently insufficient information to identify even the specific genetic variants being tested, much less the analytic validity of the approach. Nonetheless, the existing scientific technology was considered adequate to allow satisfactory accuracy and reliability of the tests for detecting genetic variation.

Regarding clinical utility, there are no published data on the long-term outcomes associated with genetic testing for CHD risk prediction. The anticipated benefit is improved identification of those at higher risk of disease, which could lead to improved clinical outcomes by virtue of increased efforts at risk reduction (through behavior change and pharmacologic treatment) and more aggressive screening for and management of manifest disease.

Among all the markers, 24 were deemed to have some degree of credibility and/or statistical significance. However, combining them in a statistical model did not provide a clinically useful stand-alone predictive test. Even the 9p21 genetic marker, when combined with traditional cardiac risk factors, provided only 0%-5% improvement in risk assessment for CHD.

Regarding clinical utility, there are no published data on the long-term outcomes associated with such testing. In theory, such tests would improve identification of those at higher risk of disease, which could lead to improved clinical outcomes by virtue of increased efforts at risk reduction (through behavior change and pharmacologic treatment) and more aggressive screening for and management of manifest disease. Potential harms must be considered, too. Among false positives – those identified incorrectly as being at increased risk for CHD – there may be unnecessarily increased anxiety and treatment-associated adverse events, without any reduction in morbidity or mortality. There also are financial costs associated with false positives. An additional risk is false reassurance for those who are at increased risk, but who are not identified by genetic testing. There is room for optimism, as early data suggest at least short-term cardiac risk reduction without clinical harms, but additional and longer-term studies are still needed.

There are notable limitations to this report. Most of the data so far were obtained in whites of European ancestry. Gene-disease associations and effect sizes may be quite different in other populations. In addition, many of the studies are relatively small and underpowered. Larger and newer studies are still identifying additional candidate genes and markers. Even more challenging is that we still do not know how to combine multiple genetic factors to develop a composite risk assessment, nor do we know how to combine a genetic risk assessment with traditional cardiac risk predications. Most models assume complete independence and simply multiply the odds ratio of each identified variant. However, this approach has not been validated and could yield falsely elevated or diminished risk scores.

The true power of genetic association studies to identify risk factors for common diseases may in fact not lie in better identification of those at increased risk. Rather, elucidation of the biological basis of such associations holds the promise of improving our understanding and eventual treatment of the underlying disease process.

So how should a physician deal with the availability of cardiac risk assessment tests? Clearly, these are best suited for people who consider themselves early adopters. The 9p21 marker appears to be the most clinically relevant marker at this point. Indeed, the risk associated with this genetic variant appears highest in younger individuals, so perhaps it might be appropriate to consider obtaining such information in a person under age 55 years who needs further encouragement to modify his or her cardiac risks.

Ideally, patients and their providers will engage in a discussion of the potential risks and benefits prior to any testing. Then, in the truest and oldest model of personalized medicine, they can decide together if pursuing such testing is appropriate on a case-by-case basis.