Pharmacogenetic testing in the face of unclear clinical efficacy: Lessons from cytochrome P450 2D6 for tamoxifen
Jeffrey Peppercorn MD, MPH,
Duke University Medical Center, Durham, North Carolina
Corresponding author: Jeffrey Peppercorn, MD, MPH, Associate Professor of Medicine, Division of Medical Oncology, Duke University Medical Center, Box #3446, Durham, NC 27710; Fax: (919) 681-0874; firstname.lastname@example.org
This study evaluated self-reported knowledge, practice, and attitudes toward commercially available cytochrome P450 2D6 (CYP2D6) pharmacogenomic testing for patients on tamoxifen for breast cancer (CYPT) among US oncologists while evidence for the use of the test was evolving.
A self-administered survey of medical oncology breast cancer specialists at National Comprehensive Cancer Network (NCCNO) centers and a random sample of community-based oncologists (CBOs) was undertaken. The survey evaluated knowledge and use of the CYP2D6 test and response to hypothetical test results.
In total, 201 of 459 (44%) oncologists responded. At a time when CYPT remained experimental, 31% of oncologists reported use of CYPT and 56% reported willingness to order CYPT outside of a clinical trial if requested by a patient. Compared to oncologists specializing in breast cancer, oncologists in community-based practice were more likely to use CYPT routinely (21% versus 11%, P < .06), to order CYPT on patient request (66% versus 44%, P < .001), and to change management for premenopausal women with intermediate metabolism (34% CBO versus 8% NCCN, P < .001). Oncologists cited data from randomized trials and professional guidelines as most influential when considering use of a genetic test.
Advances in genetics and molecular biology are leading to an increasing number of novel tests to diagnose and treat cancer and many other medical conditions. Some tests may become commercially available while clinical evidence to guide their use in practice is still emerging. Availability of promising but experimental genetic tests for use in routine clinical practice raises questions related to potential clinical harm and benefit, informed consent, and equity in access.
One area of oncology research where this presents a growing challenge is the field of pharmacogenomics, in which genetic variation in drug metabolism is evaluated to understand predictors of response and toxicity, and to guide personalization of cancer therapy. Pharmacogenomics holds the promise of identifying those patients who will most benefit from a drug, and those patients who would be best served by treatment with a different therapy, potentially improving outcomes and reducing toxicity. However, this promise must be tempered by the need to validate the pharmacogenomic test through well-designed clinical trials and to conclusively demonstrate the association between genotype and outcome before use in clinical practice.
A prime example of both the promise and the challenge of the potential therapeutic application of pharmacogenomics in recent years has been the evolving role of the cytochrome P450 2D6 (CYP2D6) test for tamoxifen metabolism. Tamoxifen, a selective estrogen receptor–modulating agent, is one of the most commonly used medications for breast cancer treatment and prevention. However, tamoxifen is not effective in all patients, and reasons for progression of disease, recurrence, or development of a new breast cancer despite tamoxifen therapy are not completely understood.
Genetic differences in metabolism of tamoxifen to the more potent estrogen receptor antagonist, endoxifen, mediated by the hepatic enzyme CYP2D6 (in addition to other enzymes) were believed to potentially play a role in determining efficacy and outcomes among patients.[2, 3] Roughly 10% of patients in most population-based studies can be characterized as genetically “poor metabolizers” (PM), generating very little endoxifen, compared to roughly 50% of the population who are classified as “extensive metabolizers” (EM), and the remaining 40% as those with intermediate metabolism (IM). Multiple studies have examined the role of genetic differences in CYP2D6 in determining breast cancer outcomes with conflicting results. In addition, 2 recent large retrospective studies suggested no clear association between CYP2D6 genotype and outcome.[6, 7] Although this remains an area of active clinical investigation, there is no clear role for use of this test outside of a clinical trial at this time.[8, 9]
As this scientific story evolved, the genetic test for CYP2D6 genotype remained commercially available. The Roche AmpliChip CYP450 Test was approved by the US Food and Drug Administration (FDA) in 2004 and a number of other labs offered CLIA-licensed CYP2D6 testing. Although deemed investigational by most insurance companies, the test was available at the time of this survey for less than $500. The test could be ordered in routine practice, patients could obtain the test themselves (as a buccal swab) without physician involvement, and consideration of routine testing was discussed at national oncology meetings. At the time of this survey, the test was experimental but available outside of clinical trials and subject to direct-to-consumer advertising.
The extent of oncologist awareness of the CYP2D6 tests, familiarity with the evidence, and clinical practice regarding testing during this period of evolving evidence is unknown. The goal of this study was to assess the use of this test outside of clinical trials and the attitudes of community-based oncologists (CBOs) and breast cancer specialists with regard to CYP2D6 testing among patients with breast cancer eligible for tamoxifen therapy.
MATERIALS AND METHODS
In 2009, we conducted an anonymous cross-sectional self-administered mailed survey of practice and attitudes toward CYP2D6 testing for tamoxifen metabolism (CYPT) outside of clinical trials among all breast cancer medical oncologists affiliated with the National Comprehensive Cancer Network (NCCNO) and a random sample of CBOs identified from the American Society of Clinical Oncology (ASCO) directory. The survey evaluated knowledge of the CYP2D6 test, use of the test outside of trials, requests for the test by patients and third parties, and response to hypothetical test results. Associations between practice setting and CYP2D6 knowledge, CYPT, and practice patterns were evaluated.
A study-specific survey consisting of 36 questions was developed and piloted among oncologists at Duke University Medical Center, Durham, NC. The final survey was mailed to all NCCNOs identified through search of the NCCN member Web sites (excluding Duke) to identify all medical oncologists specializing in breast cancer. The CBO sample was identified using the ASCO 2008 Membership Directory. We randomly selected potential subjects using the method of Helft et al with selection of eligible US-based medical oncologists identified as described in Peppercorn et al.[11, 12] Eligibility consisted of verification by survey that the participant was a US-based practicing medical oncologist taking care of at least one breast cancer patient per week. A second mailing was conducted 4 weeks after the initial mailing to improve response rate. A $2 token incentive was also included with the initial mailing. The study was approved by the Duke University Institutional Review Board.
Based on response rates in published surveys of US oncologists, we estimated the response rate to be between 40% and 50%.[12, 14] To detect significant differences in response of 20% with 80% power and 5% alpha, we needed to survey 196 oncologists. We identified 173 NCCNOs through cancer center Web sites. Assuming a response of approximately 50% among NCCNOs and 40% among CBOs, we would need to contact at least 275 physicians from the ASCO directory. Given potential for wrong addresses and misclassification of invited participants, we contacted 300 CBO.
The data were intended to be descriptive in nature, and the results of this survey are designed to be hypothesis-generating, rather than to definitively establish a link between a set of oncologists' responses and personal or practice variables. The primary outcome of interest was reported use of CYP2D6 testing as indicated by report of percent of patients on tamoxifen, for whom CYP2D6 testing is currently being ordered. Our primary hypothesis was that CBOs would be significantly more likely to offer CYP2D6 testing compared with NCCNOs. Our study was powered to detect a 20% difference in use of CYP2D6 testing between these groups.
Secondary outcomes of interest included awareness of CYP2D6 testing and currently published data in this area, prevalence of patients or third parties requesting CYP2D6 testing, willingness to order the test upon patient request, response to genotype results in hypothetical clinical scenarios, and factors considered important when considering ordering a genetic test. Descriptive statistics are reported. Fisher's exact tests and independent sample t tests were performed to test for significance. All tests of hypotheses were 2-sided.
Of 300 surveys mailed to CBOs, 14 were returned due to invalid addresses. Among 459 oncologists surveyed, 201 responded (44%), including 97 of 173 (56%) NCCNOs and 104 of 286 (36%) CBOs. Respondent's characteristics are reported in Table 1.
Table 1. Demographics of Respondents
All (N = 201)
CBO (N = 104)
NCCNO (N = 97)
Abbreviations: CBO, community-based oncologist; CYP2D6, cytochrome P450 2D6; Hem-Onc, hematologic oncology; NCCNO, oncologists affiliated with the National Comprehensive Cancer Network.
Breast cancer patients per week
Source of first information about CYP2D6 testing
Awareness and Use of CYP2D6 Testing for Patients on Tamoxifen
The vast majority of oncologists were aware of the CYP2D6 test (90%); however, awareness of the test was more frequent among NCCNOs versus CBOs (98% versus 83%, P < .001). Overall, 31% of oncologists reported having ordered the test outside of a clinical trial, and 14% reported routine use of the test in their clinical practice. Use of the test in “routine” clinical practice, as opposed to sporadic use, was defined as use in more than 10% of patients on tamoxifen. By this definition, routine use appeared more frequent among CBOs compared with NCCNOs (22% versus 11%, P = .06). Only 6% of oncologists reported use of the test in the majority of their patients on tamoxifen, with 3% among CBOs and 9% among NCCNOs, a difference that was not statistically significant. Frequency of testing is reported in Figure 1.
In contrast, during this period, 82% of oncologists reported ordering Oncotype or Mammaprint multigene prognostic assays for at least 10% of patients with endocrine receptor–positive, lymph node–negative breast cancer, and 38% reported using one of these tests for the majority of such patients. There was no significant difference between CBOs and NCCNOs in use of multigene prognostic assays.
Receipt and Response to Request for Testing From Patients
Overall, 28% of oncologists reported that patients had requested the CYP2D6 test. Requests for testing were more frequently encountered by NCCNOs compared with CBOs (33% versus 12%, P < .001). Even though reported use of the test was rare, 56% reported that they would order the test if requested by a patient (66% of CBOs versus 44% of NCCNOs, P < .001). Although we did not specifically ask oncologists to report changes in patient management following request for testing, we found that those reporting willingness to order the test on request were also most likely to change management based on test results. For example, when presented with a scenario of a premenopausal patient who is found to have a poor metabolism genotype, 76% of those reporting willingness to order the test on request report that they would change management (adding ovarian suppression to tamoxifen, switching to ovarian suppression alone, or switching to ovarian suppression plus an aromatase inhibitor), versus 36% of those who stated that they might order the test, versus 14% of those who stated that they would not order the test on request. Figure 2 displays oncologists' reported response to CYP2D6 test results according to willingness to order the test at the request of the patient. Twenty-two percent of oncologists also reported receiving requests for CYP2D6 testing from third-party insurers or pharmacies. We did not assess response to such requests. Experience with ordering and requests for testing is presented in Table 2.
Table 2. Attitudes and Practice With Regard to CYP2D6 Testing Among Oncologists
All (N = 201)
CBO (N = 104)
NCCNO (N = 97)
Abbreviations: CBO, community-based oncologist; CYP2D6, cytochrome P450 2D6; NCCNO, oncologists affiliated with the National Comprehensive Cancer Network.
Aware of the test
Ordered the test
Routinely discuss with patients
Will order if requested by patient
Understanding of the Role of CYP2D6
We assessed beliefs regarding the importance of CYP2D6 for tamoxifen metabolism to determine whether there was a correlation between the physicians' understanding of current evidence and their clinical practice. All questions had been addressed in the scientific literature at the time of the survey with ongoing research to determine conclusive results.
Among all respondents, 60% (57% CBOs, 62% NCCNOs, P = .6) believed that recurrence risk was higher for women with poor metabolism genotypes on tamoxifen. Although only 14% of physicians believed that premenopausal women with poor metabolism were likely to have better outcomes if treated with ovarian suppression and aromatase inhibitors instead of tamoxifen, CBOs were significantly more likely to hold this belief than NCCNOs (20% versus 7%, P = .008). Based on emerging data, we also asked physicians whether some antidepressant medications, such as serotonin-selective reuptake inhibitors, could reduce tamoxifen metabolism through CYP2D6. Overall, 76% believed that some serotonin-selective reuptake inhibitors could block tamoxifen metabolism (91% NCCNOs versus 62% CBOs, P < .001). There was no significant correlation between reported beliefs in the importance of CYP2D6-mediated tamoxifen metabolism as expressed by response to these questions and reported clinical practice when comparing those reporting routine use of the test versus rare/never use.
Willingness to Change Practice Based on CYP2D6 Test Results
We presented 3 hypothetical scenarios regarding management of patients on tamoxifen who obtained commercially available CYP2D6 results from an external source. For a premenopausal woman with a PM genotype, 33% would make no change, whereas 56% would change therapy, including 38% who would switch to ovarian suppression (OS) plus an aromatase inhibitor (AI), 9% who would switch to OS alone, and 8% who would add OS to tamoxifen. CBOs were more likely to switch to OS plus AI (47% versus 31%, P < .02), whereas NCCNOs were more likely to recommend no change (50% versus 18%, P < .001). For a premenopausal woman with IM genotype, 66% of respondents would make no change, 20% would change therapy, with the remainder unsure (3%) or writing in other responses (10%). CBOs were again more likely to change therapy compared to NCCNOs (34% versus 8%, P < .001), with addition of OS to tamoxifen being most frequently selected among both groups. The third case involved a postmenopausal woman with PM genotype. In this case, only 14% of respondents would not change therapy, and the vast majority of both CBOs and NCCNOs reported that they would switch to AI (82% versus 71%, P = .08). The remainder reported that they were unsure or commented they only use AI, not tamoxifen, for postmenopausal patients. None would stop therapy.
Of interest, for all scenarios, among oncologists who indicated they might change management based on CYPT results acquired by the patient, 43% had never ordered the test.
Factors Affecting Use of CYP2D6 Tests
We asked oncologists to rate the importance of factors that might drive consideration of a ordering a genetic test on a scale from 1 to 10, with 10 being the most important. The list of factors and responses is reported in Figure 3. The median importance was greater than 5 for all factors listed, suggesting that the sensitivity of this nonvalidated measure to determine which factors are most important may be limited. Evidence of survival benefit from randomized trials (mean = 9.08) and professional guideline recommendations (mean = 8.74) were rated the most important factors among all oncologists. Professional guidelines were rated more important among CBOs than NCCNOs (mean of 9.01 versus 8.47, P = .02). FDA approval was rated 8 to 10 (highest importance) among 67% of CBOs versus 44% of NCCNOs (P = .002), with means of 7.81 and 6.69, respectively (P = .002). Relatively lower levels of importance were assigned to retrospective evidence of the tests utility for prognosis, toxicity, or response. Overall, oncologists indicated that whether their colleagues were using the test (practice among peers) was considered relatively unimportant.
The CYP2D6 test is a prime example of the potential for genetic technology to become commercially available while scientific evidence to guide its use in clinical practice is still evolving. At the time of our survey, the value of this test to guide tamoxifen use was not established and clinical research was ongoing, but the test was both commercially available, advertised in the media, and requested by patients and insurers and pharmacies. Today, even with the most recent data suggesting no clear role for this test in clinical practice, some mail-order pharmacies in the United States continue to suggest that physicians consider this test and request a rationale for not ordering the test from the prescribing physician.
The survey results presented above provides evidence of the potential for use of pharmacogenetic tests outside of clinical trials, as this new era of pharmacogenomic testing presents an opportunity to consider how we may better guide practice in this area in the future. Several factors are notable. First, despite widespread awareness of the test, just under one-third of oncologists reported ever ordering it (31%), and use of the test in routine practice was rare (14%). Only a small minority of oncologists (6%) reported using this test in the majority of patients during this period. This is in contrast to much more widespread (82%) reported routine use of multigene assays of prognosis and response to tamoxifen among the same clinicians. The basis for this difference is not known, but cannot be attributable to commercial availability or duration of availability with both the Oncotype Dx test and FDA-approved CYP2D6 test becoming available in 2004. Factors that distinguish the 2 types of genetic tests include strength of the evidence (strong data supporting Mammaprint and Oncotype testing was published in 2002 and 2004, respectively)[17, 18] and recommendations in national guidelines (CYP2D6 testing was not covered in ASCO or NCCN guidelines at the time of this survey). It is also possible that oncologists were more familiar with using tumor genes and other biomarkers to guide therapy, and less comfortable with the novel field of pharmacogenetics.
The hesitancy to use the CYP2D6 test, despite its availability and some degree of patient demand, among the vast majority of oncologists seems appropriate and reassuring given the available evidence at the time of the survey. However, use of the test was twice as common among CBOs compared to breast cancer specialists at academic centers. This suggests that as novel tests become commercially available, there is a need for education targeting community oncologists, who are expected to manage a highly diverse group of malignant diseases, regarding current evidence and practice standards. Such education could take the form of manuscripts, meeting presentations, and/or provisional expert opinions such as those produced by ASCO. As shown above, professional guidelines were viewed as highly influential in guiding practice, particularly among CBOs.
Although not using the test in routine practice, a majority of oncologists (56%) reported willingness to order the test if requested by a patient, and many suggested they would change treatment based on test results. This may reflect the recent emphasis on shared decision-making, particularly in breast cancer management, and could be deemed an appropriate response to clinical uncertainty (ie, place a strong weight on patient preference when the harms and benefits of an intervention are unclear). However, at minimum, this suggests a need to clearly inform patients of the clinical uncertainty surrounding the ordering of a test and to emphasize potential harms from such practice. It also suggests a need to monitor the impact of direct-to-consumer advertising for genetic testing on clinical care. For example, if premenopausal patients were taken off tamoxifen due to CYP2D6 tests suggesting poor metabolism and placed on OS and an AI, it is possible, based on current evidence that they received inferior therapy or equivalent therapy with greater toxicity. The results of the TEXT and SOFT trials should clarify the clinical significance of such changes in therapy.
Since this survey was conducted, 2 major trials investigating the importance of CYP2D6 genotype have suggested no role for CYP2D6 testing in routine practice. In the ATAC trial, among 588 patients treated with adjuvant tamoxifen with 10 years of follow-up, there was no significant difference in breast cancer outcomes on the basis of CYP2D6 genotype. Similarly, in the BIG 1-98 trial, among 1243 women assigned to tamoxifen, there was no difference in breast cancer outcomes on the basis of CYP2D6 genotype. In fact, contrary to some earlier reports, women with PM or IM genotypes appeared to have higher rates of tamoxifen-associated side effects, suggesting that a strategy of testing for CYP2D6 or switching therapy based on symptoms is inappropriate. However, recent critique of these studies by genetics experts confirms that further investigation is still needed.[8, 9]
In this context, these survey results provide a cautionary tale. We are reminded once again of the need for adequate scientific evaluation of new medical technologies and of the need for evidence-based practice. In a setting where outcomes for too many women on tamoxifen remain inadequate, there is an understandable desire to adopt a test that promised to more rationally guide therapy. The tendency for some in clinical practice to be early versus late adopters has long been documented, and there are advantages and disadvantages to both approaches. The question is where the threshold for early adoption should be. Clearly, commercial availability alone must not be a primary consideration, and promising retrospective data must be followed by more rigorous prospective analysis. In fact, commercial availability while evidence is in development might be a key criterion for ASCO, NCCN, and others to plan educational outreach to inform practice. CBOs cite guidance from professional organizations, data from randomized controlled trials, and insurance coverage as the 3 top factors they consider when considering a genetic test. As we await definitive trial data, and as cost promises to be less and less of a prohibitive factor in the future, guidelines and other forms of educational initiatives may be the most effective way to promoted evidence-based practice.[24, 25] Because patient request appears to play a powerful role in shaping practice in this area, there may also be a role for greater efforts to provide patient-centered educational materials on controversial clinical topics.
Limitations of this study include the potential for response bias. Although more than 50% of breast cancer specialists identified chose to participate, only 36% of CBOs did so, leaving the possibility that those with greater or less familiarity with the test opted to respond. We did not adjust for multiple comparisons, and differences on the basis of practice setting should be interpreted with caution. Similarly, suggestions that clinicians would order the test if requested or change management based on results were based on hypothetical scenarios and may not correspond with actual clinical decisions.
This survey offers a rare window into the experiences and views among academic and community-based oncologists regarding a then-novel pharmacogenetic test that was commercially available and advertised in both the scientific and lay literature even as clinical evidence was in development. Our results provide reassurance that despite calls for increased CYP2D6 testing in routine practice in a variety of forums, the use of CYPT was infrequent in oncology practice. Diversity in practice patterns and knowledge suggests a role for educational initiatives targeting clinicians and patients when such tests become available, and highlights the need for further research into how novel technology is adopted in oncology and how educational interventions may improve evidence-based practice.
This work was supported by the Greenwall Foundation Faculty Scholars Program in Bioethics (to J.P.) and by a Career Development Award from the American Society of Clinical Oncology Conquer Cancer Foundation (to J.P.).