Tamoxifen is a member of the Selective Estrogen Receptor Modulators (SERMs), and by virtue of its antiestrogenic properties, particularly in breast tissue, is considered an essential component in the armamentarium of drugs used for the prevention and treatment of hormone receptor–positive (HR+) breast cancer. Early studies evaluating the efficacy of tamoxifen demonstrated that its use in adjuvant settings resulted in an approximately 40% reduction in breast cancer recurrence rates and a 30% reduction in mortality rates ( 1 ). In light of such favorable patient outcomes, medical societies recommend that tamoxifen be offered for five years to all women with HR+ breast cancer and recent American Society of Clinical Oncology (ASCO) guidelines recommend 10 years of treatment in most women for prevention of recurrence ( 2 ). Additionally, guidelines recommend at least five years of tamoxifen therapy for primary prevention in women at risk for developing invasive HR+ breast cancer. Numerous studies revealed that much of the antiestrogenic effects of tamoxifen reside in its active metabolite, endoxifen, which shows a 100-fold greater affinity for estrogen receptors than the parent compound ( 3 , 4 ). The rate-limiting step in tamoxifen metabolism to endoxifen is mediated primarily by the highly polymorphic cytochrome P450 CYP2D6 enzyme. Loss of functional genetic polymorphisms in CYP2D6 lead to the absence of functional CYP2D6 protein in approximately 5% to 10% of whites (people of European ancestry) and 1% to 2% of those of Asian and African ancestry ( 5 ). In the literature, these are commonly referred to as CYP2D6–poor metabolizers (PMs). The first study to report an association between CYP2D6 polymorphisms and endoxifen plasma concentration was published in 2003 and showed that CYP2D6 PMs exhibited lower endoxifen plasma concentrations than those with functional CYP2D6 enzyme ( 6 ).The results from this study served as an impetus to investigate the influence of CYP2D6 gene variation on clinical outcomes with tamoxifen ( 7–16 ). Early retrospective studies addressing this question were small but suggested that CYP2D6 PMs had worse outcomes with tamoxifen than those who did not carry any CYP2D6 loss of function alleles. While not all the early studies documented this association, a 2006 Food and Drug Administration advisory committee suggested the tamoxifen product label be updated to include CYP2D6 genotype as an important biomarker of tamoxifen efficacy. In 2009, a cohort study with over 1300 patients showed statistically significantly worse outcomes in CYP2D6 PMs, particularly time to recurrence, event-free survival, and disease-free survival ( 7 ). In 2012, genetic substudies of two landmark clinical trials, the Breast International Group 1–98 (BIG 1–98) and the Arimidex, Tamoxifen, Alone or in Combination (ATAC), reported on CYP2D6 associations with outcomes in the tamoxifen arms of each study (n = 1243 and n = 588, respectively). Neither study observed a link between CYP2D6 genotype and clinical outcomes with tamoxifen ( 11 , 12 ), which was thought to signal the death knell for the concept of CYP2D6- guided tamoxifen therapy. Nevertheless, the results of the ATAC and BIG 1–98 have been the subject of intense criticism among proponents for CYP2D6 genotyping because of arguments that the data were flawed because of use of formalin-fixed paraffin-embedded (FFPE) tumor tissues for CYP2D6 genotyping. The substantial deviation from Hardy Weinberg Equilibrium (HWE; which defines expected vs observed genotype frequencies in a randomly mating population) in the BIG 1–98 study (based on reported genotype frequencies, we calculate the deviation from HWE at P = 10 –92 ) and modest deviation in ATAC and others were at the core of the outcry regarding these widely heralded negative studies. In a randomly mating population, such as in the United States, Europe, and Asia, where these data arise, deviation from HWE is used to signal genotyping errors. Statistically significant deviations from HWE raise substantial concerns about the quality and accuracy of the genotype data. Chromosomal instability in breast cancer tissue at the CYP2D6 locus on chromosome 22q13 has led to concern about genotyping errors based on genotypes determined from breast cancer tissue. In this issue of the Journal, Goetz and colleagues seek to resolve the ongoing dispute regarding the appropriateness of use of FFPE breast tumor tissue for CYP2D6 ( 17 ). They genotyped nearly 1200 samples and provide clear evidence for loss of heterozygosity (ie, loss of one chromosomal allele at the CYP2D6 locus) in over a third of the HR+ FFPE tumor samples. This was accompanied by statistically significant deviations from HWE in the genotype calls. They also documented that FFPE nonmalignant tissue does not exhibit loss of heterozygosity or deviation from HWE, indicating that it is the breast tumor tissue, and not FFPE samples per se, that is problematic for use in CYP2D6 genotyping. They also report that while there was an approximately 20% discordance between FFPE tumor and buccal cell genotypes, there was 100% concordance between FFPE nonmalignant and buccal cell genotypes. In essence, the chromosomal instability in the CYP2D6 locus in breast tumor tissue means that DNA from this source is not reflecting germline DNA in many patients and in particular is not reflecting the CYP2D6 genotype in the liver. These data provide clear evidence that CYP2D6 genotyping should not be done on breast tumor samples, but rather on adjacent normal tissue, or preferably a traditional germline DNA source for genotyping such as white blood cells from blood or saliva or buccal cells. They also raise questions about the results of trials that relied exclusively on FFPE tumor tissue for genotyping, which include ATAC ( 11 ), BIG 1–98 ( 12 ), and the tamoxifen exemestane adjuvant multinational (TEAM) trial ( 13 ), which are the largest among the negative trials, and all of which had statistically significant deviations from HWE for the CYP2D6*4 genotypes. BIG 1–98 is clearly the most alarming, regarding concerns about erroneous genotypes. The original positive study also falls into the category of creating concern about the findings because of the HWE deviation ( 10 ). While the positive association studies did not exclusively use nontumor tissue for their genotyping, one cannot help but notice that the positive studies were much more likely to use nontumor tissue for genotyping and their genotyping results did not deviate from HWE or deviated with much less statistical significance (see Table 1 ). A basic tenet of genetics research is that the quality of the genotype data are confirmed before genetic association analyses with the genotype data are considered. Unfortunately, this basic tenet was violated in a substantial percentage of the CYP2D6-tamoxifen studies that comprise the current literature. Studies evaluating impact of CYP2D6 on tamoxifen efficacy in breast cancer patients with the CYP2D6 –poor metabolizer phenotype * CI = confidence interval; EM = extensive metabolizer; FFPE = formalin-fixed paraffin-embedded; HR = hazard ratio; HWE = Hardy Weinberg Equilibrium; OR = odds ratio; PM = poor metabolizer. † Included studies are those for which the *4 allele was the major loss of function allele tested, normal doses of tamoxifen were used, and substantial portions of the study population were not HR negative and for which clinical outcomes were the primary analysis, testing outcomes in PMs vs EMs. ‡ P value was calculated by editorial authors based on genotype frequencies reported in each study using Chi-squared test with 1 degree of freedom. If genotype data were not provided in the paper for editorial authors to calculate, then report in the paper by the original authors is included. Studies evaluating impact of CYP2D6 on tamoxifen efficacy in breast cancer patients with the CYP2D6 –poor metabolizer phenotype * CI = confidence interval; EM = extensive metabolizer; FFPE = formalin-fixed paraffin-embedded; HR = hazard ratio; HWE = Hardy Weinberg Equilibrium; OR = odds ratio; PM = poor metabolizer. † Included studies are those for which the *4 allele was the major loss of function allele tested, normal doses of tamoxifen were used, and substantial portions of the study population were not HR negative and for which clinical outcomes were the primary analysis, testing outcomes in PMs vs EMs. ‡ P value was calculated by editorial authors based on genotype frequencies reported in each study using Chi-squared test with 1 degree of freedom. If genotype data were not provided in the paper for editorial authors to calculate, then report in the paper by the original authors is included. The CYP2D6- tamoxifen story is complicated. Genotyping this gene is difficult in normal settings, because it has a large number of single-nucleotide polymorphisms, gene deletions, duplications, and multiplications, along with adjacent pseudogenes, all of which make determination of the CYP2D6 genotype–determined phenotype the most complicated in pharmacogenetics. The tamoxifen–breast cancer literature has unfortunately been further complicated by the fact that tumor tissue samples were commonly used and the CYP2D6 locus happens to be one that is often deleted (or an allele is deleted) in breast tumors, which can lead to genotypes that do not reflect the germline genotype and thus are erroneous genotypes. Those working in this field need to determine how to wipe the slate clean, and get the genotype data right. Pharmacological principles, the published literature (including most of the negative studies, which have point estimates suggesting risk in PMs), and a meta-analysis by the International Tamoxifen Pharmacogenetics Consortium ( 18 ), at a minimum are suggestive that CYP2D6 metabolizer status influences the efficacy of tamoxifen. Whether it does influence efficacy remains to be determined from properly conducted studies, wherein genotyping is properly done on appropriate DNA sources and other factors are appropriately considered. It is also clear that resolving issues surrounding genotype accuracy and correcting the genotypes alone will not resolve the questions about the role of CYP2D6 genotype or endoxifen on outcomes with tamoxifen. For example, a recent study suggests there is an endoxifen threshold concentration that needs to be reached in order for women to have good outcomes, and while PMs were overrepresented in the group that failed to reach that threshold it was not a perfect correlation ( 19 ). Thus, the answer to personalization of tamoxifen therapy might lie in validating these findings and then dosing tamoxifen to exceed the threshold endoxifen concentration, without a focus on CYP2D6 genotype per se. The impact of potent inhibitors of CYP2D6 (which can convert someone to a phenotypic PM) must also be considered moving forward if the truth about CYP2D6 and tamoxifen is to be revealed. The CYP2D6-tamoxifen story is not closed. In fact, there should be a reboot, with a focus on doing the genotyping and analyses correctly. Women with breast cancer deserve for the scientific community to continue to work on this question and to get it right. It is unacceptable that a woman might be placed on 10 years of therapy with a drug for which her genotype predisposes her to reduced efficacy and poor outcomes. She deserves an evidence base that can truly guide the most appropriate treatment for her. This work was supported by National Institutes of Health Pharmacogenomics Research Network grant U01-GM074492 (to JAJ). The funder had no role in the writing of the editorial nor the decision to submit it for publication. The authors have no conflicts of interest to declare.
No takes yet. Share an insight, caveat, or question.
Johnson et al. (2015) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: