Prediction of tamoxifen outcome by genetic variation of CYP2D6 in post-menopausal women with early breast cancer

Abstract

The question of whether genetic polymorphisms of CYP2D6 can affect treatment outcome in patients with early post-menopausal oestrogen receptor (ER)-positive breast cancer has been a matter of debate over the past few years. In this article we revisit the hypothesis of CYP2D6 being a potential tamoxifen outcome predictor and provide detailed insight into the ongoing controversy that prevented the CYP2D6 marker from being accepted by the scientific and clinical community. We summarize the available pharmacokinetic, pharmacodynamic and pharmacogenetic evidence and resolve the controversy based on the recognized methodological and statistical issues. The cumulative evidence suggests that genotyping for CYP2D6 is clinically relevant in post-menopausal women. This is important, because the clarification of this issue has the potential to resolve a clinical management question that is relevant to hundreds of thousands of women diagnosed with ER-positive breast cancer each year, who should not be denied effective endocrine therapy.

Tamoxifen, the kick start of personalized medicine in oncology

Tamoxifen, a selective oestrogen receptor modulator (SERM) is considered the first targeted treatment in oncology heralding the era of personalized medicine 1. Tamoxifen was approved by the Food and Drug Administration (FDA) for the treatment of metastatic breast cancer already in 1977 and is meanwhile labelled for the adjuvant treatment of early breast cancer as well as the prevention of developing breast cancer in women at high risk 2. The targeting of tamoxifen to block oestrogen stimulated breast tumour growth with long term adjuvant treatment resulted in a major improvement of the reduction of recurrences and deaths worldwide 3. Typically, adjuvant endocrine therapy involves 5 years of treatment but new data from the ATLAS trial showed that 10 years of treatment is even better than 5 years and results in a 50% reduction of mortality 5002. Recently, tamoxifen pharmacogenomic research, which investigates the prediction of clinical outcome based on the patient's constitutional CYP2D6 genotype, sparked new hopes for evidence based treatment decisions in order to identify those patients likely not to respond to tamoxifen and guide them towards alternative endocrine therapy 5003,6. In the light of a worldwide disease burden of over 1.38 million women newly diagnosed with breast cancer each year and 458 000 annual deaths 7, this provides an important prospect for the majority of these patients who have hormone receptor-positive breast cancer, specifically classified with oestrogen receptor (ER)-or progesterone receptor-positive status or both. More than two-thirds are characterized as having ER-positive status of which all premenopausal patients are treated with tamoxifen. For post-menopausal patients, who account for 75% of all diagnosed breast cancers, two equally potent endocrine therapies, tamoxifen and aromatase inhibitors (AIs) are available as standard adjuvant therapy for early breast cancer. The current goal is to guide patients with a suitable outcome predictor towards their best treatment option based on their personal capacity to respond to tamoxifen. The CYP2D6 polymorphism would be such a predictor, which by definition of the National Institutes of Health (NIH) 8 serves as a characteristic for objective measurement and evaluation as an indicator of a pharmacological response.

Scope and controversy of tamoxifen CYP2D6 pharmacogenetics

Tamoxifen pharmacogenetic studies in post-menopausal patients with ER-positive early breast cancer moved to centre stage in 2005, when investigators at the Mayo clinic in their lead study tested the polymorphic cytochrome P450 (CYP) 2D6 enzyme as a potential outcome predictor 9. This became feasible, because the metabolism of tamoxifen and action at the ER together with the in-depth knowledge of the complex CYP2D6 polymorphism, all obtained during decades of intense in vitro, animal in vivo and clinical research, are very well understood 10–12. Goetz et al. reported that CYP2D6 poor metabolizers (PMs) (based on CYP2D6*4 and *6) have less benefit from tamoxifen compared with extensive metabolizers (EMs), results that were obtained from a retrospective analysis of a small prospective monotherapy study from the USA 9. The approach utilizes the pronounced inter-individual variation of CYP2D6 phenotypes known as EMs for normal activity, intermediate metabolizers (IMs) for reduced activity, PMs for no activity and ultra-rapid metabolizers (UMs) for higher than normal activity 13,14 which determine an individuals' capacity to metabolize CYP2D6 substrates including tamoxifen. These phenotypes can be inferred from genotypes due to the highly polymorphic nature of the CYP2D6 gene. More than 100 genetic variants are known 15 of which major non-functional/null function (PM) alleles include CYP2D6 *3, *4, *5, *6, major decreased function (IM) alleles include CYP2D6 *10, *17, *41, and increased function (UM) alleles include *1xN, *2xN. Their frequencies and inter-ethnic variations as well as impact on CYP2D6 function are reviewed by Zanger & Schwab 12.

Notably, the CYP2D6 genetic marker approach, despite its popular use, must be regarded as second best to a more preferable direct biomarker approach of measuring the active metabolite concentrations in the patients' plasma for a correlation with clinical outcome. However in oncology, marker based outcome studies depend on clinical endpoints that become available only years after the completion of long term drug treatment. In the case of adjuvant tamoxifen, this may take more than 10 years from the start of treatment to the occurrence of sufficient breast events (local or distant recurrence) required for the analysis. Therefore, the major limitation is the availability of large patient cohorts with the respective clinical follow-up and the collection of their biological materials for marker analysis. A direct biomarker approach for measuring tamoxifen and its metabolites has never been feasible due to the lack of required plasma samples for the patients under investigation because their tamoxifen treatment was many years ago and did not include baseline and steady-state plasma collections. Although the same applies to the genetic approach, the limitation has been overcome by the use of genomic DNA which can be isolated from archived tissue obtained at the time of surgery and histopathological diagnosis routinely stored at respective pathology departments. The long term stability of genomic DNA and easy access from formalin-fixed paraffin-embedded (FFPE) tissue allowed the recruitment of suitable study cohorts with available long term follow-up and DNA source, which finally set the stage for the conduct of tamoxifen pharmacogenetic studies. However, these studies are retrospective due to the lack of implementation into previous clinical trial protocols and upfront collection of the required study materials. Consequently, the published studies widely differ by study design including size, end point definition and inclusion criteria, length of follow-up and DNA source (Figure 1), as well as extent of CYP2D6 genotyping and assay validation. Therefore, it is of no surprise that they also differ in their results. Within 7 years of international research and following about two dozen international publications, the association between CYP2D6 PM status and non-favourable tamoxifen outcome appeared to be non-reproducible leading to confusion and intense controversy within the scientific and clinical community 16–21.

Figure 1.

Attention must be paid when genomic DNA is isolated for pharmacogenetic investigations. DNA isolated from peripheral blood mononuclear cells represents the true germline genome and is ideal for accurate genotyping. If tumour tissue is used for genomic DNA isolation, the genomic DNA isolate usually contains a mixture of the germline and tumour genome. Because tumour cells are prone to somatic mutations, loci that are heterozygous in the germline can appear homozygous during genotyping due to loss of heterozygosity (LOH). Because this will have profound consequences for phenotype assessment such as CYP2D6 metabolizer status, genotype frequencies must be quality controlled by testing of the Hardy–Weinberg equilibrium (HWE). Deviations from HWE indicate genotyping errors which may arise from high tumour content in the DNA source tissue thereby masking the constitutional genotype. If formalin-fixed paraffin-embedded (FFPE) tissue is used it is important to avoid core punches because this will enrich for tumour cells. Rather, whole sections should be used that contain sufficient amounts of normal (stroma) cells

Yet, a large study from Germany and the USA published in the Journal of the American Medical Association in 2009 5003 verified the tamoxifen CYP2D6 outcome effect in post-menopausal early breast cancer by including 1325 patients with adjuvant monotamoxifen treatment (median follow-up of 6.3 years) into their study and performing accurate CYP2D6 phenotype assignments via a comprehensive genotyping approach. The study concluded that the tamoxifen treatment benefit for CYP2D6 EM patients is similar to that of a hypothetical AI treatment, but worse for CYP2D6 PM and deficient patients who should not receive tamoxifen, but AI upfront (Figure 2). Despite the high impact publication, an influential part of the scientific and clinical community continued to question the validity of these data and hesitated to implement CYP2D6 testing into the clinic. This controversy became even more severe following the publication of negative findings from the Breast International Group (BIG) 1–98 22 and Arimidex, Tamoxifen, Alone or in Combination (ATAC) 23 groups that reported on pharmcogenetic findings in patient subgroups of trials originally investigating the efficacy of tamoxifen vs. AI in early breast cancer 24,25. These retrospective analyses of prospectively collected patient cohorts were expected to settle the controversy. Based on their negative findings they concluded that testing for CYP2D6 has no value in clinical practice 22,23. Immediately after their publications in the Journal of the National Cancer Institute in March 2012, numerous experts in their Letters to the Editor revisited the issue and identified severe shortcomings related to the associated technology and statistical procedures 26–28. Finally, the underlying reasons for the ongoing controversy over the CYP2D6 tamoxifen pharmacogenetic issue have been explained [6, 26–28] culminating in the demand for the retraction of the BIG1-98 CYP2D6 tamoxifen study [26].

Figure 2.

Kaplan–Meier estimates of recurrence probabilities comparing tamoxifen with a hypothetical aromatase inhibitor (AI) curve (adapted from 5002). 2D6 EM indicates extensive metabolism (i.e. patients with two functional CYP2D6 alleles, including patients with ultrarapid metabolism), decreased, patients with any intermediate or poor metabolism alleles. Nonadjusted, heterogeneity-corrected Kaplan–Meier estimates for the CYP2D6 decreased and EM phenotypes as well as the entire tamoxifen cohort unselected by genotype; 95% confidence interval (CI) is shown for EM patients (shaded area). Assuming the Cox proportional hazards assumption, a hypothetical AI survival curve (red) was estimated based on a hazard ratio of 0.76 for anastrozole relative to tamoxifen 54 and the Kaplan–Meier estimate of the entire tamoxifen cohort 5002

DNA source and Hardy–Weinberg equilibrium are the key

A key issue in all genetic studies is the quality of the primary genetic data as no inference can be drawn from genotype data of low quality 26. Determining constitutional genotypes from FFPE tumour tissue is technically challenging, particularly with a complex gene like CYP2D6 that exhibits copy number variation and is flanked by two highly similar pseudogenes 27. Moreover, the potential for genotyping errors attributable to allele loss in breast cancer must be considered. The latter refers to loss of heterozygosity (LOH), a hallmark of cancer in general but in particular of breast cancer. Therefore, it is of no surprise that the chromosome 22q13 locus which harbours the CYP2D6 gene is frequently deleted in breast cancer (Figure 3A). LOH at 22q was initially demonstrated by microsatellite DNA analysis using multiple dinucleotide (CA)_n repeats for the comparison of matched normal and tumour DNA of clinical cases with 53% of the tumours being affected 29. Notably, 22q13 deletions were associated with worse prognosis in a Japanese study with 32% of the patients' tumours having 22q13 LOH 26,30. With regard to tumour subtype, LOH affected more than 25% of ER-positive invasive lobular and invasive ductal carcinomas as shown by genome-wide fractional allelic loss 31. Although these studies do not specify CYP2D6 as the deleted gene, recent data from The Cancer Genome Atlas (TCGA) dataset confirmed that the CYP2D6 locus is deleted in 35% of ER-positive tumours (Figure 3B, Perou, TCGA Breast Data, personal communication, March 20, 2013). From this it follows that genomic DNA samples isolated from FFPE breast tumour tissue and in particular tumour core samples may not be representative for the germline genome of the patients under investigation (Figure 1).

Figure 3.

(A) Schematic overview of the CYP2D6 genetic locus on chromosome 22q13 and tumour loss of heterozygostiy (LOH). (B) SWITCH plots indicating LOH at the CYP2D6 locus. 29% of all breast tumours of the The Cancer Genome Atlas (TCGA) breast cancer cohort 55 are affected by LOH. However, this somatic defect is present in 35% of ER-positive tumours (C. Perou, TCGA Breast Data, personal communication, March 20, 2013) 56

A standard reporting practice for genotyping quality is the consistency of measured genotype proportions with those expected under the Hardy–Weinberg equilibrium (HWE). The HWE states that gene frequencies and genotype ratios in a randomly-breeding population remain constant from generation to generation. Thus, the finding of severe deviation from HWE in the BIG 1–98 study with a strong allelic imbalance for the CYP2D6*4 allele (P = 2.5 × 10⁻⁹²) indicates a severe genotyping error 27,28 that may be contributed to by 22q13 LOH, but also technical issues may account for it. This renders the results and final conclusion of the BIG 1–98 CYP2D6 study and for similar reasons also the ATAC pharmacogenetic study non-conclusive in that they cannot contribute to resolve the controversy. As it stands, the most significant publication to date is the 2009 JAMA publication by Schroth et al. 5002 based on 1325 hormone receptor positive adjuvant tamoxifen treated patients and their comprehensive CYP2D6 genotypes analyzed in genomic DNA samples derived from blood and whole FFPE tissue sections for the assessment of accurate CYP2D6 phenotypes 5002,6,32.

The evidence and why CYP2D6 matters

Tamoxifen is a pro-drug

The pharmacological evidence supports the tamoxifen pro-drug theme in that tamoxifen undergoes extensive metabolism by multiple CYP enzymes resulting in at least two active metabolites 4-OH-tamoxifen and endoxifen 33,34. The formation of endoxifen primarily depends on CYP2D6 35,36. Breast cancer patients treated with tamoxifen have less 4-OH-tamoxifen in their serum, which at the standard dose of 20 mg day⁻¹ corresponds to 10 to 20% of the concentration of endoxifen, a reason why endoxifen is considered the major metabolite 37,38.

There is a strong CYP2D6 gene−dose effect

Endoxifen concentrations are highest in UM (77 nmol l⁻¹) and EM patients (36.9 nmol l⁻¹) but lowest in PM (9.9 nmol l⁻¹) patients as reported from a large prospective observational trial (P_trend = 10⁻¹⁶) 38. PM concentrations reach about 25% of those measured in EM patients and increase with increasing numbers of function alleles.

Endoxifen's activity at the oestrogen receptor is concentration dependent

An in vitro model of breast cancer cells exposed to a fixed mixture of clinical concentrations of tamoxifen and its major metabolites but varying endoxifen concentrations showed that endoxifen's effect on transcription and inhibition of proliferation was concentration dependent, with minimal effects at low concentrations (< 20 nmol l⁻¹ as in PMs) but significantly greater effects occurring at higher concentrations (40 to 60 nmol l⁻¹, as in IMs, EMs) 37.

Endoxifen is clinically relevant

High but not low endoxifen lconcentrations correlated with a 26% reduced breast cancer recurrence rate as shown in a retrospective analysis of 1370 women treated with tamoxifen in the Women's Healthy Eating and Living (WHEL) trial (HR = 0.74, 95% CI 0.55, 1.00) 39. In particular, women with an impaired CYP2D6 metabolizer phenotype (PM, IM) were more likely to be in the low endoxifen bottom quintile group with a high risk for recurrence. Breast cancer outcomes were not correlated with concentrations of tamoxifen, 4-OH-tamoxifen and N-desmethyl-tamoxifen 39.

CYP2D6 is a tamoxifen outcome predictor

Following the first report 9, the tamoxifen benefit of EMs over PMs was confirmed within a cohort of patients with adjuvant tamoxifen monotherapy (HR = 1.90, 95% CI 1.10, 3.28) which represents the largest reported study to date involving 1325 women with tumour progression status 5002. More confirmatory data came from the retrospective analysis of CYP2D6 genotype from the Austrian Breast and Colorectal Cancer Study Group 8 (ABCSG8) clinical trial which demonstrated that CYP2D6 PM patients treated with tamoxifen had significantly higher odds of a disease event compared with EM patients (OR = 2.45, 95% CI 1.05, 5.73) 40. Importantly, the study by Schroth et al. to a large extent included genomic DNA samples derived from peripheral blood mononuclear cells for genotyping without violation of HWE. Genotypes obtained from whole mount tumour sections with an average non-tumour cell content of 51% showed minor deviation from HWE (P = 0.015) indicating sufficient normal DNA in most of the isolates for mainly accurate phenotype assignments 6. Moreover, in the ABCSG8 CYP2D6 study, unlike in ATAC 23 and BIG 1–98 22, the common CYP2D6*4 allele was within HWE 40. Other recently published data from the Tamoxifen Exemestane Adjuvant Multinational (TEAM) trial, despite control for LOH, cannot contribute to the interpretation of CYP2D6 tamoxifen outcome prediction 41. A plausible explanation considered by the authors is the switch to exemestane after 2.5 years of tamoxifen and the censoring of disease-free survival at that time point, thereby preventing the insight into the question of interest, namely the likelihood of late breast cancer recurrence following long term tamoxifen.

Doubling of the tamoxifen dose increases endoxifen concentrations

A genotype-guided tamoxifen dosing approach showed that doubling the daily dose from 20 mg to 40 mg led to a significant rise in endoxifen concentrations in IM and PM patients. Whereas IM patients reached endoxifen concentrations comparable with those of EM patients receiving the standard dose, this was not the case for PMs 42. Notably, successful dose adjustment is the level of data usually required for FDA-prescribing recommendations after organ dysfunction, drug interaction or age, suggesting relevance for CYP2D6 guided tamoxifen dosing in routine clinical practice in order to reduce inter-patient variation and under dosing of patients with breast cancer 43.

CYP2D6 inhibitors block tamoxifen efficacy

CYP2D6 efficacy can be blocked by strong inhibitors such as selective serotonin re-uptake inhibitors (SSRI) including paroxetine and fluoxetine given as co-medication for the relief of tamoxifen-induced post-menopausal symptoms 44,45. The use of potent inhibitors can result in a change of phenotype from EM to PM known as phenocopying. Thus, drug interactions are a major confounding factor of the CYP2D6 phenotype and need to be avoided despite other cytochrome P450 enzymes contributing to the formation of endoxifen (CYP2C9 contributes to the formation of 4-OH-tamoxifen, which provides the source of 20 to 30% total endoxifen) 38,46. The recommendation to avoid strong CYP2D6 inhibitors during tamoxifen treatment is straightforward and requires the prescription of medications with little CYP2D6 inhibitory effect such as venlafaxine or citalopram 47. The American Society of Clinical Oncology (ASCO) recommends caution when CYP2D6 inhibitors are co-administered during tamoxifen treatment 48.

Present recommendations and prescribing information

Despite the strong mechanistic and clinical support in favour of an association between CYP2D6 genotype and tamoxifen efficacy current recommendations and prescribing information issued by regulatory authorities in the United States and Europe differ with regard to the recommendation of CYP2D6 genotyping prior to tamoxifen treatment. The personalized medicine overview of Shah & Shah very recently summarized the tamoxifen prescribing information taking into account pharmacogenetic information 49. Accordingly, the consensus of the US Clinical Pharmacology Subcommittee of the FDA Committee of Pharmaceutical Sciences in October 2006 was that the US label of tamoxifen should be updated to reflect the increased risk for breast cancer along with mechanistic data, but there was disagreement on whether CYP2D6 genotyping should be recommended. It was also concluded that there was no direct evidence of a relationship between endoxifen concentrations and clinical response 50. Consequently, the US label for tamoxifen does not include any information on the relevance of CYP2D6 polymorphism and has not been updated until now. In the European Union (EU), the prescribing information from the Drug Safety Update Bulletin from the UK Medicines and Healthcare products Regulatory Agency (MHRA) was revised in November 2010 in that concomitant use of medicines known to be potent CYP2D6 inhibitors should be avoided whenever possible in patients treated with tamoxifen for breast cancer 51. However, the evidence linking various PM genotypes and tamoxifen treatment outcomes is mixed and inconclusive. Therefore, it emphasized that there was no recommendation for genetic testing before treatment with tamoxifen 51.

Finally, the German Federal Institute for Drugs and Medicinal Products (Bundesinstitut für Arzneimittel und Medizinprodukte, BfArM) acknowledges in the updated Summary of Product Characterization of April 2013, that both drug interactions (the pharmacokinetic inhibition of the CYP2D6 enzyme by strong inhibitors) and CYP2D6 polymorphism can reduce the efficacy of tamoxifen during breast cancer treatment 52, but a clear recommendation for genetic testing has not been issued.

Conclusions

The evidence presented and discussed in this review now explains many of the discrepant findings or contradiction in the data regarding the relevance of CYP2D6 genetics for the prediction of tamoxifen outcome in post-menopausal breast cancer patients reported from previous studies. We consider the cumulative evidence sufficient to accept the CYP2D6 tamoxifen pharmacogenetic relationship in post-menopausal women and we explain under which conditions CYP2D6 genotyping has clinical relevance. Moreover, we describe how future prospective studies should be done. The clinical community should harness this information to resolve a clinical management question affecting two-thirds of women diagnosed with breast cancer, and this should not be further delayed, particularly since numerous, in part FDA approved, genotyping assays are available 53. Given that alternatives to tamoxifen exist, the failure to resolve this question is particularly egregious in that the potential exists to deny women effective endocrine therapy.

Competing Interests

All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare HB and MS report scientific collaborations with Roche Molecular Diagnostics and Siemens Healthcare Diagnostics Products GmbH in the previous 3 years. No other relationships or activities exist that could have influenced the submitted work.

This work was supported by The Bosch Foundation Stuttgart, Germany, the 7FP EU Marie Curie Initial Training Network ‘FightingDrugFailure’ (GA 238132), the Federal Ministry for Education and Research (BMBF), Germany (grants 03IS2061C and 01ZP0502) and the IZEPHA grant 21-0-0.