ESR1 and ESR2 polymorphisms in the BIG 1-98 trial comparing adjuvant letrozole versus tamoxifen or their sequence for early breast cancer

Abstract

Purpose

Estrogen receptor 1 (ESR1) and ESR2 gene polymorphisms have been associated with endocrine-mediated physiological mechanisms, and inconsistently with breast cancer risk and outcomes, bone mineral density changes and hot flushes/night sweats (HF/NS).

Methods

DNA was isolated and genotyped for six ESR1 and two ESR2 single nucleotide polymorphisms (SNPs) from tumor specimens from 3691 postmenopausal women with hormone receptor-positive breast cancer enrolled in the BIG 1-98 trial to receive tamoxifen and/or letrozole for five years. Associations with recurrence and adverse events (AE) were assessed using Cox proportional hazards models.

Results

3401 samples were successfully genotyped for five SNPs. ESR1 rs9340799(XbaI)(T>C) variants CC or TC were associated with reduced breast cancer risk (HR=0.82,95%CI=0.67–1.0), ESR1 rs2077647(T>C) variants CC or TC was associated with reduced distant recurrence risk (HR=0.69,95%CI=0.53–0.90), both regardless of the treatments. No differential treatment effects (letrozole vs. tamoxifen) were observed for the association of outcome with any of the SNPs. Letrozole-treated patients with rs2077647(T>C) variants CC,TC had a reduced risk of bone AE (HR=0.75,95%CI=0.58–0.98, P_interaction=0.08), whereas patients with rs4986938(G>A) genotype variants AA and AG had an increased risk of bone AE (HR=1.37,95%CI=1.01–1.84, P_interaction=0.07).

Conclusions

We paradoxically observed 1) rare ESR1 homozygous polymorphisms were associated with lower recurrence, and 2) ESR1 and ESR2 SNPs were associated with bone AEs in letrozole-treated patients. Genes that are involved in estrogen signaling and synthesis have the potential to affect both breast cancer recurrence and side effects, suggesting that individual treatment strategies can incorporate not only oncogenic drivers but also SNPs related to estrogen activity.

Introduction

The relationship between estrogen receptors (ER) and their ligand, the hormone estrogen (17β-estradiol), and the enzymes that synthesize it are not well understood. The ESR1 and ESR2 genes encode for ERα and ERβ, respectively, which are responsible for the effects of estrogens and are therapeutic targets of selective estrogen receptor modulators (SERMs) including the first breast cancer targeting drug tamoxifen. The ERs are complex genes with many polymorphic and splice variants. Genetically determined variants in sex steroid hormone pathways have been related to several measures of health status (i.e. circulating hormone concentrations, menstrual cycle profiles, lipids, diabetes mellitus, depressive symptoms, measures of cognition, bone mineral density (BMD) and vasomotor symptoms) in a community-based population of premenopausal women[1]. Polymorphisms in the ESR1 and ESR2 genes have been reported to be associated with multiple endocrine-mediated physiological mechanisms including lipid profile[2–6], mammographic density[7], venous thromboembolism[8, 9], cognition[10]; and inconsistently associated with breast cancer risk and outcomes and BMD changes[11–17].

The role that ESR1 and ESR2 polymorphisms play in breast cancer outcomes and the effects that they exert on BMD in postmenopausal women requires further studies to confirm any effect they may have in other populations. No pharmacogenetic studies have been published on ESR1 and ESR2 polymorphisms and the differential effectiveness or side effect profile of SERMs or aromatase inhibitors (AIs) in postmenopausal women with breast cancer.

We investigated the clinical relevance of ESR1 and ESR2 SNPs in postmenopausal women treated with tamoxifen and/or letrozole in the Breast International Group (BIG) 1-98 trial. We hypothesized that ESR1 and/or ESR2 phenotypes that potentially result in receptor activity changes could affect disease outcomes that varies by type of endocrine therapy. We also investigated potential associations between ESR1 and ESR2 SNPS and hot flushes or night sweats (HF/NS) during the first two years of treatment, and bone adverse events (AEs) during and subsequent to treatment.

Materials and Methods

Patients

The BIG 1-98 trial is a randomized, phase III, double blind study that recruited postmenopausal women with estrogen receptor (ER) and/or progesterone receptor (PgR)-positive early breast cancer. From 1998 to 2000, women were randomly assigned to receive monotherapy with letrozole (Femara®, Novartis) 2.5mg daily or tamoxifen 20 mg daily for five years, and from 1999 to 2003, to one of four arms: tamoxifen or letrozole for five years or sequential therapy comprising letrozole for two years followed by tamoxifen for three years, or tamoxifen for two years followed by letrozole for three years[18–22].

A total of 8010 women were enrolled between March 1998 and May 2003. After the initial trial results were released in 2005, patients assigned to tamoxifen monotherapy were offered the chance to switch to letrozole for the remainder of their adjuvant therapy, and 619 (25.2%) of them did so.

All participants provided written informed consent. Ethics committees and relevant health authorities approved the protocol. Trial participants were followed clinically at baseline, every 6 months for the first 5 years during blinded study drug dispensing, and yearly thereafter. AEs were recorded at study visits and graded according to the Common Toxicity Criteria (CTC) v2.0.

Tissue Collection, DNA Extraction, and ESR1 and ESR2 Genotyping

The International Breast Cancer Study Group (IBCSG), between 1998 and 2010 in accordance with institutional guidelines and national laws, carried out retrospective tissue collection. Novartis, the BIG 1-98 pharmaceutical partner, provided funding to partially cover associated institution costs. The IBCSG Biological Protocols Working Group approved this project. All material processing and genotyping were done without the knowledge of patients’ treatment assignments or outcomes. Formalin-fixed, paraffin-embedded (FFPE) primary breast cancer tissue blocks for 5786 of 8010 enrolled in BIG 1-98 patients were assessed for availability of material for translational research. An area that was representative of the invasive tumor component was identified, and one or two 1-mm cores were punched in this area. DNA was isolated for potential ESR1 and ESR2 genotyping for 3691 of 8010 trial patients, and 3401 were included in the analytic cohort (Figure 1).

Figure 1.

Flow diagram showing the derivation of the analytic cohort of 3401 patients.

Genomic DNA was extracted using the QIAamp DNA FFPE tissue kit (Qiagen, Valencia, CA) according to manufacturer’s instructions. DNA was eluted with 60 μL of sterile distilled water, quantified and quality controlled according to the 260/280 nm ratio using the Infinite M 200 NanoQuant (Tecan, Mannedorf, Switzerland) and aliquoted at a concentration of 10 ng/μL.

DNA samples were genotyped for six ESR1 (rs9340799(XbaI), rs2234693(PvuII), rs11963577, rs2077647, rs9341070, rs746432) and two ESR2 (rs4986938, rs1256049) SNPs, selected based on location and/or whether they had been evaluated in prior studies. Five of eight SNPs (4 in ESR1, 1 in ESR2) passed genotype quality assurance and were included in our statistical analyses. These SNPs included three ESR1 intronic SNPs, rs9340799 T>C(XbaI), rs2234693 T>C(PvuII) and rs11963577 C>T; and one ESR1 synonymous SNP, rs2077647 T>C; and one ESR2 SNP in the 3′UTR, rs4986938 G>A.

All samples were processed using polymerase chain reaction (PCR)-based methods and were genotyped on the GenomeLab SNPstream Genotyping System (Beckman Coulter, Brea, CA) in a Clinical Laboratory Improvement Amendments (CLIA)-compliant laboratory. All reagents were preformulated and included in the GenomeLab SNPware Reagent kit. The GenomeLab SNPstream Genotyping System Software Suite v2.3 and SNP genotyping software (ABI) were used for array imaging and genotype calling. One SNP rs2234693 was re-run on the 7900HT Fast Real-Time 200 PCR System (Applied Biosystems, Foster City, CA) using a commercially-available predesigned TaqMan assay. The optimization employed Applied Biosystems’ TaqMan probes, in conjunction with the KlearKall Mastermix from KBioscience (Beverly, MA).

The ESR1 and ESR2 genotypes were tested for deviations from Hardy-Weinberg Equilibrium (HWE). Significant violations were not observed across most sites, with some minor exceptions (Supplementary Table S1). As an additional control, we compared genotypes derived from tumor versus whole blood DNA in 120 matched samples as an extension of our recently published report[23]. All five of the ESR genotypes tested (ESR1:rs9340799, rs2234693, rs11963577 and rs2077647; ESR2:rs4986938) showed very high to complete concordance between tumor versus whole blood (97.3% to 100%) suggesting that tumor-derived genotypes accurately represented germline genotypes.

Statistical Analyses

Disease endpoints were: 1) breast cancer-free interval (BCFI): time from randomization to first occurrence of invasive breast cancer recurrence at a local, regional, or distant site, or a new invasive cancer in the contralateral breast; and 2) distant recurrence-free interval (DRFI): time from randomization to recurrence at a distant site. In the absence of an event, the endpoints were censored at last follow-up time, or in the analyses focused on 5-year monotherapy, at time of selective crossover from tamoxifen to letrozole after dissemination of the primary trial results in 2005[18]. Cox proportional hazards models were stratified by randomization option (2 or 4-arm), and prior chemotherapy (and by treatment assignment for analyses that included all patients). The models adjusted for these characteristics at randomization: local therapy (mastectomy vs. breast-conserving surgery), tumor grade (2/3 vs. 1/unknown), tumor size (>2cm vs. ≤2cm/unknown), nodal status (positive vs. negative/Nx), peritumoral vascular invasion (present vs. absent/unknown), HER2 status (positive vs. negative/unknown) and Ki-67 labeling index (≥14% vs. <14%/unknown). The proportional hazards assumption was assessed by testing for genotype-by-time interaction and using Schoenfeld residuals.

To investigate the association of ESR1and ESR2 genotypes with early onset of HF/NS, all treatment groups were included (N=3401) (Figure 1), and the endpoint was defined as the time from randomization to the first report of new or worsening HF/NS of any grade within 23 months of randomization (prior to the treatment switch in the sequential arms). The Cox proportional hazards model was adjusted for race (white vs. other), age (years, in quartiles), body mass index (<18.5,18.5–29.9,>30 kg/m²), history of HRT use (never, within last 3 months, ≥3 months ago), and baseline HF/NS (present vs. absent).

For the association of bone AEs, analyses were limited to patients assigned to five years of letrozole or tamoxifen monotherapy (N=1940) (Figure 1). Time to bone AE was defined as time from randomization to first report of grade 3–4 osteoporosis or any bone fracture during or subsequent to protocol treatment. Cox regression models were adjusted for age, BMI, prior HRT use, smoking history (yes vs. no/unknown), history of osteoporosis, bone fracture or bisphosphonate use (yes vs. no/unknown), and bisphosphonate use since randomization as a time-varying covariate.

We first assessed SNP variant effects in a genotype model using the logrank test and an additive model (df=1) that compared 0 vs. 1 vs. 2 (the number indicates minor alleles in the three variant categories: 0=common allele, 1=minor allele) using a trend test. If appropriate, dominant and recessive effects of the variant (rare) allele were also estimated, where the dominant model combined variant (rare) homozygote and heterozygote compared with the wild-type homozygote (reference), or recessive model combined variant (rare) homozygote was compared versus the combined heterozygote and wild-type homozygote (reference).

With a range of 616–3288 assessable patients, 10–36% event rates, and 13–69% prevalence of variant genotypes (assumed for calculations to be associated with increased hazard of event), there was 80% power (two-sided type I error of 0.05) to detect hazard ratios ranging from 1.2 (for n=3067, 36% event rate, 69% prevalence of variant genotype scenario) to 2.9 (for n=613, 10% event rate, 13% prevalence of variant genotype scenario) comparing variant groups versus wild-type.

Results are presented in accordance with REMARK criteria[24]. All statistical tests were two-sided. No multiple comparisons adjustments were implemented, and P-values less than 0.05 in the overall cohort, or interaction P-values less than 0.10, were considered statistically significant.

Results

Patients with DNA available for genotyping (Figure 1) were representative of the BIG 1-98 population characteristics and clinical outcomes (Table 1). Most (97%) patients were reported to be white/Caucasian, 36% had lymph node-positive disease and 70% had no previous chemotherapy.

Table 1.

Characteristics and clinical outcomes of patients in the BIG 1-98 trial according to availability of DNA for genotyping^a

	DNA for ESR1, ESR2 Genotypingb
Characteristic	No (n=4319)	Yes (n=3691)
Four-arm randomization option, %	70	86
Median follow-up, months	100	97
Postmenopausal, %	98	99
White race, %	98	97
Age, median (IQR), years	61 (56,67)	61 (56,68)
BMI, median (IQR), kg/m2	26 (23,29)	26 (23,30)
Mastectomy, %	52	33
Previous (neo)adjuvant chemotherapy, %	21	30
Lymph node-positive, %	46	36
Tumor size >2 cm, %	40	35
Tumor grade 2 or 3, %	79	79
Peritumoral vascular invasion present, %	19	19
Centrally-assessed tumor featuresc
ER-absent, %	2	1
HER2-positive, %	8	6
Ki-67 LI of immunostained cells, median (IQR), %	13 (7,18)	12 (6,19)
Clinical outcomes
Breast cancer events, %	19	14
Distant recurrences, %	15	10
Early onset of adverse hot flushes or night sweats eventsd, %	39	35
Bone adverse eventse, %	27	34

^aEligible for enrollment in the randomized, phase III double-blind Breast International Group (BIG) 1-98 trial for postmenopausal women with hormone receptor-positive operable invasive breast cancer.

IQR = interquartile range; BMI = body mass index; ER = estrogen receptor; HER2 = human epidermal growth factor receptor 2; LI = labeling index.

^bGenomic DNA was isolated from formalin-fixed paraffin-embedded primary breast cancer tissue blocks from 3,691 BIG 1-98 trial patients. Genotyping for four ESR1 and one ESR2 single nucleotide polymorphisms (SNPs) was done using polymerase chain reaction-based methods.

^cCentrally-assessed tumor features were available for a subset of the trial patients (2,654 of 4319 patients with no DNA for genotyping; 3,625 of 3691 patients with DNA available for genotyping).

^dNew or worsening hot flushes or night sweats of any grade adverse events reported during the first 23 months protocol treatment period.

^eBone adverse events include grade 3–4 osteoporosis or any grade bone fracture reported during or subsequent to the 5-year protocol treatment period.

Association between ESR1 and ESR2 genotype variants and disease outcomes

For the overall analytic cohort, 14% of patients treated with letrozole or tamoxifen, alone or in sequence, had breast cancer events. There were statistically significant associations between either BCFI or DRFI and two ESR1 SNPs. The presence of rs9340799(XbaI) variants CC or TC was associated with a reduced hazard of a breast cancer event (HR=0.82, 95% CI=0.67–1.0). A similar effect was observed for a distant recurrence (HR=0.79, 95% CI=0.63–1.0) (Table 2 and Figures 2A,B). Patients with ESR1 rs2077647 variants CC or TC had a reduced hazard of a distant recurrence (HR=0.69, 95%CI=0.53–0.90) (Figure 2C), consistent with BCFI, though not statistically significant.

Table 2.

Associations between ESR1and ESR2 genotype variants and breast cancer-free interval, distance recurrence, bone adverse events and early onset hot flushes or night sweats in the overall cohort and in the cohorts according to treatment regimen in the BIG 1-98 trial

		Overall Cohort			Monotherapy cohort 5-year Letrozole		Monotherapy cohort 5-year Tamoxifen
SNP	Genetic model	N Patients (N Events)	HR (95%CI)	P	N Patients (N Events)	HR (95%CI)	N Patients (N Events)	HR (95%CI)	Interaction Pa
Breast Cancer-Free Interval (BCFI)
rs9340799(XbaI)	0 vs.1 vs.2(C)	2902(400)	0.90(0.78,1.03)	0.14	845(106)	1.10(0.84,1.42)	812(131)	0.96(0.76,1.21)	0.52
	CC,TC vs. TT(ref)		0.82(0.67,1.00)	0.05		1.03(0.70,1.51)		0.85(0.60,1.20)	0.45
rs2234693(PuvII)	0 vs.1 vs.2(C)	3288(467)	1.06(0.93,1.20)	0.38	960(121)	1.11(0.86,1.44)	914(153)	1.00(0.80,1.25)	0.46
rs11963577	TT,TC vs. CC(ref)	2729(390)	1.09(0.81,1.47)	0.57	794(104)	1.04(0.59,1.85)	761(122)	1.49(0.86,2.57)	0.44
rs2077647	0 vs.1 vs.2(C)	2314(317)	0.88(0.75,1.03)	0.11	664(75)	0.92(0.67,1.26)	650(108)	0.93(0.71,1.23)	0.80
	CC,TC vs. TT(ref)		0.85(0.68,1.07)	0.17		0.76(0.48,1.21)		0.94(0.63,1.40)	0.43
rs4986938	0 vs.1 vs.2(A)	2138(311)	0.91(0.76,1.09)	0.32	616(81)	0.86(0.59,1.25)	616(103)	0.94(0.69,1.29)	0.46
Distant Recurrence-Free Interval (DRFI)
rs9340799(XbaI)	0 vs.1 vs.2(C)	2902(289)	0.86(0.73,1.01)	0.06	845(75)	1.03(0.76,1.41)	812(93)	0.90(0.68,1.19)	0.73
	CC,TC vs. TT(ref)		0.79(0.63,1.00)	0.05		0.94(0.60,1.49)		0.81(0.53,1.23)	0.81
rs2234693(PuvII)	0 vs.1 vs.2(C)	3288(337)	1.04(0.90,1.21)	0.58	960(86)	1.09(0.80,1.48)	914(109)	0.97(0.74,1.27)	0.57
rs11963577	TT,TC vs. CC(ref)	2729(289)	1.16(0.82,1.62)	0.40	794(73)	1.03(0.52,2.04)	761(91)	1.72(0.92,3.23)	0.44
rs2077647	0 vs.1 vs.2(C)	2314(233)	0.73(0.60,0.88)	<0.001	664(51)	0.70(0.47,1.04)	650(79)	0.83(0.60,1.16)	0.43
	CC,TC vs. TT(ref)		0.69(0.53,0.90)	0.01		0.58(0.33,1.01)		0.84(0.53,1.35)	0.22
rs4986938	0 vs.1 vs.2(A)	2138(225)	0.96(0.78,1.19)	0.72	616(54)	0.95(0.61,1.48)	616(77)	1.07(0.74,1.53)	0.56
Bone Adverse Events
rs9340799(XbaI)	0 vs.1 vs.2(C)	2902(999)	0.96(0.88,1.04)	0.34	845(285)	0.86(0.73,1.01)	812(246)	0.98(0.83,1.17)	0.15
rs2234693(PuvII)	0 vs.1 vs.2(C)	3288(1123)	1.00(0.92,1.08)	0.93	960(321)	0.91(0.78,1.07)	914(277)	1.00(0.85,1.18)	0.21
rs11963577	TT,TC vs. CC(ref)	2729(941)	0.90(0.74,1.10)	0.31	794(267)	0.84(0.57,1.25)	761(230)	1.01(0.66,1.55)	0.27
rs2077647	0 vs.1 vs.2(C)	2314(790)	0.96(0.87,1.06)	0.47	664(227)	0.82(0.68,0.99)	650(196)	1.04(0.84,1.28)	0.06
	CC,TC vs. TT(ref)		0.95(0.83,1.10)	0.53		0.75(0.58,0.98)		1.04(0.77,1.40)	0.08
rs4986938	0 vs.1 vs.2(A)	2138(729)	1.05(0.93,1.18)	0.43	616(215)	1.19(0.95,1.48)	616(183)	0.88(0.69,1.13)	0.07
	AA,AG vs. GG (ref)		1.11(0.95,1.30)	0.21		1.37(1.01,1.84)		0.93(0.67,1.30)	0.07
Early Onset Hot Flushes or Night Sweats
		Overall Cohort			2-year Letrozole		2-year Tamoxifen		Interaction Pb
rs9340799(XbaI)	0 vs.1 vs.2(C)	2902(1040)	1.04(0.96,1.13)	0.35	1459(491)	1.03(0.91,1.16)	1443(549)	1.06(0.94,1.19)	0.72
	CC vs. TC,TT(ref)		1.16(0.99,1.37)	0.07		1.06(0.84,1.35)		1.28(1.03,1.59)	0.27
rs2234693(PuvII)	0 vs.1 vs.2(C)	3288(1176)	0.98(0.90,1.06)	0.55	1653(553)	0.96(0.86,1.08)	1635(623)	0.99(0.88,1.10)	0.66
rs11963577	TT,TC vs. CC(ref)	2729(992)	0.97(0.80,1.17)	0.73	1375(471)	0.97(0.74,1.27)	1354(521)	0.98(0.75,1.29)	0.87
rs2077647	0 vs.1 vs.2(C)	2314(840)	1.04(0.95,1.15)	0.38	1156(386)	1.01(0.89,1.16)	1158(454)	1.07(0.93,1.22)	0.47
	CC,TC vs. TT(ref)		1.09(0.95,1.26)	0.23		1.12(0.91,1.38)		1.06(0.87,1.29)	0.86
rs4986938	0 vs.1 vs.2(A)	2138(765)	1.07(0.95,1.21)	0.25	1061(359)	1.12(0.94,1.32)	1077(406)	1.01(0.86,1.19)	0.61
	AA vs. AG,GG (ref)		1.29(1.04,1.61)	0.02		1.40(1.02,1.93)		1.16(0.86,1.58)	0.47

Abbreviations: BIG = Breast International Group; SNP = single nucleotide polymorphism; HR = hazard ratio; CI = confidence interval. HRs, CIs and P-values estimated using adjusted Cox proportional hazards models.

^aInteraction P-value: Wald test of treatment assignment (5-year letrozole vs. 5-year tamoxifen) by genotype interaction.

^bInteraction P-value: Wald test of treatment assignment (2-year letrozole vs. 2-year tamoxifen) by genotype interaction.

Figure 2(A, B, C).

Kaplan-Meier estimates of breast cancer-free interval (BCFI) (A) and distant recurrence-free interval (DRFI) according to ESR1 SNP rs9340799(XbaI)(T>C) variants (B) and DRFI according to ESR1 SNP rs2077647(T>C) variants (C) for patients in the BIG 1-98 trial. Hazard ratios (HR) and confidence intervals (CI) were estimated using the adjusted Cox proportional hazards model.

There were no statistically significant differential treatment effects (letrozole vs. tamoxifen) for the association of either BCFI or DRFI with any of the five SNPs in the subset of 1,940 patients assigned to a 5-year monotherapy regimen (Table 2).

Association between ESR1 and ESR2 genotype variants and bone adverse events and early onset hot flushes or night sweats

Over the 5-year treatment period and subsequent follow-up through 2010, 34% of patients reported grade 3–4 osteoporosis or any grade bone fracture. In the overall cohort, there was no prognostic association of bone AEs with any of the genotypes. In the subset of 1,940 patients assigned to a monotherapy regimen, there were statistically significant differential treatment effects (letrozole vs. tamoxifen) for the association of bone AEs with ESR1 rs2077647 (T>C) genotype variants (treatment-by-genotype interaction P=0.08; Table 2 and Figure 3A,B). For patients treated with letrozole, having the rare allele (C), either homozygous or heterozygous, was associated with a 25% reduced risk of a bone AE (HR=0.75, 95% CI=0.58–0.98), whereas for patients treated with tamoxifen, the hazard of a bone AE was not reduced (HR=1.04, 95% CI=0.77–1.40). Similarly, a statistically significant differential treatment effect was observed for bone AEs with ESR2 rs4986938(G>A) genotype variants (interaction P=0.07, Figure 3C,D). For patients treated with letrozole, having the rare allele of A, either homozygous or heterozygous, was associated with a 37% increased risk of a bone AE (HR=1.37, 95% CI=1.01–1.84), whereas for patients treated with tamoxifen, the hazard of a bone AE was not increased (HR=0.93, 95% CI=0.67–1.30).

Figure 3 (A, B, C, D).

Kaplan-Meier estimates of freedom from adverse bone events according to ESR1 SNP rs2077647(T>C) variants (A and B); and according to ESR2 SNP rs4986938(G>A) variants (C and D) for patients in the BIG 1-98 trial assigned to treatment 5-years letrozole (A and C) or 5-years tamoxifen (B and D). Hazard ratios (HR) and confidence intervals (CI) were estimated using the adjusted Cox proportional hazards model in which treatment-by-genotype interaction for rs2077647 was P=0.08 and for rs4986938 was P=0.07.

Over the first two years of protocol treatment with letrozole or tamoxifen monotherapy, 35% of patients reported HF/NS. Patients with ESR2 rs4986938 homozygous (AA) variant had a 29% increased risk of early onset of HF/NS events (HR=1.29, 95% CI=1.04–1.61; Table 2, Figure 4) compared to patients who had heterozygous (AG) or wild-type (GG) variants.

Figure 4.

Kaplan-Meier estimates of hot flushes or night sweats according to ESR2 SNP rs4986938(G>A) variants for patients in the BIG 1-98 trial. Hazard ratios (HR) and confidence intervals (CI) were estimated using the adjusted Cox proportional hazards model.

Discussion

There have been no studies evaluating the association between ESR polymorphisms in postmenopausal ER-positive breast cancer patients treated with SERMs or AIs and breast cancer outcomes. In this study, we found a statistically significant association between two ESR1 SNPs and BCFI or DRFI among postmenopausal endocrine therapy-treated patients in the BIG 1-98 trial[18]. Both rs9340799(XbaI) (variants CC or TC) and rs2077647 (variants CC or TC) were associated with a reduced hazard of a breast cancer event with the pattern of association more strongly observed for distant recurrence. With regard to HF/NS, patients with the ESR2 SNP rs4986938 variant minor allele AA reported an increased number of early onset HF/NS. In distinct contrast, there was no predictive (differential effect by treatment) association of breast cancer outcomes or early onset of HF/NS with any of the five SNPs. Of note, in the same patient population, we found that patients who had both CYP2D6 poor metabolizer and intermediate metabolizer phenotypes had an increased risk of tamoxifen-induced hot flushes compared to those with the extensive metabolizer phenotype, although no association with breast cancer outcome was found[25].

With regard to bone AEs known to occur in postmenopausal women receiving adjuvant AI therapy, we found that patients with the ESR1 SNP rs2077647 variants CC or TC had a reduced risk of a bone AE, while those with the ESR2 SNP rs4986938 rare allele (A) had an increased hazard of a bone AE when treated with letrozole and not tamoxifen.

A number of published studies have investigated the association of ESR1 and ESR2 polymorphisms and breast cancer risk with varied and equivocal results. There has been only one study published on the association of ESR1 polymorphisms and breast cancer survival[11], and although no overall association was observed between ESR1 gene polymorphisms and breast cancer, there were interactive effects of ESR1 gene polymorphisms and ER status on outcome.

The two most commonly studied ESR1 polymorphisms, rs2234693 and rs9340799 (PvuII and XbaI restriction fragment length polymorphisms (RFLPs), respectively), are found in intron 1, 50 bp apart, and have been investigated with regard to breast cancer risk with equivocal results overall being reported depending on patient ethnicity and age. In an early large-scale population-based case-control study, ESR1 PvuII and XbaI RFLPs were investigated in 1,069 Chinese breast cancer cases (~64% ER+) and 1,166 age-matched controls. The PvuII polymorphism was associated with an increased risk of breast cancer while the XbaI polymorphism was associated with a non-significantly elevated risk, mainly confined to postmenopausal women[26]. This increased risk was not observed in another study in a similar Chinese population of 614 women with breast cancer[27].

In a smaller study on 205 Korean women with breast cancer, the ESR1 rs2234693(PvuII) RFLP did not show any difference between cases and controls, however rs9340799(XbaI) RFLP, moreso in postmenopausal then premenopausal women, was found to modify individual breast cancer risk[28]. Genotyping of four ESR1 SNPs, rs746432, rs2234693(PvuII), rs9340799(XbaI) and rs1801132, was performed on 1,183 Caucasian postmenopausal women (393 breast cancer cases and 790 controls) from the Study of Osteoporotic Fractures[29]. A protective effect of SNP rs9340799(XbaI) was observed, while no statistically significant association was found for any of the other three SNPs and breast cancer risk in this postmenopausal population. In a case-control study conducted with a total of 846 pairs (388 Japanese, 79 Japanese Brazilians and 379 non-Japanese Brazilians) in pre and postmenopausal breast cancer women, none of the five ESR1 SNPs (rs2234693, rs9340799, rs1801132, rs3798577 and rs2228480) or two ESR2 SNPs (rs4986938 and rs1256049) were associated with breast cancer risk[30]. In a large meta-analysis of 1,678 breast cancer cases and 1,678 general population controls from Asian populations, the association between ESR1 XbaI and PuvII SNPs and breast cancer risk was evaluated[31]. The risk of breast cancer was not observed in premenopausal and postmenopausal women with the rs9340799(XbaI) polymorphism, however, premenopausal women with breast cancer with the rs2234693(PvuII) variant had a significantly elevated breast cancer risk, while postmenopausal women showed a non-significant increased risk.

There have also been several studies focused on ESR2 SNPs and their association with breast cancer risk. In one study, three common ESR2 polymorphisms, rs1256049 (G1082A), rs4986938 (G1730A) and rs928554 (Cx+56 A-->G), were not found to be significantly associated with breast cancer risk in 723 breast cancer cases (323 sporadic and 400 familial cases)[32]. In a meta-analysis of the two most commonly studied ESR2 polymorphisms, rs4986938 and rs2987983 (nine studies of 10,837 cases and 16,021 controls for rs4986938; 8 studies of 11,652 cases and 15,726 controls for rs1256049), rs4986938 AA/AG vs. GG was associated with a significant but small decreased breast cancer risk, while rs2987983 was not[33]. In our study, the ESR2 SNP rs4986938 was not prognostic or predictive of outcome in tamoxifen or letrozole-treated postmenopausal ER+ women.

Given the differing results in our European population, taken together, studies suggest that ethnicity may influence how ESR1 and ESR2 SNPs affect breast cancer risk and that no general conclusion can be drawn regarding the role they might play in breast cancer outcomes in either pre or postmenopausal women.

There have been relatively few studies that have investigated the role of ESR SNPs as they relate to endocrine treatment AEs. AIs profoundly reduce already low circulating estrogen levels in postmenopausal women by a further 80–90% compared with tamoxifen, which is associated with a modest increase of deleterious effects on the musculoskeletal system, a reason for treatment discontinuation[34]. In a recent study, the ESR1 SNP rs2234693(PvuII) variants TT and TC were associated with a lower risk of musculoskeletal AEs, while SNP rs9340799(XbaI) variant TT was associated with a higher risk of musculoskeletal AEs in 436 postmenopausal Chinese Han women receiving adjuvant AI[35]. However, we found no association of ESR1 SNPs rs2234693(PvuII) and rs9340799(XbaI) with bone AEs in a predominantly European population. Although the observed differences may be related to ethnicity, other factors may be at play, including perhaps the smaller number in the Chinese study.

Both ESR1 rs2234693(PvuII) and ESR2 rs4986938 influence the prevalence of hot flushes in tamoxifen-treated postmenopausal women[36]. In 297 participants after four months of tamoxifen treatment, postmenopausal women with ESR2-02 (rs4986938 in our study) GG genotypes had a 4.6-fold increase in the number of hot flush scores compared to other postmenopausal women. On the other hand, postmenopausal women with the ESR2-02 AA genotype were significantly less likely to experience tamoxifen treatment-induced hot flushes than women who carried at least one ESR2-02 G allele[36]. These results are contrary to our findings that ESR1 rs2234693(PvuII) had no effect while patients with the ESR2 rs4986938 variant minor homozygous allele (AA) had an increased risk of early onset HF/NS overall, although not significantly associated with tamoxifen treatment.

BMD has also been a concern for women on hormone replacement therapy (HRT). A few studies show that ESR SNPs can affect bone loss[17, 37, 38]. Women with the ESR1 PvuII genotypes PP and Pp have a greater risk of relatively fast bone loss after menopause and therefore, may derive more benefit from HRT[38]. A later study also found similar results and plus found ESR1 XX genotype was associated with lower rates of bone loss and hence would benefit more from HRT[17]. In a more recent study[37] women carrying the C allele of ESR1 rs2234693(PvuII) had a decreased risk of all-cause mortality with HRT. They also showed in distinct contrast that women who were homozygous for the T allele had a significantly increased risk of cancer-related mortality. The findings were similar for ESR1 rs9340799(XbaI) and ESR2 rs1271572.

Our study has limitations. We tested four ESR1 SNPs and one ESR2 SNP, but testing more ESR1 and ESR2 SNPs might have provided additional informative variants. There may be potential sources of false-positive results, including sample size or imbalance between patient groups. Patients were from heterogeneous ethnic groups, though in general, genotype frequencies corresponded well with those reported in the literature for a predominantly European Caucasian population. Nonetheless, we cannot rule out the possibility that the observed association with ESR genotypes were due to linkage disequilibrium with another causative gene and/or were the result of background genetic factors confined to an ethnically-defined subgroup.

Pharmacogenomics of breast cancer-related studies can provide information that can provide patients with the optimal therapy that will result in the best outcome, while limiting potential AEs. Our findings, together with those from previous studies, have implicated polymorphisms in ESR genes affecting outcomes and AEs. We found that two ESR1 SNPs were predictive of disease outcome, one of which was also associated with a decreased risk of a bone AE but only in letrozole-treated patients, and one ESR2 SNP was associated with an increased risk of early onset of HF/NS in postmenopausal endocrine-responsive breast cancer patients. These results provide increasing evidence that genes involved in estrogen signaling and synthesis affect outcomes and AEs, which if incorporated into a therapeutic strategy will increase both quality-of-life and survival in women with breast cancer.