VOLUME 26 䡠 NUMBER 36 䡠 DECEMBER 20 2008

JOURNAL OF CLINICAL ONCOLOGY O R I G I N A L R E P O R T

From the Division of Clinical Pharmacology,

Department of Medicine, Indiana University Estrogen Receptor Genotypes Influence Hot Flash
School of Medicine, Indianapolis, IN; Breast
Oncology Program, University of Michigan Prevalence and Composite Score Before and After
Comprehensive Cancer Center, Ann Arbor,
MI; and Breast Cancer Program, Sidney Tamoxifen Therapy
Kimmel Comprehensive Cancer Center,
Yan Jin, Daniel F. Hayes, Lang Li, Jason D. Robarge, Todd C. Skaar, Santosh Philips, Anne Nguyen,
Johns Hopkins School of Medicine, Balti-
more, MD.
Anne Schott, Jill Hayden, Suzanne Lemler, Anna Maria Storniolo, David A. Flockhart, and Vered Stearns
Submitted February 21, 2008; accepted
A B S T R A C T
August 11, 2008; published online ahead of
print at [Link] on November 17,
2008.
Purpose
Hot flashes are common and frequently lead to drug discontinuation among women prescribed
Supported in part by Pharmacogenetics tamoxifen. We determined whether genetic polymorphisms in estrogen receptors (ESRs) ␣ and ␤
Research Network Grants No. U-01
(ESR1 and ESR2, respectively) are associated with tamoxifen-induced hot flashes.
GM61373 (D.F.) and R-01 GM56898
(D.F.); Clinical Pharmacology Training Patients and Methods
Grant No. 5T32-GM-08425 (D.F.) from We determined ESR1 PvuII and XbaI and ESR2-02 genotypes in 297 women who were initiating
the National Institute of General Medical tamoxifen. One-week hot flash diaries were collected to calculate a hot flash score (frequency ⫻
Sciences; Damon Runyon-Lilly Clinical
severity) before and 1, 4, 8, and 12 months after starting tamoxifen.
Investigator award CI-3 from the Damon
Runyon Cancer Research Foundation Results
(V.S.); Fashion Footwear Foundation/QVC Approximately 80% of 297 participants reported hot flashes before or during the first year of
Presents Shoes on Sale (D.F.H.); and the tamoxifen. After 4 months of tamoxifen, premenopausal women who did not receive adjuvant
General Clinical Research Centers at the
chemotherapy had a four-fold increase in hot flash score (from 5.9 to 23.6; P ⫽ .003) compared
University of Michigan (Grant No.
M01-00042 from the National Institutes
with a 1.17-fold increase (from 19.6 to 23; P ⫽ .34) in those who received chemotherapy. In
of Health), Georgetown University (Grant premenopausal women, increased number of ESR1 PvuII and XbaI CG alleles was associated with
No. M01-RR13297 from the National higher baseline hot flash scores compared with those who had other haplotypes (P ⫽ .0026). At
Institutes of Health), and Indiana Univer- 4 months, postmenopausal women with ESR1 PvuII CC and ESR2-02 GG genotypes had 4.6 times
sity (Grant No. M01-RR00750 from the increases in hot flash scores than other postmenopausal women (56 v 12; P ⫽ .0007). Women
National Institutes of Health).
who had the ESR2-02 AA genotype were significantly less likely to experience tamoxifen-induced
Presented in part at the 108th Annual hot flashes than women who carried at least one ESR-02 G allele (hazard ratio, 0.26; 95% CI, 0.10
Meeting of the American Society for
to 0.63; P ⫽ .001).
Clinical Pharmacology and Therapeutics,
March 21-24, 2007, Anaheim, CA; and Conclusion
at the 43rd Annual Meeting of the Knowledge of menopausal status, prior chemotherapy, and ESR genotype may help predict which
American Society of Clinical Oncology, women are most likely to suffer hot flashes during tamoxifen treatment.
June 1-5, 2007, Chicago, IL.

Authors’ disclosures of potential con- J Clin Oncol 26:5849-5854. © 2008 by American Society of Clinical Oncology
flicts of interest and author contribu-
tions are found at the end of this
article. may be optional for women at high risk for breast
INTRODUCTION
Clinical Trials repository link available on cancer, it is an integral part of adjuvant breast cancer
[Link].
In women who have estrogen receptor (ESR)–rich, therapy for many women, and completion of the
Corresponding author: Vered Stearns, MD, early-stage breast cancer, tamoxifen administered prescribed treatment is optimal to reduce breast
Sidney Kimmel Comprehensive Cancer
for 5 years is associated with a 42% reduction in cancer recurrence and death.1-3
Center, Johns Hopkins School of Medicine,
Bunting-Blaustein Cancer Research Bldg, relapse and a 22% reduction in breast cancer-related Selective ESR modulators, like tamoxifen, exert
1650 Orleans St, Rm 145, Baltimore, MD death.1 Tamoxifen also reduces the risk of a new their pharmacologic effect via interaction with the
21231-1000; e-mail: vstearn1@[Link].
breast cancer by nearly one half.2 One of the most ESRs. Two primary ESRs have been identified in
The Appendix is included in the common and bothersome adverse events associated humans: ESR ␣ (ESR1) and ESR ␤ (ESR2). Both are
full-text version of this article,
available online at [Link].
with tamoxifen is hot flashes. This effect is believed genetically polymorphic in the germline. In the
It is not included in the PDF version to be secondary to a central nervous system anties- ESR1 gene, two single nucleotide polymorphisms
(via Adobe® Reader®). trogenic effect.3 Up to 80% of women prescribed (SNPs), PvuII (c.454-397T⬎C, rs#2234693) and
© 2008 by American Society of Clinical tamoxifen complain of hot flashes, and approxi- XbaI (c.454-351A⬎G, rs#9340799), have been asso-
Oncology
mately 30% rate them as severe.4 Tamoxifen- ciated with clinical phenotypes and have influenced
0732-183X/08/2636-5849/$20.00 associated symptoms may interfere with quality of high-density lipoprotein cholesterol response to es-
DOI: 10.1200/JCO.2008.16.8377 life and may result in discontinuation of the agent in trogen in menopausal women6 and bone density.7
Trial Registration: NCT00228930. more than 40% of women.5 Although tamoxifen The same SNPs have been correlated with breast

© 2008 by American Society of Clinical Oncology 5849

 Jin et al

cancer risk8 and survival.9 The haplotypes of these two SNPs have been composite score was generated by multiplying the number of mild, moderate,
associated with risk of myocardial infarction.10 Several polymor- severe, or very severe hot flashes by 1, 2, 3, and 4, respectively, and summing
phisms in ESR2 also are associated with clinical outcomes. We have the values into one score.
reported differential changes in total cholesterol and triglycerides ac-
cording to ESR2-02 genotype.11 These findings suggest that SNPs in Statistical Analysis
The primary end point of the clinical protocol was to assess the relation-
ESR1 and ESR2 may be associated with differential risk of tamoxifen-
ship between pharmacogenetics, tamoxifen metabolism, and frequency and
induced effects. severity of hot flashes associated with tamoxifen. The women were categorized
We hypothesized that genetic variants in the ESRs may play a role into seven different groups determined by menopausal status, presence of hot
in susceptibility to tamoxifen-induced hot flashes. We investigated the flashes at baseline (0 to 2 or ⬎ 2), and prior adjuvant chemotherapy, to allow
association between ESR1 and ESR2 genotypes and hot flashes in a for heterogeneity. An enrollment target of 43 participants per group (301
prospective study of a cohort of tamoxifen-treated women. participants overall) was chosen to have an 85% power to detect a .5 effect size
in the difference of hot flash score change between two genotype groups,
considering a 15% homozygous variant genotype frequency and a 15% drop-
PATIENTS AND METHODS out rate. Given that the standard deviation of the hot flash score is 9 units, this
.5 effect size is translated to the 4.5-unit difference in hot flash score change
from baseline to month 4.
Patients and Study Design Follow-up weekly hot flash scores were compared with baseline scores by
We designed an open-label, prospective observational trial to test asso- using linear regression with repeated measures. Associations between ESR
ciations between polymorphisms in candidate genes and tamoxifen pharma- genotypes and baseline weekly hot flash scores were examined in each meno-
cokinetics, adverse effects, and secondary benefits. The study design and pausal group. The comparisons were performed by using linear regression
results related to pharmacogenetic influence on tamoxifen pharmacokinetics within each menopausal status. Associations between ESR genotypes and the
and associations between SNPs in ESR1 and ESR2 with lipid levels have been changes in hot flash score from baseline to month 4 were assessed by the
previously published.11,12 interactions between genotypes and time by using linear regression with re-
Participants included women 18 years or older who were recommended peated measures.
tamoxifen for adjuvant treatment or for the prevention of breast cancer. The The ESR1 haplotypes were constructed from ESR1PvuII and ESR1XbaI
women were recruited from three academic cancer centers (Georgetown Uni- SNPs by using the PHASE2 online software ([Link]
versity, University of Michigan, and Indiana University). Women were ex- stephens/[Link]). Its association with hot flash score was tested with a
cluded if they had concurrent adjuvant chemotherapy, radiation therapy, generalized estimating equation approach,14 through its online implementa-
other endocrine therapy except ovarian suppression, or chronic corticosteroid tion ([Link]
therapy; or if they used clonidine, bellargal, or megestrol acetate for the treat- The association with genetic and clinical predictors was tested through
ment of hot flashes. Vitamin E, selective serotonin reuptake inhibitors (SSRIs), survival analyses. Time to hot flash was treated as a time-to-event outcome.
serotonin-norepinephrine reuptake inhibitors (SNRIs), or herbal use was al- Kaplan-Meier plots are displayed, and P values were calculated from log-rank
lowed, provided that the participant had been taking the agent for at least 4 tests. Odds ratios (ORs) for the presence of hot flashes predicted from ESR
weeks before study entry and intended to continue taking the agent for at least genotypes are reported; their P values were based on ␹2 tests.
the first study month. The protocol was approved by the institutional review Most of the analyses, including ␹2 tests, Wilcoxon tests, survival analyses,
boards and general clinical research centers of all three sites. Each participant and linear regression with repeated measures, were conducted in SAS 9.1(SAS
provided written informed consent. Institute Inc, Cary, NC). P values were confirmed by bootstrap algorithms.
Baseline history and physical, a comprehensive medication list, and Haplotype and haplotype-phenotype analyses were performed with pre-
clinical laboratory tests were obtained for each participant. Menopausal status scribed online software.14,15
was determined based on menstrual history. Women who were age 60 years or
older or those who had no menses during the prior 12 months or who had a
prior bilateral oophorectomy were defined as postmenopausal; women who
had regular menses before adjuvant chemotherapy or tamoxifen were RESULTS
defined as premenopausal; and the rest were defined as perimenopausal.
Participants received open-label tamoxifen 20 mg orally each day. Medical Demographic Characteristics of Subjects
history and current medications were recorded 1, 4, 8, and 12 months after Two hundred ninety-seven participants were enrolled on the
initiating tamoxifen.
study. Baseline patient characteristics were compared among the
three menopausal groups (Table 1). Baseline mean hot flash fre-
Genotyping Analysis
At baseline, a 10-mL blood sample in a heparinized Vacutainer tube quency and hot flash composite scores were significantly lower in
(Becton Dickinson, Franklin Lakes, NJ) was collected from each participant. the premenopausal group compared with the post- and perimeno-
Samples were stored in a ⫺80°C freezer. Genomic DNA was extracted by using pausal groups (P ⫽ .02).
a QIAmp DNA Mini Kit (Qiagen, Valencia, CA). ESR1 PvuII and XbaI geno-
types were determined according to the method of Herrington et al,6 with
minor modifications.11 Genotyping for ESR2-02 was performed by Taqman ESR Genotype
assays, as previously described by the National Cancer Institute Cancer Ge- ESR1 PvuII, ESRI XbaI, and ESR 2-02 genotypes were deter-
nome Anatomy Project ([Link]). The amplification and mined in 289, 287, 254 participants, respectively (Appendix Table A1,
analysis were performed by using the iCycler Real-Time Thermocycler (Bio- online only). Eight, 10, and 43 samples were not assessable for ESR1
Rad Life Science Research Group, Hercules, CA). PvuII, ESRI XbaI, and ESR 2-02 genotypes, respectively. Genotype
distributions were in Hardy-Weinberg equilibrium for the entire
Hot Flash Diaries
Participants completed validated hot flash daily diaries for 7 days at
cohort. We observed four possible haplotypes in the ESR1 gene on the
baseline and 1, 4, 8, and 12 months after starting tamoxifen.13 During each 24 basis of the ESR1 PvuII and XbaI genotype: T-A (frequency, 48.8%),
hour period, the women recorded the number of hot flashes that they had C-G (frequency, 31.8%), C-A (frequency, 16.0%), and T-G (fre-
experienced and how many were mild, moderate, severe, or very severe. A quency, 3.4%).

5850 © 2008 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

 Estrogen Receptor Genotype and Tamoxifen-Induced Hot Flashes

Table 1. Patient Characteristics at Baseline

Menopausal Status
Overall Premenopausal Perimenopausal Postmenopausal
(N ⫽ 297)ⴱ (n ⫽ 93) (n ⫽ 37) (n ⫽ 164)
Characteristic No. % No. % No. % No. % P†
Age, years ⬍ .0001
Mean 52.1 43.2 49.1 57.9
Range 29-87 29-57 40-59 35-87
BMI, kg/m2 28.0 ⫾ 0.4 27.2 ⫾ 0.7 28.1 ⫾ 1.1 28.4 ⫾ 0.5 .29
Ethnicity
White 275 92.6 87 93.8 34 91.9 154 93.6 .70
Other 19 6.5 6 6.2 3 8.1 10 6.4
Prior chemotherapy 142 47.8‡ 44 46.8 19 51.4 78 47.6 .89
Prior ET
Yes 102 34.3 4 4.3 8 21.6 90 55.5 ⬍ .001
No 147 49.5 72 77.7 26 70.3 48 29.3
Unknown 48 16.7 17 18.1 3 8.1 26 15.9
SSRI/SNRI
Yes 55 14.1 13 14.0 9 24.3 33 20.1 .18
No 230 81.8 78 84 24 64.9 128 78.0
Unknown 12 4 2 2.2 4 10.8 3 1.8
Hot flash frequency 12.7 ⫾ 1.1 8.6 ⫾ 1.6 13.9 ⫾ 2.7 15.0 ⫾ 1.7 .03
Hot flash score§ 20.9 ⫾ 2.2 12.5 ⫾ 2.8 22.7 ⫾ 5.9 25.3 ⫾ 3.2 .02

Abbreviations: BMI, body mass index; ET, estrogen therapy; SSRI/SNRI, selective serotonin or serotonin-noradrenergic reuptake inhibitor.
ⴱ
Menopausal status was unknown in three women who are, therefore, not included in the subgroup comparison.
†Comparison among different menopausal status stages.
‡One patient who had prior chemotherapy did not provide information regarding her menopausal status.
§Hot flash score equals hot flash frequency times hot flash severity.

Changes in Hot Flash Prevalence Rate and Weekly pausal, 65% of perimenopausal, and 65% of postmenopausal women
Composite Score During the First Year of reported symptoms of hot flashes (P ⬍ .001). Four months after
Tamoxifen Treatment initiation of tamoxifen therapy, premenopausal hot flash prevalence
Among the 297 participants, 286 returned their baseline hot flash increased from 36% to 61% (P ⫽ .001), whereas it increased in post-
diaries, and 250, 238, 213, and 212 diaries were available 1, 4, 8, and 12 menopausal women from 65% to 78% (P ⫽ .02; Appendix Figure A1,
months after tamoxifen treatment, respectively. Diaries were not online only).
available on all patients because of discontinuation of tamoxifen dur- As expected, the mean weekly hot flash score was significantly
ing the first year of treatment (n ⫽ 41) or because of failure to lower in premenopausal women before tamoxifen treatment (12.5 ⫾
return diaries. 2.8) compared with postmenopausal women (25.3 ⫾ 3.2, P ⫽ .02).
A higher proportion of perimenopausal and postmenopausal This difference disappeared after the first month of treatment (Table
patients reported hot flashes compared with premenopausal women 2). Hot flash scores remained elevated throughout the first year of
at all the time points. Before tamoxifen treatment, 36% of premeno- tamoxifen treatment regardless of baseline menopausal status.

Table 2. Hot Flash Score During the First Year of Tamoxifen Treatment
Score by Treatment Group
Overall Premenopausal Perimenopausal Postmenopausal
Time,
Months No. Mean ⫾ SE No. Mean ⫾ SE No. Mean ⫾ SE No. Mean ⫾ SE Pⴱ
Baseline† 286 20.9 ⫾ 2.1 94 13.9 ⫾ 2.9 36 22.7 ⫾ 4.9 156 24.7 ⫾ 3.2 0.02
1‡ 250 32.5 ⫾ 3.1 80 30.2 ⫾ 6.3 30 32.0 ⫾ 7.9 139 34.0 ⫾ 3.9 0.58
4§ 238 32.5 ⫾ 3.4 77 27.5 ⫾ 5.5 26 52.2 ⫾ 10.7 134 38.9 ⫾ 4.8 0.14
8§ 213 37.6 ⫾ 3.8 65 30.0 ⫾ 6.7 23 48.0 ⫾ 9.5 124 39.4 ⫾ 5.1 0.27
12§ 212 37.9 ⫾ 4.4 67 36.2 ⫾ 10.3 21 45.9 ⫾ 11.1 123 37.3 ⫾ 4.8 0.90

Abbreviation: SE, standard error.

ⴱ
Comparison of premenopausal v postmenopausal patients.
†All patients who returned their hot flash diaries provided information regarding their menopausal status.
‡One participant who did not provide information regarding her menopausal status returned hot flash diaries on months 1, 4, 8, and 12.
§Comparison among different menopausal status stages.

[Link] © 2008 by American Society of Clinical Oncology 5851

 Jin et al

Baseline Hot Flash Score and ESR Genotype

In premenopausal women before tamoxifen treatment, a statis- 300

tically significant association was observed between baseline weekly

hot flash score and the ESR1 XbaI and PvuII genotypes. Participants

Change in hot flash score

200
who had the ESR1 XbaI GG genotype had significantly higher hot *
flash scores than those who had AG and AA genotypes (Appendix *
Figure A2, panel A, online only). Similarly, participants who had 100
the ESR1 PvuII CC genotype had statistically significantly higher hot
flash scores than carriers of the CT and TT genotypes. Therefore, ESR1
CG haplotype was associated with higher hot flash scores in pre- 0
menopausal women. One copy more of the CG ESR1 haplotype is
associated with a 2.5-fold higher hot flash score (P ⫽ .0026 for
-100
gene-dose effect; Appendix Figure A2, panel B, online only). A similar
trend was maintained in premenopausal women after 1 and 4 months Genotype
of tamoxifen treatment, although the P values were not statistically ESR1 PvuII CC CT/TT CC CT/TT
ESR2-02 GG GG AG/AA AG/AA
significant (data not shown). N 16 46 61
In postmenopausal women, neither history of prior chemother-
apy, SSRI/SNRI use, prior hormone therapy use, nor the ESR geno- Fig 1. Effect of estrogen receptor (ESR) genotype on changes in hot flash
composite scores after 4 months of tamoxifen treatment. ESR1 and ESR2 gene
types (Appendix Figure A2, panel C, online only) predicted baseline interactions and tamoxifen-associated change in hot flash score. (F) Individual
hot flash score. However, the sample size is too small to reach firm data points; *P ⫽ .0005 for ESR1 PvuII CC and ESR2-02 GG group compared with
conclusions. The effects of these factors on the hot flash score in other haplotypes.
perimenopausal women could not be tested because of the small
sample size.
genotypes had the least increase, a significant gene-dose relationship
Tamoxifen-Induced Changes in Weekly Hot Flash was also observed (P ⫽ .0007).
Composite Score
After initiation of tamoxifen, hot flash scores reached a plateau 4 Risk of Developing Hot Flashes During the First Year
months after the initiation of tamoxifen treatment in all women and of Tamoxifen Treatment
did not change significantly after that time (Table 2 and online only Among 297 participants, 242 (81%) reported hot flashes at least
Appendix Figure A4). Because many women were prescribed SSRI/ once, either before or during the study period, and 27 (9%) did not
SNRI after 4 months of tamoxifen, and because the number of report any hot flashes in their returned diaries. The other 28 partici-
patients without diaries increased at later time points, we used the pants did not report hot flashes in their diaries, but they failed to
comparison between baseline and 4 month values as the best indica- follow-up at some point during the study; therefore, the hot flash
tor of tamoxifen-induced hot flashes. status was censored at the last follow-up time point for these partici-
In the 36 premenopausal women who had previously received pants. The risk of developing hot flashes during the study period was
adjuvant chemotherapy, hot flash scores increased from 19.6 ⫾ 4.9 at analyzed with a Cox regression model. Menopausal status, history of
baseline to 23.0 ⫾ 6.8 at 4 months after tamoxifen treatment, a change chemotherapy, and the number of ESR1 GC haplotype alleles or
that was not statistically significant (P ⫽ .57). In contrast, in 36 pre- ESR2-02 genotypes were included in the Cox regression model, which
menopausal women who did not receive chemotherapy, we observed excluded participants who reported hot flashes before tamoxifen
a significant increase in hot flash scores from 5.9 ⫾ 2.5 at baseline to treatment. Women who carried a ESR2-02 AA genotype were signifi-
23.6 ⫾ 9.4 after 4 months of tamoxifen treatment (P ⫽ .0015). cantly less likely to develop hot flashes compared with participants
who had the AG/GG genotype (hazard ratio, 0.26; 95% CI, 0.10 to
0.63; P ⫽ .001; Fig 2A). If we included participants who reported hot
Tamoxifen-Induced Changes in Weekly Hot Flash
flashes at baseline, those who had the ESR2-02 AA genotype were also
Score and Genotype significantly less likely to report hot flashes compared with partici-
The effects of ESR genotypes on tamoxifen-induced hot flashes at pants who had the AG or GG genotype (OR, 0.12; 95% CI, 0.04-0.41;
4 months were observed only in the postmenopausal group. In pre- P ⫽ .0001; Fig 2B).
menopausal women, ESR genotype did not predict changes in hot
flash score. Among postmenopausal women, no obvious association
was observed between tamoxifen-induced changes in hot flash score DISCUSSION
and ESR1 or ESR2 genotype. When interactions between ESR1 and
ESR2-02 genotypes were considered (Fig 1), carriers of ESR1 PvuII CC As previously documented in cross-sectional surveys,4,16,17 we observed
and ESR2-02 GG genotype had significantly more increases in hot that premenopausal women had a greater increase in tamoxifen-
flash scores after tamoxifen treatment than the other genotype com- induced hot flashes compared with peri- or postmenopausal
binations (P ⫽ .0005). Because women who were homozygotes for women. Prior history of chemotherapy may alter the risk of
both ESR1 PvuII CC and ESR2-02 GG genotype had the highest tamoxifen-induced hot-flashes in premenopausal women. In pre-
increases in hot flash composite scores after tamoxifen treatment, and menopausal women, an increase in the number of baseline ESR1
because the women who were not homozygotes for either of these CG haplotypes was associated with higher hot flash scores. After

5852 © 2008 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

 Estrogen Receptor Genotype and Tamoxifen-Induced Hot Flashes

group were less likely to develop hot flashes compared with those in
A 1.0 the ESR1 Pvull CT group,19 which is consistent with data in this study.
Hot Flash–Free Patients (proportion)

In a study of 51 postmenopausal Japanese women, a short number

of CA repeats of the ESR2 gene was associated with a 7.0-fold
0.8 (range, 1.25- to 35.9-fold) increased the risk of hot flashes com-
pared with women who had extremely short or long variants.20
Although intriguing, the sample size is small and the confidence
0.6 interval wide. While we did not determine the CA repeat genotype
for participants in this study, it is possible that both a short CA
repeat genotype and the ESR2 associations observed in our study
0.4 may result in a change in the expression or functionality of the
estrogen receptor beta protein, and both factors may influence the
risk for and severity of tamoxifen-induced hot flashes.
0.2
ESR2-02 AA
Genetic polymorphisms of ESR1 and ESR2 are extremely com-
ESR2-02 AG/GG plex. Greater than 1,000 SNPs have been reported in ESR1, and there
are hundreds of assumed haplotypes. The functional consequences of
0 2 4 6 8 10 12 these variants are not well understood. As a result, we chose to assess
the effects of the ESR SNPs that have already been shown to have
Time to Hot Flash (month) functional impact. The role of other important SNPs in ESR1 and
B ESR2 genes will be studied in the future, when more data become
available on the molecular functionality of SNPs in either of the estro-
AA
gen receptors.
ESR1
Xbal
v P = .63 Other potential candidate genes may explain tamoxifen-induced
GG/AG hot flashes more specifically.21 We have previously demonstrated that
CC polymorphisms in cytochrome P450 2D6 (CYP2D6) are associated
ESR1
v P = .53 with decreased conversion of the parent drug tamoxifen to its most
Pvull CT/TT
active and abundant metabolite, endoxifen.12 Our group is currently
AA developing bioinformatic tools to stratify participants by a CYP2D6
ESR2-02 v P = .0001
AG/GG
score that considers both genotype and concomitant medication.
In this study, we used subjective criteria to determine
menopausal status. Future studies that include serum estrogen
0 0.5 1.0 1.5 2.0 2.5 3.0 and pituitary hormone concentrations may help better correlate
Odds Ratio for Hot Flashes tamoxifen-induced changes in neuro-hormonal concentrations and
menopausal symptoms. We used highly validated but subjective dia-
Fig 2. Effect of estrogen receptor (ESR) genotypes on risk of developing hot ries to assess hot flash frequency and severity. Although it is possible
flashes during the first year of tamoxifen treatment. (A) Kaplan-Meier curves for that diary data may underestimate the frequency of hot flashes, we
the effect of ESR2-02 AA genotype on hot flash–free survival during the first year
of tamoxifen treatment (n ⫽ 115). Patients who had baseline hot flashes were would expect a similar degree of underestimation in all genotype
excluded. (B) Odds ratio of experiencing hot flashes before and during the first groups.22 We have also allowed SSRI/SNRI use, but these are agents
year of tamoxifen treatment according to genotype. Error bars represent 95% that may reduce the frequency or severity of hot flashes and may,
CI (n ⫽ 122).
therefore, influence our results. However, most women who initi-
ated new SSRI/SNRI prescriptions received a prescription after the
4-month visit, which somewhat reduced the bias. Finally, 20% and
initiation of tamoxifen, postmenopausal women homozygotes of 40% of women were lost to follow-up 4 and 12 months after tamox-
the ESR1 PvuII CC and ESR2-02 GG genotype had the greatest ifen treatment, respectively, a rate that is comparable with that of the
increase (55.6 to v 11.9 to 12.5; Figure 1) in hot flash scores. Most Breast Cancer Prevention Trial P-1 trial and with studies that evalu-
notably, women who had the ESR2-02 AA genotype had a signifi- ated treatments for hot flashes.4,23 The impact of these lost data points
cantly lower risk for developing tamoxifen-induced hot flashes may also influence the observed associations between ESR1 and ESR2
compared with women who had AG or GG genotypes regardless of and hot flashes, because patients who experienced worse hot flashes
menopausal status (OR, 0.12). may have dropped out at a greater rate. If specific genotypes predis-
Women who report hot flashes during menopause are more pose women to hot flashes and lead to drug discontinuation, it is
likely to complain of tamoxifen-associated symptoms compared with possible that such genes are over-represented in a drop-out group and
other women.18 This observation suggests that genetic factors in the under-represented in the studied group. Thus, we may actually under-
ESR signaling pathway may influence the risk for experiencing hot estimate the genotype effects on hot flashes.
flashes. Several small cross-sectional studies have reported associa- In conclusion, we have demonstrated that specific ESR genotypes
tions between hot flashes during natural menopause and polymor- may also influence hot flash risk and score in breast cancer patients
phisms in the ESR1 and ESR2 genes. In a study that evaluated the before and after tamoxifen treatment. If our findings are confirmed in
influence of ESR1 genotypes PvuII and XbaI on prevalence of hot other datasets or prospective studies, evaluation of risk factors for hot
flashes in 177 postmenopausal women, women in the ESR1 PvuII TT flashes may enable clinicians to provide prospective counseling and

[Link] © 2008 by American Society of Clinical Oncology 5853

 Jin et al

symptom management in women with breast cancer who are recom-

AUTHOR CONTRIBUTIONS
mended tamoxifen.

Conception and design: Yan Jin, Daniel F. Hayes, Lang Li, Anne
AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS Nguyen, David A. Flockhart, Vered Stearns
OF INTEREST Financial support: Daniel F. Hayes, David A. Flockhart, Vered Stearns
Administrative support: Daniel F. Hayes, Anne Nguyen, Jill Hayden,
Although all authors completed the disclosure declaration, the following Suzanne Lemler, David A. Flockhart, Vered Stearns
author(s) indicated a financial or other interest that is relevant to the subject Provision of study materials or patients: Daniel F. Hayes, Anne Schott,
matter under consideration in this article. Certain relationships marked with a Jill Hayden, Suzanne Lemler, Anna Maria Storniolo, David A. Flockhart,
“U” are those for which no compensation was received; those relationships Vered Stearns
marked with a “C” were compensated. For a detailed description of the Collection and assembly of data: Yan Jin, Daniel F. Hayes, Lang Li,
disclosure categories, or for more information about ASCO’s conflict of interest
Jason D. Robarge, Todd C. Skaar, Santosh Philips, Anne Nguyen, Jill
policy, please refer to the Author Disclosure Declaration and the Disclosures of
Hayden, Suzanne Lemler, Anna Maria Storniolo, David A. Flockhart,
Potential Conflicts of Interest section in Information for Contributors.
Vered Stearns
Employment or Leadership Position: Yan Jin, Eli Lilly (C); Daniel F.
Data analysis and interpretation: Yan Jin, Daniel F. Hayes, Lang Li,
Hayes, Eli Lilly (C); Anna Maria Storniolo, Eli Lilly (C) Consultant or
Advisory Role: Todd C. Skaar, Roche Diagnostics (C); Anna Maria Jason D. Robarge, Todd C. Skaar, Santosh Philips, Anne Nguyen, Anne
Storniolo, Eli Lilly (C); David A. Flockhart, LabCorp (C), Roche Schott, Anna Maria Storniolo, David A. Flockhart, Vered Stearns
Diagnostics (C); Vered Stearns, Wyeth (C), Concert Pharmaceuticals Manuscript writing: Yan Jin, Daniel F. Hayes, Lang Li, Jason D.
(C), JDS Pharmaceuticals (C) Stock Ownership: None Honoraria: Todd Robarge, Todd C. Skaar, Santosh Philips, Anne Nguyen, Anne Schott,
C Skaar, Roche Diagnostics; Anna Maria Storniolo, GlaxoSmithKline Anna Maria Storniolo, David A. Flockhart, Vered Stearns
Research Funding: Daniel F. Hayes, AstraZeneca, GlaxoSmithKline, Final approval of manuscript: Yan Jin, Daniel F. Hayes, Lang Li, Jason
Novartis, Pfizer; Anna Maria Storniolo, GlaxoSmithKline; David A. D. Robarge, Todd C. Skaar, Santosh Philips, Anne Nguyen, Anne Schott,
Flockhart, Novartis, Pfizer; Vered Stearns, GlaxoSmithKline, Novartis, Jill Hayden, Suzanne Lemler, Anna Maria Storniolo, David A. Flockhart,
Pfizer Expert Testimony: None Other Remuneration: None Vered Stearns

breast cancer: Results from the Shanghai Breast menopause-related symptoms. Menopause 11:519-
REFERENCES Cancer Study. Cancer Epidemiol Biomarkers Prev 530, 2004
12:853-859, 2003 17. Biglia N, Cozzarella M, Cacciari F, et al: Meno-
1. Effects of chemotherapy and hormonal ther- 9. Boyapati SM, Shu XO, Ruan ZX, et al: Poly- pause after breast cancer: A survey on breast cancer
apy for early breast cancer on recurrence and 15- morphisms in ER-␣ gene interact with estrogen survivors. Maturitas 45:29-38, 2003
year survival: An overview of the randomised trials. receptor status in breast cancer survival. Clin Cancer 18. Loprinzi CL, Zahasky KM, Sloan JA, et al:
Lancet 365:1687-1717, 2005 Res 11:1093-1098, 2005 Tamoxifen-induced hot flashes. Clin Breast Cancer
2. Cuzick J, Powles T, Veronesi U, et al: Over- 10. Schuit SC, Oei HH, Witteman JC, et al: Estro- 1:52-56, 2001
view of the main outcomes in breast-cancer preven- gen receptor ␣ gene polymorphisms and risk of 19. Malacara JM, Perez-Luque EL, Martinez-Garza
tion trials. Lancet 361:296-300, 2003 myocardial infarction. JAMA 291:2969-2977, 2004 S, et al: The relationship of estrogen receptor-␣ poly-
3. Stearns V, Ullmer L, Lopez JF, et al. Hot 11. Ntukidem NI, Nguyen AT, Stearns V, et al: morphism with symptoms and other characteristics in
flushes. Lancet 360:1851-1861, 2002 Estrogen receptor genotypes, menopausal status,
post-menopausal women. Maturitas 49:163-169,
4. Day R: Quality of life and tamoxifen in a breast and the lipid effects of tamoxifen. Clin Pharmacol
2004
cancer prevention trial: A summary of findings from Ther 83:702-710, 2008
20. Takeo C, Negishi E, Nakajima A, et al: Asso-
the NSABP P-1 study. Ann NY Acad Sci 949:143- 12. Jin Y, Desta Z, Stearns V, et al: CYP2D6
ciation of cytosine-adenine repeat polymorphism of
150, 2001 genotype, antidepressant use, and tamoxifen me-
the estrogen receptor-beta gene with menopausal
5. Fallowfield L: Acceptance of adjuvant therapy tabolism during adjuvant breast cancer treatment.
symptoms. Gend Med 2:96-105, 2005
and quality of life issues. The Breast 14:612-616, 2005 J Natl Cancer Inst 97:30-39, 2005
6. Herrington DM, Howard TD, Hawkins GA, et 13. Sloan JA, Loprinzi CL, Novotny PJ, et al: 21. Goetz MP, Rae JM, Suman VJ, et al: Pharma-
al: Estrogen-receptor polymorphisms and effects of Methodologic lessons learned from hot flash stud- cogenetics of tamoxifen biotransformation is asso-
estrogen replacement on high-density lipoprotein ies. J Clin Oncol 19:4280-4290, 2001 ciated with clinical outcomes of efficacy and hot
cholesterol in women with coronary disease. N Engl 14. Schaid DJ, Rowland CM, Tines DE, et al: flashes. J Clin Oncol 23:9312-9318, 2005
J Med 346:967-974, 2002 Score tests for association between traits and hap- 22. Carpenter JS, Monahan PO, Azzouz F: Accu-
7. Kurabayashi T, Matsushita H, Tomita M, et al: lotypes when linkage phase is ambiguous. Am J racy of subjective hot flush reports compared with
Association of vitamin D and estrogen receptor Hum Genet 70:425-434, 2002 continuous sternal skin conductance monitoring.
gene polymorphism with the effects of longterm 15. Stephens M, Smith NJ, Donnelly P: A new Obstet Gynecol 104:1322-1326, 2004
hormone replacement therapy on bone mineral den- statistical method for haplotype reconstruction from 23. Stearns V, Slack R, Greep N, et al: Paroxetine
sity. J Bone Miner Metab 22:241-247, 2004 population data. Am J Hum Genet 68:978-989, 2001 is an effective treatment for hot flashes: Results
8. Cai Q, Shu XO, Jin F, et al: Genetic polymor- 16. Crandall C, Petersen L, Ganz PA, et al: Asso- from a prospective randomized clinical trial. J Clin
phisms in the estrogen receptor ␣ gene and risk of ciation of breast cancer and its therapy with Oncol 23:6919-6930, 2005

■ ■ ■

Acknowledgment
We thank Claudine Isaacs, MD, Lynda Ullmer, and Ann Gallagher for their invaluable help in the enrollement and monitoring of
study participants.

5854 © 2008 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY