Original Contribution

Association of Leukocyte Telomere Length With Breast Cancer Risk: Nested

Case-Control Findings From the Shanghai Women’s Health Study

Shimian Qu, Wanqing Wen, Xiao-Ou Shu, Wong-Ho Chow, Yong-Bing Xiang, Jie Wu, Bu-Tian Ji,

Nathaniel Rothman, Gong Yang, Qiuyin Cai, Yu-Tang Gao, and Wei Zheng*

* Correspondence to Dr. Wei Zheng, Vanderbilt Epidemiology Center and Vanderbilt-Ingram Cancer Center, Vanderbilt University School

of Medicine, 2525 West End Avenue, 8th Floor, Nashville, TN 37203-1738 (e-mail: wei.zheng@vanderbilt.edu).

Initially submitted March 13, 2012; accepted for publication June 14, 2012.

Telomeres are specialized chromatin structures essential for the maintenance of chromosomal integrity and

stability. Telomere shortening has been linked to multiple aging-related diseases, including cancer. Evidence

associating telomere length with breast cancer risk—most of which has been from retrospective case-control

studies—is conflicting. We conducted a nested case-control study based on the Shanghai Women’s Health

Study (1997–2009) in which we evaluated the association of telomere length and breast cancer risk using pe-

ripheral blood samples collected before cancer diagnosis (601 cases and 695 controls). We used monochrome

multiplex quantitative polymerase chain reaction to measure relative telomere length. Multiple logistic regres-

sions were used to derive adjusted odds ratios with 95% confidence intervals as the measure of association.

Telomere length was inversely correlated with age (r =−0.22). Women with moderately long telomeres (those in

the fourth quintile) had the lowest breast cancer risk. Risk increased in a dose-response manner with decreasing

quintile of telomere length; odds ratios were 1.39 (95% confidence interval (CI): 0.95, 2.04), 1.79 (95% CI: 1.17,

2.75), and 2.39 (95% CI: 1.45, 3.92), respectively, for the third, second, and first quintiles compared with the

fourth quintile. A slightly elevated risk of breast cancer (odds ratio = 1.35, 95% CI: 0.90, 2.04), although one that

was not statistically significant, was found in the top quintile (longest telomeres). Our results support the hypoth-

esis that telomere shortening is associated with increased risk of breast cancer and suggest a possible elevated

risk associated with long telomeres.

breast cancer; biomarkers; epidemiology; genetic factors; telomere

Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio.

Telomeres, located at the ends of eukaryotic chromo-somes, are specialized chromatin structures composed ofhighly conserved tandem hexameric nucleotide repeats,TTAGGG (1). Telomeres are essential for preserving chro-mosomal integrity and stability through prohibiting karyotyp-ic aberrations, such as chromosome ends fusion, nucleolyticdecay, and aberrant recombination (2). Telomere length ismaintained by telomerase, a multisubunit ribonucleoproteinpolymerase that normally is not expressed in somatic tissuebut is elevated in germ cells and neoplastic tissues (1, 3).Telomeres undergo progressive shortening with each cell di-vision, partly because of the incomplete replication at 3′

end of chromosomes during DNA replication. When the crit-ical telomere length is reached, the signal for replicative sen-escence is activated, and the cell enters cell-cycle arrest orundergoes apoptosis (4). Short telomeres can lead to chro-mosomal instability, which can increase the rates of geneticmutations and chromosome abnormalities (4). It has beenshown that cigarette smoking, oxidative stress, and chronicinflammation might cause telomere shortening (5–8). Telo-mere shortening has been shown to be associated with multi-ple aging-related diseases, such as cancer (1). In humans, thebest example of a short telomere phenotype is dyskeratosiscongenita. Patients with this disease carry mutations in the

617 Am J Epidemiol. 2013;177(7):617–624

American Journal of Epidemiology

© The Author 2013. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of

Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.

Vol. 177, No. 7

DOI: 10.1093/aje/kws291

Advance Access publication:

February 25, 2013

by guest on February 18, 2015http://aje.oxfordjournals.org/

Dow

nloaded from

components of telomerase complex, which results in shorttelomeres, leading to bone marrow failure and an elevatedrisk of cancer (9).Studies using clinical samples obtained from patients

with cancer—including cancers of the breast, lung, andcolon/rectum—have shown that, in general, telomere lengthis shorter in cancer tissues than in adjacent tissues (10).Recently, several epidemiologic studies have evaluated theassociation of germline telomere length, measured usingDNA from peripheral blood cells or buccal cells, withcancer risk and prognosis (11, 12). Regarding breastcancer, results from previous studies have been conflicting.Of the 8 retrospective case-control studies reported to date,2 showed short telomere length to be associated with in-creased risk (13, 14), whereas 2 other studies found an ele-vated risk among women with longer telomeres (15, 16).Four studies showed no significant association (17–20). Toour knowledge, only 2 prospective studies have been con-ducted, and both studies failed to identify any statisticallysignificant associations (14, 21). Reasons for the apparentlycontradictory results remain unknown. Differences in studydesigns, sample sizes, analytic approaches, laboratory assayprotocols, and background exposures could have contribut-ed to the inconsistent findings. To clarify the associationbetween telomere length and breast cancer risk, we con-ducted a large case-control study nested within the Shang-hai Women’s Health Study using blood samples collectedbefore cancer diagnosis.

MATERIALS AND METHODS

Study population

Subjects of the present study were participants in theShanghai Women’s Health Study, a population-based cohortstudy. Detailed methodology for the Shanghai Women’sHealth Study has been published elsewhere (22). The studywas approved by the relevant institutional review boards forhuman research. Briefly, from 1997 to 2000, a total of 74,942Chinese women between the ages of 40 and 70 years whoresided in 7 urban communities of Shanghai were recruitedinto the cohort study, with a response rate of 92.7%. In-person interviews were conducted to collect information re-garding demographic characteristics, anthropometrics, usualdietary habits, physical activities, and other lifestyle factors.Of the study participants, 56,831 (75.8%) provided a bloodsample at baseline recruitment (Appendix Table 1). Buffycoat samples were stored at −80°C until DNA isolation.The cohort has been followed using a combination of bi-

ennial home visits, record linkage to cancer incidence andmortality data collected by the Shanghai Cancer Registry,and death certificate data collected by the Shanghai VitalStatistics Unit. For cohort members who were diagnosedwith cancer, medical charts were reviewed to verify the di-agnosis, and detailed information regarding pathologiccharacteristics of the cancer was obtained.Included in the current nested case-control study were 601

incident breast cancer cases identified during follow-up of thecohort to December 2009, as well as their matched controls.Incidence-density sampling method was used to select

controls from the same risk set as the index case. Cases andcontrols were individually matched on age (difference ≤730days), date (difference ≤30 days), and time (morning or after-noon) of sample collection, time interval since last meal (dif-ference ≤2 hours), antibiotic use in the past week (yes vs.no), and menopausal status at time of sample collection. Typ-ically, one control was selected for each case. Because somecases from our substudy could be matched with the controlsfor cases with other types of cancers, there were second con-trols for 91 cases and third controls for 3 cases.

Laboratory methods

Genomic DNA was extracted from buffy coats using aQIAamp DNA kit (Qiagen, Valencia, California) followingthe manufacturer’s protocol. Relative telomere length wasmeasured using a monochrome multiplex quantitative poly-merase chain reaction method described recently byCawthon (23), with minor modifications. Briefly, telomerelength assay was carried out in a 15-µL polymerase chainreaction consisting of 1x QuantiFast SYBR Green PCRMaster Mix (Qiagen), 700-nM telomere primers telg andtelc, 200-nM albumin primers albugcr1 and albdgcr1, and5-ng DNA. A multistep thermal cycling procedure wasperformed on a Bio-Rad (Hercules, California) CFX384Real-Time System. After amplification, a dissociation curvewas performed to confirm the specificity of the reaction. Foreach standard curve, 2-fold serial dilutions of a referenceDNA sample were used to produce a standard curve. Addi-tionally, a calibrator DNA (same as the reference DNA), 2negative controls, and 2 quality-control DNA samples with along and short telomere length (Telo TAGGG TelomereLength Assay kit; Roche, Indianapolis, Indiana) were includ-ed in each of the 384-well assay plates. Bio-Rad CFXmanager version 1.6 software was used to determine the rel-ative telomere length through a 2-step relative quantification.In the first step, the ratio of the telomere repeat copy numberto the single-copy gene (albumin) copy number, as ameasure of relative telomere length, was determined for eachsample based on the standard curve. In the second step, theratio for each sample was normalized to the calibrator DNAto standardize sample values across all reaction plates.Samples from each matched case-control set were assayedon the same plate. Laboratory personnel were blinded toeach subject’s case-control status. Coefficients of variationfor the interplate ratio of telomere repeat copy number to thesingle-copy gene (albumin) copy number were 15.6% and16.2% for the long and short telomere quality-controlsamples, respectively. Inter- and intraplate coefficients ofvariation of calibrator DNA samples were 12.2% and 5.3%,respectively. The mean ratio of long to short telomerequality-control samples in our assays was 2.9, very close tothe mean ratio of 2.7 determined by southern blot method(Telo TAGGG Telomere Length Assay kit; Roche), indicat-ing that our protocol to estimate telomere lengths was valid.

Statistical methods

Means and percentages of selected baseline characteris-tics for cases and controls were calculated and compared

618 Qu et al.

Am J Epidemiol. 2013;177(7):617–624

by guest on February 18, 2015http://aje.oxfordjournals.org/

Dow

nloaded from

using t tests for continuous variables and χ2 tests for cate-gorical variables. Data on relative telomere length werelog-transformed so that these data were approximately nor-mally distributed.

We compared the case-control difference of the geometricmeans of the log-transformed telomere length using a 2-way(case-control status and matched sets) analysis of variancewith adjustment for age at blood collection. Because DNAsamples for cases and controls in the same matched sets wereassayed on the same plates, interplate differences for thematched sets were accounted for automatically. To evaluatethe association between breast cancer risk and telomerelength, the ratio for telomere length was categorized intoquintiles based on distribution among controls. Odds ratiosand 95% confidence intervals were estimated using condi-tional logistic regression models to account for the matchedsets, with additional adjustment for age at blood collection.Further adjustment for other demographic characteristics andknown breast cancer risk factors did not materially alter theassociation between telomere length and breast cancer risk.Tests for linear trend were estimated using the median valuefor each telomere length quintile. A restricted cubic splinefunction was used in the conditional logistic regressionmodel to evaluate the shape of the association (24). Themodel with 4 knots was used in the analysis because thismodel had the best fit of data as demonstrated by its havingthe lowest Akaike information criterion value. Likelihoodratio tests were used to evaluate linear effect, nonlinear effect,and overall effect of telomere length on breast cancer risk.Stratified analyses were performed to evaluate potential inter-actions. All statistical tests were based on 2-sided probability.

RESULTS

Table 1 presents the distributions of selected baseline de-mographic characteristics and major risk factors for breastcancer cases and matched controls. Cases and controls werecomparable in age at blood collection, age at menopause,body mass index (BMI; weight (kg)/height (m)2), and par-ticipation in leisure-time physical activity. There weresignificant case-control differences in the distributions of ed-ucation, age at menarche, age at first live birth, and familyhistory of breast cancer. Very few women in this cohort reg-ularly smoked cigarettes (2.9%), drank alcoholic beverages(2.7%), or took hormone replacement therapy (3.3%).

Telomere length was inversely correlated with age(r = −0.22; P < 0.0001). The geometric means of telomerelength were approximately 6.6% (cases) and 4.2% (con-trols) shorter in women who were 50–59 years of age and10.9% (cases) and 9.4% (controls) shorter in those whowere 60 years of age or older compared with women whowere younger than 50 years (data not shown). With the ex-ception of BMI, no apparent association of telomere lengthwas seen for other major breast cancer risk factors listed inTable 1. Both underweight (BMI <18.5) and obesity (BMI≥30) were associated with reduced telomere length (datanot shown).

Overall, no significant difference was observed in geo-metric means of telomere length between cases and con-trols (Table 2). Among postmenopausal women, however,telomere length was significantly shorter in cases than incontrols (P = 0.0485). No difference was observed amongpremenopausal women.

Table 1. Comparison of Demographic Characteristics and Known Breast Cancer Risk Factors in Cases and Their

Matched Controls, Shanghai Women’s Health Study, 1997–2000

VariableCases (n = 601) Controls (n = 695)

% Mean (SD) % Mean (SD) P Valuea

Demographic characteristic

Age at blood collection, years 52.7 (8.8) 53.4 (9.0) 0.163

High school education or above 50.6 38.4 <0.001

Family income >30,000 yuan 18.6 17.0 0.4362

Postmenopausal 48.3 50.4 0.449

Known risk factors 0.134

Age at menopause, yearsb 49.3 (4.6) 48.8 (3.8) 0.009

Age at menarche, years 14.7 (1.8) 15.0 (1.8) <0.001

Age at first live birth, years 26.4 (4.1) 25.5 (4.2) 0.004

Number of live births 1.6 (1.1) 1.8 (1.2) 0.971

Body mass indexc 24.4 (3.5) 24.3 (3.5) 0.513

Ever exercised regularly 34.9 36.7 0.747

Ever used hormone replacement 3.0 3.3 0.002

Family history of breast cancer 4.0 1.3 0.163

Abbreviation: SD, standard deviation.a P values were derived from t tests for continuous variables or χ2 tests for categorical variables.b Among postmenopausal women.c Weight (kg)/height (m)2.

Telomere Length and Breast Cancer 619

Am J Epidemiol. 2013;177(7):617–624

by guest on February 18, 2015http://aje.oxfordjournals.org/

Dow

nloaded from

To estimate odds ratios for the association betweenbreast cancer risk and telomere length, subjects were cate-gorized into 5 groups based on quintile distribution amongcontrols (Table 3). Compared with women in the longesttelomere group (top 20%), those in the shortest telomeregroup (bottom 20%) had a statistically significant increasein the risk of breast cancer (odds ratio (OR) = 1.77, 95%

confidence interval (CI): 1.02, 3.06) (Table 3). This associ-ation was stronger in postmenopausal women (for quintile1 vs. quintile 5, OR = 2.32, 95% CI: 0.99, 5.43) thanin premenopausal women (for quintile 1 vs. quintile 5,OR = 1.33, 95% CI: 0.63, 2. 80), although the interactiontest was not statistically significant. Women in the second-longest telomere group (fourth quintile) had the lowest risk

Table 2. Case-Control Differences in Relative Telomere Length, Shanghai Women’s Health Study, 1997–2009

Subject Category No. of Subjects Log Telomere Length 95% CIa P Value

All women

Cases 601 −0.040 −0.054, −0.027

Controls 695 −0.027 −0.039, −0.014 0.1337

Premenopausal women

Cases 311 0.002 −0.016, 0.020

Controls 345 0.004 −0.013, 0.021 0.8634

Postmenopausal women

Cases 290 −0.084 −0.103, −0.064

Controls 350 −0.058 −0.076, −0.041 0.0485

Abbreviation: CI, confidence interval.a Adjusted for age.

Table 3. Adjusted Odds Ratios for the Association Between Telomere Length and Breast Cancer Risk by

Menopausal Status, Shanghai Women’s Health Study, 1997–2009

Quintile of Telomere Lengthby Participant Category

No. of Cases No. of Controls

Longest TelomereGroup

Second-LongestTelomere Group

ORa,b 95% CIa,b ORa,c 95% CIa,c

All women 601 695

5 (long) 117 139 1.00 Referent 1.35 0.90, 2.04

4 88 139 0.74 0.49, 1.12 1.00 Referent

3 119 139 1.03 0.66, 1.61 1.39 0.95, 2.04

2 129 139 1.32 0.81, 2.16 1.79 1.17, 2.75

1 (short) 148 139 1.77 1.02, 3.06 2.39 1.45, 3.92

Premenopausal women 311 345

5 (long) 70 68 1.00 Referent 1.69 0.98, 2.91

4 46 75 0.59 0.34, 1.02 1.00 Referent

3 75 72 1.05 0.60, 1.83 1.77 1.06, 2.97

2 60 71 0.99 0.51, 1.92 1.67 0.90, 3.11

1 (short) 60 59 1.33 0.63, 2.80 2.25 1.11, 4.58

Postmenopausal women 290 350

5 (long) 47 71 1.00 Referent 1.03 0.54, 1.96

4 42 64 0.98 0.51, 1.87 1.00 Referent

3 44 67 0.99 0.46, 2.11 1.01 0.56, 1.83

2 69 68 1.74 0.82, 3.70 1.78 0.97, 3.26

1 (short) 88 80 2.32 0.99, 5.43 2.38 1.18, 4.80

Abbreviations: CI, confidence interval; OR, odds ratio.a Adjusted for age at blood collection.b Using subjects in the longest telomere group, the fifth quintile was the reference group.c Using subjects in the second-longest telomere group, the fourth quintile was the reference group.

620 Qu et al.

Am J Epidemiol. 2013;177(7):617–624

by guest on February 18, 2015http://aje.oxfordjournals.org/

Dow

nloaded from

of breast cancer, and a reverse J-shaped association was ob-served using restricted cubic spline function (Figure 1; testfor nonlinearity P = 0.0127 in all women combined). Whenwomen in the fourth quintile were used as the referencegroup for risk estimate, odds ratios for multiple groups witha short telomere were statistically significant (Table 3; oddsratios are shown in the right panel). A similar pattern ofassociation was found after excluding cases diagnosedwithin the first year after blood collection.

Both BMI and physical activity level have been reportedto be related to telomere length. Therefore, we investigatedpotential interactions of telomere lengths with these 2 vari-ables, as well as with years of menstruation (as a measureof cumulative endogenous estrogen exposure), in relation tobreast cancer risk (Table 4). A short telomere was associat-ed with elevated breast cancer risk in all groups defined byBMI, physical activity level, and years of menstruation, al-though not all odds ratios were statistically significant. A4-fold elevated risk of breast cancer was observed amongoverweight postmenopausal women (BMI ≥25) who alsohad a short telomere length (lowest quintile). The test formultiplicative interaction between BMI and telomere lengthwas marginally significant (P = 0.078) (Table 4).

DISCUSSION

In the present large case-control study in which we usedgenomic DNA from peripheral blood cells collected beforecancer diagnosis, we observed a reverse J-shaped associa-tion between telomere length and breast cancer risk. Com-pared with the group of women whose telomeres were inthe 60th to 79th percentiles of length (fourth quintile), therisk of breast cancer was increased in a dose-response

manner with a decreasing quintile of telomeres to a morethan 2-fold elevated risk in women with the shortest telo-meres (bottom quintile). Women in the top quintile, whosetelomere length measured in the top 20%, also had an ele-vated risk, although the risk estimate was not statisticallysignificant. In general, our study supports the hypothesisthat telomere shortening is associated with an elevated riskof breast cancer.

It has been well documented that short telomeres are as-sociated with multiple premature aging syndromes, includ-ing ataxia telangiectasia, ataxia-telangiectasia–like disorder,aplastic anemia, Bloom syndrome, Fanconi anemia, Nijme-gen breakage syndrome, and dyskeratosis congenita (1, 9).Patients with these syndromes have an elevated risk ofcancer. Mouse models also support the notion that telome-rase deficiency and short telomere-length increase the riskof a tumor (1). Epidemiologic studies have linked shorttelomeres in peripheral blood leukocytes to elevated risksof several human cancers, including cancers of the bladder,esophagus, stomach, head and neck, ovary, and kidneys(11, 12, 25). Breast cancer has been investigated in multipleprevious studies. However, 8 of 10 previous studies wereretrospective case-control studies that used blood samplescollected after cancer diagnosis or even cancer treatment toassess telomere length (13–20). Results from these studieshave been inconsistent, perhaps because of the influence ofcancer diagnosis and treatment on telomeres (26). Indeed, arecent study in which investigators used blood samples col-lected after cancer diagnosis showed a striking inverse as-sociation between telomere length and the risks of cancersof the breast and colon/rectum (14). The association,however, was substantially attenuated when using samplescollected before cancer diagnosis (14). Four of the 8 retro-spective case-control studies showed no significant associa-tion (17–20), and 2 other studies found an elevated riskamong women with longer telomeres (15, 16). Only 2nested case-control studies of breast cancer have examinedthe association with telomere length (14, 21). Using dataand prospectively collected samples from the Nurses’Health Study, De Vivo et al. (21) reported that having ashort telomere length was related to a slightly elevated, al-though not statistically significantly so, risk of breast cancerin postmenopausal women (bottom quartile versus top,OR = 1.25, 95% CI: 0.93, 1.88). Intriguingly, the risk ofbreast cancer was lowest in the third quartile, similar towhat was found in our study. The other published prospec-tive study, which was nested within the European Prospec-tive Investigation Into Cancer and Nutrition-Norfolkcohort, showed that short telomeres were associated with anonsignificantly elevated breast cancer risk (quintile 4 vs.quintile 1, OR = 1.58, 95% CI: 0.75, 3.31) (14). The samplesize for that study, however, was relatively small (199 casesand 420 controls). Although neither of those studies report-ed statistically significant associations, the direction of as-sociation identified in both studies was consistent with thedirection found in our study.

The observation of a suggestive elevated risk of breastcancer among women in the longest telomere group is un-expected, although biologically plausible. Long telomerescould be a result of telomerase reactivation, which may

Figure 1. Association between baseline leukocyte telomere lengthand subsequent risk of breast cancer, Shanghai Women’s HealthStudy, 1997–2009. OR, odds ratio.

Telomere Length and Breast Cancer 621

Am J Epidemiol. 2013;177(7):617–624

by guest on February 18, 2015http://aje.oxfordjournals.org/

Dow

nloaded from

predispose cells to delayed senescence and increase thechance of acquiring genetic abnormalities. It has beenshown that a long telomere may be associated with an in-creased risk of some cancers (12), including lung cancer(27) and non-Hodgkin’s lymphoma (28). Specifically forbreast cancer, a previous study also found an elevated riskof cancer related to long telomeres (15). However, bloodsamples for that study were collected after patients were di-agnosed with cancer.The major strengths of our study include the large sample

size and the use of blood samples collected before diagno-sis. Very few women smoked cigarettes regularly (2.4%)(22), and thus potential confounding effects due to cigarettesmoking should be minimal. Similar to other studies, how-ever, the precision of relative telomere measurement was

not optimal, and a relatively large coefficient of variationwas observed. However, this measurement error should berandom, which tends to attenuate the association betweentelomeres and breast cancer risk. By analyzing samplesfrom the same case-control set in the same plate and usingthe monochrome multiplex quantitative polymerase chainreaction method, we should have minimized the impact ofinterplate variation and detection bias on study results. Ex-tensive data were collected in our study, which allowed usto carefully evaluate the potential influence of confounderson our results. No major confounder, however, was identi-fied in the study. Nevertheless, we cannot completely ruleout the possibility of unmeasured or residual confoundingeffects on our results. The potential for selection biasshould be small in this study because of the very high

Table 4. Odds Ratiosa for Breast Cancer by Joint Distributions of Telomere Length and Body Mass Index,b Physical Activity Level, and Years

of Menstruation, Shanghai Women’s Health Study, 1997–2009

Quartile of Telomere Length

First Second Third Fourth

OR 95% CI OR 95% CI OR 95% CI OR 95% CI

All Womenc

Body mass index

<25 1.66 1.00, 2.77 1.27 0.82, 1.96 1.00 Referent 1.12 0.72, 1.75

≥25 2.02 1.19, 3.42 1.24 0.75, 2.03 1.33 0.81, 2.19 1.18 0.67, 2.07

P for interaction 0.786

Exercise

Yes 1.58 0.92, 2.70 0.89 0.54, 1.47 1.06 0.65, 1.72 0.96 0.55, 1.69

No 1.68 1.01, 2.77 1.32 0.86, 2.04 1.00 Referent 1.04 1.67, 1.61

P for interaction 0.612

Years of menstruation

<32 years 1.56 0.87, 2.79 0.96 0.57, 1.60 1.00 Referent 1.05 0.64, 1.73

≥32 years 2.46 1.36, 4.45 1.76 1.04, 2.97 1.38 0.81, 2.35 1.37 0.77, 2.47

P for interaction 0.793

Postmenopausal Womend

Body mass index

<25 3.01 1.33, 6.81 2.56 1.24, 5.27 1.00 Referent 2.04 0.85, 4.87

≥25 4.23 1.90, 9.41 2.22 1.02, 4.82 2.74 1.25, 6.02 1.70 0.64, 4.55

P for interaction 0.078

Exercise

Yes 1.95 0.93, 4.06 1.46 0.71, 2.97 0.96 0.48, 1.91 0.97 0.40, 2.31

No 2.19 1.02, 4.71 1.33 0.67, 2.61 1.00 Referent 1.07 0.49, 2.31

P for interaction 0.967

Years of menstruation

<32 years 2.35 0.87, 6.39 0.79 0.29, 2.19 1.00 Referent 1.09 0.37, 3.23

≥32 years 3.49 1.37, 8.84 2.67 1.14, 6.30 1.47 0.61, 3.57 1.68 0.63, 4.46

P for interaction 0.383

Abbreviations: CI, confidence interval; OR, odds ratio.a Adjusted for age at blood collection.b Weight (kg)/height (m)2.c There were 601 cases and 695 controls.d There were 290 postmenopausal cases and 350 controls.

622 Qu et al.

Am J Epidemiol. 2013;177(7):617–624

by guest on February 18, 2015http://aje.oxfordjournals.org/

Dow

nloaded from

response rate (92.7%) at the baseline survey and the verylow rate of loss to follow-up (<1%).

In conclusion, our study, a large case-control study usingprediagnostic samples, revealed a reverse J-shaped associa-tion between telomere length and breast cancer risk. Ingeneral, these findings support the hypothesis that telomereshortening is associated with an elevated risk of breastcancer, but they also suggest a possible adverse effect oflong telomeres. Additional research is needed to furtherevaluate the association of telomere length and breastcancer risk.

ACKNOWLEDGMENTS

Author affiliations: Division of Epidemiology, Depart-ment of Medicine, Vanderbilt University Medical Centerand Vanderbilt-Ingram Cancer Center, Nashville, Tennes-see (Shimian Qu, Wanqing Wen, Xiao-Ou Shu, Jie Wu,Gong Yang, Qiuyin Cai, Wei Zheng); Occupational Epide-miology Branch, National Cancer Institute, National Insti-tutes of Health, Bethesda, Maryland (Wong-Ho Chow,Bu-Tian Ji, Nathaniel Rothman); and Department of Epide-miology, Shanghai Cancer Institute, Shanghai, People’sRepublic of China (Yong-Bing Xiang, Yu-Tang Gao).

Funding for this work was provided by the National In-stitutes of Health (grant R37 CA 070867). Sample prepara-tions and telomere length assays were performed at theSurvey and Biospecimen Shared Resource, which is sup-ported in part by the Vanderbilt-Ingram Cancer Center(P30 CA68485).

We wish to thank Regina Courtney for her excellent lab-oratory assistance in sample handling and preparation.

Conflict of interest: none declared.

REFERENCES

1. Blasco MA. Telomeres and human disease: ageing, cancerand beyond. Nat Rev Genet. 2005;6(8):611–622.

2. Stewart SA, Weinberg RA. Telomeres: cancer to humanaging. Annu Rev Cell Dev Biol. 2006;22:531–557.

3. Greider CW. Telomerase activity, cell proliferation, andcancer. Proc Natl Acad Sci U S A. 1998;95(1):90–92.

4. Mathon NF, Lloyd AC. Cell senescence and cancer. Nat RevCancer. 2001;1(3):203–213.

5. Epel ES, Blackburn EH, Lin J, et al. Accelerated telomereshortening in response to life stress. Proc Natl Acad SciU S A. 2004;101(49):17312–17315.

6. Oikawa S, Tada-Oikawa S, Kawanishi S. Site-specific DNAdamage at the GGG sequence by UVA involves accelerationof telomere shortening. Biochemistry. 2001;40(15):4763–4768.

7. Valdes AM, Andrew T, Gardner JP, et al. Obesity, cigarettesmoking, and telomere length in women. Lancet. 2005;366(9486):662–664.

8. von Zglinicki T . Oxidative stress shortens telomeres. TrendsBiochem Sci. 2002;27(7):339–344.

9. Alter BP, Giri N, Savage SA, et al. Cancer in dyskeratosiscongenita. Blood. 2009;113(26):6549–6557.

10. Bisoffi M, Heaphy CM, Griffith JK. Telomeres: prognosticmarkers for solid tumors. Int J Cancer. 2006;119(10):2255–2260.

11. Ma H, Zhou Z, Wei S, et al. Shortened telomere length isassociated with increased risk of cancer: a meta-analysis.PLoS One. 2011;6:e20466.

12. Wentzensen IM, Mirabello L, Pfeiffer RM, et al. Theassociation of telomere length and cancer: a meta-analysis. Cancer Epidemiol Biomarkers Prev. 2011;20(6):1238–1250.

13. Levy T, Agoulnik I, Atkinson EN, et al. Telomere length inhuman white blood cells remains constant with age and isshorter in breast cancer patients. Anticancer Res. 1998;18(3A):1345–1349.

14. Pooley KA, Sandhu MS, Tyrer J, et al. Telomere length inprospective and retrospective cancer case-control studies.Cancer Res. 2010;70(8):3170–3176.

15. Svenson U, Nordfjall K, Stegmayr B, et al. Breast cancersurvival is associated with telomere length in peripheral bloodcells. Cancer Res. 2008;68(10):3618–3623.

16. Gramatges MM, Telli ML, Balise R, et al. Longer relativetelomere length in blood from women with sporadic andfamilial breast cancer compared with healthy controls. CancerEpidemiol Biomarkers Prev. 2010;19(2):605–613.

17. Zheng YL, Ambrosone C, Byrne C, et al. Telomere length inblood cells and breast cancer risk: investigations in two case-control studies. Breast Cancer Res Treat. 2010;120(3):769–775.

18. Shen J, Terry MB, Gurvich I, et al. Short telomere length andbreast cancer risk: a study in sister sets. Cancer Res. 2007;67(11):5538–5544.

19. Shen J, Gammon MD, Terry MB, et al. Telomere length,oxidative damage, antioxidants and breast cancer risk. IntJ Cancer. 2009;124(7):1637–1643.

20. Barwell J, Pangon L, Georgiou A, et al. Is telomere length inperipheral blood lymphocytes correlated with cancersusceptibility or radiosensitivity? Br J Cancer. 2007;97(12):1696–1700.

21. De Vivo I, Prescott J, Wong JY, et al. A prospective study ofrelative telomere length and postmenopausal breast cancerrisk. Cancer Epidemiol Biomarkers Prev. 2009;18(4):1152–1156.

22. Zheng W, Chow WH, Yang G, et al. The Shanghai Women’sHealth Study: rationale, study design, and baselinecharacteristics. Am J Epidemiol. 2005;162(11):1123–1131.

23. Cawthon RM. Telomere length measurement by a novelmonochrome multiplex quantitative PCR method. NucleicAcids Res. 2009;37(3):e21.

24. Harrell F. Regression Modeling Strategies. New York, NY:Springer; 2001.

25. Prescott J, Wentzensen IM, Savage SA, et al. Epidemiologicevidence for a role of telomere dysfunction in canceretiology. Mutat Res. 2011;730(1-2):75–84.

26. Schroder CP, Wisman GB, de JS, et al. Telomere length inbreast cancer patients before and after chemotherapy with orwithout stem cell transplantation. Br J Cancer. 2001;84(10):1348–1353.

27. Shen M, Cawthon R, Rothman N, et al. A prospective studyof telomere length measured by monochrome multiplexquantitative PCR and risk of lung cancer. Lung Cancer.2011;73(2):133–137.

28. Lan Q, Cawthon R, Shen M, et al. A prospective study oftelomere length measured by monochrome multiplexquantitative PCR and risk of non-Hodgkin lymphoma. ClinCancer Res. 2009;15(23):7429–7433.

Telomere Length and Breast Cancer 623

Am J Epidemiol. 2013;177(7):617–624

by guest on February 18, 2015http://aje.oxfordjournals.org/

Dow

nloaded from

Appendix Table 1. Comparison of Demographic Characteristics and Known Breast Cancer Risk Factors in

Subjects With and Without Blood Samples, Shanghai Women’s Health Study, 1997–2009

Variable

Provided a Blood Sample at Baseline

No (n = 18,111) Yes (n = 56,831)

%Mean(SD)

%Mean(SD)

P Valuea

Demographic characteristic

Age at blood collection, years 52.0 (9.1) 52.2 (9.1) 0.0604

High school education or above 51.7 38.2 <0.001

Family income >30,000 yuan 20.9 16.4 <0.0001

Postmenopausal 48.9 49.8 0.0353

Known risk factors

Age at menopause, yearsb 48.8 (4.1) 48.6 (4.4) <0.0001

Age at menarche, years 14.9 (1.7) 14.9 (1.7) 0.0135

Age at first live birth, years 25.5 (4.2) 25.5 (4.1) 0.2047

Number of live births 1.7 (1.2) 1.8 (1.2) <0.0001

Body mass indexc 23.7 (3.5) 24.1 (3.4) <0.0001

Ever exercised regularly 33.2 36.2 <0.0001

Ever used hormone replacement 3.6 3.5 0.3128

Family history of breast cancer 1.8 1.9 0.1143

Abbreviation: SD, standard deviation.a P values were derived from t tests for continuous variables or χ2 tests for categorical variables.b Among postmenopausal women.c Weight (kg)/height (m)2.

624 Qu et al.

Am J Epidemiol. 2013;177(7):617–624

by guest on February 18, 2015http://aje.oxfordjournals.org/

Dow

nloaded from