Tamoxifen and risk of large bowel cancer in women with breast cancer

Breast Cancer Research and Treatment53: 271–277, 1999.c© 1999Kluwer Academic Publishers. Printed in the Netherlands.

Report

Tamoxifen and risk of large bowel cancer in women with breast cancer

Polly A. Newcomb, Cam Solomon, and Emily WhiteFred Hutchinson Cancer Research Center, Cancer Prevention Research Program, Seattle, WA, USA

Key words:breast neoplasms, therapy, colorectal neoplasms, estrogen antagonists, estrogen replacement therapy

Summary

Background:The increasingly consistent association between estrogen replacement therapy and colorectal cancersuggests that the anti-estrogen tamoxifen may also be associated with large bowel cancer incidence.

Methods:Women with new diagnoses of breast cancer were identified from the Surveillance Epidemiologyand End Results (SEER) Program, a set of geographically defined, population based cancer registries representingapproximately ten percent of the U.S. population. Of 85,411 women with local or regional breast cancer diagnosedfrom 1983–90, 14,984 women were reported to have received hormonal therapy and 70,427 were not known to havereceived hormonal therapy. Subsequent cancer diagnoses were identified in this cohort beginning 6 months afterinitial breast cancer diagnosis until death, or December 31, 1994. Multivariate Cox proportional hazards modelswere used to estimate the risk of developing colorectal cancer and other second cancers according to hormonaltherapy use.

Results:Over the follow-up period 793 colorectal, 2,648 contralateral breast, 506 endometrial, 250 ovarian, 98gastric, and 1,765 other cancers were identified in the study cohort. While overall there was no association betweenhormonal therapy use and colorectal cancer (relative risk (RR) 1.09, 95% confidence interval (CI) 0.88–1.35), inthe period five or more years after diagnosis, risk was increased significantly by about 50% (95% CI 1.00–2.15). Asexpected, based upon clinical trials data, cancers of the contralateral breast were significantly decreased, and cancersof the uterine endometrium were significantly increased. No other meaningful associations were observed. Whenwomen were excluded for whom hormonal therapy might represent therapy other than tamoxifen (premenopausalwomen and those who received chemotherapy), this did not meaningfully alter these estimates.

Conclusions:The results of this large population based cohort study suggest that tamoxifen therapy may modestlyincrease risk of large bowel cancer in women, but only after 5 years following initiation of breast cancer therapy.

Introduction

Several observations – both epidemiologic and bio-logic – support a role for estrogens in the etiology ofcolorectal cancer [1, 2]. Most convincing, perhaps, isthe inverse association observed between exogenoushormone use and large bowel cancer risk [3]. Thereduction in risk appears to be large – nearly 50%among current users of hormone replacement therapy[4–8]. Less consistently reported has been an inverseassociation between oral contraceptives and colorectalcancer incidence [9]. By extension, other exogenoussex hormones – such as the antiestrogen tamoxifen –may also be associated with risk [10].

Tamoxifen, already widely used by millions ofwomen in the treatment of breast cancer, has been

successfully tested in healthy women for the primaryprevention of breast cancer [11, 12]. While tamoxifenexerts an anti-estrogenic effect in the breast [13], ithas been observed to have partial or full estrogeniceffects on various tissues, including the endometrium,bone, and cardiovascular system [13–16]. It is un-known whether estrogen agonist or antagonist profilesmight predominate in the colon of hormonal therapyusers; the presence of estrogen receptors in the normaland malignant colon [17–20] could support either role.

There is some preliminary evidence to suggest thatcertain bowel and gastric cancers are increased amongusers of tamoxifen therapy. In a pooled analysis of datafrom three randomized trials of adjuvant tamoxifentherapy, gastrointestinal tumors were increased 2–3fold among tamoxifen therapy users [21]. However,

272 PA Newcomb et al.

other recent studies have not observed an association[22, 23]. While the randomized trial of Fisher et al.[22] was limited by small numbers of colorectal cancercases using hormonal therapy, the generally null studyof Curtis et al. [23] was large and population-based.The interpretation of this latter study is constrained,however, by the use of a general population compari-son group, rather than women with a history of breastcancer not using hormonal therapy. The failure to use aninternal comparison group in this study likely resultedin an exposed group with higher levels of surveillance,and different baseline risk of colon cancer [24]. To eval-uate more validly the effect of hormonal therapy in thelarge bowel, we examined the occurrence of colorectaland other cancers among a population based sample ofUS women diagnosed with breast cancer according touse of hormonal therapy.

Methods

Data were collected by nine regional registries that arepart of the Surveillance Epidemiology and End Results(SEER) Program [25] of the National Cancer Insti-tute. These geographically defined population basedcancer registries include the states of Connecticut,Iowa, Hawaii, New Mexico, Utah, and the metropoli-tan areas of Atlanta, Detroit, and the Seattle–PugetSound, representing approximately 10% of the U.S.population. For each woman diagnosed with breastcancer, information was recorded on date of diagno-sis, demographic characteristics, stage of disease andhistology, and first course of therapy (surgery, radio-therapy, chemotherapy, and hormonal therapy), basedon standard protocol. The first course of therapy in-cluded therapy initiated or planned during the 4 monthsafter diagnosis, excluding any therapy for recurrences.Vital status is updated yearly by contact with physiciansand linkage to state death files.

A cohort of 91,709 women with invasive primarybreast cancer were identified. From this group we ex-cluded women with distant disease(n = 6100) andthose for whom treatment information was missing(n = 198). Of the remaining 85,411 women, 14,984 re-ceived hormone therapy at the first course of treatment(or it was planned for subsequent therapy). The remain-ing 70,427 women were not known to have received (orplanned) hormonal therapy. Since other hormones andsteroids (which may have accompanied chemotherapy)may have also been included as ‘hormonal therapy,’ weattempted to improve the specificity of tamoxifen useby also limiting the analysis to women most likely tobe true users: women over age 50 years who did not

receive chemotherapy [23]. For this more limited studygroup of 54,821 women, 8939 women were consideredto have used hormonal therapy.

The outcome events were all subsequent diagnosesof primary cancer at least 6 months following the initialbreast cancer diagnosis(n = 6060). Data collected oneach new primary included date of diagnosis, anatomicsite, and vital status.

Statistical analysis

For each subject, follow-up time was allocated accord-ing to the date of initial breast cancer diagnosis. Foreach cancer site analysis, follow-up terminated withthe date of diagnosis of each second primary cancerof interest, date of death, or December 31, 1994. Forsome analyses, time since diagnosis was categorizedas less than 5 years(n = 76,745, including manywho also contributed to the> 5 year category), and 5or more years since the initial breast cancer diagnosis(n = 49,754).

As a measure of association we used the rela-tive risk, defined as the incidence of specified casesamong women who had used hormonal therapy di-vided by the incidence among patients who had notused such therapy. We used proportional hazards toaccount for differential followup and for the effects ofcovariates [26]. Covariates included in these modelswere age (5 year intervals), year of initial diagno-sis, stage of disease at diagnosis (local and regional),treatment (chemotherapy and/or radiation), and reg-istry. Interactions involving age were evaluated withage on a continuous scale. Subjects with unknownvalues for variables in the multivariate models wereexcluded.

Results

Overall, the prevalence of hormonal therapy use was18%. Using the more stringent definition that in-cluded only women> 50 years without a history ofchemotherapy, the prevalence of use was 16%. Com-pared to women not reported to have used hormonaltherapy, hormonal therapy users were older, morelikely to be diagnosed with regional stage disease andhave larger tumor sizes, and more likely to have recentdiagnoses (Table 1). Hormonal therapy use also variedsomewhat by geographic locale, with prevalence of usehighest in Seattle and lowest in Connecticut.

During follow up (mean duration 76.6 months) weidentified 6,060 cases of cancer, including 793 col-orectal cancers, 2,648 contralateral breast cancers, 506

Hormonal therapy and large bowel cancer273

Table 1. Selected characteristics of SEER cohort of womenwith local or regional breast cancer by use of hormonal therapy,1983–1990

HormonalTherapy Users Non-Users

Characteristics (N = 14, 984) (N = 70, 427)N % N %

Age< 50 3,100 21 19,143 2750 to 64 5,331 36 22,783 3265+ 6,553 44 28,501 40

StageLocal 5,511 37 47,267 67Regional 9,473 63 23,160 33

TreatmentChemotherapy 4,894 33 9,944 14Radiation 4,857 32 16,290 23

RegistrySan Francisco/Oakland 2,670 18 12,448 18Connecticut 1,723 11 12,220 17Detroit 2,405 16 12,770 18Hawaii 559 4 2,732 4Iowa 2,642 18 8,957 13New Mexico 982 7 3,046 4Seattle 2,967 20 8,881 13Utah 309 2 3,743 5Atlanta 718 5 5,630 8

Year of Diagnosis1983–85 3,095 21 25,500 361986–88 5,639 38 28,156 401989–90 6,250 42 16,771 24

Tumor SizeNo measurable mass 133 1 1,471 23–20 mm 6,799 49 35,827 5721–50 mm 5,754 41 22,341 3551+ 1,250 9 3,490 6

RaceWhite 13,261 89 61,850 88Black 979 7 5,029 7Other 703 5 3,215 5Unknown 41 0 333 0

endometrial cancers, 250 ovarian cancers, 98 gastriccancers, and 1,765 cancers of other sites.

Overall, in this cohort of breast cancer patients, hor-monal therapy use was not associated with colorectal

cancer incidence (RR= 1.09, 95% CI 0.88–1.35)(Table 2). This estimate differed only very slightlyfrom the model adjusted for age only (RR= 1.04,95% CI 0.86–1.26), suggesting that confounding wasunlikely to have introduced substantial bias. However,time since initial breast cancer diagnosis (and presum-ably, first use of hormonal therapy) appeared to modifythis relationship (Table 3). For colorectal cancers diag-nosed within 5 years of the breast cancer diagnosis,the relative risk was 0.89; the risk after 5 or moreyears was elevated by about 50% (RR= 1.47, 95%CI 1.00–2.15). When we excluded women for whommisclassification of hormonal therapy use may havebeen more likely (Table 3), the relative risk was alsoelevated for women with greater time since diagnosis(RR = 1.46), although the confidence limits includedone.

As expected, contralateral breast cancer occurrencewas significantly reduced among women who usedhormonal therapy (Table 2: RR= 0.79, 95% CI0.70–0.90). There was a suggestion that the risk re-duction was greater within 5 years from diagnosis(RR = 0.74) than for more than 5 years(RR =0.82). (The exclusion of women with possibly mis-specified hormonal therapy use slightly strengthenedthis association, Table 3.) The inverse relation ap-peared to increase with increasing age, with therelative risk for women over 65 years 0.64 (95%CI 0.51–0.79). However, in the more limited studygroup, the effect of age was not statistically sig-nificant, suggesting perhaps that misclassificationof hormonal therapy was less likely among olderwomen.

In contrast, cancers of the endometrium were in-creased among hormonal therapy users. The increase inrisk was about 40% overall, but was only significantlyelevated after 5 years from diagnosis (RR= 1.87,95% CI 1.25–2.82). This pattern was also observedin the restricted analyses (Table 3). Another repro-ductive tumor, ovarian cancer, was not associatedwith hormonal therapy use (RR= 0.86, CI 0.58–1.28). Exclusion of younger women and those withchemotherapy increased the relative risk for ovariancancer to 1.14 (95% CI 0.69–1.88). Overall, othercancer sites were slightly, but not statistically sig-nificantly increased among hormonal therapy users:among subgroups examined, more recent treatmentappeared to be positively associated with risk. Thispattern was observed in the complete and limited studygroups.


Table 2. Relative risk of subsequent cancer among women who used hormonal therapy relative to women who did not, among womendiagnosed with local or regional breast cancer between 1983 and 1990 by time since initial diagnosis

Cancer site All women Events< 5 years post diagnosis Events> 5 years post diagnosisN (by hormonal RR† 95% CI N (by hormonal RR† 95% CI N (by hormonal RR† 95% CI

therapy use) therapy use) therapy use)User Non- User Non- User Non-

user user user

All cancers 859 4548 1.00 (0.93, 609 2875 0.91 (0.83, 250 1673 1.08 (0.93,1.08) 0.99) 1.25

Breast 296 2070 0.79 (0.70, 222 1365 0.74 (0.63, 74 705 0.82 (0.63,0.90) 0.85) 1.06)

less than 50∗ 81 555 1.02 (0.80, 62 373 0.89 (0.67, 19 182 0.95 (0.58,1.30) 1.17) 1.56)

50–64∗ 118 717 0.84 (0.68, 84 461 0.70 (0.55, 34 256 0.95 (0.66,1.02) 0.89) 1.38)

65 and greater∗ 97 798 0.64 (0.51, 76 531 0.69 (0.54, 21 267 0.62 (0.39,0.79) 0.89) 0.97)

Colorectal 123 591 1.09 (0.88, 83 393 0.89 (0.69, 40 198 1.47 (1.00,1.35) 1.15) 2.15)

Endometrium 101 359 1.39 (1.09, 63 224 1.07 (0.80, 38 135 1.87 (1.25,1.77) 1.45) 2.82)

Ovary 33 195 0.86 (0.58, 23 119 0.83 (0.51, 10 76 0.78 (0.38,1.28) 1.32) 1.59)

Stomach 19 73 1.09 (0.63, 14 52 0.96 (0.51, 5 21 1.46 (0.51,1.88) 1.82) 4.20)

All others 209 1305 1.17 (1.02, 83 541 1.12 (0.95, 292 764 1.03 (0.80,1.34) 1.32) 1.33)

† Adjusted for age, race, stage, registry, year of diagnosis, tumor size, and treatment (radiation/chemotherapy).∗ p-value for differential effect of hormonal therapy by age of index breast cancer: all women, p= 0.02; events< 5 years, p= 0.59;events> 5 years, p= 0.14.

Table 3. Relative risk of subsequent cancer among women who had hormonal therapy relative to women who did not,among women> 50 years with breast cancer without chemotherapy, by time since initial diagnosis

Cancer site All∗ Events< 5 years post diagnosis Events> 5 years post diagnosisRR† 95% CI RR† 95% CI RR† 95% CI

All cancers 1.02 (0.92, 1.12) 0.94 (0.83, 1.05) 0.99 (0.81, 1.21)Breast 0.72 (0.59, 0.87) 0.67 (0.54, 0.83) 0.68 (0.45, 1.01)

50 to 64∗∗ 0.79 (0.60, 1.04) 0.71 (0.54, 0.94) 0.61 (0.36, 1.05)65+∗∗ 0.68 (0.53, 0.86) 0.64 (0.47, 0.88) 0.76 (0.43, 1.35)

Colorectal 1.13 (0.88, 1.44) 0.94 (0.71, 1.26) 1.46 (0.92, 2.31)Endometrium 1.40 (1.05, 1.89) 1.13 (0.80, 1.59) 1.98 (1.15, 3.40)Ovary 1.14 (0.69, 1.88) 1.19 (0.67, 2.12) 0.75 (0.28, 2.00)Stomach 1.00 (0.48, 2.06) 0.84 (0.36, 1.96) 1.20 (0.32, 4.47)All others 1.21 (1.02, 1.44) 1.20 (0.98, 1.46) 0.91 (0.65, 1.29)

∗ Number of hormonal therapy users= 8939; non-users= 45,882.† Adjusted for age, race, stage, registry site, year of diagnosis, tumor size, and treatment (radiation).∗∗ p-values for differential effect of hormonal therapy by age of index breast cancer: all, p= 0.22; events< 5 years,p= 0.50; events> 5 years, p= 0.78.


Discussion

In this prospective cohort study, we observed hormonetherapy, presumably tamoxifen, was associated witha modest increased risk of colorectal cancer amongwomen 5 or more years from initial breast cancer diag-nosis. This evaluation was based upon a large numberof cases, both with and without hormonal therapy treat-ment. Thus, chance is an unlikely explanation for ourfindings. Further we have sufficient follow-up to sug-gest that the effect of hormonal therapy is present onlyafter 5 years of initial therapy, although additional yearsof follow-up would be helpful in describing this latenteffect. Contributing to our confidence in these find-ings were the expected decreases in the occurrence ofcontralateral breast cancers and the expected increasesin endometrial cancer risk based on prior randomizedtrials and observational studies [21, 22, 26–32].

This approach to evaluating the relation betweenhormonal therapy and large bowel cancer risk has someimportant limitations. First, the ascertainment of hor-monal therapy use may be inaccurate, since it reflectsonly treatment initiated (or planned) and reported at thetime of record abstraction – about 4 months followingdiagnosis. However, it is likely that any misclassifi-cation is non-differential, since exposure informationis recorded prior to the diagnosis of any second pri-mary disease. The medical record, from which theSEER information on treatment is based, appears tobe an excellent source of information on hormonaltherapy use. In the study of Cook et al., agreementbetween pharmacy and medical records with SEER washigh with 94% sensitivity and 100% specificity [26].Our attempt to improve the classification of tamoxifenexposure by excluding from the analysis women forwhom hormonal therapy was infrequently used failedto meaningfully change our estimates of effect. Indeed,when Cook et al. evaluated the impact of these ex-clusions on the validity of ‘hormonal therapy’ in theSEER record [32] as an indicator of tamoxifen use,sensitivity was similar for the full and limited studygroups; specificity increased from 91% to 99% withthese exclusions. It is apparent that some misclassifi-cation occurred, since the effect of hormonal therapyin our study is attenuated compared to the relative risksobtained in randomized controlled trials. For example,we found the reduction in contralateral breast diseasewas about 20–30% while an overall 39% reduction inrisk was reported from meta-analyses of clinical trials[9].

A second limitation is that information on spe-cific patterns of use, including dose and duration,was unavailable in the SEER records. Both actions

of tamoxifen, reducing contralateral breast cancer andincreasing risk of endometrial cancer, appear to bedirectly related to cumulative dose [22, 32]. Duringthe time period under study, usual dose of hormonaltherapy was perhaps too limited (20 mg per day andduration of use was about 2 years) to evaluate dose,even if more complete data were available. The mod-est risk increases for endometrial cancer confirm thatthe patterns of hormonal therapy use resulted in fairlymodest cumulative doses of therapy; in clinical tri-als relative risks of 3.0–7.5 have been observed [21,22] with cumulative doses> 10 grams. In contrast,population-based studies have reported increases verysimilar to our own relative risk of 1.4 [27–31]. No doubtdifferences in compliance in a population-based studyalso account for the attenuation of effects comparedwith clinical trials.

Third, although the SEER cohort is population-based, subjects may move from the population baseor be otherwise lost to follow-up, resulting in under-ascertainment of colorectal cancer diagnoses in thestudy cohort. It is unlikely, however, that womentaking hormonal therapy will differ from non-users intheir migration patterns. Finally, surveillance bias isat least a theoretical possibility among hormonal ther-apy users. Since women who use tamoxifen therapyare more likely to have ongoing medical surveillance,they may be more likely to have colorectal lesionsdetected. Evidence for this bias would be suggestedif large bowel tumors among women using hormonaltherapy are more likely at an earlier stage of disease. Wedid not observe any stage differences in colon cancerbetween hormonal therapy users (43% localized) andnon-users (45% localized). Bias arising from lead timebias and selective detection of colorectal lesions amonghormonal therapy users would be mitigated in a study ofmortality, which could also be conducted in this cohort.

Any mechanisms involved in tamoxifen related car-cinogenesis might be through both tumor-initiating andtumor promoting actions [10, 21, 34]. Since our ob-servations of hormonal therapy effects in the bowelappear only after some induction time, tumor initiatingeffects may predominate. The demonstration of DNAadducts in a range of tissues following hormonal ther-apy use [35, 36] supports this mechanism. This appearsin contrast to the acute estrogen benefits of hormonereplacement on colorectal cancer incidence [2], butmay be consistent with a promoting effect on tumorprecursors, adenomatous polyps.

The findings of this and other studies are of interestbecause of tamoxifen’s widespread use [15], recentstudies demonstrating benefits of long term use [37,38], and its efficacy in healthy women as a method of


primary prevention of breast cancer [12]. Future stud-ies of breast cancer patients are necessary to provideessential information on variation of effects of dose,timing, and duration to clarify this relationship.

Acknowledgements

We are grateful to Barry E. Storer for statistical advice.This study was supported, in part, by grant R01 64138from the National Cancer Institute.

References

1. McMichael AJ, Potter JD: Reproductions, endogenousand exogenous hormones and colon cancer: a review andhypothesis. J Natl Cancer Inst 65: 1201–1207, 1980

2. Potter JD: Hormones and colon cancer. J Natl Cancer Inst87: 1039–1040, 1995

3. Calle EE: Hormone replacement therapy and colorectal can-cer: interpreting the evidence. Cancer Causes Control 8:127–129, 1997

4. Jacobs EJ, White E, Weiss NS: Exogenous hormones, repro-ductive history, and colon cancer. Cancer Causes Control 5:359–366, 1994

5. Newcomb PA, Storer BE: Postmenopausal hormone use andrisk of large-bowel cancer. J Natl Cancer Inst 87: 1067–1071, 1995

6. Calle EE, Miracle-McMahill HL, Thun MJ, Heath CW Jr:Estrogen replacement therapy and risk of fatal colon cancerin a prospective cohort of postmenopausal women. J NatlCancer Inst 87: 517–523, 1995

7. Grodstein F, Martinez ME, Giovannucci E: Postmenopausalhormone use and colorectal cancer in the nurse’s healthstudy. Am J Epidemiol 143: 563, 1996

8. Kampman E, Potter JD, Slattery ML, Caan BJ, EdwardsS: Hormone replacement therapy, reproductive history, andcolon cancer: a multicenter, case-control study in the UnitedStates. Cancer Causes Control 8: 146–158, 1997

9. Potter JD, Slattery ML, Bostick RM, Gapstur SM: Coloncancer a review of the epidemiology. Epidemiol Rev 15:499–545, 1993

10. Jordan VC: Tamoxifen and tumorigenicity: a predictableconcern. J Natl Cancer Inst 87: 623–626, 1995

11. Bush TL, Helzlsouer KJ: Tamoxifen for the primary preven-tion of breast cancer: a review and critique of the conceptand trial. Epidemiol Rev 15: 233–243, 1993

12. Fisher B, Constantino JP, Wickerman DL, Redmond CK,Kavanah M, Cronin WM, Vogel V, Ribidoux A, Dimitrov N,Atkins J, Daly M, Wieand S, Tan-Chiv E, Ford L, WolmarkN: J Natl Cancer Inst 90: 1371–1388, 1998

13. Love RR: Antiestrogens as chemopreventive agents inbreast cancer: promise and issues in evaluation. Prev Med18: 661–671, 1989

14. Furr BJ, Jordan VC: The pharmacology and clinical uses oftamoxifen. Pharmacol Ther 25: 127–205, 1984

15. Lerner LJ, Jordan VC: Development of antiestrogens andtheir use in breast cancer. Cancer Res 50: 4177–4189, 1990

16. Bilimona MM, Jordan VC, Morrow M: Activity in post-menopausal patients. In: Jordan VC (ed), Tamoxifen: aguide for clinicians and patients. PRR Inc., Huntington, NY,1996, pp 75–88

17. Francavilla A, Di Leo A, Polimeno L, Conte D, BaroneM, Fanizza G, Chiumarulo C, Rizzo G, Rubino M: Nu-clear and cytosolic estrogen receptors in human coloncarcinoma and in surrounding noncancerous colonic tissue.Gastroenterology 93: 1301–1306, 1987

18. McClendon JE, Appleby D, Claudon DB, Donegan WL,DeCosse JJ: Colonic neoplasms: tissue estrogen receptorand carcinoembryonic antigen. Arch Surg 112: 240–241,1977

19. Meggouh F, Lointier P, Pezet D, Saez S: Status of sex steroidhormone receptors in large bowel cancer. Cancer 67: 1964–1970, 1991

20. Di Leo A, Linsalata M, Cavallini A, Messa C, Russo F:Sex steroid hormone receptors, epidermal growth factorreceptor, and polyamines in human colorectal cancer. DisColon Rectum 35: 305–309, 1992

21. Rutqvist LE, Johansson H, Signomklao T, Johansson U,Fornander T, Wilking N: Adjuvant tamoxifen therapy forearly stage breast cancer and second primary malignancies.J Natl Cancer Inst 87: 645–651, 1995

22. Fisher B, Costantino JP, Redmond CK, Fisher ER, Wick-erham DL, Cronin WM: Endometrial cancer in tamoxifen-treated breast cancer patients: findings from the NationalSurgical Adjuvant Breast and Bowel Project (NSABP)B-14. J Natl Cancer Inst 86: 527–537, 1994

23. Curtis RE, Boice JD Jr, Shriner DA, Hankey BF, FraumeniJF Jr: Second cancers after adjuvant tamoxifen therapy forbreast cancer. J Natl Cancer Inst 88: 832–834, 1996

24. Schoenberg BS, Greenberg RA, Eisenberg H: Occurrenceof certain multiple primary cancers in females. J Natl CancerInst 43: 15–32, 1969

25. Kosary CL, Ries LAG, Miller BA, Harkey BF, Harras A,Edwards BK: SEER cancer statistics review, tables andgraphs: 1973–1993. National Cancer Institute, Bethesda,MD, 1996

26. Cox DR: Regression models and life tables. J R Stat Soc[B] 34: 187–220, 1972

27. Cook LS, Weiss NS, Schwartz SM, White E, McKnight B,Moore DE, Daling JR: Population-based study of tamoxifentherapy and subsequent ovarian, endometrial, and breastcancers. J Natl Cancer Inst 87: 1359–1364, 1995

28. Sasco AJ, Chaplain G, Amoros E, Saez S: Endometrialcancer following breast cancer: effect of tamoxifen andcastration by radiotherapy. Epidemiology 7: 9–13, 1996

29. van Leeuwen FE, Benraadt J, Coebergh JW, KiemeneyLA, Gimbrere CH, Otter R, Schouten LJ, Damhuis RA,Bontenbal M, Diepenhorst FW, van den Belt-Dusebout AW,van Tinteren H: Risk of endometrial cancer after tamoxifentreatment of breast cancer. Lancet 343: 448–452, 1994

30. Andersson M, Storm HH, Mouridsen HT: Incidence ofnew primary cancers after adjuvant tamoxifen therapy andradiotherapy for early breast cancer. J Natl Cancer Inst 83:1013–1017, 1991

31. Fornander T, Rutqvist LE, Cedermark B, Glas U, MattssonA, Silfversward C, Skoog L, Somell A, Theve T, WilkingN, Askergren J, Hjalmar ML: Adjuvant tamoxifen in earlybreast cancer: occurrence of new primary cancers. Lancet1: 117–120, 1989

32. Early Breast Cancer Trialists’ Collaborative Group: Sys-temic treatment of early breast cancer by hormonal, cyto-


toxic, or immune therapy: 133 randomised trials involv-ing 31,000 recurrences and 24,000 deaths among 75,000women. Lancet 339: 1–15, 1992

33. Cook LS, Weiss NS, Potts M: Re: Second cancers afteradjuvant tamoxifen therapy for breast cancer. [Letter]. J NatlCancer Inst 89: 657, 1997

34. Rutqvist LE: Re: Second cancers after adjuvant tamoxifentherapy for breast cancer. [Letter]. J Natl Cancer Inst 88:1497–1499, 1996

35. Han XL, Liehr JG: Induction of covalent DNA adducts inrodents by tamoxifen. Cancer Res 52: 1360–1363, 1992

36. Greaves P, Goonetilleke R, Nunn G, Topham J, Orton T:Two-year carcinogenicity study of tamoxifen in AlderleyPark Wistar-derived rats. Cancer Res 53: 3919–3924, 1993

37. Swedish Breast Cancer Cooperative Group: Randomizedtrial of two versus five years of adjuvant tamoxifen for

postmenopausal early stage breast cancer. J Natl Cancer Inst88: 1543–1549, 1996

38. Fisher B, Dignam J, Bryant J, DeCillis A, WickerhamDL, Wolmark N, Costantino J, Redmond C, Fisher ER,Bowman DM, Deschenes L, Dimitrov NV, Margolese RG,Robidoux A, Shibata H, Terz J, Paterson AH, FeldmanMI, Farrar W, Evans J, Lickley HL: Five versus more thanfive years of tamoxifen therapy for breast cancer patientswith negative lymph nodes and estrogen receptor-positivetumors. J Natl Cancer Inst 88: 1529–1542, 1996

Address for offprints and correspondence:P.A. Newcomb, FredHutchinson Cancer Research Center, 1100 Fairview Ave. N.,MP-702 PO Box 19024 Seattle, WA 98109-1024, USA;Tel:206 667 3476;Fax:206 667 5977;E-mail:[email protected]

Documents

Tamoxifen and risk of large bowel cancer in women with breast cancer