Objective: To investigate how adipose tissue alters endogenous hormone levels and may affect events at the ovarian tissue level.

Methods: We assessed current weight, weight at age 18, and adult weight change in relation to ovarian cancer risk among 109,445 participants in the Nurses' Health Study. Women reported ovarian cancer risk factors and new ovarian cancer diagnoses in biennial mailed questionnaires from 1976 to 1996. Height and weight were queried in 1976, current weight was updated biennially, and weight at age 18 was ascertained in 1980. During 20 years of follow-up and 1,703,474 person-years, 402 cases of epithelial ovarian cancer were confirmed. We used pooled logistic regression to control for age, oral contraceptive use, smoking history, parity, age at menarche, and tubal ligation.

Results: We found no evidence of an association between recent body mass index (BMI, kg/m²) and ovarian cancer risk. The multivariable relative risk for women with BMI of 30 kg/m² or higher versus BMI less than 21 kg/m² was 1.05 (95% confidence interval 0.73, 1.51). For BMI at age 18, there was no association with ovarian cancer risk overall, but a two-fold increase in premenopausal ovarian cancer risk associated with having a BMI at age 18 of 25 kg/m² or higher versus BMI less than 20 kg/m² (relative risk 2.05, 95% confidence interval 1.07, 3.93, P for trend = .01). Adult weight gain was not associated with ovarian cancer risk.

Conclusion: We found no evidence of an association between recent BMI or adult weight change and ovarian cancer risk. Higher BMI in young adulthood was associated with an increased risk of premenopausal ovarian cancer. If confirmed, these findings suggest an additional reason for avoiding adolescent obesity.

^aChanning Laboratory, Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts, USA
^bDivision of Preventive Medicine, Department of Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts, USA
^cDivision of General, Medicine, and Primary Care, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts, USA
^dDepartment of Nutrition, Harvard School of Public Health, Boston, Massachusetts, USA
^eDepartment of Environmental Science, Harvard School of Public Health, Boston, Massachusetts, USA
^fDepartment of Epidemiology, Harvard School of Public Health, Boston, Massachusetts, USA

This study was supported by grants (CA87969) from the National Institutes of Health and by a grant from the American Cancer Society (CCDA-0017901, to Dr. Fairfield).

We are indebted to the participants in the Nurses' Health Study for their continuing cooperation, and to Gary Chase, Karen Corsano, Barbara Egan, Diane Feskanich, Dorota Gertig, and Mary Louie.

(Obstet Gynecol 2002:100:288-296. © 2002 by The American College of Obstetricians and Gynecologists.)

The prevalence of obesity and overweight continues to increase markedly in the United States, particularly among young adults.¹ Health effects of obesity are numerous, including increased mortality from cancer and cardiovascular disease.^2,3 Alterations in endogenous hormones due to excess adiposity are a proposed mechanism by which obesity could influence some types of malignancies.⁴

The etiologies of ovarian cancer are still largely unknown. Reproductive experiences including pregnancies, breast-feeding, tubal ligation, and oral contraceptive use are associated with decreased risk of ovarian cancer, whereas advancing age increases risk.⁵ Ovarian stimulation from elevated gonadotropins associated with oocyte depletion is one proposed mechanism of pathogenesis.⁶ Repeated epithelial insults during ovulation, followed by higher mitotic rates during healing, may play a role in ovarian cancer initiation.⁷ High serum androgens may also stimulate ovarian epithelial proliferation and contribute to ovarian cancer risk.⁵

Women with a high body mass index (BMI) in late adolescence⁸ and at the time of attempted conception⁹ may be at increased risk of infertility, which in turn has been associated with increased ovarian cancer risk in some studies.^10-12 This relationship may be mediated by greater exposure to endogenous androgens associated with obesity.^13-15 In addition, adult weight gain and obesity in the postmenopausal period increase levels of endogenous estrogens.^16,17 The relationship between postmenopausal estrogens and ovarian cancer is unclear. Higher serum estrogens could act in ovarian cancer promotion, or alternatively could decrease gonadotropin release, in turn decreasing ovarian cancer risk.⁶

Several case-control studies have examined the relationship between obesity and ovarian cancer with variable findings.^18-24 These studies were limited by a single assessment of weight reported after disease diagnosis, although most attempted to ascertain usual adult weight before the onset of illness. Of the three prospective studies of obesity and ovarian cancer, one showed no relationship,²⁵ and another showed a 60% increase in ovarian cancer mortality for the most obese women.²⁶ In the prospective Iowa Women's Health Study, no association with current BMI was observed, but a more than two-fold increase in ovarian cancer risk was observed among women in the highest quartile of waist-to-hip ratio.²⁷

We examined ovarian cancer risk in relation to recent BMI, adult weight change, BMI at age 18, and waist-to-hip ratio among participants in the Nurses' Health Study, a prospective cohort of women followed for 20 years.

Materials and Methods

In 1976, 121,700 married, female registered nurses from eleven US states, aged 30-55, completed a mailed questionnaire on known or suspected risk factors for cancer and coronary heart disease. Follow-up questionnaires are mailed biennially to update information on risk factors and newly diagnosed diseases. The follow-up rate for the cohort through May 31, 1996, is 95% of potential person-years. The Institutional Review Board at Brigham and Women's Hospital approved this study.

We identified incident cases of ovarian cancer reported in each biennial questionnaire from 1976 to 1996. For women reporting such a history, we obtained discharge summaries and pertinent pathology reports. Trained physician reviewers unaware of exposure status established histologic type, subtype, morphology, and stage. Deaths in the cohort were identified through family members, the US Postal Service, and the National Death Index. We estimate that 98% of all deaths are ascertained.²⁸

On the 1976 questionnaire, we queried women about current height and weight. With each subsequent biennial questionnaire, women reported current weight. In 1980, we also asked women to report recalled weight at age 18. In 1986, we asked respondents to measure waist (while standing, relaxed) and hip circumference (largest point) with a tape measure, to the nearest ¹/₄ inch. We used BMI as our main measure of obesity in this analysis.

We categorized BMI into five groups: less than 21, 21-22.9, 23-24.9, 25-29.9, and higher than 30 kg/m² (National Institutes of Health clinical guidelines categorize BMI 25-29.9 as overweight and higher than 30 as obese).²⁹ Updated BMI (reported on the most recent questionnaire) was used as a measure of recent adiposity. If a nurse stopped reporting weight, the last reported weight was carried over for two questionnaire cycles, after which the nurse did not contribute person-time (to analyses of BMI and weight change) until reporting another weight. To calculate weight change from age 18, we subtracted reported weight at age 18 (in kg) from current reported weight on each biennial questionnaire. We categorized weight change into five groups: loss of 2 kg or more, loss or gain of 0-2 kg, gain of 2.1-10 kg, gain of 10.1-20 kg, or gain of more than 20 kg. For BMI at age 18, the exposure groups were: less than 20, 20-20.9, 21-22.9, 23-24.9, and higher than 25 kg/m². Too few women had BMI at age 18 higher than 30 to examine this group. Waist and hip circumference were queried once in 1986, and 52,391 women reported this information. We categorized waist-to-hip ratio in four groups: less than or equal to 0.73, 0.74-0.77, 0.78-0.82, and more than 0.83.

In a subsample of 184 women from the cohort, reported weights on a mailed questionnaire were highly correlated (r = 0.96) with technician-measured weights 6-12 months later, although averaging 1.5 kg (3.3 lb) lower than measured weights.³⁰ In a separate subsample of 140 women in the cohort, reported waist and circumferences were highly correlated with technician measurements (Pearson correlations 0.89 and 0.84, respectively). The correlation for waist-to-hip ratio was 0.62 after adjustment for age and BMI and correction for random measurement error.³¹ In a validation study among a different cohort of younger nurses participating in the Nurses' Health Study II, recalled weight at age 18 was highly correlated with recorded weights from school records (r = 0.87).³²

Using information from biennial questionnaires, we assessed exposure status for multiple known or proposed risk factors for ovarian cancer. Age was calculated from date of birth to date of questionnaire return, was updated biennially, and was divided into 5-year increments (less than 50, 50-54, 55-59, 60-64, and more than 65). Oral contraceptive use was assessed from 1976 to 1982, when use became rare because of the age of the cohort. We used duration of oral contraception use (never, less than 3 years, 3-5 years, or more than 5 years) as a covariate because this measure of exposure has been consistently associated with decreased risk in our study³³ and in others.^34,35 Parity, defined as number of pregnancies lasting at least 6 months, was queried from 1976 through 1984 and was categorized as none, one to two, three to four, or more than four. Age at menarche was assessed in 1976 and classified into age less than 12, 12, and 13 or older. Tubal ligation history (ever, never) was reported in each biennial questionnaire from 1976 to 1982 and in 1994. We assessed smoking history in every questionnaire, and categorized exposure as never, past, or current use. For all main analyses (updated BMI, BMI at age 18, weight change, and waist-to-hip ratio), the covariates were used as described above in our models. For stratified analyses to look for evidence of associations among specific groups, we collapsed variables into two categories only because of small case numbers. Updated variables (reflecting responses in the most recent questionnaires) were used in all analyses where available, as described above.

Age at first birth was assessed in 1976, whereas age at menopause and postmenopausal hormone use were assessed in 1976 with biennial updates thereafter. We queried women about history of menstrual irregularity in 1982, defined as usually having regular versus irregular menses. Lactation was queried in 1986, and family history of ovarian cancer (in mother or sister) in 1992. Physical activity was ascertained in 1980 and updated biennially using a series of questions about daily household and recreational activities, including housework, gardening, walking, tennis, and running, among others. We used a cumulative updated measure of activity grouped into five categories as a potential covariate in our models.

Potential dietary risk factors were assessed using a semiquantitative food frequency questionnaire, administered in 1980, and updated in 1984, 1986, and 1990. Details of the questionnaire and documentation of its reproducibility and validity have been published elsewhere.^36-39 We considered dietary fat, fiber, alcohol, lactose, caffeine, and total energy, all in quintiles.

We excluded women based on the following: those who did not report weight or height in 1976 (n = 1277); those reporting any diagnosis of cancer besides nonmelanoma skin cancer (n = 3148); those with a history of bilateral oophorectomy or hysterectomy with unknown number of ovaries removed; and those with history of pelvic irradiation (n = 7628). We previously reported 97% agreement between self-reported surgical menopause and medical records.⁴⁰ Because ovarian cancer may become advanced before clinical detection, we used a 2-year lag between exposure (current BMI) and incidence for each period, such that BMI in 1976 was used to predict ovarian cancer rates for the period 1978-1980. We therefore excluded person-time and cases (n = 17) from 1976 to 1978 in all primary analyses of BMI. After exclusions, 109,445 women remained in the analysis at the start of follow-up. Exclusions were updated biennially. During the 20-year period of follow-up and 1,703,474 person-years, 697 cases of ovarian cancer were reported by participants. Of these cases, 42 were nonepithelial (and not the subject of this analysis), we could not confirm 79 because of unavailable medical records or nurse refusal, 130 were rejected after reviewing medical records or contacting the nurse (most commonly these were nonmalignant ovarian tumors or other tumors with ovarian metastases), and we confirmed 446 cases (402 of which were invasive) by medical record review. All analyses use the 402 invasive cases unless otherwise noted.

We calculated person-years from the 1978 questionnaire return date to date of ovarian cancer diagnosis, death, or May 31, 1996, whichever was sooner. Incidence rates were calculated by dividing the number of cases by the number of person-years in each category of BMI. We computed relative risks (RRs) by dividing the rate for the highest four categories of BMI by that for the lowest category. We used pooled logistic regression with 2-year time increments⁴¹ to control for age, duration of oral contraceptive use, parity, age at menarche, tubal ligation, and smoking. Pooled logistic regression allows us to evaluate the strength of the relationship between a "yes/no" outcome (ovarian cancer) and an exposure (obesity), which is changing over time while adjusting for other factors. Although our main BMI analyses included a 2-year time lag, we also assessed exposure latency by conducting analyses using 4-, and 8-year time lags between BMI and ovarian cancer incidence. For example, in the 4-year time lag analysis, 1976 BMI was used to predict ovarian cancer risk for the 1980 to 1982 period, and 1978 BMI to predict risk from 1982 to 1984.

We also assessed additional factors in all our models, including analyses of BMI, BMI at age 18, weight change, and waist-to-hip ratio. Potential confounders included age at first birth, postmenopausal hormone use, and age at menopause. As these factors were not found to influence the relation between BMI and ovarian cancer, we did not include them in our final models. In analyses limited to 1980 through 1996, we examined physical activity as well as fat, fiber, alcohol, caffeine, lactose, and total energy consumption in our models. Although physical activity may be along the etiologic pathway of obesity and ovarian cancer (and therefore it may best not be included in final models, anyway), it was not found to influence the relation between BMI and ovarian cancer risk. Because the specific dietary factors also did not alter this relationship, we excluded them from our final models. We carried out stratified analyses using person-time and cases from 1982 forward to assess possible associations with menstrual irregularity. Analyses of family history of ovarian cancer and breast-feeding history were assessed using only the last 4 and 10 years of follow-up, respectively. We used interaction terms and stratified analyses to assess for differential effects (effect modification) among levels of the variables included in the final model and the other potential covariates described above. We report RRs and 95% confidence intervals (CIs), as well as tests for trend.

For all analyses, we carried out analyses with all women together and stratified by menopausal status. In the latter analysis, women were not included during intervals when their menopausal status was uncertain, but they subsequently reentered into the analysis when they reported definite menopause.

In additional analyses, we used BMI at age 18 and weight change from age 18 to current weight to predict ovarian cancer risk. Because weight at age 18 was first ascertained in 1980 as noted, these analyses include only cases and person-time from 1980 through 1996 (331 cases). Women who did not report BMI at age 18 were excluded from these analyses. We performed our main analysis with a 2-year time lag, and also carried out analyses using 4- and 6-year time lags. We also assessed recent weight change using weight in 1976 subtracted from currently reported weight (analyses included a 2-year lag). For waist-to-hip ratio, we used person-time from 1988 forward, to include a 2-year lag after ascertainment of waist and hip circumference in 1986. This analysis was limited to the 52,391 women reporting these measurements.

Only invasive epithelial cancers were considered in our main analyses. We also evaluated risk of invasive and borderline tumors together and risk of the major specific subtypes of ovarian cancer (serous, mucinous, and endometrioid).

Results

We documented 402 cases of invasive epithelial ovarian cancer during 1,703,474 person-years of follow-up. The average age at ovarian cancer diagnosis was 57 years. Women with a higher baseline BMI were less likely to use oral contraceptives than leaner women, and were less likely to use postmenopausal estrogens, or to be current smokers in 1976 (Table 1).

Recent BMI was not consistently associated with ovarian cancer risk overall (RR 1.05 for BMI 30 or higher compared with less than 21, 95% CI 0.73, 1.51), or specifically among premenopausal or postmenopausal women (Table 2). We found no evidence that other potential ovarian cancer risk factors influenced these relationships. Although we were unable to control for having a family history of ovarian cancer (it was not asked until 1992), a reported family history was uncommon and varied only modestly according to category of BMI (Table 1). There were no differential effects of BMI among categories of the other potential risk factors, including physical activity, or among the covariates in the final model. Further adjustment for weight change did not alter our findings. In analyses including borderline tumors (n = 446), our findings were similar.

We observed a two-fold increase in risk of premenopausal ovarian cancer (RR 2.05, 95% CI 1.07, 3.93, P for trend = .01) among women with the highest BMI at age 18 (25 kg/m² or more compared with less than 20 kg/m²) (Table 3). There was no apparent increase in risk of postmenopausal ovarian cancer associated with higher BMI at age 18. Among premenopausal women who never used oral contraceptives, women with the highest BMI at age 18 appeared to have a higher risk of ovarian cancer (RR 2.59, 95% CI 1.01, 6.64); among oral contraceptive users, the RR was 1.64 (95% CI 0.66, 4.06). We observed no other differential associations among strata of other risk factors, including postmenopausal hormone use, parity, history of tubal ligation, smoking history, age at menarche, and duration of oral contraceptive use. In analyses from 1982 to 1996, history of menstrual irregularity did not alter the relationships between BMI at age 18 and ovarian cancer. When we stratified women according to this history, we found similar associations between BMI at age 18 and ovarian cancer in both women with and without a history of menstrual irregularity. Analyses adjusting for adult weight change yielded similar results.

In our main analysis of weight change (with a 2-year lag), we observed an apparent decrease in risk of ovarian cancer among women who gained 20 kg or more since age 18, compared with a 2-kg or less loss or gain (RR 0.63, 95% CI 0.40, 1.01, P for trend = .006) (Table 4). However, after adding lags of 4 or 6 years into the analysis, this association essentially disappeared. Results were very similar when stratified by menopausal status. Controlling for postmenopausal hormone use did not alter the findings. In stratified analyses, we observed no differential effects of weight gain among users versus nonusers of postmenopausal hormones, although small numbers in each category of hormone use limited this analysis. Similarly, the relationships between weight change and ovarian cancer did not appear to vary by level of any other potential risk factors including oral contraceptive use. We had limited power to evaluate relationships between adult weight loss and ovarian cancer risk because so few women lost weight from age 18 during the follow-up period. In analyses of more recent weight change (from 1976 to present weight), we found similar relationships as in change from age 18. With a 2-year lag, there was a 30% lower risk of ovarian cancer among women with a gain of 20 kg or more, compared with those with a 2.0-kg loss or gain (RR 0.70, 95% CI 0.34, 1.42). With increasing lags, this association approached the null.

We found no evidence of an association between waist-to-hip ratio and ovarian cancer risk, although we had limited power to evaluate this association in analyses with a 2-year lag (n = 83). For women with the highest waist-to-hip ratio (0.83 or more) versus the lowest (0.73 or less), we found a multivariable RR of 1.12 (95% CI 0.59, 2.12, P for trend = .96). We attempted to further divide the uppermost group to include waist-to-hip ratios more than 0.89, but only four cases remained in the highest group. An analysis without a time lag revealed similar results (n = 100; for waist-to-hip ratio 0.83 or less versus lowest, the multivariable RR was 1.01, 95% CI 0.56, 1.83).

In analyses including both invasive and borderline tumors together (n = 106 premenopausal cases), we found slight attenuation of our results for BMI at age 18 among premenopausal women (for BMI higher than 25 kg/m², RR 1.74, 95% CI 0.96, 3.15) and similar results for postmenopausal women. Higher BMI at age 18 appeared more strongly associated with the serous subtype of ovarian cancer among premenopausal women (for BMI higher than 25 kg/m², RR 2.90, 95% CI 1.04, 8.08), although there were only 45 cases in this subgroup analysis. For recent BMI, we found no association with the serous subtype of ovarian cancer, similar to our finding with all subtypes combined. We had insufficient power to reliably estimate risks of recent or young adult obesity associated with the risk of endometrioid or mucinous subtypes in either menopausal group or among all women together.

Discussion

We observed no association between recent BMI and ovarian cancer risk overall or by menopausal status. Weight gain from age 18 and recent weight gain were not associated with ovarian cancer risk in analyses with time lags to exclude women with preclinical disease. Waist-to-hip ratio also did not appear associated with ovarian cancer risk. Obesity at age 18 was associated with a two-fold higher risk of premenopausal ovarian cancer, but not postmenopausal cancer.

The observed lack of an association with recent BMI in our study is consistent with findings from the Iowa Women's Health Study,²⁷ and others,^18,20,22,25 which found no association between BMI and ovarian cancer risk. We are unaware of other studies in which the relationship between weight gain and ovarian cancer risk was examined. Our finding of a modest inverse association between weight gain and postmenopausal ovarian cancer disappeared with increasing time lags (although the test for trend remains statistically significant because we include the weight loss category in the test for trend). This suggests that the observed association (without lags) is due to preclinical disease, and that weight changes due to tumor growth may antedate diagnosis by several years. Ovarian cancer is often diagnosed at a late stage, with a long period of active disease while the woman is undiagnosed. We did not observe this finding with recent BMI, but weight loss may be a more sensitive indicator of preclinical disease.

Unlike the Iowa Women's Health Study, we did not find evidence of an association between waist-to-hip ratio and ovarian cancer risk.²⁷ Although we had only 83 cases in our main analysis with a 2-year time lag (100 cases in analysis without a lag), the Iowa study included only 97 cases. In addition, although that study reported an RR of 2.3 for the highest waist-to-hip ratio (more than 0.89), there was no apparent dose-effect gradient. However, we had limited power to assess this association, and our CIs include RRs comparable with those previously reported. Also, the women in the Iowa cohort had a higher distribution of waist-to-hip ratio than did the nurses.

Although few studies have assessed adolescent or young adult BMI and ovarian cancer, our findings were similar to those of Farrow et al, who found women with the highest BMI at age 30 had a 70% increase in risk of ovarian cancer, particularly in the premenopausal period.²¹ Women who are obese in adolescence experience more menstrual irregularity⁸ and have higher endogenous androgen levels.^13-15 Oral contraceptives (which decrease ovarian cancer risk) are known to decrease ovarian testosterone.^42-44 Ovarian cancer was associated with elevated serum androstenedione and dehydroepiandrosterone in one study.⁴⁵ Our finding that higher BMI at age 18 is associated with increased ovarian cancer risk, particularly among nonusers of oral contraceptives, supports the hypothesis that endogenous androgens increase ovarian cancer risk. Lower circulating progesterone provides another possible explanation for the observed increase in risk of premenopausal ovarian cancer with young adult BMI. Progesterone is proposed as a potential protective factor for ovarian cancer.⁵ Obesity in premenopausal women is associated with an increased number of anovulatory cycles and, subsequently, lower progesterone levels.⁴⁶ Of note, our finding that young adult obesity increases risk does not support the hypothesis that epithelial injury from uninterrupted ovulation is the critical factor in ovarian cancer initiation. However, uninterrupted ovulation may still play an initiating role.⁷

Specific strengths of our study include multiple, prospective ascertainments of weight among a large cohort of women. Having multiple measures of weight minimizes the potential impact of measurement error, and also allowed us to assess associations with weight gain over time. Our use of repeated measures also allows analyses with lags between exposure and disease incidence — particularly important in these analyses because of the long preclinical disease period. The long duration of follow-up for this cohort makes it possible to assess relationships between ovarian cancer and obesity or weight gain among both premenopausal and postmenopausal women. This is particularly important given the proposed physiologic effects of adiposity (related to endogenous hormonal balance), which would be expected to differ in the ovulating versus quiescent ovary.

Potential limitations of the study include the possibility of error in self-reported weights. However, measurement error in this setting is expected to be random and would generally bias results towards the null. In addition, an earlier validation study revealed a high correlation between self-reported and technician-measured weights.³⁰ Preclinical disease may also result in changes in weight, either because of tumor-induced cachexia or weight gain due to ascites. However, our main analyses employed a 2-year time lag, and analyses using a 4-year time lag were similar. We use BMI as our main measure of adiposity in this study. Although BMI is an imperfect measure of adiposity (particularly in older individuals), it is strongly correlated with total body fat content in adults and is easy to measure in clinical settings.²⁹ We are also limited by having only a single reported height measurement in 1976. However, we expect loss of height over the course of the study to be modest for most women, and would not expect this to vary substantially according to history of ovarian cancer.

Another potential limitation is our inability to control for family history in our main analyses. However, analyses from 1992 forward (albeit with limited power) suggest that family history is unlikely to be responsible for our findings. Although other factors may have influenced our findings, we have multiple prospective measures of the known ovarian cancer risk factors. Although we did not ascertain parity after 1984, the majority of the women in the cohort would have completed childbearing by then (the youngest women were 38 at that point). In addition, the magnitude of the observed increase in premenopausal ovarian cancer risk associated with obesity at age 18 is large, making residual confounding by parity or other factors an unlikely explanation for our findings. Our cohort is mainly white, limiting generalizability to minority populations. This is particularly important because obesity is more prevalent among blacks, and incidence of ovarian cancer differs by racial group. We are also somewhat limited by relatively few ovarian cases overall in the cohort, which made subgroup and stratified analyses difficult. Also, BMI was assessed from 1976 forward, whereas BMI at age 18 was not assessed until 1980. However, analyses using BMI data from 1980 through 1996 produced similar results.

In summary, we observed no association between recent BMI or weight change and ovarian cancer. Women who were heaviest at age 18 were at increased risk of ovarian cancer in the premenopausal period. This is particularly concerning, given the current epidemic of obesity among adolescents and young adults. If confirmed, these findings suggest an additional reason for avoiding adolescent obesity.

References

1. Mokdad AH, Serdula MK, Dietz WH, Bowman BA, Marks JS, Koplan JP. The spread of the obesity epidemic in the United States, 1991-1998. JAMA 1999;282:1519-22.

2. Manson JE, Willett WC, Stampfer MJ, Colditz GA, Hunter DJ, Hankinson SE, et al. Body weight and mortality among women. N Engl J Med 1995;333:677-85.

3. Calle EE, Thun MJ, Petrelli JM, Rodriguez C, Heath CW Jr. Body-mass index and mortality in a prospective cohort of U.S. adults. N Engl J Med 1999;341:1097-105.

4. Kirschner MA, Ertel N, Schneider G. Obesity, hormones, and cancer. Cancer Res 1981;41:3711-7.

5. Risch HA. Hormonal etiology of epithelial ovarian cancer, with a hypothesis concerning the role of androgens and progesterone. J Natl Cancer Inst 1998;90:1774-86.

6. Cramer DW, Welch WR. Determinants of ovarian cancer risk. II. Inferences regarding pathogenesis. J Natl Cancer Inst 1983;71:717-21.

7. Casagrande JT, Louie EW, Pike MC, Roy S, Ross RK, Henderson BE. "Incessant ovulation" and ovarian cancer. Lancet 1979;2:170-3.

8. Rich-Edwards JW, Goldman MB, Willett WC, Hunter DJ, Stampfer MJ, Colditz GA, et al. Adolescent body mass index and infertility caused by ovulatory disorder. Am J Obstet Gynecol 1994;171:171-7.

9. Green BB, Weiss NS, Daling JR. Risk of ovulatory infertility in relation to body weight. Fertil Steril 1988;50:721-6.

10. Chen Y, Wu PC, Lang JH, Ge WJ, Hartge P, Brinton LA. Risk factors for epithelial ovarian cancer in Beijing, China. Int J Epidemiol 1992;21:23-9.

11. Booth M, Beral V, Smith P. Risk factors for ovarian cancer: A case-control study. Br J Cancer 1989;60:592-8.

12. Harris R, Whittemore AS, Itnyre J. Characteristics relating to ovarian cancer risk: Collaborative analysis of 12 US case-control studies. III. Epithelial tumors of low malignant potential in white women. Collaborative Ovarian Cancer Group. Am J Epidemiol 1992;136:1204-11.

13. Bates GW, Whitworth NS. Effect of body weight reduction on plasma androgens in obese, infertile women. Fertil Steril 1982;38:406-9.

14. Harlass FE, Plymate SR, Fariss BL, Belts RP. Weight loss is associated with correction of gonadotropin and sex steroid abnormalities in the obese anovulatory female. Fertil Steril 1984;42:649-52.

15. Pasquali R, Antenucci D, Casimirri F, Venturoli S, Paradisi R, Fabbri R, et al. Clinical and hormonal characteristics of obese amenorrheic hyperandrogenic women before and after weight loss. J Clin Endocrinol Metab 1989;68:173-9.

16. Hankinson SE, Willett WC, Manson JE, Hunter DJ, Colditz GA, Stampfer MJ, et al. Alcohol, height, and adiposity in relation to estrogen and prolactin levels in postmenopausal women. J Natl Cancer Inst 1995;87:1297-302.

17. Lipworth L, Adami HO, Trichopoulos D, Carlstrom K, Mantzoros C. Serum steroid hormone levels, sex hormone-binding globulin, and body mass index in the etiology of postmenopausal breast cancer. Epidemiology 1996;7:96-100.

18. Annegers JF, Strom H, Decker DG, Dockerty MB, O'Fallon WM. Ovarian cancer: Incidence and case-control study. Cancer 1979;43:723-9.

19. Risch HA, Weiss NS, Lyon JL, Daling JR, Liff JM. Events of reproductive life and the incidence of epithelial ovarian cancer. Am J Epidemiol 1983;117:128-39.

20. Cramer DW, Welch WR, Hutchison GB, Willett W, Scully RE. Dietary animal fat in relation to ovarian cancer risk. Obstet Gynecol 1984;63:833-8.

21. Farrow DC, Weiss NS, Lyon JL, Daling JR. Association of obesity and ovarian cancer in a case-control study. Am J Epidemiol 1989;129:1300-4.

22. Hartge P, Schiffman MH, Hoover R, McGowan L, Lesher L, Norris HJ. A case-control study of epithelial ovarian cancer. Am J Obstet Gynecol 1989;161:10-6.

23. Purdie D, Green A, Bain C, Siskind V, Ward B, Hacker N, et al. Reproductive and other factors and risk of epithelial ovarian cancer: An Australian case-control study. Survey of Women's Health Study Group. Int J Cancer 1995;62:678-84.

24. Parazzini F, Moroni S, La Vecchia C, Negri E, dal Pino D, Bolis G. Ovarian cancer risk and history of selected medical conditions linked with female hormones. Eur J Cancer 1997;33:1634-7.

25. Tornberg SA, Carstensen JM. Relationship between Quetelet's index and cancer of breast and female genital tract in 47,000 women followed for 25 years. Br J Cancer 1994;69:358-61.

26. Garfinkel L. Overweight and cancer. Ann Intern Med 1985;103:1034-6.

27. Mink PJ, Folsom AR, Sellers TA, Kushi LH. Physical activity, waist-to-hip ratio, and other risk factors for ovarian cancer: A follow-up study of older women. Epidemiology 1996;7:38-45.

28. Stampfer MJ, Willett WC, Speizer FE, Dysert DC, Lipnick R, Rosner B, et al. Test of the national death index. Am J Epidemiol 1984;119:837-9.

29. National Institutes of Health. Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults. Bethesda, MD: National Institutes of Health, National Heart, Lung and Blood Institute, 1998.

30. Willett W, Stampfer MJ, Bain C, Lipnick R, Speizer FE, Rosner B, et al. Cigarette smoking, relative weight, and menopause. Am J Epidemiol 1983;117:651-8.

31. Rimm EB, Stampfer MJ, Colditz GA, Chute CG, Litin LB, Willett WC. Validity of self-reported waist and hip circumferences in men and women. Epidemiology 1990;1:466-73.

32. Troy LM, Hunter DJ, Manson JE, Colditz GA, Stampfer MJ, Willett WC. The validity of recalled weight among younger women. Int J Obes Relat Metab Disord 1995;19:570-2.

33. Hankinson SE, Colditz GA, Hunter DJ, Willett WC, Stampfer MJ, Rosner B, et al. A prospective study of reproductive factors and risk of epithelial ovarian cancer. Cancer 1995;76:284-90.

34. Whittemore AS, Harris R, Itnyre J. Characteristics relating to ovarian cancer risk: Collaborative analysis of 12 US case-control studies. II. Invasive epithelial ovarian cancers in white women. Collaborative Ovarian Cancer Group. Am J Epidemiol 1992;136:1184-203.

35. Cancer and Steroid Hormone Study. The reduction in risk of ovarian cancer associated with oral-contraceptive use. Centers for Disease Control Cancer and Steroid Hormone Study. N Engl J Med 1987;316:650-5.

36. Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985;122:51-65.

37. Willett WC, Sampson L, Browne ML, Stampfer MJ, Rosner B, Hennekens CH, et al. The use of a self-administered questionnaire to assess diet four years in the past. Am J Epidemiol 1988;127:188-99.

38. Colditz GA, Martin P, Stampfer MJ, Willett WC, Sampson L, Rosner B, et al. Validation of questionnaire information on risk factors and disease outcomes in a prospective cohort study of women. Am J Epidemiol 1986;123:894-900.

39. Salvini S, Hunter DJ, Sampson L, Stampfer MJ, Colditz GA, Rosner B, et al. Food-based validation of a dietary questionnaire: The effects of week-to-week variation in food consumption. Int J Epidemiol 1989;18:858-67.

40. Hankinson SE, Hunter DJ, Colditz GA, Willett WC, Stampfer MJ, Rosner B, et al. Tubal ligation, hysterectomy, and risk of ovarian cancer. JAMA 1993;270:2813-8.

41. D'Agostino RB, Lee ML, Balanger AJ, Cupples LA, Anderson K, Kannel WB. Relation of pooled logistic regression to time dependent Cox regression analysis: The Framingham Heart Study. Stat Med 1989;9:1501-15.

42. Murphy A, Cropp CS, Smith BS, Burkman RT, Zacur HA. Effect of low-dose oral contraceptive on gonadotropins, androgens, and sex hormone binding globulin in nonhirsute women. Fertil Steril 1990;53:35-9.

43. van der Vange N, Blankenstein MA, Kloosterboer HJ, Haspels AA, Thijssen JH. Effects of seven low-dose combined oral contraceptives on sex hormone binding globulin, corticosteroid binding globulin, total and free testosterone. Contraception 1990;41:345-52.

44. Kuhnz W, Sostarek D, Gansau C, Louton T, Mahler M. Single and multiple administration of a new triphasic oral contraceptive to women: Pharmacokinetics of ethinyl estradiol and free and total testosterone levels in serum. Am J Obstet Gynecol 1991;165:596-602.

45. Helzlsouer KJ, Alberg AJ, Gordon GB, Longcope C, Bush TL, Hoffman SC, Comstock GW. Serum gonadotropins and steroid hormones and the development of ovarian cancer. JAMA 1995;274:1926-30.

46. Potischman N, Swanson CA, Siiteri P, Hoover RN. Reversal of relation between body mass and endogenous estrogen concentrations with menopausal status. J Natl Cancer Inst 1996;88:756-8.

Address reprint requests to: Kathleen M. Fairfield, MD, DrPH, Brigham and Women's Hospital, Harvard Medical School, Department of Medicine, Channing Laboratory, 181 Longwood Avenue, Boston, MA 02115; E-mail: kathleen.fairfield@channing.harvard.edu

Received September 17, 2001.
Received in revised form February 19, 2002.
Accepted March 14, 2002.