Anthropometric variables, physical activity, and incidence of ovarian cancer

The Iowa Women's Health Study

Authors

  • Jeffrey P. Anderson B.S.,

    1. Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, Minnesota
    Search for more papers by this author
  • Julie A. Ross Ph.D.,

    1. Division of Pediatric Epidemiology and Clinical Research, Department of Pediatrics, University of Minnesota, Minneapolis, Minnesota
    Search for more papers by this author
  • Aaron R. Folsom M.D., M.P.H.

    Corresponding author
    1. Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, Minnesota
    • Division of Epidemiology, School of Public Health, University of Minnesota, Suite 300, 1300 South Second Street, Minneapolis, MN 55454-1015
    Search for more papers by this author
    • Fax: (612) 624-0315


Abstract

BACKGROUND

Reports on the relation between anthropometric variables (height, weight) and physical activity with ovarian cancer have been inconclusive. The objective of the current study was to extend investigation of potential associations in the Iowa Women's Health Study cohort.

METHODS

The relation between self-reported anthropometric variables and incident ovarian cancer was studied in a prospective cohort of women ages 55–69 years who were followed for 15 years. Two hundred twenty-three incident cases of epithelial ovarian cancer were identified by linkage to a cancer registry.

RESULTS

No association was found overall between ovarian cancer and height, but a positive association was observed for serous ovarian cancers (relative risk [RR], 1.86 for highest quartile vs. lowest quartile; 95% confidence interval [95% CI], 1.06–3.29). Although current body mass index (BMI) was not associated with ovarian cancer, a BMI ≥ 30 kg/m2 at age 18 years appeared to be associated positively with ovarian cancer (multivariate-adjusted RR, 1.83 for BMI ≥ 30 kg/m2 vs. BMI < 25 kg/m2; 95% CI, 0.90–3.72), and this association was stronger after exclusion of the first 2 years of follow-up (RR, 2.15; 95% CI, 1.05–4.40). In a multivariate analysis, waist-to-hip ratio was associated with ovarian cancer (RR, 1.59 for highest quartile vs. lowest quartile; 95% CI, 1.05–2.40), but a linear dose response was not found. An index that combined the frequency and intensity of leisure-time physical activity was associated positively with ovarian cancer incidence (multivariate-adjusted RR, 1.42 for high activity vs. low activity; 95% CI, 1.03–1.97). This association was particularly strong for frequency of vigorous physical activity (multivariate-adjusted RR, 2.38 for > 4 times per week vs. rarely/never; 95% CI, 1.29–4.38).

CONCLUSIONS

Anthropometric variables were not major risk factors for ovarian cancer in the cohort studied; however, high BMI in early adulthood and frequent and vigorous physical activity may increase the risk of ovarian cancer among postmenopausal women. Cancer 2004;100:1515–21. © 2004 American Cancer Society.

Ovarian cancer ranks sixth in cancer incidence and fifth in mortality among women in the United States. Approximately 4% of female cancers are ovarian.1 Established risk factors for ovarian cancer include age; family history of ovarian, breast, endometrial, prostate, or colorectal cancer; inheritance of BRCA1/2 mutations; and hereditary nonpolyposis colorectal cancer syndrome. Other risk factors may include long-term, unopposed estrogen replacement therapy (ERT); infertility and/or long-term use of fertility drugs; early age at menarche; late age at menopause; and low parity. Protective factors include hysterectomy, tubal ligation, and the use of oral contraceptives.1–3

In an earlier report after 7 years of follow-up of the Iowa Women's Health Study, we found a positive association of ovarian cancer incidence with physical activity and waist-to-hip ratio (WHR) but no association with body mass index (BMI).4 We did not investigate other anthropometric variables. Recently, others have reported a positive association of ovarian cancer with height,5–8 BMI,5, 6, 9–13 BMI interacting with ERT,5 and weight history.8, 14–16 Positive, negative, and null associations of ovarian cancer with physical activity have been reported since our earlier finding.17–21

To verify our findings with longer follow-up and to potentially replicate the findings of these recent reports,4–21 we examined the association of anthropometric variables and physical activity with the incidence of ovarian cancer over 15 years of follow-up in the Iowa Women's Health Study cohort.

MATERIALS AND METHODS

The Iowa Women's Health Study Cohort

In 1986, we mailed a questionnaire to a random sample of 99,826 women ages 55–69 years who had valid Iowa driver's licenses in 1985. Of these women, 41,836 returned the questionnaire and were deemed eligible. Nonrespondents had somewhat higher incidence rates of smoking-related diseases, but associations of BMI with cancer occurrence did not differ between respondents and nonrespondents.22 Follow-up questionnaires were mailed in 1987, 1989, 1992, and 1997 to confirm participants' residence, vital status, and other characteristics. The University of Minnesota Institutional Review Board approved the study.

Questionnaire

The baseline questionnaire assessed smoking status and amount of smoking. Physical leisure activity was assessed only at baseline with a general question used by the Gallup Poll23: “Aside from any work you do at home or at a job, do you do anything regularly—that is, on a daily basis—that helps keep you physically fit?” In addition, participants answered two questions regarding how often they participated in moderate physical activity (for example, bowling, golf, light sports or physical exercise, gardening, taking long walks, etc.) and vigorous physical activity (for example, jogging, racket sports, swimming, aerobics, strenuous sports, etc.). We examined responses to these two questions individually, and we also combined them to form a three-level physical activity index (low, moderate, and high) based on frequency and intensity of activity.4 This physical activity index was associated inversely with mortality from coronary heart disease in this cohort,24 indicating that it has reasonable predictive validity.

Participants were asked to report current height and weight, from which BMI (in kg/m2) was computed. A paper measuring tape was sent to each participant so that a friend, spouse, or relative could measure the circumferences of the waist (1 inch above the umbilicus) and the hips (maximum), and WHRs were subsequently calculated. The self-measured or self-reported anthropometric measures obtained using this protocol were valid and reliable.25 Participants were asked to report their weight at ages 18, 30, 40, and 50 years and also to report their maximum weight.

Women were asked at baseline the age at which they first menstruated; whether they currently had menstrual periods (within the past year); and if not, the age at which their menstrual periods no longer occurred. In addition, they were asked whether they had undergone surgical removal of the uterus and/or one or both ovaries. Hysterectomy and full/partial oophorectomy status was updated in the 1992 follow-up questionnaire. Participants were asked at baseline whether they had ever been pregnant and, for each pregnancy (up to 10 pregnancies), their age, pregnancy duration, and outcome (live birth, stillbirth, miscarriage, ectopic, or induced abortion). Participants also were asked whether they had ever taken birth control pills, at what age they began taking them, and how long they continued to take them. However, given the age of this cohort, few women had used oral contraceptives, and use was not related to ovarian cancer risk. Information on ERT was obtained by asking women in the study cohort whether they had ever used pills, other than birth control pills, that contained estrogen or other female hormones and how long they continued to take them (with response categories of < 1 month, 2–6 months, 7–12 months, 1–2 years, 3–5 years, and > 5 years). Type of ERT was not assessed.

To ascertain prevalent cancer, participants were asked whether they had ever been diagnosed by a physician with any form of cancer other than skin cancer and to specify the site. Participants were asked whether their mother, maternal and paternal grandmothers, aunts, sisters, and daughters had cancer diagnosed, including ovarian cancer.

Follow-Up

Patients with ovarian cancer were ascertained through the State Health Registry of Iowa, part of the National Cancer Institute's Surveillance, Epidemiology, and End Results Program, using an annual computer match of participant identifiers. Primary disease site, morphology, extent of disease, and date of diagnosis were obtained for each incident case of ovarian cancer from 1986 through 2000. Only patients who were diagnosed within the state of Iowa were identified and included. The current analysis included ovarian cancers classified as common epithelial tumors by the World Health Organization Histologic Classification of Ovarian Tumors.26 Serous ovarian cancers were defined by International Classification of Diseases for Oncology (second edition) morphology codes 8441, 8442, 8460, 8461, and 8462.

We considered women to be at risk from January 1986 through December 31 2000 or until they were diagnosed with ovarian cancer, died, left Iowa, or were otherwise lost to follow-up. Women who had a diagnosis of any other cancer during follow-up were considered to be at risk of ovarian cancer and were not censored. Information on deaths occurring in Iowa was obtained from the State Health Registry. Information on deaths occurring outside Iowa was obtained from the 1987, 1989, 1992, and 1997 mailed follow-up questionnaires and, for questionnaire nonrespondents, from the National Death Index. For deaths occurring outside Iowa, the censoring date was the midpoint between the date of last contact in Iowa and the date of death. Each case in which a woman was known to have left Iowa was censored at the date of the move, if known; otherwise, censoring occurred at the midpoint between the date of last contact in Iowa and the first known date on which the woman lived outside Iowa or the end of the follow-up period. The emigration rate among this cohort has been approximately 1% per year.22

Data Analysis

We excluded women who reported a history of cancer other than skin cancer at baseline (n = 3830) or who underwent bilateral oophorectomy (n = 8064). One thousand four hundred fifty-five women reported both baseline cancer and bilateral oophorectomy. We excluded women who developed nonepithelial ovarian neoplasms (n = 16) but not women who developed borderline tumors of low malignant potential. After these exclusions, 31,381 women remained in the at-risk cohort. We ran analyses using the updated oophorectomy information from the 1992 follow-up, censoring the 1219 women who reported undergoing an interim bilateral oophorectomy. Data from the State Health Registry led to the identification of 223 incident cases of epithelial ovarian cancer over the 15 years of follow-up, whereas our previous report consisted of 97 cases over 7 years.4

Height and WHR were categorized in quartiles, and BMI was categorized by standard cut-off points (< 25 kg/m2, 25–30 kg/m2, and ≥ 30 kg/m2). We computed age-adjusted and multivariate-adjusted relative risks (RR) and their 95% confidence intervals (95% CIs) and performed statistical tests for linear trends with proportional hazards regression models using the SAS PHREG procedure (SAS Institute, Cary, NC). All reported P values are two sided. Covariates adjusted for in the multivariate analyses were those established previously as possible risk factors for ovarian cancer in this cohort: age (continuous), family history of ovarian cancer in a first-degree or second-degree relative, hysterectomy status, oophorectomy status, the number of live births (0, 1–2, 3–4, or ≥ 5), pack-years of smoking (0, 1–19, 20–39, or ≥ 40), and ERT ever use.4, 27 We tested the assumption of proportional hazards and found that it was not violated for any variables except for the number of live births, which exhibited a marginally significant interaction (P = 0.03) with follow-up in association with ovarian cancer incidence. Specifically, greater numbers of live births had an inverse association with ovarian cancer early in follow-up but not later in follow-up. Inclusion of this interaction term in multivariate models had little impact on other RRs, and therefore, it was excluded in the final analysis. Supplemental analyses, including an exclusion of the first 2 years of follow-up (because of concerns related to recent weight loss) and analyses with case groups consisting only of serous tumors or nonmucinous tumors yielded results that did not differ appreciably from the results reported previously. Exceptions are described below (see Results).

RESULTS

Over the 15 years of follow-up, we identified 223 incident epithelial ovarian cancers. Table 1 shows that there was no significant association of height with incidence of ovarian cancer (multivariate-adjusted RR: 1.00, 0.98, 1.08, and 1.12 across quartiles). When it was analyzed as a continuous variable, height (in cm) still was not associated with ovarian cancer (multivariate-adjusted RR, 1.00; 95% CI, 0.98–1.03). However, in a supplemental analysis, a positive association was observed between height and incident serous ovarian tumors (n = 105; multivariate-adjusted RR: 1.00, 1.17, 1.19, and 1.58 across quartiles). Elevated WHR at baseline was marginally associated with incident ovarian cancer (multivariate-adjusted RR: 1.00, 1.77, 1.17, and 1.59 across quartiles) after this extended follow-up. No association was found with BMI at baseline. Furthermore, in contrast with the report of Rodriguez et al.,5 we found no statistical evidence for multiplicative interaction between BMI and current ERT use (P = 0.90) with respect to incident ovarian cancer.

Table 1. Relative Risks and 95% Confidence Intervals for Ovarian Cancer in Relation to Baseline Anthropometric Variables: lowa Women's Health Study, 1986–2000
VariableNo. of casesPerson-yearsAge-adjusted RRMultivariate-adjusted RRa95% CIP valueb
  • RR: relative risk; 95% CI, 95% confidence interval.

  • a

    Adjusted for age, family history of ovarian cancer, hysterectomy status, oophorectomy status, number of live births, pack-years of smoking, and estrogen replacement therapy (ever vs. never).

  • b

    Test for trend.

Height (cm)      
 ≤ 15558110,5681.001.00Reference0.47
 156–16068129,3121.000.980.68–1.42
 161–1652953,2821.051.080.68–1.71
 > 16568120,3441.091.120.78–1.61
Body mass index (kg/m2)      
 < 2582165,2711.001.00Reference0.36
 25–3086153,4531.131.140.83–1.56
 ≥ 305594,7831.171.180.83–1.69
Waist-to-hip ratio      
 ≤ 0.7841105,0161.001.00Reference0.14
 0.79–0.8373103,7241.791.771.19–2.63
 0.84–0.8947102,5411.171.170.76–1.81
 > 0.8961100,4181.541.591.05–2.40

Data on ovarian cancer incidence according to BMI at ages 18, 30, 40, and 50 years and at maximum recorded age are reported in Table 2. High BMI (≥ 30 kg/m2) at age 18 years appeared to be associated positively with incident ovarian cancer in this cohort, although the relation was not statistically significant (multivariate-adjusted RR, 1.83 for BMI ≥ 30 kg/m2 vs. BMI < 25 kg/m2; 95% CI, 0.90–3.72). BMI at ages 30, 40, and 50 years and at maximum recorded age were not significantly associated with incidence of ovarian cancer.

Table 2. Relative Risks and 95% Confidence Intervals for Ovarian Cancer in Relation to Body Mass Index History: Iowa Women's Health Study, 1986–2000
BMI quartile (kg/m2)No. of casesPerson-yearsAge-adjusted RRMultivariate-adjusted RRa95% CIP valueb
  • RR: relative risk; 95% CI, 95% confidence interval.

  • a

    Adjusted for age, family history of ovarian cancer, hysterectomy status, oophorectomy status, number of live births, pack-years of smoking, and estrogen replacement therapy (ever vs. never).

  • b

    Test for trend.

At age 18 yrs      
 < 25195365,9341.001.00Reference0.10
 25–301936,7860.971.030.64–1.66
 ≥ 30888851.711.830.90–3.72
At age 30 yrs      
 < 25171313,8661.001.00Reference0.96
 25–303870,6380.991.050.73–1.50
 ≥ 301121,8530.931.020.55–1.88
At age 40 yrs      
 < 25135260,0401.001.00Reference0.81
 25–3070109,6801.231.260.93–1.70
 ≥ 301636,6860.850.940.56–1.58
At age 50 yrs      
 < 25109205,2591.001.00Reference0.90
 25–3083144,2961.081.070.80–1.45
 ≥ 303162,8390.941.030.68–1.54
At maximum weight      
 < 2552105,9181.001.00Reference0.34
 25–3091166,3061.111.100.77–1.56
 ≥ 3080140,2331.161.190.83–1.71

After adjustment for covariates, ovarian cancer incidence was 24% higher in women who reported at baseline that they had engaged in any physical activity versus those who engaged in no regular physical activity (Table 3). Furthermore, there appeared to be a dose-response relation going from low to moderate and from moderate to high levels of the physical activity index, with women in the high-activity group having a 1.42-fold increased risk of ovarian cancer. Women who reported engaging in vigorous activity more than four times per week had a greater-than-twofold increased risk of incident ovarian cancer.

Table 3. Relative Risks and 95% Confidence Intervals for Ovarian Cancer in Relation to Leisure-Time Physical Activity: Iowa Women's Health Study, 1986–2000
VariableNo. of casesPerson-yearsAge-adjusted RRMultivariate-adjusted RRa95% CIP valueb
  • RR: relative risk; 95% CI, 95% confidence interval.

  • a

    Adjusted for age, family history of ovarian cancer, hysterectomy status, oophorectomy status, number of live births, pack-years of smoking, and estrogen replacement therapy (ever vs. never).

  • b

    Test for trend.

Regular physical activity      
 No117237,0741.001.00Reference0.12
 Yes103170,2691.221.240.94–1.63
Physical activity index      
 Low87189,8801.001.00Reference0.03
 Moderate61112,2841.181.140.81–1.60
 High69102,9551.451.421.03–1.97
Frequency of moderate physical activity (times per wk)      
 Rarely/never4581,6991.001.00Reference0.26
 149116,3130.770.750.50–1.14
 2–469125,5880.990.980.66–1.44
 > 45785,5571.201.170.78–1.75
Frequency of vigorous physical activity (times per wk)      
 Rarely/never177335,7891.001.00Reference< 0.01
 11735,7630.910.840.50–1.43
 2–41425,6331.051.030.58–1.80
 > 41193942.232.381.29–4.38

DISCUSSION

In the current cohort of older women who were followed for 15 years, we examined the way in which ovarian cancer was related to anthropometric variables and physical activity at baseline. We found no evidence of an association between ovarian cancer incidence and height in the cohort as a whole, although the incidence of serous tumors appeared to be associated positively with height. In contrast with our overall findings, three large prospective studies with greater mean values and broader ranges of participant heights recently reported a positive association between self-reported height and total ovarian cancer. Rodriguez et al. found a statistically significant linear trend in the RR across height categories for ovarian cancer mortality in a large sample of American women who were followed for 16 years (multivariate-adjusted RR, 1.41 for ≥ 177 cm vs. 152–156 cm; 95% CI, 0.95–2.09).5 In the Netherlands Cohort Study on Diet and Cancer, Schouten et al. reported a significant, positive dose-response association between height and incident ovarian cancer over 7 years (multivariate-adjusted RR, 2.17 for > 175 cm vs. ≤ 160 cm; 95% CI, 1.14–4.13).6 A prospective study of 1.1 million Norwegian women who were followed for an average of 25 years reported a positive association between height and ovarian cancer (RR, 1.29 for ≥ 175 cm vs. 160–164 cm; 95% CI, 1.11–1.51; P for linear trend < 0.001).8 In contrast, Kuper et al. reported no association between height and ovarian cancer,28 as have other studies.9, 29, 30 The biologic plausibility of height as a risk factor for ovarian cancer has not been established. However, it has been postulated that height may be a biomarker for another exposure, such as increased childhood exposure to insulin growth factor I or other growth factors.5, 6

Some evidence was found in the current cohort to suggest an association between ovarian cancer incidence and reported BMI at age 18 years, although there were only 8 cases of incident ovarian cancer among those who reported a BMI ≥ 30 kg/m2 at age 18 years. In support of this finding, a positive association of BMI history (particularly in adolescence and young adulthood) with ovarian cancer has been reported in at least three recent studies. An Israeli case–control study found a positive relation between ovarian cancer and BMI at age 18 years (multivariate-adjusted odds ratio [OR] for the highest BMI quartile vs. the lowest BMI quartile, 1.54; 95% CI, 1.17–2.02).14 Fairfield et al. also reported that BMI at age 18 years was associated positively with premenopausal ovarian cancer (multivariate-adjusted RR, 2.05 for BMI ≥ 25 kg/m2 vs. BMI < 20 kg/m2; 95% CI, 1.07–3.93).16 Engeland et al. reported a positive association between BMI in adolescence (ages 14–19 years) and ovarian cancer in a prospective cohort of Norwegian women (RR, 1.56 for ≥ 85th percentile vs. 25th through 74th percentile; 95% CI, 1.04–2.32).8 Schouten et al. reported no appreciable association between incident ovarian cancer and BMI at age 20 years.6 We found no association between ovarian cancer and historic BMI at any age other than age 18 years in the current cohort.

Studies on the relation of current BMI with ovarian cancer have been inconclusive. Positive associations,5, 6, 9–14, 29, 31 negative associations,7 and a lack of associations,8, 15, 16, 28, 32–34 as in the current study, have been reported. A previous prospective study suggested that obesity is a risk factor for ovarian cancer mortality only in ERT users.5 We did not confirm this finding, although, admittedly, our power to test interactions was limited. WHR had not been used frequently as an index of obesity in previous ovarian cancer studies. An Italian case–control study reported a positive association between WHR and ovarian cancer (multivariate-adjusted OR, 1.45 for the highest category [≥ 0.89] vs. the lowest category [≤ 0.76]; 95% CI, 1.07–1.96).15 In a previous report on the current cohort after 7 years, a positive dose-response association between WHR and ovarian cancer was found (multivariate-adjusted RR, 2.33 for highest quartile [> 0.89] vs. lowest quartile [< 0.78]; 95% CI, 1.21–4.51).4 The association between high WHR and incident ovarian cancer has become weaker with 8 more years of follow-up, with less evidence of a dose-response association.

Consistent with previous findings from the current cohort,4 in the current study, it was found that physical activity—particularly frequent vigorous activity—was an independent risk factor for incident ovarian cancer. Although this finding may be attributable to chance, as there were relatively few cases in the highest exposure categories, it has held true (with some attenuation) after more than double the length of follow-up and more than twice the number of incident cases relative to the previous report. The only other prospective report of a relation between ovarian cancer and physical activity, from the Nurses' Health Study cohort, corroborated our findings. Bertone et al. found a greater risk of ovarian cancer among women who reported more frequent (rather than less frequent) vigorous activity (multivariate-adjusted RR, 1.84 for women with 20 to < 30 metabolic equivalent task-hours [MET-hours] per week vs. women with < 2.5 MET-hours per week; 95% CI, 1.12–3.02), although the association lacked a dose-response relation.20 Other studies have found negative associations17–19 and/or a lack of associations19, 21 between physical activity and ovarian cancer. The mechanism by which the intensity and frequency of physical activity may influence ovarian carcinogenesis is speculative. According to the ‘incessant ovulation’ hypothesis, intense physical activity should disrupt menstrual cycling and, thus, should be protective.35 Alternatively, frequent vigorous activity may promote tumorigenesis in the ovary by stimulating gonadotropin production.36 Furthermore, reporting more physical activity may be characteristic of health-conscious participants who are more likely to seek health care, increasing the likelihood of detection of ovarian neoplasms.4

Some limitations must be considered when interpreting the results of the current study. First, the study population consists largely of postmenopausal Caucasian women; thus, our results may not be generalizable to the population at large. Second, we observed only 223 incident cases of epithelial ovarian cancer in 15 years of follow-up; thus, we had limited power to detect weak associations and/or potential effect modification. Third, we excluded women who had undergone bilateral oophorectomy or who had cancer at baseline based on self-report; we could not verify reports with medical records; thus, overreporting or underreporting and subsequent misclassification may have occurred. Finally, physical activity variables were taken from responses at baseline, and actual intensity and amount of recreational physical activity among individuals may have fluctuated over the follow-up period.

In conclusion, we found no association between ovarian cancer and height or current BMI. However, height may be a risk factor for serous ovarian cancers. In addition, some evidence of a positive relation of ovarian cancer with increasing WHR persists in this cohort, and BMI at age 18 years appears to be associated positively with ovarian cancer. The previously reported association between leisure-time physical activity, especially frequent vigorous activity, and incident ovarian cancer holds after 15 years of follow-up in the Iowa Women's Health Study cohort. These findings suggest that current anthropometric variables are not major risk factors for ovarian cancer, but high BMI in young adulthood may be a risk factor. Frequent and vigorous recreational physical activity also may increase the risk for ovarian cancer among postmenopausal women.

Acknowledgements

The authors thank Ching-Ping Hong for providing technical assistance.

Ancillary