Anthropometric risk factors for ovarian cancer in the NIH-AARP Diet and Health Study

Objective Identifying potentially modifiable risk factors for ovarian cancer is essential for prevention because this cancer is predominantly detected at a late stage. Here, we estimated the relations of general adiposity and measures reflecting body fat distribution to the risk of epithelial ovarian cancer. Methods We ascertained 683 ovarian epithelial cancers (343 high-grade serous, 141 non-high grade serous) among 145,575 women, aged 50–72 years (median follow-up 12.6 years), from the National Institutes of Health—American Association of Retired Persons (NIH-AARP) Diet and Health Study. Using Cox models, we estimated confounder-adjusted hazard ratios (HRs) and 95% confidence intervals (CI) for associations of overall ovarian cancer, high-grade serous and non-high-grade serous carcinoma with body mass index, waist circumference, hip circumference, waist–hip ratio, waist–height ratio, body adiposity index, body shape index, and abdominal volume index. Results Anthropometric measures were unrelated to overall ovarian cancer, high-grade serous cancer, and non-high-grade serous cancer. For example, the HR for overall ovarian cancer per standard deviation increment of body mass index at baseline was 0.98 (95% CI 0.88–1.10). Similar associations were observed with measurements of body fat distribution. Conclusion These results do not indicate that adult adiposity is associated with ovarian cancer risk in post-menopausal women. Supplementary Information The online version contains supplementary material available at 10.1007/s10552-020-01377-y.


Introduction
Ovarian cancer is the fifth most common cause of cancer death in women in North America and western Europe [1]. Meta-analyses and pooled observational studies suggest that obesity may be positively related to ovarian cancer risk [2][3][4][5]. The World Cancer Research Fund/American Institute for Cancer Research concluded that the evidence for the link between obesity and increased ovarian cancer risk is probable [6], and an umbrella review graded the evidence as suggestive [3]. A growing body of research further suggests that body composition plays an important role in site-specific cancer development [7][8][9]. Previous studies indicate that anthropometric measures of general obesity and body fat distribution may be differentially related to ovarian cancer risk; however, results have been conflicting [4,5,[10][11][12]. We conducted a cohort study among post-menopausal women using data from the National Institutes of Health-American Association of Retired Persons (NIH-AARP) Diet and Health Study to provide further insights into the association between body fatness and subtype-specific epithelial ovarian cancer risk by comparing indicators of general obesity and body fat distribution.

Study population
The NIH-AARP Diet and Health Study is a prospective cohort study of persons in the U.S. [13]. At baseline (1995)(1996), 3.5 million AARP members aged 50-71 years who resided in six states (California, Florida, Louisiana, New Jersey, North Carolina, and Pennsylvania) and two metropolitan areas (Atlanta, Georgia; and Detroit, Michigan) were invited to complete a questionnaire on demographic, diet, and lifestyle characteristics. Questionnaires were completed satisfactorily by 566,398 participants. Six months after completing the baseline questionnaire, a second questionnaire was mailed to living participants who did not have a self-reported colon, breast, or prostate cancer at baseline to collect additional information. Self-reported weight and height were collected on the baseline questionnaire. Self-reported waist circumference and hip circumference were assessed on the second questionnaire. We excluded male participants (n = 339,666); those with unknown cancer or previous diagnosis of cancer other than non-melanoma skin cancer at baseline before completion of the second questionnaire (n = 16,300); those with bilateral oophorectomy and unknown oophorectomy status (n = 57,047); those with no information on height, weight, waist circumference, or hip circumference (n = 95,764); and subjects with body mass index less than 18.5 kg/m 2 or more than 65 kg/m 2 (n = 3,528). Our final analytical datasets included 145,575 women for the analysis on body mass index, 60,999 for waist circumference, 60,826 for hip circumference, 60,597 for waist-to-hip ratio, 60,999 for waist-to-height ratio, 60,826 for body adiposity index, 60,999 for body shape index, and 60,597 for abdominal adiposity index. The Special Studies Institutional Review Board of the U.S. National Cancer Institute approved the study [13]. All participants gave informed consent by virtue of completing and returning the baseline questionnaire.

Assessment of anthropometric measures
Self-reported height and weight were obtained from the baseline questionnaire. Body mass index was calculated as weight (kg) divided by the square of height (in meters). In the second questionnaire, participants were instructed to measure their waist circumference and hip circumference using a tape measure to the nearest 0.25 inch while standing. Waist circumference was to be measured 1 inch above the navel if this was not the waistline. Hip circumference was defined as the largest circumference between the upper edge of the pelvis and the femur. Waist-hip ratio was calculated by dividing waist circumference (cm) by hip circumference (cm), and waist-height ratio was calculated as waist circumference (cm)/height (cm). The body adiposity index was calculated as hip circumference (cm)∕height 1.5 (m) − 18 [ 14]. The body shape index was based on waist circumference (cm)∕body mass index 2 3 × height 2 (m) [15]. The abdominal volume index was also quantified [16]:

Statistical analysis
Cox proportional hazards regression was used for estimating adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for overall ovarian cancer and ovarian cancer subtypes. Time of study entry was age at baseline (or second questionnaire) and exit time was age at cancer diagnosis or the last date at which follow-up was considered complete. Athropometric measures were modeled as a continuous and categorical metric. After confirming that the linearity assumption was met by testing cubic spline transformations [25], HRs were estimated per standard deviation (1-SD) increase in anthropometric measures. Body mass index (kg/m 2 ) was further grouped as normal weight (18.5 to < 25), overweight (25 to < 30), and obese (≥ 30). Waist circumference and other anthropometric measures were categorized according to quartiles. Models were stratified by 5-year age groups to minimize departure from proportionality and adjusted for education, race, smoking, alcohol consumption (modeled continuously using restricted cubic splines [25]), parity, age at menarche, family history of ovarian cancer, oral contraceptive use, and menopausal hormone therapy.
The improvement in predictive accuracy after adding anthropometric measures to a null model (including race, age, education, smoking, alcohol consumption, parity, family history of ovarian cancer, oral contraceptive use and menopausal hormone therapy) was evaluated in terms of explained variation (R 2 ) [26], the Bayes Information Criterion (BIC), and model discrimination using Harrell's C-index [25] derived from flexible parametric models [27]. p values for the difference between Harrell's C indices of models with and without anthropometric indicators were computed using the method proposed by Antolini et al. [28]. We used 1,000 bootstrap replications to perform internal validations and to correct R 2 , BIC, and Harrell's C-indices for optimism [25].
In a sensitivity analysis, we used regression calibration for self-reported body mass index, waist circumference, and hip circumference to assess possible regression dilution bias [29]. Because replicate measurements were not available, we applied published reliability coefficients [30][31][32][33], ranging from 0.5 to 0.9. A further threat to the validity of our estimates is potential unobserved confounding by undiagnosed ovarian cancer (often referred to as "reverse causation" [34]) if these conditions are symptomatic enough to induce a change in body weight. We, therefore, assumed 3 years of minimum latent period required for weight change due to unobserved disease to affect the outcome and excluded events that occurred during this time [35]. The statistical analysis was performed using Stata 15.1.

Results
In the analytical sample of 145,575 women, the mean (SD) age at baseline was 61.8 (5.4) years. During a median followup time of 12.6 years, participants contributed 1,897,323 person-years and 683 ovarian cancer (343 high-grade serous) cases occurred. The baseline characteristics of the analytical sample are provided in Table 1.
Sensitivity analysis indicated that HRs could have been attenuated towards the null because of measurement error in self-reported anthropometric measurements. For example, the unadjusted HR for per SD of waist circumference and overall ovarian cancer was 1.08 (95% CI 0.91-1.27), but after accounting for potential regression dilution bias, assuming an attenuation factor of 0.7, the HR was 1.11 (95% CI 0.91-1.30) (Supplementary Table 1). The associations were virtually unchanged when events occurring during the first three years of follow-up were excluded (Supplementary Table 2).

Discussion
This study examined the association of indicators of general obesity and body fat distribution with ovarian cancer risk using a large U.S. prospective cohort study of post-menopausal women. We found no association of anthropometric measures with the risk of overall ovarian cancer, high-grade serous, or non-high-grade ovarian cancers. Overall, our analysis does not support the hypothesis that central adiposity or measures of body fat distribution improve the prediction of ovarian cancer risk in post-menopausal women. A larger body of research examined the association between obesity and ovarian cancer risk. However, the findings of more than 30 epidemiologic studies have been weak and mixed [36]. Several meta-analyses and pooled analyses reported weak positive associations between adult body mass index and ovarian cancer risk, noting substantial between-study heterogeneity with weaker associations in prospective than case-control studies [2,6,37]. A systematic review from the World Cancer Research Fund and the American Institute for Cancer Research included 28 prospective studies on ovarian cancer and calculated a summary relative risk for a 5-unit increment in body mass index of 1.07 (95% CI 1.03-1.11) [2,6]. The effect size was similar in the post-menopausal group but it was less precise (relative risk per 5 kg m −2 = 1.07; 95% CI 1.00-1.14). Results from the 2013 Ovarian Cancer Association Consortium [4] pooled analysis of case-control studies found that the positive association with body mass index was stronger in premenopausal women. The heterogeneity of findings reported here and previously could be explained by menopausal status and higher statistical efficacy of meta-analysis. Indeed, our HRs for body mass index and waist circumference overlap with the HR's CIs of the World Cancer Research Fund/ American Institute for Cancer Research meta-analysis [2,6], which indicates that there might be chance variation of estimates. Few studies have examined how different measures of body fat distribution are related to ovarian cancer and its subtypes [10][11][12]. Existing cohort studies found no association of waist circumference and waist-hip ratio and ovarian cancer risk [2,6,10]. Similar to previous studies [4,5], the present study found no notable differences between histotypes. In contrast, a Mendelian randomization study suggested that obesity might cause non-high-grade but not high-grade ovarian cancer [20].
The present study has several limitations. It relied on self-reported anthropometric data and potential measurement error could have attenuated the observed associations. Anthropometric measurements were taken when most study participants had reached menopause. However, pre-menopausal anthropometric risk factors might be more strongly related to ovarian cancer risk. A Mendelian randomization study suggested a positive association with higher body mass index with risk of ovarian cancer among pre-menopausal women, but not for post-menopausal women [38]. Unfortunately, we could not test whether the association between anthropometric markers and ovarian cancer risk was modified by menopause status. More studies on body  weight and body fatness changes over the lifecourse are warranted [39,40]. Our study also lacked updated information on anthropometric measurements during follow-up.
Another drawback is the low number of cases by ovarian cancer subtype and a lack of statistical power to test for effect modification.   Code availability Code is available upon request from the corresponding author.

Conflict of interest No potential conflicts of interest were disclosed.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.