Body Size, Weight Cycling, and Risk of Renal Cell Carcinoma among Postmenopausal Women: The Women’s Health Initiative (United States)

Although obesity is an established risk factor for renal cell carcinoma, the possible effect of central adiposity and long-term variation in weight has yet to be established. The authors studied 140,057 women aged 50–79 years enrolled in the Women’s Health Initiative in the United States to examine the role of obesity, especially abdominal obesity, and weight cycling in relation to risk of renal cell carcinoma among postmenopausal women. Cox models were used to estimate relative risks and their corresponding 95% conﬁdence intervals. During an average of 7.7 years of follow-up through September 12, 2005, a total of 269 incident cases of renal cell carcinoma were identiﬁed. Central adiposity, as indicated by waist-to-hip ratio, was an important risk factor for developing renal cell carcinoma (highest vs. lowest quartile: relative risk ¼ 1.8, 95% conﬁdence interval: 1.2, 2.5; p for trend ¼ 0.0003). Moreover, women who had experienced weight cycling more than 10 times were at 2.6 times (95% conﬁdence interval: 1.6, 4.2) increased risk compared with women whose weight was stable. Results add evidence that obesity, particularly central adiposity, is associated with an increased risk of renal cell carcinoma among postmenopausal women. Furthermore, they indicate that weight cycling is independently associated with further increased risk of this

strongly associated with abdominal obesity and visceral fat deposition (5,6). However, to our knowledge, only one prospective study has examined the possible association between abdominal obesity and RCC (7,8), in which a positive association was shown.
Intentional weight loss is frequently followed by unintentional regain, thus leading to weight cycling. Indeed, weight cycling is the most common weight pattern in middle-aged women in the United States, with about 35 percent having at least one episode of intentional weight loss followed by regain of more than 10 pounds (4.5 kg) (9). Weight cycling has been associated with increased mortality in three large population-based studies (10)(11)(12). It has been suggested that one adverse effect of weight cycling could be the subsequent redistribution of body fat to the upper body (13)(14)(15), although this was not found in all studies (16,17). One casecontrol study in Sweden suggested that repeated weight cycling over time may increase the risk of RCC independently of obesity (18). However, this finding was not confirmed when international data from a series of coordinated case-control studies were pooled (19).
Most data on weight and weight change as risk factors for RCC have been from case-control studies. Thus, the possibility of bias due to differential misclassification of exposure (weight, weight fluctuations) between cases and controls or selection bias could not be ruled out in these studies. The Women's Health Initiative, a large prospective study, collected detailed information on body measurements, weight cycling, and potential confounders. We analyzed data from the Women's Health Initiative cohort to help understand the role of body size, body fat distribution (waist circumference, hip circumference, and waist-to-hip ratio (WHR)), and weight cycling in relation to risk of RCC among postmenopausal women.

The Women's Health Initiative
The Women's Health Initiative is an ongoing, ethnically and geographically diverse, multicenter clinical trial and observational study designed to address some of the major causes of morbidity and mortality in postmenopausal women (20 (36,282 women). Participants in the observational study were 93,676 women who were screened for the clinical trials but proved to be ineligible or unwilling to participate or were recruited through a direct invitation for screening into the observational study. All participants in the Women's Health Initiative gave informed consent and were followed prospectively. Details of the scientific rationale, eligibility requirements, and baseline characteristics of the participants in the Women's Health Initiative have been published elsewhere (20)(21)(22)(23)(24)(25).

Measurements of exposure
All exposures in our analyses were collected at baseline for all subjects. During the baseline clinic visit, trained and certified staff performed anthropometric measurements, including height, weight, hip and waist circumferences, and blood pressure. Weight was measured to the nearest 0.1 kg, and height was recorded to the nearest 0.1 cm. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. Waist circumference at the natural waist or narrowest part of the torso, and hip circumference at the maximal circumference, were measured to the nearest 0.1 cm. WHR was computed as the ratio of these two measurements.
Weight changes during participants' adult life were obtained by self-report questionnaire and were categorized as weight stayed stable (within 10 pounds), steady gain in weight, lost weight as an adult and kept it off, and weight has gone up and down again by more than 10 pounds. Those who answered that their weight had gone up and down again by more than 10 pounds were considered to have experienced weight cycling, and they were asked a separate question assessing how many times weight cycling occurred when the participant was not pregnant or sick. Choices were 1-3, 4-6, 7-10, 11-15, and more than 15 times. We combined the last two categories in our analysis because of a small proportion of women in these two groups.
Information on demographic characteristics, medical history, and personal habits (lifestyle) was obtained by interview or by self-report using standardized questionnaires that included age at enrollment (<55, 55-59, 60-64, 65-69, 70-74, 75), smoking status (never, past, current), alcohol intake (nondrinker, past drinker, <1 drink/month, 1 drink/ month-<1 drink/week, 1-<7 drinks/week, 7 drinks/ week), parity (0, 1, 2, 3, 4, 5), history of hypertension, history of diabetes while not pregnant, history of kidney or bladder stones, history of oral contraceptive use, and physical activity (metabolic-equivalent tasks per week were computed as the product of days per week, minutes per day, and the metabolic-equivalent task value for each activity). Dietary intake was obtained by using a validated food frequency questionnaire based on instruments previously used in large-scale dietary intervention trials (26,27).

Follow-up and ascertainment of cases
Women in the clinical trial were followed through regularly scheduled examinations to ensure timely ascertainment of updated medical histories. All women participating in the clinical trial were expected to attend annual clinic visits, with intermediate 6-month mail, phone, or clinic contacts. The observational study participants were contacted annually by mailed self-administered questionnaires in all years except year 3, when the questionnaires were filled out at the clinic visit to obtain updates of their medical histories and selected exposure data. The completion rate of annual questionnaires for the observational study was 93-96 percent. In this study, all subjects were followed up until September 12, 2005. Initial reports of cancer were ascertained by selfadministered questionnaires, and all cases of cancer were confirmed by review of medical records and pathology reports by physician adjudicators at the clinical centers. We focused our analyses on RCC and did not consider renal pelvis cancers because of their different etiology (28).

Study population
The following participants were excluded from the original cohort of 161,808: 14,849 women who had a history of cancer (except nonmelanoma skin cancer) at baseline; 668 women who had no follow-up time; and 6,234 women for whom values were missing for the main exposures and confounders (including weight, height, waist circumference, hip circumference, weight change, smoking, hypertension, and diabetes). Finally, 140,057 women remained for further analysis.

Statistical analyses
Cox proportional hazards regression models were used to estimate the association between exposure and risk of RCC, stratified by study cohort (participation in the observational study or clinical trials, and different treatment assignments for all three clinical trials). In the multivariable models, we initially considered many potential confounders listed in table 1. However, in our final models, we adjusted for only those covariates associated with RCC with a significance level of 0.2 after adjusting for age, which included age, smoking status, hypertension, oral contraceptive use, and total energy intake.
Anthropometric variables were treated as both continuous and categorical (in quartiles) in the regression models, except for BMI, which was categorized according to established clinical cutpoints: <25, 25-<30, 30-<35, and 35 kg/m 2 . Tests for trend were performed by creating a continuous variable from the medians of the categories. Interactions between exposures and cohort type (clinical trial/ observational study) and between different exposures were tested by entering into the model multiplicative interaction terms; no significant interactions were found. In addition, we conducted sensitivity analyses by excluding the first 2 years of follow-up to reduce the possibility that some cases of RCC might have been prevalent at baseline.
The assumption of proportional hazards was tested based on the cumulative sums of Martingale residuals with a Kolmogorov-type supremum test (29), where 1,000 realizations were used. We tested all exposure variables of interest (all p values were larger than 0.3) and all potential confounding variables (the lowest p value was 0.08 for fruit and vegetable intake), which indicates that the proportional assumption for each variable in our study was satisfied.

RESULTS
During an average of 7.7 years of follow-up, 993,029 person-years were accrued, among which a total of 269 incident cases (121 from the clinical trial and 148 from the observational study) of RCC were identified. Characteristics of selected variables at baseline by quartiles of WHR are shown in table 1. Compared with women with a lower WHR, those with a higher WHR tended to be older, non-White, and less educated. Women with a higher WHR were also more likely to be smokers, to have a higher total energy intake, to be less physically active, and to have a higher prevalence of diabetes, hypertension, and kidney or bladder stones, but they were less likely to be alcohol drinkers or to use oral contraceptives. WHR was positively associated with BMI, waist or hip circumference, weight gain, and weight cycling. In contrast, women with a lower WHR were more likely to keep a stable weight or to lose weight and keep it off.

Risk factors for RCC
The relative risk of RCC increased 17 percent (95 percent confidence interval (CI): 8 percent, 27 percent) progressively with increasing age in every 5-year increment. The age-adjusted risk of RCC was elevated significantly for women who were former smokers (relative risk (RR) ¼ 1.3, 95 percent CI: 1.0, 1.7), especially for current smokers (RR ¼ 1.7, 95 percent CI: 1.1, 2.6) compared with never smokers; in women with hypertension (RR ¼ 1.4, 95 percent CI: 1.1, 1.8); and in women in the highest quartile of total energy intake (RR ¼ 1.5, 95 percent CI: 1.1, 2.2) compared with women in the lowest quartile of total energy intake. Women of American Indian or Alaskan Native origin appeared to have a higher risk (RR ¼ 2.8, 95 percent CI: 0.9, 8.8) compared with White, non-Hispanic women, but this finding was based on only three cases of RCC in this ethnic group. Other variables, including education, physical activity, fruit and vegetable intake, kidney or bladder stones, diabetes, parity, and alcohol intake, did not show a significant association with risk of RCC (data not shown). Women who were oral contraceptive users appeared to have a slightly lower risk (RR ¼ 0.8, 95 percent CI: 0.6, 1.1), which did not reach conventional values for significance. This variable was included in multivariable models because the p value was <0.2.

Anthropometric measures
All anthropometric variables but hip circumference were positively and significantly associated with age-adjusted risk of RCC (table 2). After further adjustment for potential confounders, the associations were somewhat attenuated but were still statistically significant. In the multivariate model, women whose BMI was 35 kg/m 2 had a 60 percent excess risk compared with women of normal weight, and the risk increased 3 percent (95 percent CI: 1 percent, 5 percent) per unit increase in BMI. For women in the highest versus lowest quartile of WHR, the relative risk was 1.8 (95 percent CI: 1.2, 2.5). When WHR was analyzed as a continuous variable, we observed a 24 percent (95 percent CI: 14 percent, 35 percent) increase in risk per 0.1-unit increase in WHR.
To distinguish the effect of BMI from central adiposity, we included BMI and WHR in the same model. WHR remained as an independent predictor (highest vs. lowest quartile: RR ¼ 1.6, 95 percent CI: 1.1, 2.3), whereas the effect of BMI became weaker and nonsignificant (data not shown).

Weight cycling
Compared with stable weight, neither steady gain in weight nor weight loss was significantly associated with risk of RCC. However, following multivariate adjustment, weight gain and loss of more than 10 pounds were significantly associated with an increased risk of RCC, although further adjustment for covariates including WHR weakened the association. Excess risks increased with number of times that weight cycled 10 pounds or more, and women who had changed weight more than 10 times were at 2.6 times increased risk compared with women whose weight remained stable (table 3). The results were similar when BMI was substituted for WHR in the multivariable model (data not shown).
Results remained similar when we excluded the first 2 years of follow-up. For example, in the multivariate model, women whose BMI was 35 kg/m 2 had a relative risk of 1.7 (95 percent CI: 1.1, 2.8) compared with that for women of normal weight (BMI <25 kg/m 2 ); for women in the highest versus lowest quartile of WHR, the relative risk was 1.8 (95 percent CI: 1.2, 2.7); and women who had experienced weight cycling more than 10 times were at 2.7 times (95 percent CI: 1.6, 4.6) increased risk compared with women whose weight was stable. In addition, the results from the observational cohort were consistent with those for the entire cohort (data not shown).

DISCUSSION
In this large prospective study with an average of 7.7 years of follow-up, we observed that obesity, particularly central adiposity as indicated by a high WHR, was an important risk factor for developing RCC in postmenopausal women. Moreover, weight cycling, measured as number of weight gains and losses of more than 10 pounds, was an independent predictor of risk of RCC, with a significant trend with increasing number of substantial weight changes.
Our observation of increased risk of RCC with high BMI is in agreement with most previous studies (30); increased risk with high WHR is also consistent with one previous study (7,8). However, the underlying mechanisms for the link of general and abdominal obesity with risk of RCC remain elusive. Several plausible biologic mechanisms have been suggested. Increasing BMI and a high WHR are accompanied by elevated levels of fasting serum insulin and free insulin-like growth factor-I in both men and women (31). Insulin and free insulin-like growth factor-I interact with and regulate the synthesis and bioavailability of sex steroids (32,33) that may affect renal cell proliferation and growth (34). In our study, we observed that abdominal adiposity contributed to the risk of RCC among postmenopausal women beyond that attributable to the aspects of obesity captured by BMI, which suggests that central obesity may play an important role in the etiology of RCC among postmenopausal women. Previous studies have shown that general obesity is more related to sex steroid hormone metabolism, whereas central adiposity is more related to insulin-glucose and growth factor metabolism (35,36). Since upper body obesity strengthens risk of the metabolic syndrome, our results raise the possibility that weight may be associated with RCC through mechanisms related to the metabolic syndrome. The metabolic syndrome entails a constellation of features including hypertension, dyslipidemia, decreased insulin sensitivity, and increased levels of biologically active insulin-like growth factor (36). General and abdominal obesity might also be associated with RCC via hypertension or diabetes, which have been suggested as risk factors for RCC (37,38). However, in our study, the risk of RCC associated with abdominal obesity appears to be independent of hypertension and diabetes, suggesting that weight may also influence RCC through different mechanisms. Obese individuals have been reported to have a higher glomerular filtration rate and renal plasma flow independent of hypertension (39), which may increase the risk of kidney damage.
We observed that weight cycling is an independent predictor of risk of RCC. To our knowledge, the association with weight cycling has not been previously investigated based on prospective data, and two case-control studies have provided equivocal results (18,19). Weight cycling has been hypothesized to have deleterious metabolic, behavioral, and health consequences. Some evidence suggests that weight fluctuation has independent adverse effects on cardiovascular and all-cause mortality (10,12,40). However, the mechanism(s) by which weight cycling might influence renal carcinogenesis is not clear.
One possibility for how weight cycling might influence renal carcinogenesis may be via mechanisms similar to those for abdominal obesity. Some early studies (41) suggested that weight cycling could have an adverse impact on body composition and body fat distribution, which, through its effect on insulin sensitivity, could also increase the risk of developing type 2 diabetes (42, 43), a risk factor for development of RCC (38). Some epidemiologic studies (14, 15), but not others (16,44), observed that weight cycling was associated with redistribution of body fat to the upper body compartments, which strengthens the risk of the metabolic syndrome (45)(46)(47). Epidemiologic studies have also shown that weight cycling was associated with increased incidence of hypertension (45,46,48), another risk factor for RCC. Another possibility is that frequent weight cycling may result in kidney damage independent of hypertension, by inducing metabolic or functional changes within the renal tubules that increase susceptibility to carcinogens or promoting agents.
The most important strength of our study is the prospective design, in which information about exposures was collected before RCC was diagnosed. Other strengths include the large size of the cohort; the reasonably large number of cases; the high prevalence of obesity, central adiposity, and weight cycling; and detailed information on potential confounders. One limitation is that many exposures, particularly the number of weight cycles, were based on self-report. There is a possibility of misclassification of the number of times that a person has experienced weight cycling, but, since the data were collected before RCC was diagnosed, any misclassification should be nondifferential, which would bias our results toward the null. Another limitation is that central adjudication of RCC cases was not performed; thus, there may also be some misclassification among the cases. However, this misclassification is likely to be nondifferential with respect to obesity and weight cycling, which would again make our results more conservative.
In conclusion, our study adds further evidence that obesity, particularly central adiposity, is associated with an increased risk of RCC among postmenopausal women. Furthermore, we found strong evidence that repeated weight change throughout adult life, namely, weight cycling, is associated with further increased risk of this malignancy.