Pelvic Floor Disorders, Diabetes, and Obesity in Women: Findings from the Kaiser Permanente Continence Associated Risk Epidemiology Study

Kaiser Permanente is a large, prepaid, managed health care plan that serves >3 million residents in southern California. The Epidemiology of Prolapse and Incontinence Questionnaire (EPIQ) was developed to assess the prevalence of PFDs in a sample of women from this racially and ethnically diverse population. Survey development, pilot testing, and survey methods have been described elsewhere (29–31). Briefly, the EPIQ was developed and validated in English and Spanish to assess the presence or absence of AI, OAB, SUI, and pelvic organ prolapse in a community-dwelling population.

After approval by the institutional review board, samples of 3,050 women in each of four age strata (25–39, 40–54, 55–69, and 70–84 years) were selected from the Kaiser Permanente Southern California (KPSC) membership who had an address on file with the health plan. Surveys in English and Spanish were mailed with a cover letter, small incentive, and postcard to opt-out or request additional information, followed by a second survey mailing, a reminder telephone call, and a third survey mailing to women in the youngest age strata (31). Of 12,200 surveys mailed, 4,458 (37%) were returned. Data were collected from April 2004 through January 2005

Women were screened for PFDs based on their responses to stem questions plus their degrees of bother, as indicated on a visual analog scale. Positive and negative predictive values and 95% CIs for the detection of specific PFDs were 88% (75–95) and 87% (76–93) for SUI, 77% (59–88) and 90% (81–95) for OAB, and 61% (48–73) and 91% (80–96) for AI, respectively. AI included flatal, solid, and/or liquid incontinence (30).

We linked survey respondents to the KPSC Diabetes Case Identification Database, which uses an algorithm to identify members who have a high probability of having diabetes (32,33) based on at least one of the following criteria: 250.XX ICD-9 hospital diagnosis, a prescription for insulin or other oral hypoglycemic agents, A1C ≥6.7%, or a fructosamine test result ≥280 μmol/l. Women with gestational diabetes (ICD-9 code 648.8) and no other criteria were not included. For this analysis, women who were identified as having diabetes based only on the A1C threshold had to meet or exceed 7.0% to increase the sensitivity of the algorithm. We assumed that the majority of these women have type 2 diabetes, as it comprises 85–95% of all adults with diabetes (34).

To characterize the women with diabetes, information about current treatment (insulin and/or oral hypoglycemic agents), based on the most recent prescription(s) filled before survey completion and the results of the A1C measured closest to the time of survey completion (±6 months), were extracted from the pharmacy and laboratory databases, respectively. All laboratory tests were conducted at a single laboratory operated by KPSC.

Age was calculated in completed years on the date of survey completion, and BMI was calculated as weight in kilograms divided by the square of height in meters and dichotomized into nonobese (<30 kg/m²) or obese (≥30 kg/m²). Smoking was categorized as never smoked, past smoker, or current smoker. Chronic lifting was defined as repetitive lifting of >9 kg regularly for >1 year. Caffeine consumption was defined as more than one cup of caffeinated beverage per day. Presence or absence of neurological disease, lung disease or asthma, history of depression, hysterectomy, menopause status (yes/no/don’t know), and hormone exposure (never/past/present) were assessed using survey data.

To adjust for the known associations between pregnancy, mode of delivery, and PFD as previously described (31), we defined the nulliparous group as those women who had never been pregnant or only delivered a baby ≤2 kg. The cesarean birth group was defined as having been delivered by one or more cesarean births and no vaginal births exceeding 2 kg. Vaginally parous women were defined as having one or more vaginal deliveries exceeding 2 kg birth weight regardless of history of cesarean births. Parity was modeled as a continuous variable.

Of 4,458 EPIQ surveys returned, we excluded women sequentially for the following reasons: insufficient data to categorize into one of the three birth groups (n = 289), insufficient information to assess at least one of the PFDs (n = 66), and insufficient information to calculate BMI (n = 141), for a final sample size of 3,962 subjects.

Statistical analyses were performed with SAS version 8.02 (SAS Institute, Cary, NC). Power and sample size calculations were based on the primary study objectives to assess the prevalence of each PFD and to identify the risk of vaginal delivery compared with cesarean births (31). We assessed the differences between groups of women using χ² tests for categorical variables and Student’s t tests for continuous variables.

Each PFD (SUI, OAB, or AI) was expressed dichotomously as “present” or “absent.” Women for whom we did not have information to assess presence or absence were excluded from the models for that outcome. Among women with information to assess the presence or absence of at least one of these PFDs, we created a summary variable labeled “any PFD.” Significance was evaluated using a two-sided P value of <0.05. Logistic regression analysis was used to calculate the odds ratios and 95% CIs for the associations between diabetes and obesity and each and any PFD.

Multiple logistic regression models were constructed for all women in the study sample. We assessed the contributions of diabetes and obesity to the likelihood of having each and any PFD after controlling for other known risk factors. Women were categorized hierarchically as nonobese/nondiabetic (reference), nonobese/diabetic, obese/nondiabetic, and obese/diabetic.

Once all of the variables were entered into the model, we removed covariates that were no longer significant in the multivariate model and had no impact on the primary variable of interest except for age (modeled as a continuous variable), race/ethnicity, mode of delivery, and parity, which remained in every model.

The median age of the women studied was 56.6 years, and the racial/ethnic distribution was 62% white, 19% Hispanic, 10% black, 8% Asian/Pacific Islanders, and 1% other or unknown race (Table 1). Ten percent (n = 393) of the women in the sample had diabetes. Compared with women without diabetes, we found that women with diabetes were significantly more likely to be older, African American or Hispanic, obese, parous, postmenopausal, and to have had a hysterectomy, a history of depression, a neurological condition, or lung disease. The prevalence of the PFDs was 15% SUI, 13% OAB, and 25% AI, and 35% had any PFD (Table 2).

Women with diabetes were significantly more likely to have each or any PFD than women without diabetes (Table 2). Of the women with diabetes, over half (56%) were obese; 17% were on insulin, 63% were treated with oral hypoglycemic agents only, and 20% were not on any diabetes medications. Over two-thirds (n = 271) had an A1C test in the 6 months before or after their survey completion, with a mean value of 7.0%. Of these women, 24% were in borderline control (7.0–8.5%) and 12% were in poor control (>8.5%). Women with diabetes were 90% more likely to have SUI or OAB, 50% were more likely to have AI, and 68% were more likely to have any PFD than women without diabetes (Table 3).

Obese women were over twice as likely to experience SUI and OAB, >40% were more likely to have AI, and 92% more likely to have any PFD than women who were not obese (Table 3). When we restricted our analysis to women with diabetes, as shown at the bottom of Table 3, we found that being obese was positively associated with all conditions, but the relationship with AI was not significant.

When we examined the associations between other common risk factors for PFDs (shown in Table 1) and each and any PFD, we found that age, race/ethnicity, smoking status, mode of delivery, parity, hormone use, menopause, previous hysterectomy, history of depression, neurological disease, lung disease, and caffeine consumption were significantly associated with each and any PFD, with the following exception: caffeine consumption was not associated with OAB (data not shown).

When diabetes and obesity were combined hierarchically into a four-category exposure variable (nonobese/nondiabetic [reference], nonobese/diabetic, obese/nondiabetic, and obese/diabetic), we found that the unadjusted odds of having SUI, OAB, AI, or any PFD progressively increased with each category (Table 3). There was no statistical interaction between having diabetes and being obese for any of the four outcomes (data not shown).

After controlling for age, race/ethnicity, mode of delivery, and other known risk factors for PFDs that were significant in the bivariate analysis, we found that women categorized as obese/diabetic had the highest probability of having SUI, AI, and any PFD, whereas women who were obese/nondiabetic were as likely as obese/diabetic women to have OAB (Table 4). Women categorized as nonobese/diabetic did not differ significantly in their prevalence of OAB, AI, or any condition than nonobese/nondiabetic women (reference), whereas nonobese/diabetic women were significantly more likely to have SUI than nonobese/nondiabetic women.

In our sample of community-dwelling women, we found that being obese, regardless of having diabetes, increased the likelihood of having a PFD compared with nonobese women. The prevalence of SUI, AI, and any PFD increased in the following manner: nonobese/nondiabetic (lowest), nonobese/diabetic, obese/nondiabetic, and obese/diabetic (highest), while women who were obese, regardless of whether they had diabetes, were most likely to have OAB.

Our approach to these analyses differed from others, as we directly examined the associations between PDFs and diabetes with or without obesity using women with neither condition as the reference group instead of examining the association between one of these conditions while controlling for the other (9–11). We were able to examine these associations across three different conditions, whereas many reports (7,9,10,12,28) limit their analysis to one condition, and unlike some studies (7,12), we were able to include premenopausal women in our cohort. As with most other studies, we found an association between PFDs and both diabetes and obesity.

While studies (14,15) of the relationship between AI and diabetes have had conflicting results, we found that AI was associated with having diabetes among obese women only, whereas the relationship between AI and diabetes in women who were not obese was not statistically significant.

The strength of this study includes using a carefully validated instrument to assess a spectrum of PFDs in a large, racially and ethnically diverse population distributed across a wide age range including obese and nonobese women. In addition, we were able to characterize the women in our sample with diabetes by linking clinical information about glycemic control and diabetes treatment regimen to the survey responses closest to the time of the survey.

Our response rate was lower than anticipated despite considerable effort to increase it, particularly among younger health plan members. We found that younger members were hardest to reach; the likelihood of not having a valid address on file decreased with age, from 11% of 25- to 39-year-old subjects to 3% of 70- to 84-year-old subjects. When we compared women in the final analytic sample (n = 3,962) with all other women originally surveyed (n = 8,238), 10% of the women in the sample and 11% of the remaining women had diabetes (P < 0.05). Among women with diabetes, there was no difference in mean A1C percent (P = 0.76) nor a difference in the racial/ethnic distribution (P = 0.26) when women in the analytic sample were compared with all others originally surveyed. Data on the prevalence of obesity were not available for comparison. Given that our overall prevalence of obesity and diabetes was consistent with what we would have anticipated given national estimates, we do not believe that our response rate biased the result of this study.

As this was a secondary analysis of data gathered primarily to evaluate the associations between pregnancy, mode of delivery, and PFDs (31), we did not have enough power to assess the relationship between glycemic control, diabetes treatment, and PFDs. Finally, we could only examine associations between prevalent PFDs and obesity and diabetes without information on the temporal sequence the onset of these conditions, since this was a cross-sectional study.

The findings from this study suggest that being obese may be a modifiable risk factor for PFDs. Women who are obese, regardless of whether they have diabetes, are more likely to have SUI, OAB, and AI, whereas nonobese/diabetic women had similar odds of each and any PFD as nonobese/nondiabetic women. Other published studies have suggested that weight loss may reduce the prevalence of incontinence among this group of high-risk women. Given the aging of the population, the increased prevalence of obesity, and the concurrent increase in the prevalence of diabetes in the U.S., women and health care professionals should be made aware of the associations between PFDs and obesity and diabetes. Women who are obese, regardless of whether they have diabetes, should be advised that they may be more likely to develop a PFD associated with their weight and should be encouraged to adopt patterns of physical activity and dietary intake to promote healthy weight loss and maintenance of a healthy weight.

This study was funded by R01 HD41113. Analyses were funded by Kaiser Permanente Direct Community Benefit funds.

Parts of this article were presented in abstract form at the 66th annual meeting of the American Diabetes Association, Washington, DC, 9–13 June 2006.

The authors acknowledge the contribution of Richard Contreras, Stephen Derose, and Vicki Chiu.