Urinary Incontinence and Diabetes in Postmenopausal Women

This cross-sectional analysis used baseline data from a prospective evaluation of UTI risk among postmenopausal women, conducted within a group model health maintenance organization with ∼450,000 members (Group Health Cooperative [GHC]). Women aged 55–75 years were eligible to participate if they had no natural menstrual cycle in the preceding 12 months, resided in the Pierce, King, or Snohomish counties of Washington State, and had been enrolled in GHC for at least 1 year. Members of GHC are similar in age and ethnicity to women in the surrounding Seattle-Tacoma-Bremerton region, but have slightly higher educational attainment and less representation in the highest extremes of income distribution. Compared with the U.S. population, there are fewer blacks, higher educational levels, and less representation in the lowest extreme of income distribution (18).

Women were randomly selected using GHC enrollment files. The GHC diabetes registry was also used to enrich the population with diabetic women, who were randomly selected and frequency matched according to age with the main study group. Exclusion criteria for all potential participants included residential nursing care, restriction to a wheelchair, dementia or a severe psychiatric disorder (as determined by the recruiter or reported by a participant or household member), indwelling or intermittent urinary catheterization, end-stage renal disease requiring dialysis, active malignancy other than skin cancer, acute cystitis in the preceding 90 days, or chronic antibiotic use. The Human Subjects Committees of GHC and the University of Washington approved all procedures.

Data from a detailed interview and research clinic visit at baseline, in conjunction with computerized laboratory, hospital, and pharmacy records (for identification and verification of diabetes), were used to identify and measure potential risk factors.

The baseline interview assessed general medical and surgical health information, estrogen exposure (including oral estrogen with or without progesterone and vaginal estrogen cream), history of UTI, and diabetes status. Women who reported having diabetes were asked to complete a supplemental questionnaire. BMI was obtained from an HMO mammography surveillance database, based upon self-reported height and weight, using the measurement most temporally proximate to study enrollment. During the baseline research clinic visit, women submitted clean-catch midstream urine specimens that were cultured for aerobic bacteria. Vaginal introital swabs were also obtained and cultured for aerobic gram-negative rods using previously described methods (19). Lastly, postvoid residual bladder volume was measured using a portable ultrasound device (BladderScan BVI 2500+), a method that is highly correlated with catheterized volumes (20).

At baseline, diabetes presence was ascertained by self-report of physician diagnosis or inclusion in the GHC diabetes registry. This registry captures information monthly from laboratory, pharmacy, and hospital discharge summary databases, continuously updating the diabetes status of enrollees as described elsewhere (21). The presence of diabetes in women who were identified for the study from the GHC diabetes registry was confirmed by record review.

Blood was drawn for fasting glucose determination at the research clinic visit to identify previously undiagnosed diabetes. If the result was >125 mg/dl, the woman was considered to have diabetes. All women who met any criterion for diabetes underwent high-performance liquid chromatography testing of HbA_1c (Variant II Hemoglobin A1c; Bio-Rad, Hercules, CA). Women who reported age of diabetes onset at <30 years, one or more episodes of ketoacidosis, and continuous insulin use since diagnosis were considered to have type 1 diabetes. Retinal disease and neuropathy were assessed by self-report.

Because of the variety in incontinence definitions in the literature, we approximated a validated and internationally recommended measure of incontinence severity, the Sandvik Index (9,22), to define a primary outcome (severe incontinence), as well as a less specific assessment (any incontinence in the past month) The Sandvik Index incorporates both amount and frequency of incontinence. It is based on three levels of incontinence amount (drops, small splashes, more), which may be dichotomized into two levels: 1 = drops (from our survey, few drops/damp underwear) or 2 = more (moderately wet underwear/completely wet underwear/outer clothes wet or leakage to the floor). The amount (1 or 2) is multiplied by four levels of frequency (1 = less than once a month, 2 = few times a month, 3 = few times a week, 4 = every day and/or night) to yield a score of 1–2 (slight), 3–4 (moderate), or 6–8 (severe) (22). We allocated our measures of incontinence amount and frequency during the past month to those defined by the Sandvik Index to determine the number of women with severe incontinence.

Stress incontinence was considered to be loss of “control of your urine when you laugh, cough, or during physical activities,” whereas urge incontinence was defined as loss of “control of your urine because you feel the urge to urinate but cannot reach the bathroom in time.” Affirmative responses to both questions were considered mixed incontinence. To assess a woman’s subjective perception of whether her incontinence is a problem, we used an unvalidated measure: “How would you rate your difficulty controlling urination?” Response options were no problem or mild, moderate, or severe problem.

First, we characterized the study group using bivariate analyses (χ² tests and Fisher’s exact tests when appropriate) describing the relationship between potential risk factors for incontinence and outcome measures. Next, we further characterized incontinence among women with and without diabetes.

To evaluate the relationship between aspects of diabetes and incontinence, we constructed individual multiple logistic regression models for each diabetes characteristic, adjusting for age and factors that were associated with both diabetes at baseline and severe incontinence. These included education and history of UTI in the past year. A high proportion of missing income data precluded adjustment for this variable, which was associated with both diabetes and incontinence. To demonstrate confounding of diabetes by BMI, models were constructed with and without this covariate. Similarly, models were constructed with and without another confounder, the physical function score from the 36-item short-form health survey (SF-36).

A total of 1,017 women were eligible and agreed to participate in the study, including 799 women without diabetes and 218 women with diabetes (65 randomly recruited from the GHC enrollment file plus 153 randomly selected from the diabetes registry). Because of the 2-year, prospective nature of the original study of UTI risk, participation rates were 34% among women without diabetes and 26% among women with diabetes. A refusal questionnaire was completed by 59% of women refusing to participate. Women declining participation were more likely to be widowed, have less than a college education, report nonwhite ethnicity, be nonusers of hormone replacement therapy, and have no history of prior UTI (P < 0.05 for all). In all, 24% (289 of 1,192) of refusers responding to the refusal questionnaire had diabetes compared with 21% (218 of 1,017) of participants (P = 0.05), but incontinence data were not obtained.

Table 1 summarizes participant characteristics. Mean age was 64 years, and a majority of participants were white, not sexually active, and taking oral estrogen. Overall, 21% of the sample had diabetes. Three subjects met the criteria for type 1 diabetes and were included with other women with diabetes for analysis purposes. Among the 218 women with diabetes, 30% had it for ≥10 years, 48% used oral hypoglycemia medication, 19% used insulin, 49% reported symptoms of peripheral neuropathy, and 17% reported retinopathy (Table 1).

Sixty percent of the total sample reported any incontinence in the past month. Table 2 demonstrates that women with and without diabetes had similar prevalence and duration of incontinence. Prevalence in the total sample was 17% for stress incontinence, 10% for urge incontinence, and 32% for mixed incontinence. Women with diabetes had less stress and urge incontinence, whereas 40% had mixed incontinence. Of the total sample, 8% had severe incontinence and 14% described difficulty controlling urination as a moderate to severe problem. Women with diabetes were significantly more likely to report urinary incontinence symptoms (P ≤ 0.06) (Table 2).

Seventy-seven percent (65 of 84) of women with severe incontinence also reported moderate to severe difficulty controlling urination, compared with 21% (77 of 369) of women with moderate and 2% (5 of 216) with slight incontinence (P < 0.0001). Among women reporting stress incontinence, 5% (9 of 174) had severe incontinence and 9% (16 of 174) had moderate to severe difficulty controlling urination. For urge incontinence, the corresponding numbers were 12% (13 of 106) and 21% (22 of 106). For mixed incontinence, the results were 19% (62 of 324) and 33% (108 of 324), respectively (P < 0.0001 for all).

Table 1 summarizes the relationships between exposures and incontinence outcomes. None of the diabetes characteristics were associated with any incontinence in the past month. Diabetes at baseline, diabetes treatment type, duration, glucose control, peripheral neuropathy, and retinopathy were associated with severe incontinence (Table 1). Both incontinence outcomes were associated with physical function from the SF-36, a history of bladder or urinary surgery (among 102 women who had surgery, 46 reported doing so for urinary incontinence), and history of UTI in the past year (Table 1).

In Table 3, multiple regression models for each diabetes characteristic demonstrated positive associations between the characteristic and severe incontinence, except for glucose control. Odds ratios (ORs) increased from 1.0 for no diabetes to 1.7 for diet-treated diabetes, 2.1 for pill-treated diabetes, and 2.3 for insulin-treated diabetes. Similarly, ORs increased from 1.0 to 1.8 for women with diabetes without peripheral neuropathy, to 2.3 with peripheral neuropathy, and from 1.5 for women with diabetes without retinopathy to 2.8 for those with retinopathy. Once these models were adjusted for BMI, the odds of severe incontinence for each diabetes characteristic decreased, and none remained statistically significant, although the increasing trends in odds ratios persisted (Table 3). Additional adjustment for physical function from the SF-36 diminished the odds of association and was not statistically significant (data not shown).

In this sample of postmenopausal women, with supplemental sampling of women with diabetes, 60% reported recent leakage of urine. The large Norwegian Epidemiology of Incontinence in the County of Nord-Trøndelag study, which used the Sandvik severity index, found a lower prevalence of any incontinence among similarly aged women (26–30%) but a similar prevalence of severe incontinence (7–12% compared with 8% in our study) (23). Prevalence measures of “any incontinence” vary widely throughout the literature. Multiple studies of community- dwelling peri- and postmenopausal women in the U.S. have high incontinence prevalences (7,17,24,25). Populations with higher prevalence of incontinence may have more risk factors for incontinence (e.g., higher BMI) and women in the U.S. may be influenced by direct-to-consumer advertising for incontinence pharmaceuticals, resulting in greater incontinence awareness and reporting. Additionally, participants in this prospective evaluation of UTI risk in a population enriched with women with diabetes might be more inclined to report incontinence, contributing to our high prevalence.

Although the prevalence of “any incontinence” is typically quite variable between studies, measures of severe incontinence are more consistent and range from 3 to 17% (26). We found that risk factors for women with any incontinence differ considerably from those for women with severe incontinence. Because women with severe incontinence are more likely to seek medical attention (27), risk factors for severely incontinent women are the most clinically relevant.

The prevalence of any urinary incontinence was very high and not different between women with and without diabetes. However, women with diabetes were disproportionately represented among those with more severe voiding symptoms. None of the diabetes characteristics (duration, treatment type, glucose control, or complications) were associated with “any incontinence in the past month,” but all were positively associated with severe incontinence in unadjusted models. Additionally, these characteristics (except glucose control) were independently associated with severe incontinence in multivariable models not adjusted for BMI. Although studies have found that diabetes is associated with measures of “any incontinence” (24), we did not. This measure, for our cohort, may not have identified clinically significant incontinence. Prior work suggests that the onset of diabetic cystopathy is insidious and that incontinence symptoms do not appear until late in the course of the disease (10,28). This may explain why these differences are demonstrable among more extreme measures or with assessment tools that require severe enough symptoms to be more objectively quantified.

In multiple regression models before adjustment for BMI, the presence of diabetes at baseline, treatment type, duration, peripheral neuropathy, and retinopathy were all independently associated with severe incontinence. Increasing trends in the relative odds of severe incontinence within each of these characteristics supports the hypothesis that diabetes treatment type, neuropathy, and retinopathy may play a role in the development of severe incontinence or reflect underlying disease severity. However, BMI in this population of diabetic women ultimately confounds these effects. Fifty-six percent of women with diabetes in our sample had BMIs ≥30 kg/m², compared with 21% of women without diabetes (P < 0.0001). Further research is needed to determine whether diabetes characteristics of women with type 1 diabetes (who are less likely to be overweight) or of women with more advanced type 2 diabetes are risk factors for incontinence independent of BMI.

Prior studies demonstrated higher postvoid residual bladder volume among women with diabetes (28), and it has been associated with diabetic cystopathy in selected populations (29). Postvoid residual bladder volume was higher among women with diabetes than among those without. However, the mean volume of 49 ml is not close to the range typically considered clinically significant (100–200 ml), and we found no relationship between postvoid residual bladder volumes and any or severe incontinence.

UTIs are more common among women with diabetes and are routinely associated with incontinence, supported in this study for any and severe incontinence outcomes. Women with Escherichia coli vaginal colonization or asymptomatic bacteriuria were more likely to have severe incontinence, which suggests a possible mechanism.

Although parity is an important risk factor for urinary incontinence (30), our results are consistent with the observation that the effects diminish and become negligible with older age (7,17). Multiple randomized controlled trials suggested worsening or no effect of estrogen upon urinary incontinence (31,32). However, estrogen cream was recommended in a 1996 guideline for incontinence treatment (33) and is increasingly used for vaginal symptoms since the Women’s Health Initiative discouraged routine oral estrogen use (34). Estrogen cream use was positively associated with any incontinence in our sample. Confounding by indication is possible. We found that 72% (51 of 71) of women using estrogen cream in the past year did so because of symptoms of vaginal dryness; however, further evaluation of this purported treatment is warranted.

Our results are cross-sectional, limiting the ability to fully evaluate temporal sequence, particularly for diabetes and BMI. Participant accrual from a study of UTI may have resulted in a higher-than- expected prevalence of urinary incontinence. Additionally, the prospective nature of the original study resulted in low participation rates, particularly among women with diabetes. If nonparticipants with diabetes were more ill, our estimate of incontinence in this group would be artificially low. Incontinence outcomes were self-reported; because women with diabetes may be less aware of mild incontinence symptoms (28), our ability to detect differences would be limited. Self-reported height and weight typically result in underestimation of BMI (35), thus the effect of BMI in this population may be greater than our analysis indicates. Similarly, diabetes duration, peripheral neuropathy, and retinopathy were also self-reported measures and may also be underreported, limiting our ability to detect their effects.

Study strengths include enrichment of this large population-based sample with women with diabetes and detailed information regarding diabetes characteristics. The approximation to a validated measure of incontinence severity demonstrated associations that were not apparent using a more general measure. Another strength was the ability to evaluate and adjust for a wide variety of confounding factors, including clinical measures of postvoid residual bladder volume, bacterial vaginal colonization, asymptomatic bacteriuria, and UTI, that are pertinent for women with diabetes.

In summary, urinary incontinence affects a majority of postmenopausal women. Women with diabetes have more severe incontinence and are more likely to use pads, be unable to completely empty the bladder, be unaware of leakage, and have discomfort with urination. Postvoid residual bladder volume plays no role in urinary incontinence among community-dwelling postmenopausal women, whereas BMI is the predominant factor associated with severe incontinence among diabetic women and is potentially modifiable with intervention.

This study was funded by the National Institutes of Health Grant RO1 DK43134 and is the result of work supported with resources from and the use of facilities at the VA Puget Sound, Seattle, WA. S.L.J. is a VA Health Services Research and Development fellow.