We analyzed data from 1529 women enrolled in the Study of Women’s Health Across the Nation (SWAN) who reported no incontinence at baseline and followed them through 6 annual follow-up visits (1995–2002) (). SWAN is a multi-center, multiethnic, prospective study of the menopause transition (8
) funded by the National Institute of Aging (NIA). Seven clinical sites (Boston, Massachusetts; Chicago, Illinois; the Detroit area, Michigan; Los Angeles, California; Newark, New Jersey; Pittsburgh, Pennsylvania; and Oakland, California) recruited a total cohort of 3302 women. Eligibility criteria for entry into the SWAN cohort were age 42–52 years and self-identification as one of five racial/ethnic groups (African American, Hispanic, Chinese, Japanese, Caucasian). Exclusion criteria included inability to speak English, Spanish, Japanese, or Cantonese, no menstrual period in greater than 3 months before enrollment, hysterectomy and/or bilateral oophorectomy prior to enrollment, and current pregnancy, lactation, or hormone use. The institutional review boards at all sites approved SWAN, and all women gave informed consent to participate.
Flow Diagram Showing Identification of Women Included in Analytic Cohort
A self-administered questionnaire assessed incontinence at baseline and at each follow-up visit. Based on response to the question: “In the past year/since your last study visit, have you ever leaked even a small amount of urine involuntarily?”, we classified frequency of incontinence as “almost daily/daily” (daily), “several days per week” (weekly), “less than one day per week” (monthly), “less than once a month” or “none.” We defined any incontinence as incontinence occurring at least monthly. We considered incontinence occurring less than once a month as clinically insignificant and subject to higher misclassification and thus combined this category with “no incontinence.” We categorized type of incontinence as “stress” if participants reported leakage with “coughing, laughing, sneezing, jogging, jumping, with physical activity or picking up an object from the floor” or as “urge” if participants reported leakage “when you have the urge to void and can’t reach the toilet fast enough.” We defined “any” incontinence as either stress or urge symptoms. Women who reported new onset incontinence at any of the 6 annual follow-up visits were compared to women who did not develop incontinence over the same time frame.
SWAN classified menopausal status annually from menstrual bleeding patterns. Pre-menopause was less than three months of amenorrhea and no menstrual irregularities in the previous year; early peri-menopause was less than three months of amenorrhea and some menstrual irregularities in previous year; late peri-menopause was three to 11 months of amenorrhea; and postmenopause as 12 consecutive months of amenorrhea. Women who used hormone therapy (oral contraceptives or systemic estrogen and/or progestin) prior to the last menstrual period were considered to have an “unclassifiable” menopausal status at the time of use. Similarly, we could not classify the menopausal stage of women who underwent hysterectomy as data on whether one or both ovaries remained in situ could not be confirmed. To assess whether change in menopausal status was associated with the development of incontinence, we created a variable with four mutually exclusive categories comparing the status in the previous year to the current year, ie: change in menopausal stage (eg: from pre- to early peri-menopause); started hormones (from any stage to hormone use); or stopped hormones (from hormone use to none) compared with no change in stage or hormone use.
We calculated body mass index (BMI) as weight in kilograms/(height in meters)2
based on measurements taken annually by certified staff who used calibrated scales and a stadiometer. Socioeconomic status was approximated by level of difficulty paying for basics (food, heat and shelter). Interviewers obtained self-reported medical histories, smoking history and medication use. Each year SWAN used the same questions from the Center for Epidemiological Studies-Depression scale (9
) (for characterizing our cohort, we defined depressive symptoms as a score of 16 or above), the Medical Outcomes Study Social Support Survey (10
), the Life Stressors and Social Resources Inventory (11
), and the Psychiatric Epidemiology Research Interview (12
). SWAN measured anxiety symptoms by a summed score of days in the past two weeks in which certain symptoms were experienced (grouchiness or irritability, feeling tense or nervous, pounding or racing heart, feeling fear for no reason); for characterizing our cohort, we defined anxiety symptoms by a score of 4 or more (13
). At year one only, SWAN combined responses to questions assessing sensitivity to physical sensations into a Symptom Sensitivity Scale (14
Our fixed or baseline covariates included baseline age, race/ethnicity, BMI, diabetes, hypertension, self-reported diagnosis of fibroids, parity, marital status, socioeconomic status, education, social support, general health and symptom sensitivity. Time dependent co-variates were menopausal status, hormone use prior to the final menstrual period, new self-reported diabetes and hypertension, smoking status (never, ever, current). We created variables to represent change in certain characteristics by subtracting values in the current year from values in the previous year: weight change (per pound), waist to hip ratio change (per 0.1 units), anxiety and depressive symptoms score change (per one unit), change in the number of stressful life events (per one event), change in social support scores (per one unit), and overall health status change on a five-point scale.
Drop outs were those deceased, who discontinued the study voluntarily, or who could not be contacted after missing two or more visits at the 6th annual visit. When a woman was missing data on frequency and type of incontinence from one or two visits, we imputed values as follows. If the missing value occurred at year 6, we imputed by using the value at the previous visit. If women reported no incontinence in the years previous and subsequent to a missing response, we assumed no incontinence in those missng years. If a woman was missing incontinence data in the one to two years previous to a first report of incontinence we randomly assigned her missing values to either no incontinence or the frequency and/or type of incontinence in that subsequent year. We imputed incontinence frequency for 324 women (13.5%) and type for 18 women (0.2%). When weight was missing for one or two visits, the we imputed values for 88 visits (1.2%) as the mean between the two known values. In the same way, we imputed waist circumference values for 138 visits (1.8% of all visits). For all other independent variables, we dropped missing data from the analysis. We excluded New Jersey year 6 data from our models because that site had disproportionate and systematic loss of the Hispanic and Caucasian participants during that year.
We compared proportions and means of each variable for women who were continent at baseline and who remained in or dropped out of the study using the t- and chi-squared tests. For our survival analysis, we used discrete proportional hazards models (15
). First, in our main models, we evaluated whether menopausal stage at the annual visit concurrent with the first report of incontinence and other time dependent factors were associated with the development of monthly or more any incontinence compared with no development of any incontinence. Here we included a variable to account for whether a woman had advanced to another menopausal stage or had started or stopped hormones. We created similar separate models for stress and urge incontinence. For stress incontinence, our comparison group was those women who had no development of stress incontinence and for urge incontinence it was those women who had no development of urge incontinence. Second, we evaluated menopausal stage in the year previous to the first report of incontinence, controlling for the same covariates. Finally, we examined whether concurrent menopausal stage and change in status was associated with the development of more frequent incontinence by modeling weekly or more incontinence compared to no development of incontinence. The candidate covariates described above were chosen based on the literature, a priori hypotheses and/or were associated with the outcome in univariable analysis at p < 0.10. We used SAS 9.1, SAS Institute Inc., Cary, NC, USA.
We evaluated proportional hazards assumptions by plotting log(−log(survival)) functions for various groupings of the data based on the time independent covariates and estimated survival functions for each group using the Kaplan-Meier estimates. Since the proportional hazards held regardless of the covariate groupings, the assumption appeared to be adequate for all the fixed and time-dependent covariates. We tested model stability by running models with and without imputed data and respectively forcing in menopausal status, change in status and age without significant changes in the point estimates of the other variables. Correlation matrices showed no significant co-linearity among our independent variables, including between menopausal status and age. We chose our final models based on the lowest Akaike Information Criterion (AIC) which indicate the best fit for discrete proportional hazards models.