Participants were women who enrolled in an ancillary study of the SWAN at the Chicago site. This study (the “SWAN Fat Patterning Study”) was designed to investigate the impact of the menopausal transition on the accumulation of VF. SWAN is a seven-site multiethnic longitudinal study of women transitioning through menopause, featuring ongoing annual interviews. Women were eligible for SWAN if they were between the ages of 42 and 52, not pregnant or breastfeeding, had an intact uterus and at least one ovary, had menstruated within 3 months, and were not using oral contraceptives or hormone therapy. By design, the Chicago site recruited only non-Hispanic white and black women. A unique feature of the Chicago SWAN site is that it is a population-based design, which drew on a complete community census to recruit a sample of black and white women with a 72% participation rate. These women were recruited in a way that featured comparability on socioeconomic status within the black and white women, thus minimizing any confound between ethnicity and socioeconomic status. Details of SWAN recruitment and the study protocol have been reported elsewhere (23).
Women enrolled in the Chicago SWAN Fat Patterning Study between August 2002 and December 2005 coincident with their annual SWAN follow-up visit. They were eligible if they did not have a history of diabetes, chronic liver disease, renal disease, anorexia nervosa, alcohol or drug abuse, were not currently pregnant or planning to become pregnant, and had not undergone surgical menopause (hysterectomy and/or bilateral oophorectomy). Because of equipment limitations, women with breast implants, hip replacements, or weight exceeding 299 pounds could not participate.
Of the 386 eligible Chicago SWAN participants, 77% enrolled in the Fat Patterning Study. Because many SWAN participants were postmenopausal by the time the Fat Patterning Study began, we recruited additional pre- and perimenopausal women (65% of those eligible from the census) who were screened as part of the original SWAN recruitment effort but were too young to participate in 1996. These women were younger than the previously recruited women were, but did not differ in BMI, age-adjusted total fat or VF, education, or depressive symptoms. The final cohort consisted of 435 women (200 black and 235 white). Due to missing data on VF (due to equipment malfunction) (n = 3), new surgical menopause (n = 23), new hormone use (n = 45), or missing covariates (n = 5), 359 women (166 black, 193 white) were included in the current analyses.
At entry into the parent SWAN study and at each annual assessment, all SWAN participants completed a standard protocol; full details are provided elsewhere (23). Covariates of interest for the present analyses were measured as part of the annual SWAN assessment coincident with recruitment to the SWAN Fat Patterning Study for the 297 SWAN participants. The 138 women recruited uniquely to the Fat Patterning Study completed the same protocol as the SWAN participants. For all women, data presented here represent the baseline visit for the SWAN Fat Patterning Study.
All aspects of the study were approved by the Rush University Medical Center's institutional review board, and all women provided written, informed consent for their participation.
Assessment of VF. VF in the abdominal cavity was assessed by CT at High Tech Medical Park located within 9 miles of the Chicago SWAN study site. All abdominal CT scans were conducted by a trained technician using a General Electric Lightspeed VCT scanner (General Electric Medical Systems, Milwaukee, WI), with the participant in the supine position and arms folded across her chest. Following a scout view, a single 10-mm thick image of the abdomen at the L4-L5 vertebral space was obtained. Images were stored on optical disks and transferred to the reading center at the University of Colorado Health Sciences Center for analysis. Scans were read using a software developed by the reading center (RSI, Boulder, CO) and used in large cohort studies (24,25). Scans were read by a trained radiologist, blind to the participants' clinical or demographic characteristics. The radiologist defined total abdominal fat area using a cursor to delineate the area within the muscle wall surrounding the abdominal cavity (26). VF was defined as all adipose tissue within this area with an attenuation range between −190 and −30 Hounsfeld units (26).
Assessment of total body fat. Total body fat mass was assessed with whole body dual-energy X-ray absorptiometry scans using a General Electric Lunar Prodigy scanner (GE-Lunar, Madison, WI). Dual-energy X-ray absorptiometry scans use two X-ray energy sources, which enable separation of body mass into fat mass, lean tissue mass, and bone mineral content. Dual-energy X-ray absorptiometry scans, completed the same day as the CT scans at High Tech Medical Park, were performed with a participant in the supine position, arms by her side, wearing only a hospital gown. Scans were analyzed using GE-Lunar enCORE software (GE-Lunar). For data analyses, total body fat was quantified as the percent of fat in the total body habitus to represent the amount of fat for a given body size; total percent fat was calculated as total fat mass divided by total mass (total fat mass, total lean mass, and bone mineral content). Due to equipment malfunction, five women were unable to complete a dual-energy X-ray absorptiometry scan as part of the baseline study visit. Because BMI and total fat are very highly correlated (r = 0.83) in our sample, a regression equation was estimated to predict total body fat from BMI, and this value was then used in analyses.
Assessment of covariates.
Menopausal status: Bleeding criteria were used to characterize menopausal status as premenopausal (normal cycling), early perimenopausal (irregular cycles but bleeding within the past 3 months), late perimenopausal (irregular cycles with bleeding in the past 11 months but not within the last 3 months), and postmenopausal (no menses for at least 12 months). This self-reported definition was validated against follicle-stimulating hormone levels in that postmenopausal women who had higher follicle-stimulating hormone levels than peri- and premenopausal women (geometric mean = 86.3, 29.7, and 14.7, respectively; all pairwise comparisons P < 0.001).
Age was calculated as the difference between exam date and self-reported date of birth. Race was self-reported as black or white. The highest educational degree was self-reported at the screening visit: high school or less, some college, college degree, or graduate school.
All participants underwent exams, which included interviews, anthropometry, questionnaires, and a blood draw for the assessment of sociodemographic factors, cardiovascular risk factors, and reproductive hormones. Due to budgetary constraints, the blood assays, including high-density lipoprotein, were analyzed every other year only.
The Framingham risk score (FRS) was calculated using data on age, smoking status, total serum cholesterol and high-density lipoprotein cholesterol, resting blood pressure, and use of antihypertensive medication (27). Total cholesterol and high-density lipoprotein cholesterol were analyzed on EDTA-treated plasma using standard methods, previously described (28,29). Resting blood pressure was measured with a mercury sphygmomanometer, using a standard protocol following at least a 5-min rest with participants seated. Blood pressure was measured in the right arm, using an appropriately sized cuff. Two sequential blood pressure readings were obtained, 2 min apart. Use of antihypertensive medication was self-reported annually in SWAN and confirmed via medication review. For the calculation of the FRS, lipids were taken from the previous year, if necessary.
Depressive symptoms were assessed in SWAN with the 20-item Center for Epidemiological Studies Depression Scale (CES-D) (30), a validated scale with good test–retest reliability in ethnically diverse samples (31), and used extensively in epidemiological studies (32). A score of ≥16 on the CES-D is indicative of clinically significant symptomatology (32); therefore, we modeled the CES-D as a dichotomous predictor, CES-D score ≥16 vs. <16 as the referent category.
Hormones: Phlebotomy was performed in the morning following an overnight fast. Subjects were scheduled for venipuncture on days 2–5 of a spontaneous menstrual cycle (in cycling women) within 60 days of the anniversary of the baseline examination date. All assays were performed on the ACS-180 automated analyzer (Bayer Diagnostics, Tarrytown, NY) using a double-antibody chemiluminescent immunoassay with a solid phase anti-immunoglobulin-G conjugated to paramagnetic particles, anti-ligand antibody, and competitive ligand labeled with dimethylacridinium ester. The estradiol (E2) assay modifies the rabbit anti-E2–6 ACS-180 immunoassay to increase sensitivity, with a lower limit of detection of 1.0 pg/ml. Serum testosterone (T) concentrations were determined by competitive binding of a dimethylacridinium ester–labeled T derivative to a rabbit polyclonal anti-T antibody premixed with monoclonal anti-rabbit immunoglobulin-G antibody immobilized on the solid phase paramagnetic particles. Inter- and intra-assay coefficients of variation were 10.5% and 8.5%, respectively, with a limit of detection of 2.19 ng/dl. The two-site chemiluminescent assay for serum SHBG concentrations involved competitive binding of dimethylacridinium ester–labeled SHBG to a commercially available rabbit anti-SHBG antibody and a solid phase of goat anti-rabbit immunoglobulin-G conjugated to paramagnetic particles. Inter- and intra-assay coefficients of variation for SHBG were 9.9% and 6.1%, respectively, with a limit of detection of 1.95 nmol/l. Serum E2 concentrations were measured with a modified, off-line ACS-180 (E2–6) immunoassay. Inter- and intra-assay coefficients of variation averaged 10.6% and 6.4%, respectively, over the assay range, and the lower limit of detection was 1 pg/ml. Duplicate E2 assays were conducted with results reported as the arithmetic mean for each subject, with a coefficient of variation of 3–12%. All other assays were single determinations. Bioavailable testosterone was calculated as T (ng/dl) × 100/(28.84 × SHBG (nmol/l)).
Physical activity: Physical activity was measured at the SWAN baseline visit and follow-ups 3, 5, 6, and 9 via self-report with an adapted version of the Kaiser Physical Activity Survey, which was initially adapted from the Baecke physical activity questionnaire (33), assessing frequency of sports, nonsports leisure time, and household/child-care activities. An activity score was created by summing across domains, with a higher score indicating greater activity. For the current analyses, data from the Kaiser Physical Activity Survey was administered at the concurrent SWAN visit except for ∼20% of the sample for whom Kaiser Physical Activity Survey values were obtained from their most recently available prior visit.
Insulin resistance (homeostasis model assessment for insulin resistance) was calculated from glucose and insulin using the homeostasis model (34).
We present additional variables for descriptive purposes. Smoking was assessed in SWAN by questions on ever smoking, amount smoked and quit date. BMI was calculated as weight in kilograms divided by height in meters squared. Standardized protocols were used to measure weight, height and waist circumference. Height was measured without shoes using a stadiometer. Weight was measured without shoes and in light indoor clothing using scales that were calibrated on a monthly basis to a standard. Waist circumference was measured with the respondent in nonrestrictive undergarments.
Data analyses. All analyses were conducted using PC-SAS, version 9.1 (SAS Institute, Cary, NC). We used descriptive statistics to characterize participants on fat measurements, reproductive hormones, and all other covariates included in our analytic models. For reproductive hormones, SHBG, and insulin resistance, we report median and interquartile range, because the distributions are skewed. To assess the bivariate relation of hormones to VF, we present Spearman's rank correlations without and with adjustment for percent body fat. We used linear models to examine independent correlates of VF. First, we developed a base model by including all covariates significantly correlated with either VF or bioavailable testosterone, using a liberal P value of 0.2. We did not include insulin resistance in this model, because it is not in the causal pathway from testosterone to VF. Initially, we included smoking in the models, but it did not survive in the base or in any subsequent model and was thus excluded in the final presentation. Inspection of residual plots revealed that a transformation of VF was necessary, and we chose to use the logarithm (to base 10). Therefore, we present geometric means in the subgroup comparisons. Our final base model included percent body fat, age, race, FRS, dichotomous depression, and physical activity. To test our hypothesis that postmenopausal women have more VF, we added an indicator for postmenopausal status to this base model. To test hypotheses regarding the association of testosterone and estradiol with VF, we added both hormones to the base model. To evaluate the possible role of insulin resistance in mediating the association between bioavailable testosterone and VF, we included homeostasis model assessment for insulin resistance as a covariate. To test the hypothesis that postmenopausal women have more bioavailable testosterone, we used a linear model adjusting for all covariates in the base model. Age, FRS, homeostasis model assessment for insulin resistance, physical activity, and hormone measures were modeled continuously in the current analyses whereas race (with black ethnicity as the referent) and depression (CES-D, referent <16) were modeled as binary variables.