Subjects

In all, 39 postmenopausal women from the Birmingham, AL, region were recruited through study advertisements, enrolled, and randomly assigned by the hospital research pharmacy, using a block design of 6, to consume a daily shake supplement containing either soy protein plus isoflavones or an isocaloric casein placebo containing no isoflavones (supplements and placebo donated by Revival Soy, Kernersville, NC). Inclusion for our study mandated that female subjects must be between the ages of 45-60 years, be amenorrheic for at least 12 months, have a documented follicle-stimulating hormone >30 mIU/mL, and have a body mass index between 30-40 kg/m ² .

Volunteers were excluded from our study if they were shown to have diabetes mellitus based on a 75-g oral glucose tolerance test, consumed a strict vegetarian or low-fat diet, consumed a diet high in fiber or soy based on a standard dietary screening questionnaire, or if there was a weight change of >10 pounds within the prior 12 months. Other exclusion criteria were the regular consumption of vitamin and mineral supplementation greater than the recommended daily allowance, regular participation in an exercise program (more than twice weekly), cigarette smoking, hormone replacement or selective estrogen receptor modulator therapy within the prior 12 months, known casein/milk allergy, moderate to excessive alcohol consumption (>2 drinks daily), or a known estrogen-dependent neoplasia.

A total of 50 volunteers were assessed for eligibility in the General Clinical Research Center (GCRC) after telephone screening, and 39 met all eligibility criteria. Those who did not meet criteria had diabetes mellitus diagnosed by the 2-hour oral glucose tolerance test (3), had a follicle-stimulating hormone level too low (4), had clotting irregularities (2), or refused to proceed with the study when presented with more complete information (2). In all, 39 women were randomized, and 33 completed the entire 3-month study and were analyzed. Of the 6 drop-outs, 3 dropped out as a result of unrelated illness, 2 dropped out because of discomfort during the initial overnight hospitalization, and 1 dropped out due to a work/school schedule conflict.

Shake supplements

The composition of the shakes was as follows: 120 calories, 2.5 g fat, 7 g carbohydrates, 600 mg calcium, 500 mg phosphorus, 320 mg sodium, 560 mg potassium, and 3 mg iron. The soy-containing shakes had 20 g soy protein plus 160 mg isoflavones (96 mg available as aglycones). The isocaloric placebo contained 20 g casein protein and no isoflavones. This supplement is well studied and has been used in other trials. Volunteers mixed the powdered shakes with water in the morning and consumed half of the shake in the morning with breakfast and the other half in the evening with dinner. All were in regular contact with the GCRC dietician, both prior to enrollment in the study, and throughout the study’s course. Subjects’ weights were checked 2 weeks postenrollment, at 1 month, and at 2 months. If the weight differed by >2.3 kg from the original weight, the dietician was consulted to instruct the volunteer further on weight maintenance. Compliance with the supplements was established by measuring serum isoflavone levels at baseline, 4 weeks, 8 weeks, and 12 weeks. The study was approved by the institutional review board and the GCRC at the University of Alabama at Birmingham, and all subjects gave their consent to participate.

Insulin secretion and insulin-stimulated glucose disposal

Three days prior to testing, volunteers were given a standardized diet to follow in which meals contained 55% carbohydrates, 30% fat, and 15% protein. After 3 days, volunteers presented to the GCRC for an overnight admission. On the subsequent morning, following a 12-hour fast, we performed a hyperglycemic clamp as previously described by DeFronzo et al. Briefly, a large-bore intravenous catheter was started in an antecubital vein to gain access for an infusion of 20% dextrose. A second intravenous catheter used for drawing arterialized blood samples was placed in the contralateral hand, and was kept in a 60°C hot box. When possible, this catheter was placed in a retrograde manner.

Beginning at 7:00 AM, a primed, constant infusion of [6,6- H ₂ ]glucose (prime 16.5 μmol/kg, infusion 18.3 μmol/kg/h) was started (time 0 minutes). Blood was sampled at time points 90, 100, 110, and 120 minutes to measure glucose concentrations and enrichment. A priming dose of 20% dextrose was then started at time point 120 minutes (240 mg/kg; 15 minutes duration) followed by a variable rate infusion. First-phase insulin secretion was determined by drawing blood samples every 2 minutes beginning at time point 120 minutes, and continuing through 136 minutes; blood plasma levels of insulin, C-peptide, and plasma glucose were measured. Blood samples were then drawn every 5 minutes from time point 140-240 minutes and glucose levels were determined. Second-phase insulin secretion was determined through plasma levels of C-peptide and insulin levels measured at 15-minute intervals beginning at 140-240 minutes.

Tissue glucose uptake was estimated by averaging the glucose infusion rate from 210-240 minutes. Insulin sensitivity was estimated by dividing the average glucose infusion rate by the average insulin level over the same time period. The rate of appearance of glucose was calculated as described previously. This is an index of endogenous glucose production. Total insulin-stimulated glucose disposal is estimated by determining the mean dextrose infusion used to maintain hyperglycemia during the last 30 minutes of the clamp, and the residual endogenous glucose production. Peripheral glucose disposal and endogenous glucose production were summed to determine total glucose disposal.

As described previously, insulin secretion rates were determined using plasma C-peptide levels with deconvolution analysis and linear regularization with use of a 2-compartment model using standard parameters for C-peptide kinetics. C-peptide kinetics were assumed to be biexponential, and the exponential parameters were taken to be equal to the mean values, as validated previously.

Inflammatory markers

Inflammatory markers and adipokines were measured using a multiplexing xMAP technology with a Luminex 100 Analyzer (Luminex Corp, Austin, TX), an immunoassay combining the principle of a sandwich enzyme-linked immunosorbent assay with fluorescent beads precoated with analyte-specific antibodies. The subsequent color signals were processed and quantified for each analyte.

Body composition and body fat distribution

Determination of regional and total fat mass, percent fat, and lean mass was achieved using dual energy x-ray absorptiometry. A Lunar DPX-L densitometer (Lunar Corp, Madison, WI) was used for scanning, and these scans were analyzed by using the Lunar Version 1.3 DPX-L extended analysis program for body composition.

Computed tomographic (CT) scans were used to measure subcutaneous abdominal fat, visceral fat, and total abdominal fat, as we have previously described, and as initially described by Sjöström et al. Scans were taken at the L4/L5 vertebral disk space at an attenuation range of –190 to –30 Hounsfield U, and analyzed offline using NIHimage software (National Institutes of Health, Bethesda, MD). Subcutaneous abdominal fat was calculated by subtracting visceral abdominal fat from total abdominal fat.

Biochemical analysis

During the clamp, plasma glucose levels were analyzed with an automated analyzer using the glucose oxidase method (YSI Instruments, Yellow Springs, OH). Serum glucose and C-peptide concentrations were determined using a Tosoh AIA-600 II analyzer (South San Francisco, CA), which employs an immunoenzymatic method utilizing fluorescence. Serum lipid levels (total cholesterol, high-density lipoprotein, low-density lipoprotein [LDL], and triglycerides) were measured at baseline and at 3 months using an automated system for direct measurements based on the change in absorbance at 520 nm (Synchron DxC 800; Beckman Coulter Inc, Brea, CA). Within and between run coefficients of variation for all assays were <4%.

Serum isoflavones

Determination of serum isoflavones is described previously. Briefly, serum levels of genistein, daidzein, glycitein, dihydrodaidzein, O-desmethylangiolensin, and equol were measured by liquid chromatography-multiple reaction ion monitoring-mass spectrometry. All sera samples were run independently in duplicate, and the areas of the isoflavone peaks were normalized by ratio using internal standards, and were compared to normalized areas for isoflavone standards. Intraassay coefficients of variation were 3-10% at the concentrations observed in this study.

Physical activity

The Minnesota Leisure Time Activity Questionnaire was used to self-report physical activity at the beginning and at the end of the study. This validated questionnaire was used to determine whether activity levels between groups changed over the course of the 3-month study.

Food intake

A 4-day food record was collected for each subject at the beginning and at the end of the study and analyzed by registered dieticians from the GCRC. To reflect free-living food intake during the study, dietary intake data were collected using Nutrition Data System for Research software versions 2006 and 2007, developed by the Nutrition Coordinating Center, University of Minnesota, Minneapolis, MN. Final calculations were completed using Version 2007. The Nutrition Data System for Research time-related database updates analytic data while maintaining nutrient profiles true to the version used for data collection.

Statistical analysis

Primary outcome variables were changes in visceral, subcutaneous, and total abdominal fat by CT scan. Statistical analysis included the calculation of means with SD and medians with interquartile ranges (IQR) for baseline, and for the differences between the 3-month and baseline time periods. Two-tailed Student t tests or Wilcoxon rank sum tests were used to test for associations where appropriate. Proportional data were compared by the use of χ ² test or Fisher’s exact test. To test for differences by race, we included a race-treatment interaction term in the linear regression model for each outcome variable. Effect size by race was reported only for those where the interaction term was significant. Data were analyzed with software (SAS, version 9.1; SAS Institute, Cary, NC). Statistical significance was set at P < .05 and all tests were 2-sided.