Baseline characteristics of subjects enrolled in the trial are shown in Table . Thirty-six percent of subjects were European-American, 34% African-American, 20% Hispanic-American, and 10% other or unknown. Adherence to dietary prescription was measured by counting meal replacement packages consumed as a proportion of the number of packages that should have been consumed had the subject used exactly the prescribed amount of soy product, and was expressed as a percentage. Mean group adherence did not differ statistically between the groups and ranged between 48 and 70% in both groups, except for the final month of the study when it was lower (43–48%). Regression analysis showed that in both groups amount of weight loss during the trial was positively related to degree of adherence (p = 0.0013).
Measured Baseline Subject Characteristics (Mean and SD = standard deviation)*
The results for weight change are presented in detail below. Following these, results for the remaining variables are presented in summary. Details of all analyses and results for all study variables, as well as figures, can be obtained upon request.
Figure shows the average change in weight from baseline at each visit for all subjects who made the visit. For the first 4–8 weeks the declines in weight appear to be similar at both levels of caloric restriction but between weeks 12 and 20 the difference between the 1200 and 1500 calorie allocation for the first 12 weeks seems to be reflected in a greater weight loss for the more aggressively restricted group. T-tests showed that the difference only reaches statistical significance among the newly recruited subjects at the 8 week follow-up visit (P = 0.047). This greater weight loss is retained until the end of the study in the subjects from Study 1 but not in those newly recruited. Newly recruited subjects in the 12/15/18 group achieved their largest mean weight loss at the 16 week visit (6.44 kg) while those in the 15/18 group had the largest mean weight loss at the 28 week visit (6.57 kg).
Figure 1 Changes in weight for subjects from Study 1 and newly enrolled for the 12/15/18 and 15/18 groups* *The left Y axis is the scale of weight change for subjects in study 2 only, while the right Y axis is the scale of weight change for subjects who were in (more ...)
The best fitting mixed effects model for weight change is,
ΔWT = 2.12 - 0.209Time + 0.005 (Time)2 - 0.058Base + 0.926GRP + 2.117IN1 + S
where ΔWT is the predicted weight change from the fitted model and S is a random effect that is estimated for each subject. The independent predictor variables, Time, (Time)2, Base and IN1 were significant (α = 0.05), but GRP was not (P = 0.171). No higher order interactions were found to be significant. Time ranged from a minimum of 2.86 weeks (first follow-up visit) to a maximum of 49 weeks (the final follow-up visit).
We interpret the model as follows. Subjects tend to lose weight over time (-0.209 Time
) but the loss is less pronounced at later visits. Subjects with larger baseline weights tended to lose more weight. Newly recruited subjects lost, on average, 2.1 kg more than subjects from Study 1, not surprising since subjects who had been in study 1 entered study 2 already 2.9 kg below their study 1 baseline weight having lost some weight as the control group in the previous study [10
Body Fat and Waist Circumference
The mixed effects models for total body fat change indicated that, as for body weight, subjects tend to lose total body fat over time but the loss is less pronounced at later visits due to a positive quadratic effect of time. The rate of change in body fat over time depended on whether subjects had been in the previous study. There were no significant differences between the two treatments on loss of either body fat or total body fat expressed as a percentage at any single follow-up visit.
The best mixed effects model for waist circumference change indicated that subjects' waist circumference decreased over time but the decrease was less pronounced at later visits. Subjects with larger initial waist circumference showed larger decreases and subjects in the 12/15/18 group experienced, on average, an additional 1.729 cm of waist circumference decrease versus subjects in the 15/18 group. Significant differences were observed between the two treatments at week 4, 8, 12, and 16. In each case, reductions were greater in the 12/15/18 group.
The mixed effects model for cholesterol indicates that there is an initial overall decrease in cholesterol of a magnitude related to baseline value and that, over time, the cholesterol tends to increase but at a decreasing rate. Subjects in the 12/15/18 group tended to experience, on average, an additional 7.64 mg/dl reduction in total cholesterol compared with subjects in the 15/18 group.
The models for HDL and LDL cholesterol change, as for total cholesterol, showed an initial decrease of a magnitude related to baseline value followed by a regain over time, but at a decreasing rate. For LDL, newly recruited subjects decreased 6.764 mg/dl more than subjects from Study 1. A significant interaction indicates that the rate of LDL change over time depends on the baseline value.
The mixed effects model for HDL/ Total Cholesterol change had only one significant term, the baseline value, with a coefficient indicating that subjects with higher baseline values showed larger decreases in HDL/TC. T-tests on changes from baseline show that, at all time points throughout the study, the HDL/TC improvement from baseline was highly significant (p < 0.002).
Triglyceride changes were not significantly related to any of the individual variables in the mixed effects model. Significant interaction effects indicate that the relation of the baseline value to the amount of triglyceride change depended on the group and whether subjects had served as controls in the prior study.
The mixed effects models for blood pressure change indicate that subjects' blood pressure decreased over time but the decrease is less pronounced at later visits. Subjects with higher baseline values showed larger decreases. Newly recruited subjects tended to decrease more than subjects who had been in Study 1, and subjects in the 12/15/18 group tended to experience a greater blood pressure reduction than subjects in the 15/18 group. By t-test we observed significant differences between the groups on diastolic blood pressure at weeks 16 and 28 with the 12/15/18 group producing the greater reduction.
Differences from baseline
On the whole, we observed significant differences on the primary and secondary outcomes when we compared changes at each follow-up clinic visit with baseline values. This pattern is reflected in the significance of the time variable coefficient in most mixed model analyses. Exceptions were observed on triglycerides and SBP, where the pattern of findings was less consistent. The change from baseline was particularly evident in newly recruited subjects; however, in general, it was also apparent when the analysis was carried out on only Study 1 subjects. The smaller effect seen in subjects who had previously served as controls may be a consequence of the fact that they had already lost an average of 2.9 kg during the previous 3 months.