Table 3.
Modeling results: GLM estimated change at Y4 in lumbar spine and total hip BMD regression coefficients for predictor variables
| By arm
mean (SE) |
Predictor variable |
|||||
|---|---|---|---|---|---|---|
| Response variable | DSE | ILI | Change in fitness (METs) β (SE) | Change in weight (kg) β (SE) | Change in HbA1c (%) β (SE) | Y1 change in outcome β (SE) |
| Men | ||||||
| L/S BMD % change/year baseline–Y4 | 0.96 (0.19) | 0.98 (0.19) | 0.005 (0.04) | −0.02 (0.007) | 0.09 (0.05) | 15.4 (1.4) |
| P value | 0.835 | 0.884 | 0.005 | 0.042 | <0.001 | |
| Total hip BMD % change/year baseline–Y4 | −0.53 (0.16) | −0.53 (0.16) | −0.02 (0.03) | 0.03 (0.006) | 0.01 (0.04) | 15.4 (1.4) |
| P value | 0.931 | 0.475 | <0.001 | 0.738 | <0.001 | |
| Women | ||||||
| L/S BMD % change/year baseline–Y4 | −0.65 (0.20) | −0.68 (0.19) | -0.04 (0.04) | −0.008 (0.007) | 0.05 (0.03) | 16.5 (1.3) |
| P value | 0.759 | 0.426 | 0.254 | 0.087 | <0.001 | |
| Total hip BMD % change/year baseline–Y4 | −1.27 (0.19) | −1.01 (0.18) | −0.02 (0.04) | 0.05 (0.006) | 0.09 (0.03) | 16.7 (1.4) |
| P value | 0.005 | 0.595 | <0.001 | 0.005 | <0.001 | |
Dependent variables were L/S and hip BMD. Independent variables were intervention arm (DSE vs. ILI). Predictor variables were change in fitness (METs), weight change, change in HbA1c, Y1 change in outcome. Also adjusted for were the following covariates: TZD and insulin use at baseline, age, ethnicity, study center (%), menstrual status (for female models only), duration of diabetes, baseline BMI, proteinuria (%), smoking status, baseline fitness, baseline A1C, and baseline level of outcome measure. L/S, lumbar spine.