TABLE 3

Changes associated with change in adiponectin

PlaceboMetforminLifestyle
Univariate correlations
    Change in weight−0.24P <0.001−0.29P <0.001−0.42P <0.001
    Change in waist−0.17P <0.001−0.19P <0.001−0.35P <0.001
    Change in 1/FI0.11P = 0.00070.08P = 0.020.15P <0.001
    Change in insulinogenic index−0.04P = 0.23−0.02P = 0.54−0.01P = 0.75
Multivariate modeling
    Baseline adiponectin (3 μg/ml)−0.21 ± 0.04P <0.001−0.08 ± 0.05P = 0.10−0.15 ± 0.05P = 0.005
    Baseline 1/FI (0.03 100 μU/ml)−0.02 ± 0.06P = 0.780.04 ± 0.06P = 0.520.13 ± 0.06P = 0.03
    Baseline insulinogenic index (100 μU/mg)−0.06 ± 0.06P = 0.26−0.03 ± 0.06P = 0.680.01 ± 0.07P = 0.94
    Baseline weight (10 kg)−0.01 ± 0.02P = 0.58−0.05 ± 0.03P = 0.10−0.12 ± 0.03P <0.001
    Change in weight (5 kg)−0.31 ± 0.05P <0.001−0.43 ± 0.05P <0.001−0.49 ± 0.04P <0.001
  • Data are correlation coefficient of coefficient ± SE. Pearson correlation coefficients for change in adiponectin are presented for change in relevant potential univariate determinants after adjustment for baseline weight. Coefficients in multivariate modeling were derived from linear regression modeling, reported per ∼SD for each variable. Higher values of inverse fasting insulin (1/FI) reflect greater insulin sensitivity; higher insulinogenic index (II) values reflect better β-cell function. Change values are year 1 minus baseline. In multivariate modeling, neither 1/FI or II contributed significantly to change in adiponectin in any treatment group.1/FI, inverse fasting insulin.