Table 2—

Multivariate regression models showing coefficients (P value) of testosterone (bioavailable and total) and SHBG as independent variables

Models (age included in all models)Independent variable
BAT*TTSHBG
Dependent: fasting insulin−0.035 (0.03)−0.029 (<0.0001)−0.013 (<0.0001)
 Adjusting for SHBG−0.028 (0.04)−0.016 (0.02)
 Adjusting for CT-TAF−0.004 (NS)−0.013 (0.01)−0.008 (<0.0001)
 Adjusting for CT-TAF and SHBG−0.004 (NS)−0.003 (NS)
 Adjusted for CT-IAF−0.018 (NS)−0.017 (0.01)−0.008 (<0.0001)
 Adjusted for CT-IAF and SHBG−0.016 (NS)−0.009 (NS)
 Adjusted for DEXA-TBF (kg)§−0.010 (NS)−0.012 (0.02)−0.009 (<0.0001)
 Adjusted for DEXA-TBF and SHBG−0.006 (NS)0.002 (NS)
Dependent: HOMA-IR−0.039 (0.01)−0.033 (<0.0001)−0.015 (<0.0001)
 Adjusting for SHBG−0.031 (0.03)−0.017 (0.02)
 Adjusting for CT-TAF−0.007 (NS)−0.016 (<0.0001)−0.009 (<0.0001)
 Adjusting for CT-TAF and SHBG−0.006 (NS)−0.005 (NS)
 Adjusted for CT-IAF−0.022 (NS)−0.019 (<0.0001)−0.009 (<0.0001)
 Adjusted for CT-IAF and SHBG−0.009 (NS)−0.011 (NS)
 Adjusted for DEXA-TBF (kg)§0.001 (NS)−0.015 (0.006)−0.01 (<0.0001)
 Adjusted for DEXA-TBF and SHBG0.002 (NS)0.0001 (NS)
  • Data are β (P value).

  • *

    * Bioavailable testosterone was a significant predictor of fasting insulin and HOMA-IR, but not independent of abdominal fat by CT or total body fat by DEXA;

  • the associations between total testosterone and fasting insulin and HOMA-IR were no longer independent of body fat after adjusting for confounding by SHBG;

  • inclusion of testosterone (total, bioavailable, or free) in the models did not diminish the significance of SHBG as a predictor of fasting insulin and HOMA-IR;

  • §

    § DEXA-TBF (kg) denotes total body fat in kilograms by DEXA. Using TBF expressed in percentage (%) yielded similar β and P values for the independent variables in the models. BAT, bioavailable testosterone; IAF, intra-abdominal fat, TAF, total abdominal fat, TT, total testosterone.