TABLE 3

Poisson regression analyses of the ability of candidate indexes of IR to predict incident diabetes, with combined analysis using data from the SAHS, the MCDS, and the IRAS

IndexUnadjusted
Adjusted for age, sex, systolic blood pressure, HDL cholesterol, and BMI
AROC curve
Top 10% vs. bottom 90%
AROC curve
Top 10% vs. bottom 90%
%RankP*RRRank%RankRRRank
-ISI0, 12078.515.84181.613.131
-ISIgly_a74.02<0.00013.40379.621.799
-SiM73.130.244.22279.432.112
-ISIgly_b71.540.0813.296.578.061.854.5
-QUICKI71.550.283.296.578.151.854.5
-Stum_nodem71.060.793.171077.871.7111
-BFSI69.670.273.051177.391.7510
-ISI-2h68.080.253.38478.741.828
-McAuley67.890.912.561377.1101.5312
In(FI)67.8100.992.5516.577.0111.5313.5
-Stum_wdem67.4110.962.571276.5131.1317
-ISI67.2120.723.296.577.581.854.5
BMI66.4130.612.561475.3190.7019
-Raynaud66.1140.852.5516.576.9121.5313.5
FIRI63.6150.00093.296.576.3141.854.5
HOMA-IR62.8160.993.20976.0151.827
FI59.717<0.00012.551575.9161.5315
IGR55.718<0.00011.811875.5181.1716
IGR-2h55.1190.961.641975.6171.0618
  • Ordering of indexes in this table were maintained based on ranking in the unadjusted AROC column. For consistency in the direction of the associations in the tables, we have taken the negative of indices that estimate SI (versus insulin resistance).

  • *

    * P value from algorithm developed by DeLong et al. (27), showing significance of difference in AROC for each index from one directly above it (i.e., having the next highest AROC). BFSI, Bennett’s fasting insulin sensitivity index; FI, fasting insulin; IGR, insulin/glucose ratio; Stum_wdem, Stumvoll index with demographic variables.