Table 3

Receiver operating characteristic analysis of lipid parameters in the prediction of diabetes mellitus during follow-up

β95% CIAUC (95% CI)Cut-off pointSensitivity (%)Specificity (%)Positive predictive value (%)Negative predictive value (%)
Reference model0.672 (0.646 to 0.700)0.05559.2166.899.0796.71
TC/HDL-C0.140.08 to 0.200.685 (0.659 to 0.712)*0.04376.2352.528.1997.51
Non-HDL (mg/dL)1.720.94×10−3 to 2.51×10−30.685 (0.659 to 0.711)*0.05857.4470.289.7396.73
TG (mg/dL)1.75×10−30.92×10−3 to 2.58×10−30.684 (0.658 to 0.710)*0.04574.4254.338.3297.44
TG/HDL-C0.030.01 to 0.050.680 (0.654 to 0.706)*0.04573.4554.248.2097.34
LDL-C/HDL-C0.200.09 to 0.310.680 (0.654 to 0.707)0.04672.9155.218.3297.34
HDL-C (mg/dL)−13.37−0.02 to −5.29×10−30.679 (0.652 to 0.705)0.04672.6354.328.1497.26
TC (mg/dL)2.91×10−30.23×10−3 to 5.61×10−30.674 (0.647 to 0.701)0.05264.3262.918.8196.93
LDL-C (mg/dL)1.27×10−3−1.66×10−3 to 4.19×10−30.672 (0.646 to 0.699)0.04868.5158.048.3397.06
  • β (95% CI) were tested in multivariable logistic regression models that added lipid parameter into the reference model that contained conventional risk factors (age, gender, waist circumference, body mass index, systolic blood pressure, diastolic blood pressure, C reactive protein, plasma glucose, cigarette smoking and alcohol drinking). The AUCs were compared between the reference model and multivariable regression models of each lipid parameters.

  • *P<0.05.

  • AUC, the area under the receiver operating characteristic curve; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; TC, total cholesterol; TG, triglyceride.