Predictive Potential of Twenty-Two Biochemical Biomarkers for Coronary Artery Disease in Type 2 Diabetes Mellitus

Table 3

Performance of different models for CAD prediction in type 2 diabetes patients.

	BA (%)		NMI (%)
Model 1 (full): classical factors + emerging biochemical factors	72.85%	0.002	20.52%	0.001
Model 2: classical factors only	57.14%	0.01	9.29%	0.003
Model 3: emerging biochemical factors only	66.38%	0.001	13.08%	0.003

BA: balanced accuracy. NMI: normalized mutual information.
Model 1: LDL: HDL ratio, methionine, reduced glutathione, L-arginine : asymmetric dimethyl-L-arginine, nitrate plus nitrite, n-acetyl--glucosaminidase, gender, and BMI.
Model 2: LDL : HDL ratio, gender, and BMI.
Model 3: methionine, reduced glutathione, L-arginine : asymmetric dimethyl-L-arginine, nitrate plus nitrite, and n-acetyl--glucosaminidase.
^* value based on 2,000 permutations. Age was not a significant predictor, yielding to virtually identical results after inclusion of this variable in the models (data not shown).