Table 3.

Model characteristics of each model:

Model details	Full model (with Lipoprotein (a))	Reduced model (the original Framingham)	Difference and Test
Likelihood Ratio Chi-square	136 (P < 0.001)	130 (P < 0.001)	Chi-Square = 6.53 P = 0.011
AIC	685.5	692
BIC	695.5	702.1
-2 Log Likelihood	5593	5605
AUC of receiver operating characteristic	0.833 (0.796–0.863)	0.827 (0.794–0.859	0.006 (P = 0.200, DeLong test)
Discrimination slope	0.173	0.160	IDI = 0.014 IDI = 8.75%
NRI			19.57%

AIC: Akaike Information Criterion; Measures of model complexity, weighting the additional information of a model against its entropy; a lower value indicates an increased global fit

BIC: Bayesian information criterion: It is a criterion for model selection, closely associated with AIC. BIC, like AIC, is penalty-based measure, with increased penalty (compared to AIC) for the increasing number of parameters; a lower value indicated a better model fit

AUC: Area Under the Curve using receiver operating characteristic curve statistics; Measures of discrimination; a higher value indicates better discrimination of events vs. non-events

IDI: Integrated Discriminant Improvement; Measures the difference between the discrimination slopes of two models before and after the addition of Lipoprotein (a). Discrimination slope in the binary context is defined as the difference between mean predicted probabilities of events and non-events; a higher value indicates a larger improvement

NRI: Net Reclassification Improvement; Measuring the percent of reclassified subjects, among events and non-events; a higher value indicates better implication of the added marker