Table 3.

The first three columns show Bonferroni corrected p-values (across three tests) for each of the tested metrics. For those labs that met the above criteria we assessed at what point the cases differentiated themselves from the controls. Using the validation data the c-statistic for predicting MI.

Lab testa	LRT Overall	LRT Among cases	LRT Among controls	Optimal fit	Misclassification Rateb	c-Statistic
Albumina	<0.001	<0.001	0.084	28 days	0.384b	0.635b
Calcium	0.072	0.210	1.000	—	0.468	0.551
CO2	0.964	1.000	1.000	—	0.474	0.547
Creatinine	<0.001	0.187	0.303	—	0.473	0.546
Ferritin	0.156	0.003	1.000	—	0.468	0.550
Hemoglobina	<0.001	<0.001	0.425	28 days	0.430b	0.591b
Iron transferring saturationa	<0.001	0.024	0.703	168 days	0.459b	0.572b
Phosphorousa	0.033	<0.001	1.000	14 days	0.459	0.546
Platelet counta	<0.001	0.012	1.000	14 days	0.470	0.556
Potassium	1.000	0.240	1.000	—	0.475	0.541
White blood cell counta	<0.001	<0.001	0.933	56 days	0.434b	0.588b

^aMet all three analytic criterion.

^bSignificant (p <0.05) improvement over model with just demographic factors (base misclassification rate=0.470; c=0.541).