NMR-based prediction of cardiovascular risk in diabetes

Multivariate analysis of 1H-NMR spectra of blood sera was reported previously to predict angiographically defined advanced coronary artery disease (CAD) with90% accuracy and specificity. The analysis depended mainly on the major lipid regions of the spectra, but many variables, including gender and drug treatment, affect lipid composition and are potential confounders. We have determined the predictive power of the same methodology for angiographically defined CAD using plasma samples from groups of male patients, classified by statin treatment, who had normal coronary arteries (NCAs) or CAD. Predictions for NCA and CAD groups were only 80.3% correct for patients not treated with statins and 61.3% for treated patients, compared with random correct predictions of 50%. A confidence limit of99% was achieved for 36.2% of predictions for untreated groups and 6.2% for treated groups. Detection of CAD by 1H-NMR with99% confidence was therefore very weak compared with angiography.