Computationally estimated apolipoproteins B and A1 in predicting cardiovascular risk

OBJECTIVE: Apolipoproteins B (apoB) and A1 (apoA1) may be better markers of atherosclerosis than serum lipids. We used computational methods to estimate apoB and apoA1 from serum total cholesterol, HDL-cholesterol and triglycerides and tested their clinical value in comparison to measured apoB and apoA1 values. METHODS: ApoB and apoA1 were measured with standard methods and estimated based on neural network regression models in 2166 young adult with data on carotid artery intima-media thickness (cIMT). RESULTS: Correlations between estimated and measured apoB and apoA1 were r = 0.98 and r = 0.95, respectively. ApoB/apoA1-ratio (both measured and estimated) associated with cIMT in multivariable models, and predicted cIMT at all levels of LDL-cholesterol concentration. Strong correlations between the estimated apolipoproteins and those measured from fasting samples were replicated in over 15,000 Caucasian subjects (r = 0.93-0.96 for apoB and r = 0.91-0.92 for apoA1). Correlations with cIMT were replicated in over 2000 individuals. Estimated apoB/apoA1-ratio calculated from non-fasting lipids in over 20,000 individuals in the INTERHEART study was better than any of the cholesterol measures for estimation of the myocardial risk. CONCLUSIONS: Serum cholesterol, HDL-cholesterol and triglycerides can be used to compute clinically useful estimates of apoB and apoA1. Using this methodology, estimates of apolipoproteins could be routinely added to laboratory reports to complement lipoprotein lipids in risk assessment.