Why Non-HDL Cholesterol is Preferred over Apolipoprotein B-100 (Apo B)

Abstract

IMPORTANCE Most studies have found that apo B-100 is a superior marker for Coronary risk (ASCVD) than non-HDL cholesterol (C). Usually, studies use multivariant analysis to compare indexes with single-point odds or risk ratios. In multivariant analysis when variables are highly correlated, they are difficult to interpret and the lesser may be excluded. As a result, effect sizes cannot be well compared. Receiver operator characteristic (ROC) curves provide a visual portrait of the accuracy and the diagnostic sensitivity and specificity at each decision level so that relative discrimination of each variable can be well compared. Since non-HDLC has distinct economic value, it is important to compare clinical value in an appropriate format. OBJECTIVE To compare outcomes from ROC analysis with routine one-point logistic regression. DESIGN, SETTING, AND PARTICIPANTS Lipoprotein variables alone and after correction for non-lipoprotein risk factors were compared from patients with and without significant ASCVD undergoing coronary angiography.  MAIN OUTCOME MEASURES The variables were assessed by standard logistic regression alone and by ROC curve analysis. RESULTS Although non-HDLC and apo B were stronger markers than LDLC, when examined by logistic regression, as a result of very strong collinearity, non-HDLC appeared weaker than LDLC in the presence of apo B, based on p-values. This was true when analyzed with and without non-lipid risk factors. When analyzed by ROC analysis, apo B and non-HDLC showed stronger C-statistics than LDLC and total C. At an appropriate apolipoprotein/lipid, decision level apo B showed about 6.1% greater specificity than non-HDLC. But, after adjustment for non-lipid risk factors, the c-statistics for apo B and non-HDLC were 0.64 and 0.63, respectively and there was little difference in specificity at a standard selected decision value. CONCLUSION AND RELEVANCE Except for persons with acquired or genetically determined hypercholesterolemia, the ten-year risk is calculated from an algorithm that includes non-lipid risk factors similar to those examined here. Based on this data, when assessed by the AHA/ACC ten-year screening algorithm, it is likely that non-HDLC would provide greater economic value than would apo B with similar clinical efficacy. Non-HDLC should be utilized as the preferred lipid marker.