Comparison of Conventional Lipoprotein Tests and Apolipoproteins in the Prediction of Cardiovascular Disease: UK Biobank

This prospective UK Biobank analysis in 346,686 participants compared the predictive utility of ApoB, non-HDL-C, and LDL-C for incident cardiovascular disease events, examining concordance, discrimination, and reclassification when these measures disagreed.

All three measures showed similar associations with incident CVD: hazard ratios per 1 SD increase of 1.23 for ApoB, 1.21 for non-HDL-C, and 1.20 for LDL-C — statistically indistinguishable in a general population sample. Adding ApoB to a conventional lipid model (total and HDL cholesterol) produced no meaningful improvement in C-index. Discordance analyses showed that when ApoB and non-HDL-C disagreed, risk tracked the concordant marker rather than ApoB specifically.

This large study — the most powerful population-level dataset to test the ApoB vs. non-HDL-C vs. LDL-C hierarchy — reaches a somewhat different conclusion from statin trial concordance analyses: in an unselected general population, ApoB’s predictive advantage over non-HDL-C is minimal.

The apparent contradiction with meta-analyses showing ApoB superiority (Sniderman et al., 2011; Thanassoulis et al., 2014) may reflect population differences. UK Biobank participants are generally healthier (healthy volunteer selection bias), have lower triglycerides and lower prevalence of insulin resistance, and thus show less ApoB-LDL-C discordance than high-risk clinical populations. In populations where discordance is common — diabetics, high-triglyceride patients, metabolic syndrome — ApoB’s advantage is larger.

This study is important evidence that in a low-to-moderate-risk general population, non-HDL-C may be sufficient as a lipid risk marker. It does not challenge ApoB’s superiority in the specific high-risk, high-discordance populations where ApoB-guided decisions change clinical management.

We rate the evidence strong. A very large UK Biobank analysis demonstrating that ApoB offers no meaningful predictive advantage over non-HDL-C in an unselected general population — important context for understanding when ApoB adds value beyond simpler lipid measures.