Discordance Analysis of VLDL-C and ApoB in UK Biobank and Framingham Study: A Prospective Observational Study.

BACKGROUND: Recent observational and Mendelian randomization analyses have reported significant effects of VLDL-C (very-low density lipoprotein cholesterol) on risk that is independent of ApoB (apolipoprotein B). We aim to determine the independent association of VLDL-C and ApoB with the risk of new onset cardiovascular events in the UK Biobank and Framingham Heart Study cohorts. METHODS: We included 294 289 UK Biobank participants with a median age of 56 years, 42% men, and 2865 Framingham Heart Study participants (median age, 53 years; 47% men). The residual resulting from regressing VLDL-C on ApoB expresses the portion of VLDL-C not explained by ApoB, while the residual from regressing ApoB on VLDL-C expresses the portion of ApoB not explained by VLDL-C. Cox proportional hazards models for atherosclerotic cardiovascular disease incidence were created for residual VLDL-C and residual ApoB. Models were analyzed with and without high-density lipoprotein cholesterol (HDL-C). Furthermore, we investigated the independent effects of VLDL-C after accounting for ApoB and HDL-C and of HDL-C after accounting for ApoB and VLDL-C. RESULTS: In the UK Biobank, ApoB was highly correlated with VLDL-C (r=0.70; P<0.001) but weakly negatively correlated with HDL-C (r=-0.11; P<0.001). The ApoB residual and the VLDL-C residual were significantly associated with new-onset atherosclerotic cardiovascular disease (hazard ratio [HR], 1.08 and 1.05, respectively; P<0.001). After adjusting for HDL-C, the ApoB residual remained similar in magnitude (HR, 1.10; P<0.001), whereas the effect size of the VLDL-C residual was reduced (HR, 1.02; P=0.029). The independent effect of HDL-C (after accounting for ApoB and VLDL-C) remained robust (HR, 0.86; P<0.0001), while the independent effect of VLDL-C (after accounting for ApoB and HDL-C) was modest (HR, 1.02; P=0.029). All results were consistent in the Framingham cohort. CONCLUSIONS: When adjusted for HDL-C, the association of VLDL-C with cardiovascular risk was no longer clinically meaningful. Our residual discordance analysis suggests that adjustment for HDL-C cannot be ignored.

Duke Scholars

Author Michael J Pencina Biostatistics & Bioinformatics, Division of Biostatistics