Apolipoprotein B versus non-high-density lipoprotein cholesterol: contradictory results in the same journal

This editorial refers to ‘Discordance of apolipoprotein B with low-density lipoprotein cholesterol or non-high-density lipoprotein cholesterol and coronary atherosclerosis’, by X. Su et al., https://doi.org/10.1093/eurjpc/zwac223. LDL cholesterol (LDL-C) remains the premier maker of cardiovascular risk in clinical care, notwithstanding that all the major guidelines accept that apolipoprotein B (apoB) and non-HDL cholesterol (HDL-C) are more accurate indices of risk and more precise guides to therapy than LDL-C. In this issue, Su et al.1 report that apoB was a more accurate predictor of the risk of coronary atherosclerosis, as estimated by coronary computed tomographic (CT) angiography, than LDL-C and non-HDL-C. Discordance analysis was their primary statistical method. Discordance analysis compares markers when their predictions differ and this makes it a more powerful tool than conventional statistical methods to compare the predictive powers of highly correlated variables such as LDL-C, non-HDL-C, and apoB.2 While this is the first report comparing these markers using CT angiography, multiple other reports, also using discordance analysis, have previously demonstrated the superiority of apoB over non-HDL-C.3–9 On the other hand, in the same journal, Helgadottir et al.10 report the results of an extensive Mendelian randomization analysis, which also, in part, applies discordance analysis, which demonstrates that non-HDL-C is a more accurate marker of cardiovascular risk than apoB. As well, they conclude that apoB particles containing more cholesterol are more atherogenic than apoB particles containing less cholesterol. Two articles, the same journal, one pro-apoB, the second anti-apoB. All studies have their strengths and weaknesses. A strength of the study by Su et al.1 is that it involves one of the major peoples of the world. There are a limited number of previous studies comparing these markers in Chinese people, but these also support the superiority of apoB over non-HDL-C.7,11 Another strength is that this is the first report in which coronary CT was the endpoint comparing apoB and non-HDL-C. Other analyses comparing apoB and non-HDL-C have been based on clinical endpoints or coronary calcification. The present study, therefore, significantly extends the evidence in favour of apoB. On the other hand, the weaknesses of the present of the study by Su et al.1 are that the effects of treatment may not be fully accounted for and the differences in the frequency of disease amongst the groups are disconcertingly low. The study by Helgadottir et al.10 has many strengths. The investigators are acknowledged experts and Mendelian randomization is an extraordinarily powerful tool to identify causal relations. Moreover, their results were unequivocal. When the effects of apoB were accounted for, the effects of non-HDL-C remained statistically significant. When the effects of non-HDL-C were accounted for, the effects of apoB were not statistically significant. The results cannot be more clear. Nevertheless, the results are a product of the genetic instruments selected as estimates of apoB and non-HDL-C and are only as valid as the genetic instruments that were created. In their study, a panel of 235 alleles, all of which were identified because they affect the concentration of apoB, were used to estimate non-HDL-C. Given the high correlation (0.9 or >) between apoB and non-HDL-C, the potential limitation of this approach is that an allele identified by its relation to apoB may relate to non-HDL-C only indirectly by virtue of its association with apoB. In fact, the vast majority of the coefficients for apoB and non-HDL-C are almost all identical or virtually identical. Indeed, only ∼20 are substantially different and none is directionally different. The authors note that the principal findings do not change if the major effect variants are excluded. Given the similarity in coefficients, this is puzzling. However, there are other issues. Previous Mendelian randomization analyses have not compared apoB with non-HDL-C directly but have shown that apoB is superior to triglycerides, LDL-C, and HDL-C as a marker of cardiovascular risk.12–15 These findings may be relevant. Horizontal pleiotropy in Mendelian randomization analyses occurs when the biological effect of a marker is related not to the marker itself, but to another marker, whose concentration is highly correlated with the marker selected for study. Thus, the relation between triglycerides and Very Low Density Lipoprotein (VLDL) cholesterol, while not perfect, is highly predictable, and it is generally acknowledged that VLDL cholesterol very likely accounts for whatever atherogenic risk is statistically associated with triglycerides. Thus, by virtue of horizontal pleiotropy, the previous Mendelian randomizations did include the cholesterol, although estimated indirectly, in the apoB particles. Thus, while a weakness of previous Mendelian randomization analyses was that VLDL cholesterol was not estimated directly; conversely, a strength would be that an independent instrument was used to estimate plasma triglycerides. Finally, while their results indicate that cholesterol-enriched apoB particles were more atherogenic than cholesterol-depleted apoB particles, the non-HDL-C/apoB ratio in FOURIER patients, all of whom had coronary artery disease, was 1.45 vs. 1.59 in the UK Biobank. That is, apoB particles were cholesterol-depleted, not cholesterol-enriched, in diseased patients compared with the normal population. This raises a fundamental question in analyses of causality. The strength of Mendelian randomization is that it is less subject to confounding than conventional epidemiological methods, which can only adjust for the effects of recognized confounders but are blind to the effects of those that are unrecognized. By adjusting for confounders, it is presumed that the true relation between a marker and a clinical outcome can be accurately estimated. But what if the final pathophysiological impact of the factor being studied is, at least in part, influenced by the pathophysiological impact of another downstream factor? Mendelian randomization eliminates the possibility of this effect. For example, assuming that cholesterol is the major atherogenic component of an apoB particle, the trapping of a cholesterol-rich apoB particle within the arterial wall should be worse than the trapping of a cholesterol-poor apoB particle. But, if genetic variation increased the trapping of cholesterol-poor apoB particles more than cholesterol-rich apoB particles, this would increase the atherogenic risk associated with apoB compared with non-HDL-C. By eliminating all confounders, a more valid comparison between two markers should be possible. But eliminating all confounders may create an invalid in vivo comparison. This appears to be an example of Lask’s anomaly, which states that ‘systems cannot be viewed simultaneously as wholes made up of parts and as parts made up of wholes: both views are valid, but the observer must choose one at a time’.16 As a scientist, it is the individual parts and their roles in determining the whole that matter most. As a clinician, it is the whole, the interaction of all the parts, not just the individual parts, that matters most. As a proponent of apoB, evidence that contradicts one’s beliefs must not be dismissed or disregarded. Nevertheless, I would make four points. First, Mendelian randomization is a powerful, but not an all-powerful, tool, and the superiority of apoB has been demonstrated by a variety of analytical methods, including Mendelian randomization. Second, if Helgadottir et al. are correct, an individual with a normal LDL-C but a high apoB is not at increased cardiovascular risk, whereas a patient with a high LDL-C but a normal apoB is. But this does not square with the indisputable reality that cholesterol-depleted apoB particles are much more common in patients with arteriosclerotic cardiovascular disease than cholesterol-enriched apoB particles. Moreover, their conclusions stand at variance with many other reports, which also should not be dismissed. Third, apoB is measured more accurately and precisely than LDL-C and non-HDL-C. Therefore, as a practical clinical tool to judge the risk and the adequacy of therapy, apoB is superior to LDL-C and non-HDL-C. The fourth and final point—the point I am most certain of—is that apoB is not all-informative. There must be factors that influence the entry of apoB particles into the wall and there must be factors that influence their binding to the arterial wall. Learning the names and biological properties of these factors is essential if we are to improve our characterization of those at risk. The need to learn more is imperative. None declared.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.