ApoB: The Cholesterol Number Your Panel Probably Didn't Print

The full story

Your last lipid panel printed four numbers: total cholesterol, HDL cholesterol, LDL cholesterol, triglycerides. That is the standard panel, as the American Heart Association describes it. The AHA also describes apolipoprotein B as part of an "advanced lipid panel" - a polite way of saying it is not on the default order set and nobody will run it unless you ask.

The case for asking is mechanical. Every atherogenic lipoprotein particle - chylomicron remnant, VLDL, LDL - carries exactly one apoB molecule. Measure apoB and you have counted the particles. Measure LDL cholesterol and you have weighed the cargo inside a subset of them, and the amount of cargo per particle varies from person to person. A review in the Journal of the American Heart Association makes the point that this variability is precisely what creates discordance between the two numbers, and notes that in 2019 the European Society of Cardiology and European Atherosclerosis Society stated that apoB is a more accurate marker of cardiovascular risk than either LDL-C or non-HDL-C.

What the genetics do to the argument

Mechanism is cheap. The reason apoB gets taken seriously is that when you interrogate the genome, the cholesterol measurements stop being significant and apoB does not.

A Mendelian-randomisation study in JAMA analysed 654,783 participants including 91,129 coronary heart disease cases, and compared two entirely different routes to lower lipids: triglyceride-lowering LPL variants and LDL-C-lowering LDLR variants. Per 10 mg/dL lower apoB, the two produced almost identical reductions in coronary risk - odds ratios of 0.771 and 0.773. Then the analysis adjusted for apoB, and the associations of triglycerides and LDL-C with coronary disease went to null (odds ratios 1.014 and 1.010, both P=0.19), while apoB survived intact at 0.761. The genetic route mattered less than the number of particles it removed.

A multivariable Mendelian-randomisation analysis in PLoS Medicine, using UK Biobank data against 60,801 coronary cases and 123,504 controls, found the same shape. In single-exposure models everything looked dangerous: LDL-C at an odds ratio of 1.66 per standard deviation, triglycerides at 1.34, apoB at 1.73. Put all three in together and only apoB kept a robust effect (1.92, confidence interval 1.31 to 2.81), while the LDL-C estimate reversed direction entirely (0.85, 0.57 to 1.27). A reversal like that is not evidence that LDL cholesterol is harmless. It is the statistical signature of two exposures measuring overlapping things, with one of them the better proxy for the thing that matters.

Discordance: when the two numbers disagree

Discordance is where the abstraction becomes a clinical event. In JAMA Cardiology, an analysis spanning 389,529 primary-prevention participants in UK Biobank plus 40,430 statin-treated patients from the FOURIER and IMPROVE-IT trials found that apoB, non-HDL-C and triglycerides each predicted incident myocardial infarction when assessed on their own - and that when they were assessed together, only apoB remained associated, at an adjusted hazard ratio of 1.27 per standard deviation. Once apoB was accounted for, the triglyceride-to-LDL-C ratio, a popular proxy for particle quality, was no longer associated with infarction at all.

The most concrete illustration is in the Journal of the American College of Cardiology: 13,015 statin-treated adults in the Copenhagen General Population Study, followed a median of eight years. High apoB and high non-HDL-C predicted all-cause mortality and myocardial infarction. High LDL-C did not. And in the discordant cases - high apoB alongside low LDL-C - the hazard ratios were 1.21 for all-cause mortality and 1.49 for infarction, while the mirror image, high LDL-C with low apoB, carried no excess risk whatsoever. When the two numbers disagree, this is the evidence that the particle count is the one telling the truth.

Where they are most likely to disagree

A review in Current Cardiology Reports is specific about it. ApoB is a direct measure of the circulating number of atherogenic lipoproteins, and the assay can be standardised across laboratories worldwide - so a result from one lab is comparable with a result from another. Discordance analysis makes apoB the more accurate risk marker, and it is superior to non-HDL-C in particular in mild-to-moderate hypertriglyceridaemia (175 to 880 mg/dL), in diabetes, obesity or metabolic syndrome, and when LDL-C is below 70 mg/dL. That last condition is the one that catches people already on treatment: the lower LDL-C goes, the less it has left to tell you.

There is a ceiling on the substitution, though, and it runs the other way. In 4347 lipid-clinic patients with pre-treatment fasting profiles, reported in the Journal of Clinical Lipidology, the correlation between non-HDL-C and apoB fell progressively as triglycerides climbed, and above roughly 4.0 to 5.0 mmol/L (about 354 to 443 mg/dL) non-HDL-C kept rising while apoB plateaued - producing divergent risk categorisation. In severe hypertriglyceridaemia the two are not interchangeable in either direction.

What the evidence does not show

It does not show that adding apoB to a risk calculator finds people the calculator was missing. In MESA, among participants aged 45 to 84 free of atherosclerotic disease and not taking lipid-lowering drugs, higher apoB was associated with coronary artery calcium prevalence, incidence and progression, and discordantly high apoB tracked calcium progression - but apoB added only modest predictive value for calcium beyond LDL-C or non-HDL-C. A prospective cohort in the American Journal of Cardiology, 7117 participants in the China Health and Nutrition Survey with 207 cardiovascular events over six years, found discordantly high apoB associated with higher risk (odds ratios of 1.38 and 1.40 against LDL-C and non-HDL-C respectively) and then reported what adding apoB did to the discrimination of the national risk score: it moved the area under the curve from 0.788 to 0.790. That is a rounding error wearing a lab coat.

Both things are true at once, and holding them together is the whole skill. ApoB can be the variable carrying the causal signal and still add almost nothing to a population risk model, because in most people the cheaper numbers correlate with it closely enough. It is the minority in whom they do not correlate where a measurement earns its keep. A National Lipid Association expert panel consensus statement put apoB in exactly that box in 2011, alongside LDL particle concentration, hs-CRP, Lp-PLA2, Lp(a) and lipoprotein subfractions: adjuncts for refining risk assessment or deciding whether to intensify treatment, not first-line screening.

How to actually get the number

• Ask for it by name. "ApoB" or "apolipoprotein B" - it sits on an advanced lipid panel, not on the standard four, whether you are ordering through a clinician or a mail-in service.
• Order it with the standard panel, not instead of it. Discordance is only visible if you hold both numbers at once, and you need the triglycerides to know how much weight to put on non-HDL-C.
• Expect a portable result. Because the assay can be standardised across laboratories, an apoB value travels between labs better than most of the markers sold as advanced.
• Know when it is most informative. Mild-to-moderate hypertriglyceridaemia, diabetes, obesity or metabolic syndrome, or an LDL-C already under 70 mg/dL. If your triglycerides are severely elevated, apoB and non-HDL-C part company and should not be swapped for one another.

What apoB is not is a verdict. It is a sharper version of a number you already have, and the honest framing is the one the lipid specialists used: an adjunct that changes the picture in the subset of people whose cheap numbers are lying to them, and confirms the picture in everybody else. That is still worth one line on a requisition form. It is not worth a quarterly ritual, and anybody selling you the number as a life expectancy is selling you something the evidence has not agreed to.