The narrow, live disagreement — NOT apoB-vs-LDL-C. Both camps agree apoB-containing particles are causal for ASCVD and that apoB / non-HDL-C beat LDL-C as risk markers in the discordant stratum (-> LDL ApoB and Cumulative Exposure). What they clash on is a finer question that only becomes answerable once apoB and non-HDL-C are separated: for two people with the same apoB particle count, does the cholesterol those particles carry (non-HDL-C) still move CAD risk? The held nucleus view (Marston) says no — number is the whole story. An independent MR (Helgadottir) says yes — and that the number, holding content fixed, does not move risk. — this framing (which leg is live, what both camps concede) is the wiki’s; the source-attributed claims sit in the sections below.
Underlying question
apoB (particle number) and non-HDL-C (cholesterol content of all apoB particles) are ~0.9 correlated, so they usually move together and cannot be told apart. The question is which one carries the causal signal when they are forced apart — which is also which secondary target to steer (ESC/EAS 2019 prefers apoB; both beat LDL-C). — the framing of the underlying question is the wiki’s; the high apoB/non-HDL-C correlation is well-established lipid-panel background, not a figure from either source.
The parameter table — is this a real joined issue?
The comparison Marston makes and the comparison Helgadottir makes are the same contrast (apoB particle number vs non-HDL-C cholesterol content), reached by different designs that give opposite verdicts — a genuine type-D clash, not two different questions.
| Parameter | Marston 2022 (number camp) | Helgadottir 2022 (content camp) | Same quantity? |
|---|---|---|---|
| Contrast tested | apoB vs non-HDL-C vs TG for MI | apoB vs non-HDL-C for CAD | YES — both isolate apoB-number vs non-HDL-C-content |
| Design | observational, mutual adjustment: UK Biobank (n=389,529) + 2 statin RCTs (n=40,430) | genetic MR: 235 variants, exploiting 82/235 (35%) with discordant effects on the two traits | NO — the crux of the friction (confounding present vs largely removed) |
| Verdict, apoB held | «only apoB was associated (adjusted hazard ratio [aHR] per 1 SD, 1.27; … 1.15-1.40)»; non-HDL-C fell to NS (Marston et al., 2022) | apoB beta = -0.040 (P=0.69) once non-HDL-C added, non-HDL-C beta = 0.44 (P=3.9x10^-5) once apoB added (Helgadottir et al., 2022) | Opposed verdicts on the SAME contrast |
| Estimand | adjusted HR per SD, observed levels | MR beta / log-OR per SD, genetically-predicted lifelong | different metric, same conceptual target |
| Outcome | myocardial infarction | coronary artery disease | ~same (ASCVD hard events) |
| apoB particles causal? | yes | yes — «The causal contribution of apolipoprotein B (apoB) particles to coronary artery disease (CAD) is established» (Helgadottir et al., 2022) | YES — the agreement floor |
The fourth column is YES on the contrast and on the causal floor; the one NO (design) is why the
verdicts differ, not a sign they answer different questions. Issue joined -> file the tension.
View A — particle NUMBER carries the risk, content does not (Marston 2022; the held nucleus view)
Entering apoB, non-HDL-C and TG simultaneously in a large cohort, only apoB survived. Marston’s conclusion is that particle count is the causal quantity independent of what the particles carry:
«risk of MI was best captured by the number of apoB-containing lipoproteins, independent from lipid content (cholesterol or TG) or type of lipoprotein (LDL or TG-rich). This suggests that apoB may be the primary driver of atherosclerosis and that lowering the concentration of all apoB-containing lipoproteins should be the focus of therapeutic strategies.» (Marston et al., 2022)
This is the held view on LDL ApoB and Cumulative Exposure and LDL Lowering and Cardiovascular Events (particle NUMBER, not TYPE or CONTENT, carries the risk), corroborated in the number camp by Richardson’s and Zuber’s multivariable MR that apoB beats LDL-C.
View B — cholesterol CONTENT carries the risk, number does not (Helgadottir 2022)
Helgadottir separates the two traits with genetics rather than statistical adjustment. 82 of 235 apoB variants have discordant effects on non-HDL-C vs apoB (different cholesterol mass per particle), splitting into non-HDL-C-main-effect (N=47) and apoB-main-effect (N=35) groups. Regressing each group’s CAD effect on its apoB effect gives the load-bearing proportionality test:
«the increase in log(OR) per SD change in apoB was 71% greater for non-HDL-C main- effect variants than apoB main-effect variants, showing that CAD risk conferred by apoB is dependent on the associated cholesterol amount. In contrast, non-HDL-C effects did not associate differently with CAD risk between the main-effect groups (P =0.56)» (Helgadottir et al., 2022)
So CAD risk tracks non-HDL-C consistently, but the risk per apoB particle changes with how much cholesterol the particle carries — the signature of content, not count, being causal:
«for individuals with equal levels of non-HDL-C, the number of apoB particles it is carried on does not influence the development of CAD.» (Helgadottir et al., 2022)
Helgadottir directly attributes the observational apoB signal (View A) to confounding — the explicit joining of the issue:
«Some epidemiological studies indicate that among people with the same non-HDL-C levels, the ones with higher apoB particle count would be at greater CAD risk. However, our study indicates that higher observed risk would not be due to apoB par ticle concentration, but because of confounding with other risk factors.» (Helgadottir et al., 2022)
Independence is clean: deCODE (Iceland) + Danish DBDS, with no Ference, Ala-Korpela, Marston, Zuber, or Richardson among the authors — so this is a genuinely outside-the-programme dissent from the apoB-particle-number camp, which is what gives the tension its weight (it is not a same-school artifact).
Hidden insight
— this section is the wiki’s own reasoning across the two sources; no single source states it. The robust, multi-source result — apoB/non-HDL-C beat LDL-C — is untouched. What this exposes is that the finer claim layered on top of it, that particle NUMBER carries risk independent of cholesterol CONTENT, rests on observational mutual-adjustment of two ~0.9-correlated traits, which cannot separate them cleanly. A design built to break the correlation — genetic variants with discordant effects — flips the verdict onto cholesterol content. This is the Surrogate Outcomes / MR-vs- observational discriminator biting inside the apoB story: co-adjusting two collinear candidates is not the same as identifying which is causal, and the wiki should hold the number-vs-content leg as contested, not settled.
Decision relevance
- The apoB-vs-LDL-C decision is unaffected. Both camps say apoB / non-HDL-C beat LDL-C in the metabolically-impaired stratum; the measure-beyond-LDL-C rule on LDL ApoB and Cumulative Exposure stands.
- The secondary-target choice (apoB vs non-HDL-C) is now contested. Helgadottir argues non-HDL-C — free, on every standard lipid panel (total cholesterol minus HDL-C) — captures the causal signal at least as well as a separate apoB assay: «the guidance of non-HDL-C target levels are expected to better capture risk related to apoB-containing particles» (Helgadottir et al., 2022). If it holds, the extra apoB draw buys little over non-HDL-C. This does not overturn the ESC/EAS apoB preference — it is one MR against guidance plus a lineage — but it removes the «content is irrelevant» certainty and makes non-HDL-C a defensible target.
- The therapy-monitoring corollary: Helgadottir’s «clinical benefit … proportional to the reduction in non-HDL-C, but not necessarily proportional to the reduction in apoB» (Helgadottir et al., 2022) predicts that a therapy which lowers apoB without proportionally lowering cholesterol content would under-deliver — a testable, decision-relevant divergence from the number-camp expectation.
Status and what would resolve it
Single MR (however strong the design) vs a large observational analysis backed by the apoB-particle-number
programme — held unresolved, confidence: medium. The clairvoyant-testable resolver: an independent
replication of the discordant-variant proportionality test (does CAD risk track non-HDL-C but not apoB
particle number when the two are genetically separated?), and/or a trial contrast of an apoB-lowering
agent whose non-HDL-C effect is disproportionately small — whichever way the outcome moves settles which
quantity to target. Until then the number-vs-content leg is a filed contest, and the broader apoB causal
verdict stands.
(inferred from Helgadottir et al., 2022; Marston et al., 2022)