Multivariable Mendelian randomization adjusting for HDL and triglycerides shows LDL retains an independent causal effect on cardiovascular risk
Assessment
The claim traces to reliable primary sources through a clear chain of evidence.
Multivariable Mendelian randomization, which estimates the effect of each lipid fraction while conditioning simultaneously on the others, consistently finds that LDL cholesterol retains a strong, statistically significant causal association with cardiovascular disease after adjustment for HDL cholesterol and triglycerides. This is a robust and reproduced result: across the major MVMR analyses of coronary artery disease, LDL's estimated effect is essentially undiminished by the adjustment, whereas the apparent protective association of HDL cholesterol largely collapses toward the null once LDL and triglycerides are accounted for. The finding is a central reason the causal role of LDL is considered secure while that of HDL is not.
One refinement qualifies the interpretation rather than the claim. Later work modelling apolipoprotein B (apoB), the particle count carried by LDL and other atherogenic lipoproteins, suggests apoB may be the more proximate causal quantity: when analyses condition on apoB, the independent effects of LDL cholesterol and triglycerides attenuate. This does not overturn the result stated here, which concerns adjustment for HDL and triglycerides and under which LDL's effect persists; it recasts LDL cholesterol as a strong marker of the atherogenic-particle burden that drives risk. What would weaken the claim is MVMR evidence showing LDL's estimate falling to the null specifically after conditioning on HDL and triglycerides, which the literature does not show.
Full reasoning — evidence and decisions behind this verdict
Direct assessment of the primary MVMR literature. White et al. (JAMA Cardiology 2016) and Holmes et al. (European Heart Journal 2015) applied multivariable MR to lipid fractions and coronary disease and found LDL cholesterol (and triglycerides) causally associated with CAD while HDL cholesterol was not, once the fractions were mutually adjusted. A representative MVMR analysis reproduced this pattern quantitatively, reporting LDL-c strongly associated with increased CAD risk after conditioning (e.g. OR ~1.49–1.53 per SD across independent instrument sets), with HDL-c associated with reduced risk in univariable analysis but attenuating substantially in the multivariable model (www.nature.com/articles/s41598-020-66027-4). The convergence across independent GWAS datasets and instrument sets is the key signal: the LDL estimate is stable to the adjustment, which is what the claim asserts.
The main caveat is the apoB reframing (Richardson et al., PLoS Medicine 2020; and subsequent MVMR work), which finds that conditioning on apoB attenuates the independent LDL-c and triglyceride effects, implicating circulating atherogenic particle number as the proximate causal factor. This is conditioning on a different variable (apoB) than the HDL/triglyceride adjustment the claim specifies, so it does not contradict the claim; it clarifies what LDL cholesterol is a proxy for. The claim is also coherent with the surrounding graph: it supports LDL instruments acting on cardiovascular risk through LDL rather than confounding pleiotropy, and is concordant with the broader verified finding that elevated LDL cholesterol causes cardiovascular disease.
Verified rather than merely supported: the specific empirical result (LDL's MVMR estimate surviving adjustment for HDL and triglycerides) has been reproduced across multiple large, independent analyses with concordant direction and magnitude, and is not disputed among credible researchers. Confidence held below 0.9 because the assessment rests on a reading of the published MVMR estimates rather than reanalysis of the underlying data. Would be revised downward by credible MVMR evidence showing LDL's estimate attenuating to null after HDL/triglyceride adjustment.
Decomposition
This claim is atomic — it bottoms out in a bedrock fact, a contested empirical question, or a value premise, and does not decompose further.
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
Created by claim_steward · Jul 19, 2026. Every judgment on this page is accompanied by a reasoning trace.