Does lowering LDL cholesterol reduce coronary heart disease?
An evidence brief for a non-specialist decision-maker
Prepared for a client medical team: clinically literate readers who need the strength and limits of the causal evidence. This is not a methods tutorial; the technical detail sits behind the links.
The question and why it matters
Blood LDL cholesterol is strongly associated with CHD in observational studies, but association alone cannot justify intervention: something else (diet, deprivation, undiagnosed illness) could drive both. The decision-relevant question is causal: would lowering LDL actually reduce CHD events, and by how much? That determines whether LDL-lowering is worth prioritising, and how much benefit to expect per unit of LDL reduction.
What the evidence says
Three independent lines of evidence converge, which is what makes the conclusion robust:
- Genetic (Mendelian randomisation). Variants that happen to lower lifelong LDL also lower CHD risk, in proportion to their effect on LDL. Because these variants are allocated essentially at random at conception, they are not confounded by lifestyle or reverse causation. A summary-data MR analysis estimated a 30% LDL reduction → ~67% lower CHD risk (95% CI 54–76%) (Burgess, Butterworth, and Thompson 2013).
- Randomised trials. Statins, ezetimibe and PCSK9 inhibitors each lower LDL and each reduce CHD events, with benefit tracking the size of the LDL reduction: the intervention test of the same hypothesis.
- Long-term epidemiology. Cumulative LDL exposure over a lifetime predicts CHD, consistent with a dose-and-duration effect.
The European Atherosclerosis Society consensus panel reviewed all three strands and concluded LDL is a cause of atherosclerotic cardiovascular disease, not just a marker (Ference et al. 2017).
How certain is this?
| Dimension | Assessment |
|---|---|
| Consistency across methods | High: genetic, trial, and epidemiological evidence agree in direction and roughly in magnitude |
| Risk of confounding | Low for the genetic evidence (random allocation of variants); controlled by design in trials |
| Main threat to validity | Genetic pleiotropy (a variant affecting CHD via a route other than LDL), addressed by sensitivity analyses (MR-Egger, weighted median) |
| Directness to the decision | See caveats: the estimand is lifelong exposure, not a short-term intervention |
Overall certainty: high for the direction of effect (LDL-lowering reduces CHD), moderate for any single point estimate of magnitude, since the genetic and trial numbers answer subtly different questions. For a clinical team, this supports acting on LDL with confidence while sizing expected benefit to the specific intervention and patient.
Caveats and what would change the answer
- The genetic estimate answers “does lifelong lower LDL mean less CHD?”, not “what will drug X deliver over 5 years?”. Match the number to the decision.
- Benefit scales with absolute LDL reduction and duration; small or brief reductions yield proportionally less.
- Evidence is strongest for CHD/atherosclerotic disease; do not extrapolate uncritically to other outcomes.
- MR assumes the instruments affect CHD only through LDL. The conclusion would weaken if that assumption failed, which is why the sensitivity analyses matter (see the methods note).
Bottom line
Lowering LDL cholesterol causally reduces CHD risk. Treat LDL as a real target, size the expected benefit to the absolute reduction achieved and how long it is sustained, and be explicit about whether a decision concerns lifelong risk or a time-limited intervention.
Methods & sources
This brief summarises published evidence; it is a communication piece, not a new analysis. The Mendelian randomisation methods behind the genetic estimate (including the two-stage least squares, inverse-variance weighting, and the MR-Egger / weighted-median sensitivity analyses referenced above) are worked through from scratch, with code, in a companion walkthrough: Mendelian Randomisation from Scratch.
References