Five Trials, Five Different Questions: Design Heterogeneity Among Lipoprotein(a)-Lowering Cardiovascular Outcomes Trials After Lp(a)HORIZON.

完整摘要 · 英文原文

Lipoprotein(a) (Lp(a)) is a genetically determined risk factor for atherosclerotic cardiovascular disease (ASCVD), widely regarded as causal on the basis of genetic and epidemiological evidence, for which no approved Lp(a)-lowering therapy exists. In September 2026, Novartis announced that Lp(a)HORIZON, evaluating the antisense oligonucleotide (ASO) pelacarsen and the most clinically advanced of five phase 3 cardiovascular outcomes trials (CVOTs) targeting Lp(a), had not met its primary composite endpoint. Key details, including the magnitude of Lp(a) lowering achieved, the result in the pre-specified higher-Lp(a) subpopulation, the effect estimate, and the safety profile, remain unpublished pending congress presentation. This review compares the design of all five phase 3 Lp(a)-lowering CVOTs, identified through a curated bibliography, a PubMed search for design and rationale publications, and searches of the ClinicalTrials.gov registry, and situates them within the wider landscape of Lp(a)-directed therapy. The four trials still running are not replications of the first. They differ from Lp(a)HORIZON and from one another in mechanism (ASO, small interfering RNA (siRNA), or oral small molecule), dosing interval, Lp(a) eligibility threshold, sample size, stage of disease at which they intervene, and, critically, in the composition of the primary endpoint: the two olpasiran trials use a coronary-only composite that excludes stroke, whereas Lp(a)HORIZON and, by its own nomenclature, MOVE-Lp(a) include it. A coronary-only composite and a stroke-inclusive composite do not estimate the same quantity, so effect estimates from these trials cannot be directly compared without accounting for endpoint composition, population, and follow-up. Five non-exclusive explanations for the neutral result are set out, together with the evidence bearing on each and what would distinguish them. A single neutral trial does not resolve the Lp(a) question, and the readouts expected between 2027 and 2031 will not resolve it jointly unless these design differences are held in view when the results are compared.