Network meta-analysis ranks lipid-lowering drugs for primary cardiovascular prevention by risk level

A network meta-analysis pooling 29 randomised controlled trials and 218,657 participants has compared lipid-lowering drug classes head to head for the prevention of first cardiovascular events, and the answer it returns depends sharply on where a patient sits on the cardiovascular risk spectrum. The analysis, published in eClinicalMedicine, is the first to bring contemporary lipid-lowering therapies, including PCSK9 inhibitors and bempedoic acid, into a single comparative framework restricted to people without a prior coronary event, and to separate that population by baseline atherosclerotic cardiovascular disease risk.

The researchers searched MEDLINE and Embase from database inception to 25 July 2026 for randomised trials of lipid-lowering therapy in adults without a prior myocardial infarction or acute coronary syndrome. From 13,952 records initially identified, 505 full-text articles were assessed and 29 trials were included. Each trial reported major adverse cardiovascular events as a 2-point, 3-point or 4-point composite endpoint, and because the components of these composites varied between trials, the team assigned each to standardised definition groups. Effect estimates were used exactly as published for 17 trials, while equivalent hazard ratios were derived from summary statistics for the remaining 12, including five trials reconstructed with the log-rank observed-minus-expected method of Parmar and colleagues.

The treatment network included no lipid-lowering therapy as the reference, low-, moderate- and high-intensity statins graded according to the 2018 AHA/ACC guideline, ezetimibe, fibrates, omega-3 fatty acids, cholestyramine, bempedoic acid, and combination nodes pairing statins with eicosapentaenoic acid, evolocumab, ezetimibe or fibrates. A random-effects frequentist network meta-analysis was performed in R with the netmeta package, using the DerSimonian and Laird method for between-trial variance and hazard ratios on the logarithmic scale. Notably, no network contained a closed loop, meaning coherence between direct and indirect evidence could not be tested statistically, and every estimate rested on indirect pathways.

For the 3-point MACE network, which covered 15 trials and 127,399 participants, evolocumab added to a statin showed the largest reduction against no lipid-lowering therapy, with a hazard ratio of 0.500 (95% confidence interval 0.367 to 0.682, p < 0.001). Critically, this combination also retained a significant effect against its own statin comparator in direct within-trial contrasts from VESALIUS-CV and FOURIER, with a hazard ratio of 0.737 (0.653 to 0.832). Bempedoic acid (0.640, 0.442 to 0.926), EPA plus statin (0.657, 0.443 to 0.976), high-intensity statin (0.686, 0.549 to 0.857) and moderate-intensity statin (0.811, 0.691 to 0.952) also reached significance, while fibrates and omega-3 fatty acids did not.

Stratification by baseline risk changed the picture substantially. Among the high ASCVD risk group, drawn from eight trials and 43,614 participants, evolocumab plus statin remained the top-ranked intervention (hazard ratio 0.590, 0.489 to 0.712), with bempedoic acid (0.640, 0.484 to 0.847) and high-intensity statin (0.800, 0.693 to 0.924) also significant. EPA plus statin, fibrates, moderate-intensity statin and omega-3 fatty acids were not. In low-to-moderate risk populations, covering seven trials and 83,785 participants, high-intensity statin ranked first with a hazard ratio of 0.530 (0.374 to 0.750), followed by moderate-intensity statin at 0.760 (0.637 to 0.906), while omega-3 fatty acids remained non-significant.

The authors caution that the apparently larger relative benefit of high-intensity statin at low-to-moderate risk compared with high risk reflects the different trials contributing to each node rather than true effect modification. The low-to-moderate risk estimate derives from JUPITER, which enrolled participants based on elevated high-sensitivity C-reactive protein and was terminated early for benefit, a pattern that tends to inflate effect sizes and may not represent unselected populations. The high-risk node instead draws on SPARCL. Between-trial variance in the high-risk stratum was estimated as zero, informed by only one of seven comparisons, so the narrower confidence intervals there should not be read as stronger underlying evidence.

In the 2-point MACE network of ten treatment groups, evolocumab plus statin (0.474, 0.287 to 0.785), high-intensity statin (0.650, 0.439 to 0.962) and moderate-intensity statin (0.736, 0.637 to 0.850) were significant against no therapy. Evolocumab plus statin again held its advantage against its statin comparator (0.730, 0.616 to 0.865), whereas EPA plus statin did not differ from low-intensity statin (0.820, 0.551 to 1.221). EPA plus statin, ezetimibe, fibrates, low-intensity statin, cholestyramine and omega-3 fatty acids all missed significance. In the high-risk stratum, all three active treatments were significant, while in low-to-moderate risk populations only moderate-intensity statin reached significance despite EPA plus statin ranking first.

The 4-point MACE network spanned twelve treatment groups, and all treatments except omega-3 fatty acids significantly reduced events against no lipid-lowering therapy. High-intensity statin led monotherapies at 0.560 (0.457 to 0.686), with moderate-intensity statin at 0.654, ezetimibe at 0.660, low-intensity statin at 0.700, bempedoic acid at 0.700 and fibrates at 0.810. Among add-on therapies, only evolocumab plus statin retained significance against its own statin comparator (0.768, 0.663 to 0.891); EPA plus statin did not differ from moderate-intensity statin (0.880, 0.702 to 1.103) and fibrates plus statin did not differ from high-intensity statin (1.040, 0.795 to 1.361). The high-risk stratum of this network was disconnected, so comparisons were synthesised separately by pairwise meta-analysis with no common ranking scale.

A structural feature of the evidence base shapes how these rankings should be read. Add-on therapies reach the no-treatment reference only through a statin comparator, so their network estimates carry the statin effect within them, and a combination node ranking above its own statin anchor is partly an artefact of that construction. The authors make this explicit: neither EPA nor fibrates added to statin therapy conferred significant benefit over their own statin comparator in any network, meaning their apparent effects against no therapy are attributable to the background statin. Omega-3 fatty acids ranked below no lipid-lowering therapy among high-risk participants, with a hazard ratio of 1.05 (0.89 to 1.23). The omega-3 evidence base is also narrowed by the exclusion of the STRENGTH trial, the largest neutral trial of omega-3 carboxylic acids, whose composite could not be mapped to the standard definitions.

Six pre-specified sensitivity analyses required 71 re-estimations across the networks, and the pooled hazard ratio changed in 13 instances. Reattaching every add-on therapy to a statin comparator of the next lower intensity raised the evolocumab estimate by 13 to 18 percent depending on the network, showing sensitivity to how the statin background is graded. Omitting SPARCL or JUPITER moved the 3-point estimates materially, and reclassifying the REPRIEVE trial changed EPA and moderate-intensity statin estimates by 20 percent. Certainty of evidence, assessed with the CINeMA framework, was moderate for only 4 of 27 comparisons, low for 5 and very low for 18. Sixteen trials were judged at low risk of bias, eleven raised some concerns, and two, ALLHAT and EWTOPIA 75, were at high risk overall. The authors conclude that statin intensity remains the first decision in primary prevention, that evolocumab offers a genuine additional reduction at high risk, and that bempedoic acid provides an option where statins are not tolerated, while calling for direct head-to-head trials in primary prevention populations.

Subject of Research: Comparative effectiveness of lipid-lowering drug classes for primary cardiovascular prevention across baseline risk strata

Article Title: Comparative effectiveness of lipid-lowering therapies for primary cardiovascular prevention: a network meta-analysis of randomised controlled trials

Article References: Gnanavelou, M., Ho, G. J. K., Ng, C. H., Tioh, J. K., Tham, E. K. J., Chan, T. Y. H., Yeo, L. Y. Y., Sasikumar, N. A., Yang, H. S., Lee, W. X., Chew, N., Yeung, W., Yeo, K. K., Moon, J. H., Mamas, M. A., Mehta, A., Tanaka, A., Sperling, L., & Huang, W. (2026). Comparative effectiveness of lipid-lowering therapies for primary cardiovascular prevention: a network meta-analysis of randomised controlled trials. eClinicalMedicine, 101, Article 104300. https://doi.org/10.1016/j.eclinm.2026.104300

Image Credits: AI Generated

DOI: 10.1016/j.eclinm.2026.104300

Keywords: network meta-analysis, primary prevention, statins, evolocumab, bempedoic acid, PCSK9 inhibitors, omega-3 fatty acids, MACE, cardiovascular risk, lipid-lowering therapy, eClinicalMedicine, hazard ratio