Enhancing statistical power in lipid-lowering therapy studies through optimization of the MACE endpoint composition: a model-based meta-analysis approach
Alina Volkova (1,2), Boris Shulgin (3), Cristina Leon (1), Gabriel Helmlinger (4), Kirill Peskov (1,2,3), Victor Sokolov (1,2)
(1) Modeling and Simulation Decisions FZ - LLC, Dubai, UAE, (2) Marchuk Institute of Numerical Mathematics, Moscow, Russia, (3) Research Center of Model-Informed Drug Development, Sechenov First Moscow State Medical University, Moscow, Russia, (4) Biorchestra Co., Ltd., Cambridge MA, USA
Introduction: Major adverse cardiac events (MACE) are used to assess the efficacy and safety of cholesterol-lowering compounds in late-stage clinical trials, with varying definitions and composite determinations of the MACE endpoint across studies [1]. While composite endpoints such as MACE may improve statistical efficiency, they also come with limitations such as increased complexity and potential masking of treatment benefits. A careful balance of statistical efficiency, clinical relevance, and component compatibility is essential for achieving optimal patient benefits in clinical trials.
Objectives: This study aimed at developing a model-informed methodology for the optimization of the composite MACE endpoint, based on a model-based meta-analysis performed across anti-hypercholesterolemia trials of statin as well as anti-proprotein convertase subtilisin/kexin type 9 (PCSK9) therapies.
Methods: Mixed-effects meta-regression modeling was implemented to analyze standalone MACE outcomes, incorporating predictors such as therapy type, population demographics, baseline and change over time in lipid biomarkers. Randomized clinical trials of statins and anti-PCSK9 therapies up to June 28, 2022 were identified through a systematic review and curation process in PubMed and ClinicalTrials.gov databases, following PRISMA guidelines [2]. Next, model-predicted averages of the effect size for a composite outcome and a population of interest, along with event frequencies in the control group, uncertainty in model parameters and pre-defined trial durations were used to calculate sample size of clinical studies required to achieve statistical significance of treatment benefit relative to the reference group.
Results: In total, 54 studies including 270,471 patients were collected, reporting 15 different single cardiovascular events, e.g. mortality from different causes, stroke and its subtypes, myocardial infarction and its subtypes, and others. Treatment-mediated decrease in low-density lipoprotein cholesterol (LDLc), baseline levels of remnant and high-density lipoprotein cholesterol (HDLc) as well as non-lipid population characteristics and type of therapy were identified as significant covariates for 10 of the 15 outcomes. Following meta-regression modeling, different compositions of 3- and 4-point MACE for statins and anti-PCSK9 therapies were compared, to select one with minimal population size. For the commonly used 3-point MACE, which consists of non-fatal myocardial infarction and stroke and cardiovascular mortality, the recommended population size at -30 mg/dL LDLc reduction was determined as 3515, for patients with low baseline HDLc. Contrary to common belief, it was found that the inclusion of additional outcomes did not necessarily lead to an increase in statistical power. Thus, introducing frequently occurring coronary revascularization in MACE composition was shown to significantly reduce sample size by more than half, while considering unstable angina and heart failure did not provide significant benefits in enrollment requirements.
Conclusions: A quantitative tool was developed to optimize the composite MACE endpoint following statin and anti-PCSK9 therapies, by minimizing the sample size required to achieve a statistically significant therapeutic effect, following meta-regression modeling of single MACE components. This approach can be applied to improve the design of future clinical studies in dyslipidemia as well as other cardiovascular research areas.
References:
[1] Armstrong, P. W., and Westerhout, C. M. (2017). Composite End Points in Clinical Research: A Time for Reappraisal. Circulation 135, 2299–2307. doi: 10.1161/CIRCULATIONAHA.117.026229.
[2] Moher, D., Liberati, A., Tetzlaff, J., Altman, D. G., and The PRISMA Group (2009). Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Medicine 6, e1000097. doi: 10.1371/journal.pmed.1000097.
