Guidelines recommend both ticagrelor and prasugrel as P2Y12 inhibitors for patients with acute coronary syndrome (ACS) undergoing percutaneous coronary intervention (PCI), but direct comparative evidence remains uncertain and guideline recommendations differ. This multinational observational study aimed to compare the real-world comparative effectiveness and safety of ticagrelor versus prasugrel in adults with ACS treated with first PCI, focusing on 1-year clinical outcomes.
The investigators conducted a retrospective new-user cohort study across seven claims and electronic health record databases spanning multiple countries. Prespecified diagnostics (covariate balance, empirical equipoise, and systematic error assessment) were used to determine which databases were suitable for inclusion in the primary meta-analysis. Database-specific results were combined using a Bayesian random-effects meta-analysis when diagnostic criteria were met.
Adults with ACS undergoing first PCI who initiated either ticagrelor or prasugrel were eligible. Patients with prior major ischemic or hemorrhagic events or those on oral anticoagulants were excluded. The primary outcome was 1-year major adverse cardiovascular events (MACE), defined as the composite of all-cause mortality, acute myocardial infarction (AMI), or stroke. Secondary outcomes included net adverse clinical events (NACE) and the individual components of MACE, as well as ischemic and hemorrhagic events and gastrointestinal bleeding.
Propensity scores were developed using large-scale L1-regularized logistic regression to control for confounding and were applied through stratification. Prespecified diagnostics assessed covariate balance between treatment groups, empirical equipoise across treated populations, and potential systematic error in each database; only databases meeting these criteria contributed to the primary pooled analysis. Database-specific hazard ratios (HRs) for time-to-event outcomes were estimated and then synthesized using a Bayesian random-effects meta-analysis approach to account for between-database heterogeneity.
Of the seven participating databases, three met the prespecified diagnostic criteria and were included in the primary meta-analysis. These three comprised 133,718 patients in total: 109,639 initiators of ticagrelor and 24,079 initiators of prasugrel. In the pooled analysis, the hazard ratio for 1-year MACE comparing ticagrelor with prasugrel was 1.28 (95% credible interval [CrI], 0.89–1.88). Point estimates for secondary and component outcomes likewise did not show statistically credible differences: NACE (HR 1.23, CrI 0.88–1.75), all-cause mortality (HR 1.17, CrI 0.78–1.77), cardiovascular mortality (HR 1.23, CrI 0.81–1.87), ischemic events (HR 1.28, CrI 0.88–1.90), hemorrhagic events (HR 1.01, CrI 0.72–1.39), AMI (HR 1.30, CrI 0.88–1.94), stroke (HR 1.09, CrI 0.73–1.58), and gastrointestinal bleeding (HR 1.04, CrI 0.77–1.41).
The study reports that sensitivity analyses using alternative definitions of time at risk and an alternative propensity score approach (matching) produced consistent findings with the primary analysis. A post hoc meta-analysis restricted to the two U.S. commercial claims databases produced a pooled HR for 1-year MACE of 1.49 (95% CrI 1.05–2.10), indicating a statistically credible higher risk with ticagrelor in that restricted sample. The authors highlight these contrasting aggregated signals and emphasize the influence of database context on pooled effect estimates.
Substantial between-database heterogeneity was observed in the pooled analyses. Only three of seven databases met prespecified diagnostics for inclusion in the primary meta-analysis, which underscores variability in data sources, patient populations, coding, and other contextual factors that can affect comparative-effectiveness estimates. The study is observational and retrospective, with limitations inherent to nonrandomized designs including potential residual confounding despite large-scale propensity adjustment. The report is a preprint and has not been peer reviewed; the authors explicitly note the need for further investigation into context-specific comparative effectiveness and safety before drawing definitive practice-changing conclusions.
In this prespecified multinational observational study of patients with ACS undergoing PCI, no statistically credible difference in 1-year MACE was observed between ticagrelor and prasugrel in the primary pooled analysis. However, meaningful heterogeneity across databases—and a post hoc signal from U.S. databases—suggests that comparative effectiveness may vary by setting. The findings support the need for further study, ideally including additional data sources and careful consideration of context, before changing guideline recommendations.
Authors reported multiple competing interests, including industry relationships and research support; two study authors are employees and shareholders of Johnson & Johnson. Institutional review boards at participating institutions approved the study. Patient-level data used for the analyses cannot be made publicly available; aggregated results may be available on request subject to data partner approvals. Analytic code is publicly accessible through the study repositories cited in the manuscript. The manuscript is posted as a preprint and has not undergone peer review.