Methods for analyzing recurrent and terminal events considering composite and competing endpoints in clinical trials: a practical overview

Q: What is the clinical evidence from this study?

Study design: Review. Population: Cardiovascular trials with recurrent, competing, and terminal time-to-event endpoints. Intervention: Statistical methods for analyzing recurrent and terminal events considering composite and competing endpoints.

Alidan Duoerkongjiang; Antonia Zapf; Paulus Kirchhof; Ann‐Kathrin Ozga

doi:10.1186/s12874-026-03008-7

September 12, 2026BMC Medical Research MethodologyOpen Access

Methods for analyzing recurrent and terminal events considering composite and competing endpoints in clinical trials: a practical overview

View Full Paper

Ask AI

Bookmark

Share

Key result

Trial design requires aligning statistical methods for recurrent and terminal events with the target estimand.

Why the study?

Methods for recurrent time to event analysis considering terminal events are not often applied in clinical studies due to uncertainties about model assumptions and result interpretation.

Key Points

To provide a comprehensive overview comparing model assumptions and interpretations of statistical methods for analyzing recurrent time-to-event and terminal outcomes in clinical trials.
Reviewed counting process models (Andersen-Gill, Prentice-Williams-Peterson, Lin-Wei-Ying-Yang), frailty models, rate models, and generalized pairwise comparisons (win ratio, win odds).
Applied and compared each methodology using data from the EAST-AFNET 4 cardiovascular trial evaluating composite and competing endpoints, providing reproducible R code.
Treatment effects for the composite endpoint favored treatment across all models, ranging from 0.71 for the negative binomial model to 0.91 for win odds (inverse win odds).
Repeated event analyses showed a hazard ratio of approximately 0.75 compared to 0.79 in the primary time-to-first-event analysis, with an event count incidence rate ratio of 0.71.
In pairwise comparisons, the proportion of wins was 20.18% for treatment versus 15.38% for control, with generalized pairwise comparisons showing a proportion in favor of treatment of 0.0674 (95% CI: -0.0635 to 0.196).

PICO

P

Population

Cardiovascular trials with recurrent, competing, and terminal time-to-event endpoints

I

Intervention / Comparator

Statistical methods for analyzing recurrent and terminal events considering composite and competing endpoints

Limitations

Standard composite endpoint models (e.g., Andersen-Gill, LWYY) assume equal clinical relevance across components and may mask heterogeneous or opposing component-specific treatment effects.
Several weighted hazard ratio and weighted composite approaches lack readily available implementation in standard statistical software and require a priori weight specification.
Joint frailty models can face convergence difficulties when analyzing datasets with few events and are generally limited to two recurrent events and one fatal event.
Unweighted composite models (e.g., Andersen-Gill, LWYY, PWP) assume equal clinical relevance of components and can mask opposing component-specific treatment effects.
Many weighted composite endpoint methods (e.g., parametric and non-parametric weighted all-cause hazard ratios, Wei-Lachin approach) are not readily available in standard statistical software.
Joint frailty models may experience convergence difficulties with few events and are complex to extend beyond two recurrent events and one terminal event.
Pairwise comparison methods (e.g., Win Ratio, Generalized Pairwise Comparisons) are sensitive to the prespecified endpoint hierarchy and can be biased by ties or censoring.

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

Provides a decision-guided framework comparing model assumptions, estimands, and software implementations to help clinical trialists select appropriate statistical methods for recurrent, competing, and terminal cardiovascular endpoints.

Abstract

Time to event endpoints like time to myocardial infarction, relapse, or death are often of primary interest in clinical trials. To describe the whole disease process and gain a complete picture of treatment effects, methods for recurrent time to event analysis considering terminal events were proposed, i.e. methods considering all events that occur per individual. However, these methods are not often applied in clinical studies. This might be due to the researchers' uncertainties about underlying model assumptions and interpretation of results. We therefore aim to give a comprehensive overview of methods for the analysis of multiple time to events comparing model assumptions and interpretation. We illustrate the methods using a cardiovascular trial and provide the corresponding R code.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Duoerkongjiang et al. (2026) conducted a review in Cardiovascular trials with recurrent, competing, and terminal time-to-event endpoints. Statistical methods for analyzing recurrent and terminal events considering composite and competing endpoints was evaluated. Statistical method selection for recurrent, competing, and terminal time-to-event endpoints must be guided by the specific scientific question, target estimand, and clinical relevance of components.

synapsesocial.com/papers/6aa7b3850926e14a848b298a https://doi.org/10.1186/s12874-026-03008-7

View Full Paper

Ask AI

Bookmark

Share

Methods for analyzing recurrent and terminal events considering composite and competing endpoints in clinical trials: a practical overview

Key Points

PICO

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Methods for analyzing recurrent and terminal events considering composite and competing endpoints in clinical trials: a practical overview

Key Points

PICO

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading