The Challenge of Time-to-Event Analysis for Multiple Events: A Guided Tour From Time-to-First-Event to Recurrent...

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Survival Analysis

Background:

• Clinical trials frequently compare treatments using time-to-event endpoints.
• Recurrent and absorbing events require advanced statistical methods beyond simple time-to-first event models.
• Analyzing all observed events increases statistical power and provides a comprehensive disease picture.

Purpose of the Study:

• To provide a stepwise guidance for extending time-to-first event models to complex multistate methodology.
• To incorporate recurrent and absorbing events in clinical trial analyses.
• To compare non- and semiparametric methods and their interpretations.

Main Methods:

• Extension of simple time-to-first event models to multistate methodology.
• Consideration of non- and semiparametric statistical methods.
• Simulation studies to investigate model assumptions (e.g., Markov property) and censoring impacts.
• Application to data from the Interdisciplinary Network Heart Failure trial.

Main Results:

• Multistate models allow incorporation of recurrent and absorbing events for increased power.
• Non-Markov models offer interpretable summary measurements like state occupation probability and average length of stay under random censoring.
• Partly conditional transition rates can be estimated, differing from hazards, with implications explored via simulation.

Conclusions:

• Multistate methodology provides a robust framework for analyzing complex time-to-event data in clinical trials.
• Understanding model assumptions, particularly the Markov property, is crucial for correct interpretation.
• Novel approaches enable more nuanced interpretation of treatment effects, especially in non-Markovian settings.