Multivariate meta-analysis model for the difference in restricted mean survival times

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Health Economics

Background:

• Restricted mean survival time (RMSTD) difference is an absolute treatment effect measure in time-to-event outcomes.
• Meta-analyses of randomized controlled trials (RCTs) often exclude studies with shorter follow-up durations than the chosen time horizon (τ).
• Existing methods may not fully utilize available data when follow-up times vary across trials.

Purpose of the Study:

• To develop and evaluate an individual patient data multivariate meta-analysis model for RMSTD at multiple time horizons.
• To enable the synthesis of data from RCTs with varying follow-up durations.
• To improve the statistical precision and efficiency of meta-analyses for time-to-event outcomes.

Main Methods:

• Introduction of a multivariate meta-analysis model for RMSTD incorporating within-trial covariance.
• Estimation of RMSTD at multiple time horizons (e.g., 12, 24, 36 months).
• Comparison of the proposed multivariate model against univariate meta-analysis methods via simulation studies.

Main Results:

• The multivariate meta-analysis model demonstrated superior statistical performance with a smaller mean squared error compared to univariate methods across all time points.
• The model effectively synthesizes data by borrowing strength from multiple time points, accommodating varying follow-up durations.
• Application to RCTs comparing transcatheter aortic valve replacement (TAVR) with surgical aortic valve replacement (SAVR) showed TAVR patients lived longer on average.

Conclusions:

• The proposed multivariate meta-analysis model offers a robust and statistically efficient approach for synthesizing RMSTD from RCTs with time-to-event outcomes.
• This method enhances the utilization of available data, particularly when trial follow-up durations differ.
• The findings support the use of this advanced statistical technique for more comprehensive and precise treatment effect estimation in clinical trials.