Objectives/Goals: Our study’s primary objective is to develop a multivariate random effects meta-analysis model that estimates risk ratio at a series of clinically relevant time points. Our goal in doing so is to address gaps in traditional hazard ratio (HR) meta-analyses due to observed time-varying treatment effects. Methods/Study Population: Our model estimates pooled risk ratios (RRs), the ratio of cumulative event probabilities, at predefined timepoints (e.g., 12, 24, 36 months). We derive the variance of the log-RR for each trial at each time point by applying the delta method to the Kaplan-Meier (KM) estimator function. Furthermore, we will derive the covariance to account for the correlation between RR estimates at different time points within the same trial. We would then like to incorporate these derivations of within-trial variance and covariance at each available time milestone into our multivariate model to increase the precision of our pooled RR estimated for both Overall Survival (OS) and Progression-Free Survival (PFS). Results/Anticipated Results: We identified 7 trials by consulting 5 systematic reviews published through 2021-2025 to ensure use of updated data. We reconstructed the pseudo-Individual Patient Data of these trials by applying PDFs of published KM curves to a validated survival curve reconstruction algorithm. For the outcome of overall survival, the max amount of follow-up time recorded is 108 months. On the other hand, there are 96 months of available follow-up for the outcome of progression-free survival. In a univariate random effects meta-analysis of HR for OS, the pooled HR was found to be 0.81 (CI: 0.74 – 0.89). Additionally, the pooled HR for PFS was found to be 0.68 (CI: 0.59 – 0.79). For both outcomes, we anticipate the multivariate random effects analysis of Risk Ratio to yield more precise confidence intervals than those of HR. Discussion/Significance of Impact: The proposed multivariate meta-analysis model will improve precision of treatment effect estimates which will add to the ongoing evaluation of immunotherapies as a current standard of care. This is significant to translational science as it provides time-specific treatment effect estimates that could be more interpretable in a clinical setting.
Reddy et al. (Wed,) studied this question.