Individual participant data meta‐analysis of intervention studies with time‐to‐event outcomes: A review of the methodology and an applied example

Abstract

Many randomized trials evaluate an intervention effect on time‐to‐event outcomes. Individual participant data (IPD) from such trials can be obtained and combined in a so‐called IPD meta‐analysis (IPD‐MA), to summarize the overall intervention effect. We performed a narrative literature review to provide an overview of methods for conducting an IPD‐MA of randomized intervention studies with a time‐to‐event outcome. We focused on identifying good methodological practice for modeling frailty of trial participants across trials, modeling heterogeneity of intervention effects, choosing appropriate association measures, dealing with (trial differences in) censoring and follow‐up times, and addressing time‐varying intervention effects and effect modification (interactions).We discuss how to achieve this using parametric and semi‐parametric methods, and describe how to implement these in a one‐stage or two‐stage IPD‐MA framework. We recommend exploring heterogeneity of the effect(s) through interaction and non‐linear effects. Random effects should be applied to account for residual heterogeneity of the intervention effect. We provide further recommendations, many of which specific to IPD‐MA of time‐to‐event data from randomized trials examining an intervention effect.We illustrate several key methods in a real IPD‐MA, where IPD of 1225 participants from 5 randomized clinical trials were combined to compare the effects of Carbamazepine and Valproate on the incidence of epileptic seizures.

Cohort profile: the Women’s Health Accelerometry Collaboration

“…The stratified model allows the baseline hazards for the two cohorts to differ. 31 However, the hazards of the exposure groups are assumed to be proportional, which will be tested using Schoenfeld residuals. We will censor follow-up time on the date of the cancer diagnosis, the date of death, or the date of last contact.…”

ABO blood types and sepsis mortality

“…Pooling of results among the three studies was done by merging the three cohorts and applying a Cox model stratified for centre and adjusted for age, sex, known ischemic heart disease, previous stroke and whether patients had a pulmonary infection or not. This allows the baseline hazard to vary between the cohorts and thus account for between-study heterogeneity in baseline hazard [ 20 ]. To assess the between-study variation the I 2 statistics are presented, and we tested for significant heterogeneity using the Cochran Q-test.…”

Cervical chordomas: multicenter case series and meta-analysis

“…In order to quantitatively compare the effect of GTR and STR on PFS and OS, we performed individual participant data metaanalysis of time-to-event outcomes following the recommendations of de Jong et al [24] We opted for an individual participant data meta-analysis as the studies presented data on an individual participant level, and an individual participant data metaanalysis offers several advantages compared to "classical" meta-analysis based on aggregated data (such as the possibility of better modelling time-to-event outcomes, and the assessment of intervention-covariate interactions at the participant level).…”

“…Heterogeneity was assessed by estimating median hazard ratios, corresponding to the median relative difference in the hazard of the occurrence of the outcome variable when two identical participants from two randomly selected different studies are compared. [24] To identify potential sources of heterogeneity, we built models including covariates and intervention-covariate interactions. In particular, we tested the age and sex of the participants, the number of operated levels, the location of the lesion (upper cervical chordomas, corresponding to all lesions including at least a segment at the level of C2 or above, and lower cervical chordomas corresponding to those below the level of C2 without involving C2).…”