Skip to Content

Meta-analysis of studies with competing risks outcomes

Research Objectives

Survival outcomes in randomised clinical trials are frequently subject to the occurrence of competing risks. E.g., observing disease progression may be precluded by death. Treatment effects are usually reported as hazard ratios based on Cox models for the hazard of a combined endpoint, for the risk specific hazard of interest, for all risk specific hazards, or for the subdistribution hazard of the event of interest. However, the given information does typically not suffice to conclude on the cumulative event probabilities.

Description of work

The challenge is twofold: First, we investigate to which extent information can be recovered from published reports for identifying cumulative event probabilities. This typically requires some simplifying parametric assumption. Second, we develop methodology for meta-analyses for evaluating effects in terms of the cumulative event probabilities. This typically requires some formal synthesis of results for all risk specific hazards. This methodology does not only apply to meta-analyses based on published reports, but also when a company aggregates individual patient data across several clinical trials. The methodology is, e.g., relevant for detecting a beneficial treatment effect on progression outcome, which might be blurred in analyses of a combined `progression-free survival' outcome both in single studies and in meta-analyses.

Host Institution: University Medical Center Freiburg

about seo