King's College London

Meta-analysis of rare event studies has recently become a subject of controversy and debate. We willargue and demonstrate in this paper that the occurrence of zero events in clinical trials or cohort studies,even if zeros occur in both arms (the case of a double-zero trial), is less problematic, at least from astatistical perspective, if the available statistical tools are applied in the appropriate way. In particular,it is neither necessary nor advisable to exclude studies with zero events from the meta-analysis. In termsof statistical tools, we will focus here on Mantel–Haenszel techniques, mixed Poisson regression andrelated regression models.