Background: Meta-analyses including trials with binary outcomes are very common in the systematic reviews of effectiveness of therapeutic or preventive interventions. When zero events occur in either or both arms, usually called sparse events, of at least a trial, this may cause problems in the potential analysis methods and hence lead to unreliable conclusion of the reviews.

Objectives: We investigated the handling of sparse event data in meta-analyses within the Cochrane systematic reviews

Methods: We recruited 47 systematic reviews including at least a trial with sparse event data from the Cochrane Database of Systematic Reviews (CDSR) 2005, issue 1. We, ML and PP, independently assessed the reviews under unblinded conditions using pre-established extraction forms. We collected the information of sparse event data, choices of its continuity correction, analysis approaches and sensitivity analysis concerning with the sparse event data.

Results: We found only one of 47 systematic reviews ( 2.1 %) pooled the sparse data from trials included by using the exact method based on exact probability theory. No recruited systematic review undertook sensitivity analysis for identifying the effect of sparse data on the review results. We found variety characteristics of sparse data in the reviews. We will present their details and provide discussion on the further steps and suggestions at the Cochrane Colloquium.

Conclusions: Little attention to the effect of sparse event data on the Cochrane systematic reviews is seen. We need to encourage the reviewers to be aware of effect of sparse data on the review conclusions.