The Selection of Summary Statistic Measures for Dichotomous Data in Meta-Analysis

The selection of summary statistics is very important for the data calculation and interpretation of its results in a meta-analysis. Odds ratio （OR） and relative risk （RR） are commonly used for dichotomous data in randomized controlled trials （RCTs） based meta-analysis. This paper introduces basic concepts and investigates the characteristics, difference and applicability of RR and OR in meta-analysis. The selection of RR and OR for a RCT based meta-analysis is very important, but the selection criteria and report specification still need to be improved. RR is more cautious in combined effect. Compared with RR, OR may cause quantitative （overestimated） and even qualitative differences for combined effect in a meta-analysis. In this paper, optimization selections of OR and RR according to different study characteristics and data situations were recommended, and other factors also should be taken into account. More in-depth discussions still need further studies.