Investigating heterogeneity in meta-analysis of studies with rare events

Abstract In many meta-analyses, the variable of interest is frequently a count outcome reported in an intervention and control group. Single- or double-zero studies are often observed this type data. Given setting, well-known Cochran’s Q statistic for testing homogeneity becomes undefined. paper, we propose two statistics risk ratio, particularly application case rare events meta-analysis. The first one chi-square statistic. It constructed based on information conditional probability number treatment group given total events. second likelihood ratio statistic, derived from logistic regression models allowing fixed random effects ratio. Both proposed well defined even situation single-zero studies. simulation study, tests show performance better than traditional test terms I error power under common event situations. However, as newly still unsatisfactory very suggest bootstrap approach that does not rely asymptotic distributional theory it shown performs error. Furthermore, empirical meta-analyses used to illustrate methods.