4.3.4 Considerations for the meta-analysis of dichotomous data

For meta-analyses, computation of the logarithm (log) of the RR or the log of OR, or the RD from each individual study may be used or the number of events and the total number of participants for each group. RR and RD may be computed for any experimental study (RCT) or quasi-experimental study or cohort studies. Odds ratios may be computed for any study design (experimental, quasi-experimental, cohort, case-control, or analytical cross-sectional studies). Fleiss (1994) discussed the statistical properties of the OR and concluded that the OR is the preferred effect size for the computation phase of the meta-analysis of binary data regardless of the study design of the studies. However, the OR is not easily interpretable. Therefore, reviewers should be careful in providing correct explicit interpretation of the odds ratios computed in meta-analysis. Reviewers should provide the results expressed using both absolute (RD) and relative (RR) effect sizes for meta-analysis of binary data. Reviewers should provide correct explicit interpretation of the computed effect sizes.