Have you tried plugging the study and the zeros in your meta-analysis software? If you are using RevMan, it might be that it would permit both arms having 0 or no outcome (please try this there first to confirm). In RevMan, whenever a value of 0 is plugged in the events column in a treatment arm, it will automatically substitute the value of 0.5 for mathematical/statistical computations.

Another issue you have to contend with is the possible reasons for such studies reporting zero events for the outcome(s) of interest in both arms. Are these studies in question eligible in your review inclusion criteria to begin with? Do these studies had the appropriate outcomes and follow-up period to ascertain those outcomes? With the information you shared, I am considering right away that those studies might be (outrageously) underpowered or that the follow-up period is too short to detect clinically-relevant outcomes, among others. You also have to review if your eligibility criteria is appropriate especially in the face of such studies being included.

You have to be transparent on these concerns in the manuscript results and discussion. If reconsideration would lead to a modification of your eligibility criteria in the middle of review analysis phase, then this must be specified in the manuscript as well.

Continuity correction is an acceptable method to deal with zero events in both arms.

you could do the meta analysis on risk differences (rather than ratios) if data available

see

https://www.statsdirect.com/help/meta_analysis/risk_difference.htm

Aedrian Abrilla

Thank you for your reply. I used STATA software. Is it possible to manually substitute the value of 0.5 for zero?

Min Seo Kim

Thank you for your reply, Kim. But even with the use of continuity correction option, the studies in which the incidence of the event was zero in both arms were excluded from the analysis. (I used STATA software.)