Comparison of continuity corrections and peto method in meta-analysis of rare events

When we look at the researches in the field of health in recent years, it is seen that there is an increase in the number of studies on the same or similar subjects, the quality of the studies varies and different and even opposite results are reached between the studies. Meta-analysis is a method developed to obtain a statistically common result from the results obtained from independent studies conducted on the same subject or the same subject at different times and places, taking into account the inconsistencies and the reasons for the differences. The aim of this study is to show which method is more successful in meta-analysing rare events so that health decisions can be made with more reliable prediction results. In studies on rare events, due to the small number of events, high variability is observed among the results and various statistical difficulties arise in the analyses. In this context, the necessity of continuity adjustment and which methods provide reliable results is an important issue of debate. In the study, Mantel Haenszel Method (MH) and Peto odds ratio method were considered and simulation study was carried out by considering different event probabilities, equilibrium status in groups, number of studies included in meta-analysis and sample sizes of studies, especially under fixed and random effect models. In order to determine the method that provides the closest prediction to the reality with the simulation study, the performance of the methods was evaluated on the basis of bias and mean square error criteria. As a result of the simulation study, it was observed that the Peto method is preferable in the fixed effect model because it has lower and similar median values in the fixed effect model, and similarly, the MH method also gives similar median values in the fixed effect model. However, the MH method was found to be more successful in terms of performance (bias and RMSE mean values) than the Peto method in both fixed and random effect models. Therefore, the MH method, which uses continuity correction in rare events, was found to be appropriate.