Is the Freeman‐Tukey double arcsine transformation a reliable approach? for proportion meta-analysis

AGUSTÍN CIAPPONI; BARDACH, A.E.; DEMIÁN GLUJOVSKY; CASTELLANA NOELIA

Congreso; 27th Cochrane Colloquium, London, UK; 2023

Cochrane

Background: Proportion meta-analyses are frequently used in epidemiology to estimate theburden of disease. They are usually based on transformed proportions using Freeman‐Tukeydouble arcsine transformations (FTT). Schwarzer et al. generated some controversy becausethey considered that this method produces seriously misleading results and proposed thegeneralized linear mixed models (GLMM) as a more elaborate approach.[1] However, Suhail etal., using the same set of studies, reanalyzed the data and concluded that the FTT is the mostreliable approach and remains the preferred transformation in proportion meta-analysis.[2]Objectives: To compare the reliability and robustness of FTT with GLMM in a large set ofproportion meta-analyses.Results: Results will be shown at the colloquium.Methods: We will conduct GLMM[1] and FTT over a large set of proportions from a livingsystematic review and meta-analysis about safety, immunogenicity, and effectiveness ofCOVID-19 vaccines for pregnant people (https://safeinpregnancy.org/lsr/) applyingrecommended safeguards[2]:a) avoiding the use of the average of the double arcsine and its variance for synthesis;b) using the inverse of the variance of the pooled FTT proportion; andc) modifying the confidence intervals to prevent numerical inaccuracies.Conclusions: It is important to verify the best approach to undertake proportions meta-analysis, which is critical for estimations in epidemiology and decision-making.References1. Schwarzer G, Chemaitelly H, Abu-Raddad LJ, Rücker G. Seriously misleading results usinginverse of Freeman-Tukey double arcsine transformation in meta-analysis of singleproportions. Res Synth Methods. 2019;10: 476–483.2. Doi SA, Xu C. The Freeman-Tukey double arcsine transformation for the meta-analysis ofproportions: Recent criticisms were seriously misleading. J Evid Based Med. 2021;14: 259–261.