Methods for meta-analysis and meta-regression of proportions: concepts and tutorial with Stata command metapreg

Abstract

Background: Despite the widespread use of meta-analysis of proportions, its rationale, certain theoretical and methodological concepts are poorly understood.The framework of logistic regression is well-established and provides a natural and optimal model for meta-analysis, network meta-analysis, and meta-regressionof proportions. Nonetheless, classical tools for meta-analysis of proportions based on approximation to the normal distribution continue to dominate. Methods: We explain the rationale and concepts essential in understanding statistical methods for meta-analysis of proportions. We then present metapreg; a flexible model-based tool with advanced statistical procedures to perform ameta-analysis, network meta-analysis, and meta-regression of proportions in Stata. The tool implements the logistic and the logistic-normal regression, andmarginal standardization to yield population-averaged estimates. Finally, we demonstrate using metapreg with data from four published meta-analyses; a)cure rate of treatment for cervical pre-cancer using cold coagulation, b)incomplete excision of cervical precancer as a predictor of treatment failure, c)effect of geographical area on the protective effect of BCG vaccination againsttuberculosis and d) response rate to treatment of acute mania in adults withdiverse drugs. Conclusion: metapreg is an efficient and user-friendly statistical tool dedicatedt o meta-analysis, network meta-analysis and meta-regression of binomial data. It allows both continuous and categorical explanatory covariates including relevant interactions to explain the variability in the proportions.

Global and regional estimates of genital human papillomavirus prevalence among men: a systematic review and meta-analysis

“…Meta-analyses were conducted with Stata procedures metaprop, metapreg, and metan. 24 , 25 Freeman-Tukey arcsine transformation of the prevalence was used to normalise variance; 95% CIs around the study-specific and pooled prevalences were computed on the basis of the score-test statistic. The percentage of total variation due to interstudy heterogeneity was evaluated using the I 2 measure.…”

Global Estimates of Diabetic Retinopathy Prevalence and Progression in Pregnant Individuals With Preexisting Diabetes: A Meta-analysis

“…Abbreviations: DR, diabetic retinopathy; FTT, Freeman-Tukey double arcsine transformation; GLMM, generalized linear mixed modeling; NPDR, nonproliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy; T1D, type 1 diabetes; T2D, type 2 diabetes. a N = the study-specific total subjects with relevant data. b Random intercept logistic regression model using Stata commands: metapreg event denom, studyid (study) by (diabetesTypeString) model (random) ci (wilson) prediction.…”

“… b Random intercept logistic regression model using Stata commands: metapreg event denom, studyid (study) by (diabetesTypeString) model (random) ci (wilson) prediction.…”

“…2 Methods | Detailed methodology around study selection and statistical methods has been described in our original article. 1 In addition, here, the Stata IC (Stata Corp) version 15.1 additional module command metapreg 5 was used to construct random intercept logistic regression models (a relevant GLMM, as described by Stijnen et al 6 ) for the metaanalytic outcomes of interest published previously. 1 The Stata command metapreg was also used to estimate prediction intervals.…”

Mechanical Thrombectomy for In-Hospital Onset Stroke: A Comparative Systematic Review and Meta-Analysis

“…We opted not to perform the Egger's test due to the limited number of studies included, which was fewer than ten. 13,14 Stata software (Version 17.0; StataCorp., College Station, TX, USA) was used for all statistical analyses, and the "Metapreg"usermade Stata package 15 and built-in "meta" command were utilized for the proportion and OR meta-analyses of the outcomes and baseline characteristics. Furthermore, the MetaMedian package 16 in R (version 4.2.1; R Foundation for Statistical Computing, Vienna, Austria) was employed for the meta-analysis of temporal workflow checkpoints.…”