MetaBayesDTA: Codeless meta-analysis of diagnostic test accuracy, with or without a gold standard

Introduction: Statistical models for the meta-analysis of test accuracy have historically required a high level of specialised knowledge to implement. The necessary level of expertise has recently increased further, due to the development and recommendation to use more sophisticated methods; such as those in Version 2 of the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy. This paper describes a web-based application that extends the functionality of previous applications, making many advanced analysis methods more accessible. Methods: We sought to create an extended, stand-alone, Bayesian version of MetaDTA, which (i) has the benefits of previously proposed applications and addresses key limitations of them, (ii) is accessible to researchers who do not have the specific expertise required to fit such models, and (iii) is suitable for experienced analysts due to its user-friendly interface. We created the application in R using Shiny and coded the models using Stan. Results: We created MetaBayesDTA (https://crsu.shinyapps.io/MetaBayesDTA/), which allows users to conduct meta-analysis of test accuracy, with or without assuming a gold standard. The application addresses several key limitations of other applications. For instance, for the bivariate model, one can conduct subgroup analysis, univariate meta-regression, and comparative test accuracy evaluation. Meanwhile, for the model which does not assume a perfect gold standard, the application can account for the fact that different studies in a meta-analysis use different reference tests. Conclusions: Due to its user-friendliness and broad array of features, MetaBayesDTA should appeal to a wide variety of applied researchers. We anticipate that the application will encourage wider use of more advanced methods, which ultimately should improve the quality of test accuracy reviews.