metandi: Stata software for statistically rigorous meta-analysis of diagnostic accuracy studies

Background: Many approaches to the meta-analysis of diagnostic accuracy are currently in use. Consensus is building that statistically rigorous methods involving hierarchical models are necessary to ensure valid results. Such methods correctly handle the correlation between sensitivity and specificity and the binomial distribution of the data within each study. There are two such models, the hierarchical summary ROC (HSROC) model and the bivariate random-effects model, which have been shown to be equivalent when no covariates are fitted, as well as in certain cases with covariates. Aim: To develop a user-written module metandi (Harbord 2008) for the statistical software package Stata that performs meta-analysis of diagnostic accuracy studies without covariates and displays the results in both bivariate and HSROC parameterisations, as well as on a graph. Results: The user-written command gllamm and the (faster) official command xtmelogit introduced in Stata 10 can both be used to fit the bivariate model: the corresponding HSROC parameter estimates can also be produced after some extra work. The metandi module provides a straightforward interface in which a single command fits the model and (optionally) graphs the results, e.g.: . metandi tp fp fn tn, plot Conclusion: Increasing accessibility of statistically rigorous methods will increase their use and facilitate appropriate analyses.