Metadta: a Stata command for meta-analysis and meta-regression of diagnostic test accuracy data - a tutorial

Area of Science:

• Biostatistics
• Medical Informatics
• Epidemiology

Background:

• Existing statistical packages lack comprehensive tools for meta-analysis of diagnostic test accuracy, particularly for multivariate regression.
• Manual fitting of regression models and processing estimates for diagnostic meta-analysis are time-consuming and complex.
• A user-friendly program is needed to streamline these statistical procedures for the scientific community.

Purpose of the Study:

• To introduce metadta, a statistical program designed for pooling diagnostic accuracy test data in Stata.
• To implement advanced statistical models, including bivariate random-effects, fixed-effects, and meta-regression.
• To provide a user-friendly interface for complex diagnostic meta-analysis calculations and visualizations.

Main Methods:

• The metadta program in Stata was developed to pool diagnostic accuracy data.
• It supports bivariate random-effects and fixed-effects models, with options for meta-regression.
• The program quantifies heterogeneity using an I² statistic and generates tables, forest plots, and summary receiver operating characteristic (SROC) plots.

Main Results:

• Metadta generated pooled sensitivity (0.77) and specificity (0.91) for telomerase in diagnosing bladder cancer without continuity correction.
• The program assessed the relative accuracy of HPV testing using self-collected versus clinician-collected samples for cervical precancer detection.
• Analysis indicated comparable sensitivity for HPV testing on self-samples versus clinician-samples in detecting cervical precancer.

Conclusions:

• The metadta program integrates advanced statistical procedures to bridge the gap between statisticians and systematic reviewers.
• It aims to enhance the accessibility and application of appropriate statistical methods in diagnostic meta-analysis.
• Metadta is expected to promote wider adoption of robust statistical techniques in the field.