Refining research outcomes: Lessons from the setup of an endpoint review committee and radiological review in tuberculosis observational diagnostic studies

Most tuberculosis diagnostic validation studies use microbiological or combination reference standards to define tuberculosis. We convened an Endpoint Review Committee (ERC) to define outcomes in ERASE-TB, a longitudinal cohort study across three high tuberculosis-burden countries, aimed at validating novel diagnostics for early detection. Herein, we describe processes and outcomes of the ERC. ERASE-TB enrolled 2,109 household contacts of people with tuberculosis who were followed up 6-monthly up to 24 months with clinical, microbiological, and radiological assessments at each visit. Any participants with a chest X-ray suggestive of tuberculosis or a positive Xpert MTB/Rif Ultra were investigated and eligible for endpoint review. For these, the clinical presentation, radiological, and microbiological results were reviewed independently by two study clinicians; any discordant endpoint categorisations were assessed by the ERC (first individually; if discordant, during a consensus meeting). Tuberculosis outcomes relied on predefined definitions (confirmed, likely, possible, unlikely). The ERC comprised four members: a tuberculosis programme manager, clinicians, and a radiologist. Semi-structured interviews (n = 4) with ERC members explored experiences and challenges in tuberculosis classification. A total of 96 clinical summaries underwent review, 55 were agreed internally, with the majority being categorised as confirmed tuberculosis (34/55, 61.8%). Among the remaining 41, the ERC members agreed on only 14 (34.1%) classifications, with the majority being classified as confirmed tuberculosis (10/14, 71.4%). The discordant 27 were discussed at the consensus meeting; 9/27 were classified as likely tuberculosis (33.3%), and 7/27 as possible and unlikely (25.9%). Qualitative interviews highlighted the complexity of tuberculosis diagnosis, value of longitudinal measurements, and tensions between decision making for research as compared to clinical purposes. Standardized tuberculosis classification frameworks, longitudinal data, and real-time expert review can improve diagnostic accuracy and comparability across studies. Future research should integrate structured ERC processes prospectively to refine tuberculosis case definitions and ensure robust clinical and research outcomes.