Diagnostic and prognostic accuracy of the Mycobacterium tuberculosis host response 3-gene cartridge among tuberculosis household contacts in Mozambique, Tanzania, and Zimbabwe: a prospective, longitudinal, diagnostic and prognostic accuracy cohort study

Background: Tuberculosis elimination is constrained by symptom-based and sputum-dependent diagnostic strategies, which miss asymptomatic disease and are difficult to deploy in community settings. Household contacts of people with tuberculosis are a priority population for screening and preventive therapy, but existing tests have poor prognostic ability. We aimed to evaluated host-response assays for screening and prognostic use in household contacts of people with tuberculosis. Methods: In this prospective, longitudinal, diagnostic and prognostic accuracy study, we recruited people aged 10 years and older who lived with a person diagnosed with tuberculosis in Mozambique, Tanzania, or Zimbabwe. Household contacts who had taken antimycobacterial antibiotics within the past 4 weeks were excluded. Participants had real-time Cepheid Xpert Mycobacterium tuberculosis Host Response (MTB-HR) testing and clinical, radiological, and microbiological tuberculosis screening every 6 months for up to 24 months. The primary outcomes were the diagnostic accuracy of MTB-HR obtained within 30 days of a confirmed or likely tuberculosis diagnosis at any baseline or follow-up visit, and the prognostic ability of MTB-HR for incident tuberculosis using MTB-HR results obtained 1–6 months, 6–12 months, and 1–12 months before incident tuberculosis diagnosis. Tuberculosis diagnoses were established by an endpoint review committee and we assessed discrimination using the area under the receiver operating characteristic (AUROC) curve. The study was registered with ClinicalTrials.gov (NCT04781257) and is completed. Findings: Between March 8, 2021, and March 23, 2023, we screened 2109 household contacts and enrolled 2079 for analysis (1294 [62·2%] female and 785 [37·8%] male). In the diagnostic analysis (41 household contacts with tuberculosis), the AUROC was 0·86 (95% CI 0·79–0·92). The prognostic analysis included 29 people with incident tuberculosis during the 1–6-month interval, 19 people for the 6–12-month interval, and 39 people for the 1–12-month interval, yielding AUROCs of 0·80 (0·71–0·89), 0·64 (0·53–0·76), and 0·71 (0·62–0·79), respectively, at optimised cutoffs. For the 6-month prediction at the optimised cutoff, the positive predictive value was 7·5% (95% CI 4·9–11·4). Interpretation: MTB-HR did not meet the 2025 WHO target product profile criteria for screening or prognostic use; however, its positive predictive value for incident tuberculosis was higher than that of currently used tests. These findings support a potential role for MTB-HR in screening and prevention strategies. Funding: The second European and Developing Countries Clinical Trials Partnership (EDCTP2) programme.