Abstract
BACKGROUND: World Health Organization (WHO) tuberculosis (TB) screening guidelines recommend computer-aided detection (CAD) software for chest radiograph (CXR) interpretation. However, studies evaluating their diagnostic and prognostic accuracy are limited. METHODS: We conducted a prospective cohort study of household contacts of rifampicin-resistant TB in South Africa. Participants underwent baseline CXR and sputum investigation (routine [single spontaneous] and enhanced [additionally 2-3 induced]) for prevalent TB and follow-up for incident TB. Three CXR-CAD software products (CAD4TBv7.0, qXRv3.0.0, and Lunit INSIGHT v3.1.4.111) were compared. We evaluated their performance to detect routine and enhanced prevalent and incident TB, comparing performance with blood tests (Xpert MTB host-response, erythrocyte sedimentation rate, C-reactive protein, QuantiFERON) in a subgroup. RESULTS: 483 participants were followed up for 4.6 years (median). There were 23 prevalent (7 routinely diagnosed) and 38 incident TB cases. The AUC ROCs (95% CIs) to identify prevalent TB for CAD4TBv7.0, qXRv3.0.0, and Lunit INSIGHT v3.1.4.111 were .87 (.77-.96), .88 (.79-.97), and .91 (.83-.99), respectively. More than 30% with scores above recommended CAD thresholds who were bacteriologically negative on routine baseline sputum were subsequently diagnosed by enhanced sputum investigation or during follow-up. The AUC performance of baseline CAD to identify incident cases ranged between .60 and .65. Diagnostic performance of CAD for prevalent TB was superior to blood testing. CONCLUSIONS: Our findings suggest that the potential of CAD-CXR screening for TB is not maximized as a high proportion of those above current thresholds, but with a negative routine confirmatory sputum, have true TB disease that may benefit intervention.