ObjectiveTo compare the diagnostic performance of laboratory assays on the ultrasound-guided core needle biopsy samples for diagnosis of extra-pulmonary tuberculosis (EPTB) in HIV-positive and HIV-negative patients.MethodsA total of 217 patients suspected to have EPTB underwent lesion biopsy from 2017 to 2020. Results of laboratory tests on the biopsy and non-biopsy samples were collected with clinical data for retrospective analysis of test utility. The calculated diagnostic accuracy of the tests was stratified according to the specimen types and HIV status.ResultsThe cohort contained 118 patients with a final positive diagnosis of extrapulmonary tuberculosis (EPTB group, 54.4%) and 99 finally diagnosed as without TB (non-EPTB group, 45.6%). The risk factor for EPTB was HIV co-infection (OR 2.22, 95% CI 1.17-4.28, p = 0.014). In biopsy samples, GeneXpert (Xpert) showed higher sensitivity (96.6% [91.6-98.7], p < 0.0001) than culture (56.1% [47.0-64.9]). Regardless of HIV status, Xpert had the highest sensitivity (>95%) and specificity (nearly 100%) of any methods. In non-biopsy samples, only T-SPOT.TB (T-SPOT) showed higher sensitivity than culture (90.9% [62.3-99.5] vs 35.3% [17.3-58.7], p = 0.0037). Furthermore, the sensitivities of Xpert were lower in non-biopsy samples (60.0% [23.1-92.9], p = 0.022) than in biopsy samples (100% [86.7-100]). Even in smear-negative biopsy samples, Xpert still had higher sensitivity than culture and retained high specificity (100% [95.7-100]).ConclusionSuperior performance of Xpert in diagnosing EPTB was observed regardless of HIV status and specimen types. Nevertheless, the biopsy samples still substantially facilitated the accurate diagnosis of extrapulmonary tuberculosis.