Diagnostic accuracy of three molecular assays and culture in smear-negative bronchoalveolar lavage fluid: a retrospective study using routinely collected clinical data

To assess the diagnostic accuracy of TB-DNA, SAT-TB, Xpert MTB/RIF (Xpert), and mycobacterial culture in smear-negative bronchoalveolar lavage fluid (BALF), and explore the additional yield of assay combinations. This single-centre retrospective diagnostic-accuracy study used routinely collected clinical and laboratory data from inpatients at Hangzhou Red Cross Hospital, China, between July 2016 and September 2021. Patients with smear-negative BALF and available results for all four tests were included. All tests were performed on aliquots of the same BALF specimen during routine clinical care. Diagnostic accuracy was evaluated against a composite reference standard (CRS) based on the final clinical diagnosis. Exploratory analyses assessed all two-, three-, and four-assay combinations, with positivity defined as at least one positive constituent test. Xpert rifampicin-resistance results were evaluated separately against phenotypic drug susceptibility testing (pDST). Among 6,560 patients, 3,849 had pulmonary tuberculosis (PTB) and 2,711 had alternative pulmonary diagnoses. Sensitivities of TB-DNA, SAT-TB, Xpert, and culture were 48.4%, 38.0%, 52.1%, and 30.7%, respectively; corresponding areas under the receiver operating characteristic curve were 0.740, 0.689, 0.758, and 0.653. Specificities ranged from 99.6% to 100.0%. Xpert plus TB-DNA achieved 61.4% sensitivity and 99.1% specificity. Adding SAT-TB and culture increased sensitivity by 3.5 percentage points to 64.9%, while specificity decreased to 98.9%. Among 1,002 patients with paired Xpert and pDST results, Xpert detected rifampicin resistance with 97.8% sensitivity, 97.0% specificity, and a kappa value of 0.841. Xpert showed the highest CRS-based sensitivity among the four individual tests in smear-negative BALF and provided concurrent rifampicin-resistance information. Combining assays increased diagnostic yield, although the additional gain beyond Xpert plus TB-DNA was modest. These exploratory findings support further evaluation of assay combinations but do not establish an optimal testing strategy. Prospective studies incorporating costs, turnaround time, and effects on patient management are needed.