Diagnostic accuracy evaluation of the conventional and molecular tests for Spinal Tuberculosis in a cohort, head-to-head study.

Early diagnosis of spinal tuberculosis (TB) is hampered by the flaws of conventional tests. The aim of this study was to assess the value of new and existing molecular tests in a prospective, head-to-head cohort study. Specimens were consecutively collected from spinal TB suspects in four hospitals in Beijing, China. Smear, culture, histopathology, Xpert MTB/RIF (Xpert), and MeltPro TB assays were performed in parallel using the same specimen from each patient. Drug-susceptibility testing (DST) was conducted on the isolates recovered. In total, 438 suspects were recruited; 319 of them were diagnosed with spinal TB according to the composite reference standard (CRS), which was composed of clinical, laboratory, histopathological, and radiological examinations and 18 months of follow-up. Based on conventional testing, 74.29% of patients were classified as confirmed cases, which increased to 90.6% when Xpert outcomes were integrated. Further, 76.60% of probable and 45.71% of possible cases were re-classified as confirmed cases with Xpert. Xpert (85.27%) produced higher sensitivity than histopathology (73.04%), MeltPro TB (57.68%), culture (51.72%) and smear (24.45%) (all P <0.001). Xpert was 100% concordant with phenotypic DST regarding rifampicin resistance detection. The sensitivity and specificity of MeltPro TB for rifampicin resistance detection were 100% and 97.96%, respectively, and 95.00% and 93.88% for isoniazid resistance detection. New molecular tests demonstrated excellent efficiency for spinal TB diagnosis in this cohort study, so their application as initial diagnostic tools would greatly increase the proportion of confirmed cases and dramatically reduce the delay of appropriate treatment. An updated laboratory testing algorithm of the disease is desirable.