Background: The key to achieve better clinical outcomes in treating vertebral osteomyelitis (VO) is to obtain a prompt and reliable bacterial culture. Detecting Mycobacterium tuberculosis (M. tuberculosis) in spine infection is difficult owing to slow growth of mycobacteria and the presence of fewer mycobacteria at the affected sites. The aim of this study was to validate the effectiveness of a 2-step multiplex polymerase chain reaction (M-PCR) assay in detecting M. tuberculosis in formalin-fixed paraffin-embedded (FFPE) vertebral specimens of VO. Methods: Through the Spine Operation Registry of the authors' institution, 43 patients with a diagnosis of VO and undergoing open spine surgery at the authors' hospital were retrospectively reviewed. Thirty-two patients, 12 TB- and 20 non-TB VO, who had a confirmed spine microbial culture at the affected site and a compatible histo-pathological finding were enrolled for study. Bacterial DNAs were extracted from vertebral FFPE blocks, the 2-step M-PCR assay was developed to identify bacterial or fungal infection first and then to identify mainly M. tuberculosis and Staphylococcus aureus of Gram-positive bacteria. All the FFPE vertebral specimens were blinded prior to the 2-step M-PCR study. Results: The non-TB VO cases were used as a control group. The current 2-step M-PCR assay in detecting M. tuberculosis was positive in 11 out of 12 TB VO and was positive in 2 out of 20 non-TB cases. The detection of M. tuberculosis was false negative in 1/12 (8.3%) and false positive in 2/20 (10%). It indicated a 91.7% of the sensitivity, 90% of the specificity, 84.6% of the positive predictive value, and 94.7% of the negative predictive value. The accuracy rate was 90.1%. Conclusion: Our results indicated that the current 2-step M-PCR assay is sensitive and effective in detecting M. tuberculosis in FFPE vertebral specimen from patients with VO. Potentially, the 2-step M-PCR can be applied as a quick and valuable supplementary tool for increasing the diagnostic accuracy of TB spondylitis.