The diagnosis of skeletal tuberculosis by polymerase chain reaction

OBJECTIVE: To determine the potential of the polymerase chain reaction (PCR) for detecting Mycobacterium tuberculosis in skeletal samples by comparing results obtained by 1) Ziehl Neelsen staining, Lowenstein-Jensen and Bactec culture, 2) histopathology and clinical findings at the level of agreement, sensitivity and specificity. DESIGN: Cross sectional study. SETTING: Department of Medical Microbiology, Orthopaedics and Anatomical pathology, University of the Orange Free State Bloemfontein, South Africa. SUBJECTS: 45 consecutive patients were extensively investigated, 30 patients with clinical presumptive active tuberculosis and 15 with other pathology. RESULTS: Detection using culture could confirm only three of the 26 clinically diagnosed tuberculosis cases while PCR detection confirmed disease in 15 cases. The use of PCR increased the confirmation of clinically probable tuberculosis from 14 using standard laboratory techniques and histology to 18 of 26 cases. Calculated sensitivity and specificity for PCR employing culture as the "gold standard" were 100% (with 95% CI 29.2; 100.0) and 71.4% (55.4; 84.3), which due to low detection levels, basically excludes culture as a standard for statistical analysis. Sensitivity and specificity for PCR using histology as the "gold standard" were 78.6% (49.2; 95.3) and 87.1% (70.2; 96.4) respectively with positive and negative predictive values of 73.3% (44.9; 92.2) and 90% (73.5; 97.9) respectively. Positive agreement between PCR and histology was 0.64 (0.4; 0.9) indicating fair agreement. CONCLUSION: Although numbers in the study were too low to effectively draw statistically valid conclusions the importance of the relevance of PCR for rapid detection of low numbers of acid-fast bacilli and confirmation of mycobacterial infection in spinal biopsies has been established.