Abstract
Nucleic acid amplification techniques for the diagnosis of tuberculosis (TB) are rapidly being developed. Scant work, however, has focused on pericardial TB. Using cryopreserved specimens from a prior study of pericarditis, we compared PCR to culture and histopathology for the diagnosis of tuberculous pericarditis in 36 specimens of pericardial fluid and 19 specimens of pericardial tissue from 20 patients. Fluid and tissue were cultured on Lowenstein-Jensen and Middlebrook solid media and in BACTEC radiometric broth. Tissue specimens were stained with hematoxylin-eosin, Ziehl-Neelsen, auramine O, and Kinyoun stains and were examined for granuloma formation and acid-fast bacilli. PCR was performed with both fluid and tissue with IS6110-based primers specific for the Mycobacterium tuberculosis complex by published methods. Sixteen of the 20 patients had tuberculous pericarditis and 4 patients had other diagnoses. TB was correctly diagnosed by culture in 15 (93%) patients, by PCR in 13 (81%) patients, and by histology in 13 of 15 (87%) patients. PCR gave one false-positive result for a patient with Staphylococcus aureus pericarditis. Considering the individual specimens as the unit of analysis, M. tuberculosis was identified by culture in 30 of 43 specimens (70%) from patients with tuberculous pericarditis and by PCR in 14 of 28 specimens (50%) from patients with tuberculous pericarditis (P > 0.15). The sensitivity of PCR was higher with tissue specimens (12 of 15; 80%) than with fluid specimens (2 of 13; 15%; P = 0.002). In conclusion, the overall accuracy of PCR approached the results of conventional methods, although PCR was much faster. Therefore, PCR merits further development in this regard. The sensitivity of PCR with pericardial fluid was poor, and false-positive results with PCR remain a concern.