Objective: To evaluate the accuracy of bacteriological methods and of polymerase chain reaction (with primers specific for IS6110 of the Mycobacterium tuberculosis complex) in testing sputum samples from indigenous (Amerindian) and non-indigenous patients. Methods: A total of 214 sputum samples (154 from indigenous patients and 60 from non-indigenous patients) were analyzed in order to determine the accuracy of smear microscopy (direct and concentrated versions) for acid-fast bacilli, culture, and polymerase chain reaction. Results: Both microscopy methods presented low sensitivity in comparison with culture and polymerase chain reaction. Specificity ranged from 91% to 100%, the concentrated acid-fast smear technique being the least specific. Nontuberculous mycobacteria were isolated three times more frequently in samples from indigenous patients than in those from non-indigenous patients. False-positive and false-negative polymerase chain reaction results were more common in the indigenous population. Conclusion: Positivity and isolation of nontuberculous mycobacteria in the acid-fast smear in conjunction with polymerase chain reaction positivity raise the following hypotheses: nontuberculous mycobacteria species with DNA regions homologous to, or even still possessing, the M. tuberculosis IS6110 exist in the Amazon; colonization of the oropharynx or of a tuberculous lesion accelerates the growth of the nontuberculous mycobacteria present in the sputum samples, making it impossible to isolate M. tuberculosis; A history of tuberculosis results in positivity for M. tuberculosis DNA. The absence of bacteriological positivity in the presence of polymerase chain reaction positivity raises questions regarding the inherent technical characteristics of the bacteriological methods or regarding patient history of tuberculosis.

Keywords: Tuberculosis, pulmonary/diagnosis; Mycobacterium tuberculosis, Polymerase chain reaction/methods; Indians, south american