With the objective to evaluate PCR-mediated detection of Mycobacterium tuberculosis DNA as a diagnostic procedure for diagnosis of tuberculosis in individuals attending ambulatory services in Primary Health Units of the City Tuberculosis Program in Rio de Janeiro, Brazil, their sputum samples were collected and treated with a DNA extraction procedure using silica-guanidiniumthiocyanate. This procedure has been described to be highly efficient for extraction of different kind of nucleic acids from bacteria and clinical samples. Upon comparing PCR results with the number of acid-fast bacilli, no direct relation was observed between the number of bacilli present in the sample and PCR positivity. Part of the processed samples was therefore spiked with pure DNA of M. tuberculosis and inhibition of the PCR reaction was verified in 22 out of 36 (61%) of the samples, demonstrating that the extraction procedure as originally described should not be used for PCR analysis of sputum samples.