On the basis of previously published PCR primer sequences, we have designed a sensitive system for detecting DNA of the Mycobacterium tuberculosis complex (MTB) in patient sputum samples which employs a fast and simplified sample preparation method appropriate for routine diagnostic testing. In order to evaluate the accuracy of the PCR assay, we performed a prospective study with 103 patients, comparing PCR results with culture results of samples obtained from a parallel culture assay as well as with subsequent culture results. Using two MTB-specific PCR primer systems, we found 48 of 49 tuberculosis (Tb) patients to be PCR positive (PCR sensitivity, 0.98). Sixteen of 54 presumably non-Tb patients showed amplifiable MTB DNA (specificity, 0.7). The study demonstrates that for diagnostic applications of MTB PCR two MTB-specific primer pairs should be used. MTB infection is extremely unlikely in cases of MTB PCR-negative samples: with our method for the exclusion of active Tb, the validity of one PCR assay seems to be equivalent to those of at least three culturing procedures. Positive PCR results do not necessarily reflect active MTB infection. It remains to be shown whether positive PCR results in Tb-negative patients mean false-positivity, an early laboratory finding which predicts a subsequent reactivation of a prior Tb infection, or whether asymptomatic patients may carry PCR-amplifiable MTB DNA without any clinical relevance. It is important to point out that the validity of PCR results in clinical studies depends on the use of contamination controls parallel to all PCR steps and the simplicity of the DNA extraction method as well as on the specificity of the PCR results.