A competitive PCR assay involving the use of bronchoalveolar lavage (BAL) samples for the diagnosis of invasive pulmonary aspergillosis (IPA) was developed. For this purpose, a 1-kb mitochondrial DNA fragment of Aspergillus fumigatus was sequenced. The primers used allowed amplification of A. fumigatus, A. flavus, A. terreus, and A. niger DNAs but not DNAs of other fungi and yeasts. BAL samples from 55 consecutively enrolled patients were tested. Three samples were excluded because of failure of correct amplification of the internal competitive control. Of 28 immunocompromised patients, 6 were PCR positive; 3 died of IPA and their BAL cultures yielded A. fumigatus; and 3 were culture negative and did not develop IPA. Of 15 human immunodeficiency virus-positive patients and 9 immunocompetent patients, 5 and 4, respectively, were both PCR positive and culture negative, and none developed aspergillosis. Thus, PCR confirmed IPA in three patients but gave positive results for 25% (12 of 49) of the patients who did not develop aspergillosis. The predictive value of PCR-positive results seems low for patients at risk for aspergillosis. Moreover, the risk of contamination of reaction buffers or biological samples with Aspergillus conidia seems high and has to be weighed in regard to the potential diagnostic benefit of PCR testing as a routine procedure.