Attempts were made to improve the PCR-based detection of Trypanosoma cruzi in blood samples, primarily for screening blood donors. Samples were obtained from candidate donors who were reactive in one or two of three serological tests for Chagas disease (and therefore considered 'indeterminate') or in all three tests (3+). Each sample was then examined using three different, PCR-based techniques: 'PCR-I' (in which the target DNA is a nuclear repetitive sequence); 'PCR-II' [amplifying a conserved region of the T. cruzi kinetoplast DNA (kDNA)]; and 'PCR-III' (a new strategy in which the target kDNA is amplified by 'nested' PCR). Among the samples from 3+ individuals, PCR-I, PCR-II and PCR-III amplified two (3.8%) out of 52, four (4.5%) out of 88, and 27 (25.7%) out of 105 samples tested, respectively. Seven, 69 and 70 samples from 'indeterminate' subjects were tested by PCR-I, PCR-II and PCR-III, respectively; there was not a single positive result by PCR-I or PCR-II, but three (4.3%) of the samples tested by PCR-III were positive. In a reconstruction experiment, in conditions in which PCR-I and PCR-II could not detect 10,000 parasites/ml, PCR-III was able to detect one parasite/ml. Although all three PCR-based strategies examined had rather poor sensitivities, PCR-III was far more sensitive than PCR-I or PCR-II.