We evaluated the polymerase chain reaction (PCR) and hybridization procedures for diagnosis of Lassa fever. Primers were derived from a region of the small RNA segment of Lassa virus coding for the glycoprotein. Serum samples stored for a 14-year period from patients in Sierra Leone, West Africa were examined retrospectively. Blinded samples were then tested prospectively. Eighty-eight virus isolation-negative control sera were negative by PCR and hybridization. In the retrospective study, virus was isolated from 51 of 98 specimens from patients with Lassa fever, and 33 of these were positive for Lassa virus RNA by PCR, and 42 by PCR and hybridization. Fifteen were positive by PCR and hybridization but isolation-negative, and nine were positive by isolation but PCR/hybridization-negative. Thirty-two were negative by all methods (sensitivity by PCR/hybridization compared with virus isolation 0.82, specificity 0.68). In a prospective blinded study of 195 patient sera, 51 were positive by PCR and virus isolation, and 24 were PCR positive but virus isolation-negative (sensitivity 0.66, specificity 0.71). After hybridization, 66 virus isolation-positive sera were positive. The sensitivity was 0.86 and the specificity was 0.59, and the probability of false-positive results compared with virus isolation was 32%, (chi 2 = 21.9, by McNemar's test). Since some specimens may not have contained viable virus, we re-analyzed the data of individual patients using laboratory-confirmed case definitions for Lassa fever. All specimens from patients in whom Lassa fever was excluded by serologic tests were negative by PCR/hybridization.(ABSTRACT TRUNCATED AT 250 WORDS)