Due to the early administration of antibiotics, meningococcal disease is increasingly difficult to diagnose by culturing. Laboratory studies have shown PCR to be sensitive and specific, but there have been few clinical studies. The objectives of this study were to determine the diagnostic accuracy and clinical usefulness of meningococcal PCR through a prospective comparison of real-time PCR, nested PCR, and standard culturing of blood and cerebrospinal fluid (CSF). The setting was a tertiary-care pediatric hospital in Australia, and the participants were 118 children admitted with possible septicemia or meningitis. The main outcome measures-sensitivity, specificity, and positive and negative predictive values-were compared to a "gold standard " fulfilling clinical and laboratory criteria. For 24 cases of meningococcal disease diagnosed by the gold standard, culturing of blood or CSF was positive for 15 (63%), nested PCR was positive for 21 (88%), and real-time PCR was positive for 23 (96%). The sensitivity, specificity, and positive and negative predictive values of real-time PCR (the most sensitive test) for all specimens were, respectively, 96% (95% confidence interval, 79 to 99%), 100% (95% confidence interval, 96 to 100%), 100% (95% confidence interval, 85 to 100%), and 99% (95% confidence interval, 94 to 100%). Of 54 patients with suspected meningococcal disease at admission, 23 had positive PCR results. Only one PCR specimen was positive in a patient thought unlikely to have meningococcal disease at admission. Blood PCR remained positive for 33% of patients tested at up to 72 h. Real-time PCR has high positive and negative predictive values in this clinical setting, with better confirmation of cases than nested PCR. Targeting patients for PCR based on admission criteria appears to be practical, and the test may remain useful for several days after the start of antibiotic administration.