OBJECTIVE: To compare the polymerase chain reaction (PCR) results with conventional culture results for the diagnosis of Bordetella pertussis infections. METHODS: PCR and culture were performed in the course of a large vaccine efficacy trial in Germany on specimens taken from 7153 children less-than-or-equal2 years of age with cough illness lasting >6 days, and laboratory results were compared with clinical data also obtained from the patients. Calcium alginate nasopharyngeal swabs were taken for culture and clinical data were obtained from patients. Swabs were inoculated on charcoal horse blood agar plates containing cephalexin, and then discarded. The agar plates were preincubated for 2 days at the physician's office and then shipped to the culture laboratory at the University Children's Hospital in Munich, Germany, for diagnosis of B. pertussis and B. parapertussis infections. In this laboratory, Dacron swabs for PCR were taken from each culture by a wide sweep over the culture. Swabs for PCR were stored in NaCl and sent weekly to the PCR laboratory at the University Children's Hospital in Basel, Switzerland, for PCR diagnosis of B. pertussis infections. RESULTS: B. pertussis was identified by culture in 3% (213/7153) and by PCR in 7.6% (546/7153) of the specimens. Therefore, PCR increased the identification rate of subjects with B. pertussis infection 2.6-fold. Clinical characteristics were considered according to the type of laboratory findings: group 1 consisted of 209 culture-positive and PCR-positive subjects, and group 2 of 337 culture-negative but PCR-positive subjects. Group 2 subjects were significantly more likely to have mild or atypical clinical symptoms of whooping cough than were group 1 subjects. By analyzing the PCR results of group 2 subjects semiquantitatively, it could be shown that the degree of PCR positivity correlated with the severity of the clinical symptoms of whooping cough in the patient. CONCLUSIONS: PCR identified many pertussis cases with mild or atypical clinical symptoms that were not identified by culture. Semiquantification of PCR products revealed that the less positive the PCR result, the higher was the failure rate in diagnosing pertussis by culture, and, in addition, the less typical were the clinical symptoms in the patient.