Clinical validation of a polymerase chain reaction assay for the diagnosis of pertussis by comparison with serology, culture, and symptoms during a large pertussis vaccine efficacy trial.

Abstract

To assess the diagnostic sensitivity and specificity of a Bordetella pertussis polymerase chain reaction (PCR) assay using nasopharyngeal (NP) specimens from subjects with cough illnesses participating in a large pertussis vaccine efficacy trial. From 1991 to 1994, we conducted a large pertussis vaccine efficacy trial in Germany to determine the efficacy of the Lederle/Takeda acellular pertussis component diphtheria-tetanus toxoids in comparison with the Lederle whole-cell component diphtheria-tetanus toxoids vaccine. In the final year of the follow-up period of this trial, a second NP specimen for PCR, in addition to a culture specimen and blood for specific serology (enzyme-linked immunosorbent assay), was collected by use of a Dacron swab in subjects or family members with cough illnesses >/=7 days duration or in subjects with exposure to a cough illness in a household member to establish a diagnosis of B pertussis infection. Oligonucleotide primers (pTp1 and pTp2) that amplify a 191-bp-sized DNA fragment from the pertussis toxin operon, which is specific for B pertussis, were used. The PCR-amplified products were visualized by dot blot analysis followed by hybridization with a digoxigenin labeled probe and rated as 1+, 2+, or 3+ in comparison with positive controls representing approximately 1 to 10, 11 to 50, and >50 B pertussis organisms, respectively. In the present analysis, we compare PCR findings with those of serology, culture, positive household contact, and clinical characteristics of cough illnesses. Of 392 subjects with NP specimens obtained for PCR, 376 also had NP specimens collected for culture and 282 had serum specimens. PCR and culture were positive in 86 (22%) and 23 (6%) subjects, respectively. Of the positive PCR specimens, 40 were rated 3+, 32 were rated 2+, and 14 were rated 1+; 3+ positive specimens were more prevalent among DT recipients compared with pertussis vaccine recipients. Illnesses in subjects with 3+ positive PCR results were more typical of pertussis than were those in subjects with 2+ and 1+ positive results with a mean duration of cough of 48 days versus 43 and 42 days, respectively; presence of paroxysms, whoop or vomiting in 38% versus 17% and 10%, respectively; and a clinical diagnosis of definite or probable pertussis by the investigators of 26% versus 7% and 4%, respectively. Using serologic evidence of infection as the standard, sensitivity of PCR was 61%, and specificity was 88%. For 3+ positive PCR results, the respective values were 42% and 97%. Our findings demonstrate that PCR is more sensitive than conventional culture for the diagnosis of pertussis. They also demonstrate a high specificity of PCR when serology with or without other confirmative criteria (culture and household contact) is used as the reference. Analysis of semiquantitative PCR results revealed that subjects with a 3+ PCR more frequently experienced typical illness compared with patients with 1+ or 2+ PCR. Although specific serologic study remains a necessity in pertussis research its modification for diagnosis in the clinical setting results in low sensitivity and specificity. Therefore, because PCR is more sensitive than culture and is easy to perform, it is a useful addition in the clinical setting.