Editorial Commentary: The "How" of Polymerase Chain Reaction Testing for Bordetella pertussis Depends on the "Why"

(See the Major Article by Rodgers et al on pages 322–31.) For most of its almost 400-year history as an identified human disease [1], whooping cough has been a clinical syndrome measured by passive reporting of cases with classic symptoms. Since its identification as the cause of the whooping cough syndrome by Bordet and Gengou a little more than 100 years ago, culture of Bordetella pertussis has been the reference standard for diagnosis, but even with optimal specimen collection and laboratory handling, the sensitivity of culture is low and timeliness poor [1]. In the 1990s, randomized clinical trials of pertussis vaccines in children and adults [2] led to new insights, by adding serologic tests and polymerase chain reaction (PCR) to the toolbox for detection of B. pertussis where direct comparison with control subjects was possible. First, the spectrum of clinical symptoms associated with pertussis infection was widened, in both children [3] and adults [4], to include mild and even asymptomatic cough illness. Second, inclusion of milder illness in the case definition substantially lowered vaccine efficacy estimates, more for acellular than for whole-cell vaccines [5]. Third, the whooping cough syndrome could be associated with other organisms, particularly adenovirus, in children [6] or in adults with no detectable infectious cause [4]. In the 20 years to 2010, wider application of serology in many countries has led to much greater recognition of pertussis in adults (who present later and may have less typical symptoms) and PCR has been applied increasingly in outpatient settings in children (who present earlier and for whom venipuncture is a barrier). Increased PCR use has been driven by the availability of commercial kits, and quantum increases in sensitivity with the use of insertion sequence (IS) 481, with its high copy number in B. pertussis as the target [7]. There is evidence that higher test sensitivity creates a feedback loop, stimulating clinician ordering practices [8] and greatly amplifying outbreak size [9]. However, the use of PCR has been controversial, especially outside reference laboratories, because of concerns about false-positives due to contamination, either at the time of collection or in the laboratory [10]. Attention has also turned to the appropriateness of case definitions [11]—full circle for the whooping cough syndrome. The available diagnostic tests, primarily culture, PCR, and serology, have sensitivity and specificity that vary by age, time from symptom onset, and both vaccination and exposure history [1]. This makes the clinical setting of crucial importance in 3 broad categories. The first is respiratory illness in infants <3 months of age; recent experience in California showed that pertussis was often considered only after several presentations in this age group [12]. In this age group, given high morbidity and occasional mortality, early detection and hence sensitivity rather than specificity is the priority, for both case definitions [11] and diagnostic tests. A PCR method with high sensitivity and same-day results such as IS481 is thus ideal, because addition of more specific PCR targets reduces sensitivity. The second category is evaluation of sporadic coughing illness in older children or adults, who are almost always either previously immunized against or exposed to B. pertussis. In this scenario, the major concern is the potential to infect vulnerable infants, because risk of severe illness in the individual is low [9], so IS481 is again appropriate and serologic testing should be undertaken if symptom duration is >3 weeks [11]. The third category is evaluation of outbreaks of coughing illness, either at community level or in closed settings, such as a school or residential institution [13]. Here, emphasis on specificity of diagnosis is paramount, at least in index case(s), especially if vaccine effectiveness is being evaluated. In this issue of Clinical Infectious Diseases, a detailed account is given of a community outbreak of pertussislike illness in Ohio [14]. Rodgers and colleagues from the CDC and their Ohio collaborators, identified Bordetella holmesii in 29% and B. pertussis in 71% of 164 cases among a convenience sample of 298 nasopharyngeal specimens available for retesting, which was 42% of the total of 703 IS481-positive cases. Based on various criteria, a total of 918 cases were reported in the outbreak over a 12-month period, an all-age incidence 80/100 000 population. This was almost 4-fold higher than the average all age incidence of 23/100 000 reported in a similar time period in California, although an incidence of up to 138/100 000 was reported at county level in that State [15]. Identification of B. holmesii was by detection of a specific insertion sequence—IS1001-like—and failure to detect the pertussis toxin sequences ptx-Pr and ptxS1, which are present in B. pertussis. In a small subset of 14 cases, serologic testing confirmed B. pertussis by high anti-pertussis toxin immunoglobulin G levels in 5, but the more specific PCR sequences (ptxS1 or ptxA-Pr) were detected in only 3 of them. This elegant series of laboratory studies established that this was a mixed outbreak of B. holmesii and B. pertussis; Japanese investigators reported a similar situation contemporaneously [16]. Other important findings regarding B. holmesii cases included the facts that although the clinical presentation was similar to that of B. pertussis, cases were clustered in older adolescents and were associated with a shorter duration of cough illness when treated with macrolides. Only 5 hospitalizations occurred among 160 cases of known etiology, all in infants, 4 due to B. pertussis and one to B. holmesii. Among 11–18-year-olds, 60% of B. holmesii case patients had received Tdap vaccine, compared with 44% of B. pertussis case patients, consistent with less or no vaccine effectiveness against B. holmesii. Unanswered questions for B. holmesii include its overall prevalence in cases of pertuss is like illness, whether it is found in asymptomatic persons, and whether multitarget PCR capable of detecting it should become routine laboratory practice. With respect to overall prevalence, B. holmesii was originally identified as a cause of bacteremia in immunocompromised persons and then occasionally found in respiratory specimens [17], but it has emerged as a more prevalent organism in France [18], Japan [16], and now Ohio [14]. B. parapertussis is also likely to be more common than previously recognized [19, 20]. It is now well established that waning of immunity is more rapid with acellular than with whole-cell vaccines [21, 22]. Less rapid waning of immune protection against B. pertussis with whole-cell vaccines may be related to the wider range of antigenic responses they induce, with evidence from a mouse model that whole-cell vaccines, but not acellular vaccines, may also afford cross-protective immunity against B. parapertussis [23]. If this is applicable to humans, it is plausible that long-standing acellular vaccine use could also be a factor in increasing the prevalence of symptomatic infection with a broader range of Bordetella species or antigenic variants of B. pertussis [24]. With respect to laboratory capacity, in Australia only 7% of laboratories that participate in the national quality assurance program report using multitarget PCR assays able to distinguish between B. pertussis and B. holmesii, a situation that may also pertain elsewhere. Not surprisingly, when B. holmesii DNA was included in a sample sent for testing in the 2007, 2010, and 2011 programs, 94%, 86%, and 90% of participating laboratories, respectively, reported this sample positive for B. pertussis by PCR (G. James, personal communication, September 2012). Although perhaps not essential for nonspecialist laboratories, it is important that referral laboratories receiving large numbers of specimens from suspected cases of pertussis are able to identify B. parapertussis and B. holmesii routinely, both to measure and monitor their true prevalence and for use in outbreak situations, where vaccine effectiveness often comes into question. Multiplex PCR incorporating IS1002 with IS1001 and IS481 [25] may be more practical for this purpose than the complex methodology used in the study of Rodgers et al [14]. However, the single most important issue in pertussis diagnosis remains diagnostic awareness, an essential precondition for diagnostic testing. Nowhere is this more critical than for infants who have received ≤2 doses of pertussis vaccine, in whom the criteria for suspicion should be broad and the threshold for testing low [11, 12]. Wide availability of single-target IS481 PCR using real time methods, which minimize the potential for contamination [26], may be the most cost-effective means to ensure ready availability of PCR for initial, rapid identification of potential B. pertussis in laboratories receiving respiratory specimens from young infants. However, even in this setting, careful interpretation of the clinical significance of a positive PCR for B. pertussis is essential, because a recent reevaluation of cases of sudden infant death with positive PCR has shown that this does not establish B. pertussis as the cause of death without appropriate histopathologic correlates [27]. This is an important issue for future studies monitoring the impact of initiatives such as maternal immunization against severe pertussis in early infancy. Potential conflicts of interest. Both authors: No reported conflicts. Both authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.