Virus Coinfection is a Predictor of Radiologically Confirmed Pneumonia in Children with Bordetella pertussis Infection

This article is published with digital features, including a summary slide, to facilitate understanding of the article. To view digital features for this article go to https://​doi.​org/​10.​6084/​m9.​figshare.​13252535.

Pertussis, caused by Bordetella pertussis (B. pertussis), results in significant morbidity and mortality in children. Outbreaks of pertussis have been reported over the past decades, and disease re-emergence has been the focus of international attention [1]. Liu et al. estimated that the percentage of under-5 childhood deaths due to pertussis was 0.9% in 2000–2015 [2]. Yeung et al. also estimated that there were 24.1 million pertussis cases and 160,700 deaths from pertussis in children younger than 5 years in 2014 [3]. In China, the reported morbidity of pertussis declined from 100–200 per 100,000 persons before the 1970s to less than 1 per 100,000 persons in 2012 (2517 in total) due to the introduction of vaccines in the 1970s [4]. However, the reported number of pertussis reached 3426 and 6373 cases in 2014 and 2015, respectively.

Due to limited laboratory infrastructure to support routine pertussis testing and under-recognition of pertussis among young infants, who may present with atypical symptoms, estimating the pertussis disease burden is challenging [5]. Most studies on the disease burden of pertussis selectively enrolled children with predefined symptoms, such as paroxysmal or prolonged cough, whereas infants might present with atypical symptoms according to previous observational studies [6, 7]. There have been few studies on the incidence of pertussis in infants using broad inclusion criteria [8, 9]. However, in China, the prevalence and clinical presentation of pertussis infection in children using broad inclusion criteria have never been thoroughly investigated.

Furthermore, as previous studies have reported, coinfection was quite common in B. pertussis infection. Frassanito et al. found that, in Italy, infants hospitalized for B. pertussis infection commonly have respiratory viral coinfections (47.2%) [10]. A study from Beijing, northern China, revealed a high coinfection rate of 40.1% [11]. Our previous study also found a high coinfection detection rate of 59.4% [12]. To date, the relationship between the clinical severity and B. pertussis infection status with versus without respiratory viruses remains uncertain.

To address these issues, this study aimed to prospectively investigate the prevalence of pertussis in southeast Chinese children hospitalized with lower respiratory tract infection (LRTI) during an outbreak. Furthermore, we aimed to compare the clinical characteristics and the severity of illness in B. pertussis polymerase chain reaction (PCR)-positive children with respiratory virus coinfections to those with a single B. pertussis infection.

Three of the important findings reported in our study are as follows: (1) the burden of B. pertussis-associated hospitalization in children was high in Suzhou, China; (2) coinfections were common in B. pertussis-associated LRTI, with HRV being the most commonly coinfected virus; and (3) no differences in clinical severity were observed between children with virus coinfection compared to those with single B. pertussis infection.

Our B. pertussis-associated infection rate (6.1%) in LRTIs was higher than those of previous studies [8, 9, 17] but lower than those reported by other studies [18, 19]. The discrepancies in the rates of B. pertussis-associated infection may result from different health systems in differing regions, differences in age groups, and laboratory assays. A Soweto, South Africa, study enrolled infants < 1 year of age hospitalized with LTRI and identified a rate of 2.3% [8]. Another study from Cape Town, South Africa, reported a 7% prevalence for the detection of B. pertussis in children < 13 years hospitalized with LRTI [18]. In China, a recent study from Shenzhen reported a rate of 4.51% [20] in children < 14 years [20].

In the present study, 29.2% of all PCR-positive pertussis cases were identified in children who received 3 doses of vaccine. Marshall et al. also reported that 18.8% of the PCR-positive pertussis cases had received more than 2 doses of vaccine [21]. In 2014, Zhang et al. conducted a Cochrane review of efficacy of one-, two-, and multicomponent vaccines in children. They found that the efficacies of multicomponent vaccines in preventing severe and mild pertussis were 84~85% and 71~78%, respectively [22]. Lower immunization rates among adolescents and adults and evolution of B. pertussis have been attributed to the high incidence rates of pertussis despite the children being fully vaccinated [5].

In our study, HRV was the most commonly coinfected virus among children with B. pertussis-associated LRTI, which is in line with previous studies [10, 21, 23]. Other studies have found that RSV was the most commonly coinfected virus [11, 24, 25]. The heterogeneity of social demography, differences in study methodology, and the year of study may influence the documented coinfected viruses. The high prevalence of HRV in B. pertussis LRTI highlights its important role in pertussis in our region. Our recent study also reported that HRV was the most common virus presenting with a pertussis-like syndrome [26]. Previous studies have found that patients shed HRV for weeks after infection, and could be detected in symptomatic and asymptomatic children [27]. However, in our study, the prevalence of HRV was significantly higher among the B. pertussis PCR-positive children than that among the PCR-negative children (25.7% vs. 8.2%, P < 0.01). We believe that this result demonstrates the real high HRV coinfection burn in B. pertussis LRTI.

We have reported that virus coinfection was associated with developing pneumonia but not with severe illness. To our knowledge, studies focusing on the clinical outcome of virus coinfection in pertussis have resulted in divergent findings. Marshall et al. designed a pertussis severity score system, and found that the presence of coinfection was independently associated with severe disease [21]. Kaczmarek et al. investigated pertussis-related pediatric intensive care unit (ICU) admissions, and reported that patients with coinfection had a longer length of ICU stay [24]. A recent study in China found that children with virus coinfection did not experience a more severe disease than those with a single pertussis infection [25]. Variation in age group selection, definition of severity, and adjustment in multivariable analyses may affect the findings.

To investigate whether there were patients who presented with atypical pertussis-like symptoms, we used broad inclusion for investigating B. pertussis-associated LRTI. Our study found that 72.3% of the pertussis PCR-positive patients fulfilled the CDC case definition of pertussis. Soofie et al. reported that only 38% of pertussis PCR-positive cases fulfilled the CDC case definition [8]. Alaso, Phadke et al. reported that only 27.3% met the CDC case definition [9]. This calls to attention that different definitions might be used to determine the true epidemiologic burden of pertussis. The wide use of PCR assays as a diagnostic method for the detection of pertussis might help to provide a new definition for this disease.

An important strength of our study was the use of extensive comparison groups (single vs. coinfection HRV/non-HRV virus, 4 weeks–3 months vs. 4 to < 24 months), each with adequate sample sizes, although our sample size did not allow subgroup analyses for all individual viruses or for more age subgroups to be included. Our study also has some limitations. First, we only assessed the presence or absence of coinfected viruses; we did not measure the virus load, which might have better elucidated the association between the virus and disease severity. Second, as is also the case for other studies, this study was performed at a single tertiary referral hospital and the outpatients with mild LRTI were not included in our study, which limits the generalizability of our findings. Third, we only included patients older than 4 weeks; we did not include neonates, which is unfortunate as many of the severe cases of pertussis occur in them.

We used a two-target assay (ptxS1 and IS481) that allowed identification of B. pertussis, although a false diagnosis of pertussis could be made because IS481 is also found in B. holmesii [15]. However, previous studies have found that B. holmesii infections are rare and mostly occur in older children and adults. Valero-Rello et al. reported that B. holmesii was detected in only 0.9% (only in patients > 4 years old) of patients suspected of pertussis [28]. Burgos-Rivera also reported a low positive rate of 0.66% [29]. Fraj et al. investigated 1844 children with respiratory infections and found that none had B. holmesii infection [30]. A study from Spain re-evaluated 391 B. pertussis-positive (IS481-based) samples by using multiplex RT-PCR. They found that B. holmesii was positive in only 4.1% of the patients with a median age of 9 years old [31]. As we only included children < 2 years old and performed PCR testing on a large number of patients (n = 4287), we determined that the two-target assay was acceptable and profitable for evaluating the burden of B. pertussis in our region, although most of the Chinese clinical laboratories have limited multitarget PCR capacity.

In conclusion, B. pertussis infection contributed to a high proportion of LRTI hospitalizations among southeast Chinese children. Equivalent clinical severity was observed between children with virus coinfection and single B. pertussis infection, although children coinfected with virus presented more frequently with pneumonia than those with single B. pertussis infection.