Incidence and severity of pertussis hospitalisations in infants aged less than 1 year in 37 hospitals of six EU/EEA countries, results of PERTINENT sentinel pilot surveillance system, December 2015 to December 2018

Introduction PERTINENT is a pilot active surveillance system of infants hospitalised with pertussis in six European Union/European Economic Area countries (37 hospitals, seven sites). Aim This observational study aimed to estimate annual pertussis incidence per site from 2016 to 2018 and respective trends between 2017 and 2018. Pertussis cases were described, including their severity. Methods We developed a generic protocol and laboratory guidelines to harmonise practices across sites. Cases were hospitalised infants testing positive for Bordetella pertussis by PCR or culture. Sites collected demographic, clinical, laboratory data, vaccination status, and risk/protective factors. We estimated sites’ annual incidences by dividing case numbers by the catchment populations. Results From December 2015 to December 2018, we identified 469 cases (247 males; 53%). The median age, birthweight and gestational age were 2.5 months (range: 0–11.6; interquartile range (IQR): 2.5), 3,280 g (range: 700–4,925; IQR: 720) and 39 weeks (range: 25–42; IQR: 2), respectively. Thirty cases (6%) had atypical presentation either with cough or cyanosis only or with absence of pertussis-like symptoms. Of 330 cases with information, 83 (25%) were admitted to intensive care units including five deceased infants too young to be vaccinated. Incidence rate ratios between 2018 and 2017 were 1.43 in Czech Republic (p = 0.468), 0.25 in Catalonia (p = 0.002), 0.71 in France (p = 0.034), 0.14 in Ireland (p = 0.002), 0.63 in Italy (p = 0.053), 0.21 in Navarra (p = 0.148) and zero in Norway. Conclusions Incidence appeared to decrease between 2017 and 2018 in all but one site. Enhanced surveillance of hospitalised pertussis in Europe is essential to monitor pertussis epidemiology and disease burden.


Introduction
Most severe cases of pertussis (whooping cough) occur below 5 years of age. Worldwide, it causes substantial mortality in infants (85,900 estimated deaths in 2014) [1]. Sixty-three percent of cases aged less than 1 year reported to the European Centre for Disease Prevention and Control (ECDC) in 2017 required hospitalisation [2]. Severe complications include pneumonia, seizures, encephalopathy and death.
In neonatal infection, the cough with the characteristic whoop might be absent. The initial finding for Bordetella pertussis infection is frequently apnoea [3]. Since infants and neonates often have a clinical presentation that is atypical or similar to several other respiratory diseases, Vittuci et al. support a routine pertussis laboratory diagnosis in all infants aged less than 3 months with acute respiratory symptoms [4]. Adolescents (≥ 11 years of age) and adults (≥ 18 years of age) can develop a mild, often undiagnosed, form of the disease and represent a reservoir of transmission for infants.
Pertussis is one of the least controlled vaccinepreventable diseases in European Union/European Economic Area (EU/EEA) countries [5]. The number of pertussis cases reported to ECDC increased since 2011 despite a primary series of acellular pertussis vaccine coverage exceeding 90% in most countries [6]. The last peak incidence year occurred in 2012 with 42,500 reported cases, impacting 19 of 28 countries with different magnitude [5]. It affected adolescents, adults and particularly children too young to be vaccinated or to have completed the primary series. This increase may be explained by improved diagnostic methods, increased disease awareness, waning of acellular-vaccine-induced protection, or a lower vaccine effectiveness (VE) due to bacterial mutation [7].
At EU/EEA level, the routine pertussis surveillance systems are heterogeneous. Under-diagnosis and undernotification also differ across EU/EEA countries [8]. Even though a majority uses the EU case definition [9,10], differences in laboratory procedures, completeness of reporting and differences in disease awareness still remain [8]. Most case definitions do not distinguish between all Bordetella species that can have similar respiratory presentation and include indifferently B. pertussis, B. parapertussis, B. holmesii and B. bronchiseptica. Bordetella species may not have the same pathogenicity and may not be all directly targeted by pertussis-containing vaccine [3]. Therefore, existing surveillance systems make comparison of pertussis immunisation strategies and estimates of pertussis incidence at EU/EEA level difficult. Following the 2012 outbreak, an ECDC consultation resulted in a recommendation to conduct studies in the EU/EEA to measure the burden of pertussis including infant hospitalisations [11]. In recent years, some countries have enhanced surveillance of pertussis at hospital [12][13][14] or population level [15].
In 2015, ECDC initiated PERTINENT, 'Pertussis in Infants European Network', a hospital-based active pilot surveillance system in seven study sites using the same pertussis surveillance protocol. The surveillance system has two main objectives. The first is to identify the trends in incidence of laboratory-confirmed pertussis in hospitalised infants (≤1 year old) for the respective study sites. The second is to estimate VE using the test-negative design. In the current study, annual incidences per site are estimated from 2016 to 2018 as well as the change in incidence in 2018 compared   to 2017. Pertussis cases are described by severity and other characteristics. VE results are not presented as VE will be assessed once the required sample size is reached.

Study design and setting
The PERTINENT coordination selected seven study sites (Czech Republic; France; Ireland; Italy; Norway; Catalonia, Spain and Navarra, Spain) willing to participate and able to comply with the generic PERTINENT sentinel surveillance protocol [16] and laboratory guidelines [17] and to estimate the participating hospitals' catchment population. We organised site visits and a laboratory workshop to ensure the harmonisation of site-specific protocols allowing pooling of sites' data.
From December 2015 to September 2016, study sites progressively implemented active surveillance in 41 hospitals. A large proportion of hospitals were situated in France (n = 21 hospitals) and other countries had one to six participating hospitals. Each site complied with the local ethical procedures. In May 2018, the number of participating hospitals was reduced to 37 after withdrawal of four Norwegian hospitals ( Table  1). All sites use the acellular pertussis vaccine for the primary series in infants, but national vaccine recommendations and primary schedules vary across sites (Table 1 and supplementary Table S4).

Case identification and recruitment
The study population consisted of all infants aged less than 1 year, likely to be hospitalised in one of the participating hospitals if developing pertussis-like symptoms.
To maximise the sensitivity of the surveillance, we raised hospital physicians' awareness of pertussis clinical presentation [3] and asked them to test all infants presenting at hospital with pertussis-like symptoms. 'Typical' pertussis presentation was defined either by a presence of apnoea; or by a cough associated with at least one of paroxysms, whoop or post-tussive vomiting. When physicians suspected pertussis even though some typical symptoms were missing, pertussis was considered as 'atypical'.
We identified all infants attending the hospital who were tested for pertussis and invited their parents to participate in the study. When required by the local  ethical committee, parents or legal guardians were requested to provide an informed consent. We excluded all patients with missing or pending laboratory results, testing positive to other Bordetella species than B. pertussis or whose legal guardian was unwilling to participate or unable to communicate and give consent. All laboratory-confirmed B. pertussis cases aged less than 1 year at the time of hospitalisation were included in the study (Figure 1).

Definitions
We defined a laboratory-confirmed B. pertussis case as an infant attending one of the participating hospitals (irrespective of the length of stay), aged less than 1 year and testing positive for B. pertussis by PCR (DNA detection of B. pertussis using PCR or real-time PCR in a nasopharyngeal aspirate or swab) or culture (isolation of B. pertussis from the prior-mentioned clinical specimen) regardless of the clinical criteria.
We defined a severe case as a case admitted to the intensive care unit (ICU).
We defined a likely source of infection as a person with a cough who had contact with the case in the 7 to 20 days before the date of symptom onset of the case. For each pertussis vaccine dose, we defined an infant as vaccinated if she/he had received the dose of interest > 14 days before symptom onset. Unvaccinated infants were those who had not received any dose or who had received the first dose ≤ 14 days before symptom onset.

Laboratory methods
To ensure an accurate identification of the Bordetella species, the PERTINENT laboratory guidelines recommend a diagnostic algorithm for DNA detection of Bordetella including a series of three PCRs: a triplex real-time PCR targeting IS481 gene (in B. pertussis, holmesii and some bronchiseptica strains), pIS1001 (B. parapertussis) and RNase P as the human internal control; followed by two confirmatory singleplex tests targeting ptxA-Pr (B. pertussis) and hIS1001 (B. holmesii) genes [17].

Data collection
Using a standard questionnaire, all sites collected a common set of information: demographic, epidemiological, clinical, laboratory data, vaccination status of the infant and household members, risk and protective factors and suspected source of infection. The list of potential sources of infection included close relatives and caregivers. Each study site translated the questionnaire available in English in its country's language. Data were collected through review of clinical case notes, extraction from patient registries or, if not available, interviews with parents or legal guardians.

Denominators
The denominator was the estimated population of infants aged less than 1 year likely to receive care at hospitals participating in the study. The methods to estimate the hospitals' catchment population varied across sites (Table 1). Throughout the study period, we adjusted the catchment population to the loss of the four Norwegian hospitals.

Analysis
We described cases by age, clinical presentations, risk and protective factors, severity. We used median for continuous variables (age in months, gestational age and weight at birth) and frequencies for categorical variables. We estimated the incidence by site and year. In four sites, the protocol was implemented during 2016 and we estimated 2016 incidence prorata temporis. We calculated incidence rate ratios (IRR) for 2018 compared with 2017.
We used Fisher's exact test to compare clinical signs and symptoms by age group and characteristics of cases admitted to ICU with those not admitted to ICU.

Ethical statement
The planning, conduct and reporting of the study was in line with the Declaration of Helsinki [18]. Ethical approval was not needed in Navarra as the PERTINENT study came under the umbrella of the mandatory surveillance system.
The number of reported cases by month of symptom onset ( Figure 2) was highest in August 2016 (n = 29) and in June 2017 (n = 29). Excluding the first months of progressive surveillance implementation, less cases were observed at the transition between years, such as December 2016 (n = 5), January 2018 (n = 6) and December 2018 (n = 2), but not always, as for example June 2018 (n = 5).
The highest number of cases was reported in the age group 0-2 months (n = 287), with a peak in the second month of life (n = 133), then the number decreased by age in months (Figure 3).

Potential source of infection
Information on the potential source of infection was available for 441 cases (94%). Infants' sibling was mentioned for 128 cases (31%), the mother for 106 cases (25%), the father for 82 cases (20%) and the grandparents for 41 cases (10%) ( Table 2).

Severity
Five (1%) of 466 cases with information died. They were aged from 2 weeks to 10 weeks and not vaccinated against pertussis. In two of the five cases, the mother was vaccinated 5 to 10 years earlier. For one case, the mother was vaccinated 5 days before delivery. Vaccination status of the mother was unknown for the remaining two cases.
Among infants eligible for vaccination, seven of the 25 ICU cases (28%) and 65 of the 163 non-ICU cases (40%) were vaccinated with at least one dose of pertussis vaccine (p = 0.279). None of the ICU cases and 35 of the non-ICU cases (21%) had received two or more doses of pertussis vaccine (p = 0.005).

Other Bordetella species
Five of the seven sites provided information on other Bordetella species. We identified 40 respiratory infections caused by other Bordetella species, including 17 B. parapertussis (cases aged 6 to 51 weeks) and four B. holmesii (cases aged 7 to 14 weeks). Of the 17 B. parapertussis cases, two did not have typical pertussislike symptoms but only cough and three of 12 B. parapertussis cases with information were admitted to ICU. Patients infected with B. holmesii had pertussis-like symptoms and did not require ICU admission. Among infants eligible for vaccination, eight of 13 B. parapertussis cases and two of three B. holmesii cases were vaccinated with at least one dose of pertussis vaccine.

Discussion
Over the three pilot years of the PERTINENT active hospital-based surveillance system, we identified 469 laboratory-confirmed B. pertussis cases. We observed a higher incidence in 2017 and a decrease in 2018 in all but one study site. In 2016 and 2017, most of the cases were reported during summer. One of four pertussis laboratory-confirmed hospitalised cases was admitted to ICU. The highest proportion of ICU cases was in infants aged 0-3 months. Five infants died, all were unvaccinated: either too young to be vaccinated or in the month of life targeted for the first dose of the primary schedule.
Despite standard protocols proposed, the PERTINENT pilot surveillance still has limits to consider when examining the findings. Different methods were used by study sites for estimating hospital catchment population. This may have biased the measured incidence and made annual comparison between sites difficult. However, those methods did not change over time and allowed to compute and compare incidence by sites. During the pilot phase, one country had to decrease the number of participating hospitals and we adjusted the incidence denominator accordingly.
Among the five sites that provided information on other Bordetella species, the proportion of patients with other Bordetella species among those testing positive for any Bordetella species was 11% (40/357) on average, ranging from 0 of 25 to 5 of 18. Two sites were not able to differentiate B. pertussis from the other Bordetella species that may induce a similar respiratory presentation upon infection. This likely decreased the specificity of the laboratory testing and increased reported incidences.
PERTINENT laboratories did not so far sequence B. pertussis isolates, which prevented detection of changes in the pathogen. Isolates were stored in optimal conditions to be analysed later on.
Severity of the disease remains difficult to interpret because of different severity ascertainment and clinical practices by country and by hospital. Hospitalised cases are usually severe pertussis cases, however, the probability of being hospitalised for pertussis-like symptoms is heterogeneous across sites due to different referral to hospital practices. In France, any infant aged less than 3 months with pertussis-like symptoms will be admitted regardless of the severity of the disease. In contrast, a large proportion of Norwegian infants will first attend the 'out-of-office emergency primary care' services (legevakt) and only severe cases will be transferred to hospitals. This likely affected the comparison of clinical signs and incidence rate of hospitalised severe cases between sites.
The likely source of infection was ascertained by family interviews, which may require caution in interpreting the results. Caregivers were not reported as probable sources for any of the reported pertussis cases. As parents self-reported the likely source of infection, they may have better remembered signs of coughing in the household members than among the infant's caregivers. Additional questions, laboratory confirmation in suspected sources of infection, different study designs with increased data completeness are needed to identify source of infection.
Taking the above limitations into account, our results suggest a decrease in 2018 in pertussis incidence We suggest that the summer peak observed in reported cases might possibly reflect the seasonality of the disease. It is unlikely that this was due to improved diagnosis in those months as the PERTINENT surveillance system was stable over time. An increase in pertussis cases during summer was previously reported. In the Netherlands where pertussis is a statutory notifiable disease, the annual peak incidence of notifications for all age groups (0-4; 5-12; 13-18 and 19-99 years) between 1996 and 2006 was in August [19]. More recently, using the notifiable infectious disease reporting system in China from January 2004 to May 2018, Wang suggested a seasonality in pertussis cases and a summer peak with a maximum in August [20].
Thirty cases did not have a typical pertussis clinical presentation including four with at least one clinical sign not documented and five cases with documented absence of all pertussis-like symptoms (reason for hospitalisation unknown). The EU case definition for pertussis was revised in June 2018 [21] to draw attention on atypical symptoms in adults, adolescents or vaccinated children. Our results may suggest the existence of atypical pertussis in infants [3] and highlight the need to raise clinicians' awareness about possible under-diagnosis of pertussis in that age group. The World Health Organization and United States Centers for Disease Control and Prevention pertussis case definitions do not include isolated apnoea or cyanosis in the clinical criteria for pertussis surveillance. In our study, pertussis cases aged less than 3 months were more likely to present with apnoea or cyanosis, which supports the inclusion of these clinical criteria in the pertussis case definition.
Our results suggest that cases admitted to ICU were younger and less vaccinated than non-ICU cases. In the PERTINENT hospital-based sentinel network, four pertussis deaths were reported in 2017 across the seven study sites. As pertussis cases identified in the study were followed up during hospitalisation, we expect no under-reporting of deaths. Among pertussis cases aged less than 1 year reported to TESSy, there were three pertussis deaths in 2017 across the 29 EU/EEA reporting countries [2]. This may suggest that detection of pertussis hospital death is more sensitive in the PERTINENT system. As described in other systems, deaths may be under-ascertained in routine hospital based surveillance in EU/EEA countries [22,23].
The most likely source of infection reported by the parents was firstly the patient's siblings followed by the mother. Recent studies have also shown an increased risk of transmission to siblings of primary cases [24] and a shift in the source of infection from the mother to the siblings [25]. This may be a consequence of vaccinating the mother either before, during or after pregnancy (cocooning strategy), therefore preventing transmission to infants.
Bordetella species can be isolated from both nasopharyngeal swabs or aspirates but a 15% gain in the isolation rate can be obtained by using aspirates in neonates and infants [26]. In our study, we reached a good quality of specimen collection with a high proportion of nasopharyngeal aspirates (82%). We identified 40 infections caused by other Bordetella species, including four with B. holmesii, which is rarely isolated in infants [27]. Even though other Bordetella species are not directly targeted by pertussis-containing vaccine, the later may also induce some cross-immunity for specific Bordetella species [28].

Conclusions
This pilot project shows that enhanced pertussis surveillance in Europe is possible. The generic protocol presented some challenges and efforts by all partners were needed to improve data quality and laboratory procedures but we believe this allowed to pool sites' data to better describe hospitalised laboratoryconfirmed pertussis cases, as these were recruited using the same criteria across six EU/EEA countries. However, a larger sustained project is needed with additional countries to ensure representativeness in Europe and a particular emphasis on harmonisation of laboratory methods. In the future, this surveillance network should allow monitoring emergence of atypical pertussis presentation, identifying upcoming pertussis epidemic cycles and comparing incidence over time in Europe according to immunisation strategies. It will also allow measuring the effectiveness of infants' and mothers' vaccination.