Systematic community-and hospital-based surveillance for enterovirus-D 68 in three Canadian provinces , August to December 2014

DM Skowronski 1 , C Chambers 1 , S Sabaiduc 1 , M Murti 2 , R Gustafson 3 , S Pollock 4 , D Hoyano 5 , S Rempel 2 6 , S Allison 7 , G De Serres 8 , JA Dickinson 9 , R Tellier 10 , K Fonseca 10 , SJ Drews 11 , C Martineau 8 , F Reyes-Domingo 6 , T Wong 12 13 , P Tang 1 , M Krajden 1 1. 1. British Columbia Centre for Disease Control, Vancouver, British Columbia, Canada 2. Fraser Health Authority, Surrey, British Columbia, Canada 3. Vancouver Coastal Health Authority, Vancouver, British Columbia, Canada 4. Interior Health Authority, Kelowna, British Columbia, Canada 5. Vancouver Island Health Authority, Victoria, British Columbia, Canada 6. Public Health Agency of Canada, Ottawa, Ontario, Canada 7. Northern Health Authority, Prince George, British Columbia, Canada 8. Institut national de santé publique du Québec, Québec, Canada 9. University of Calgary, Alberta, Canada 10. Alberta Provincial Laboratory, Calgary, Alberta, Canada 11. Alberta Provincial Laboratory, Edmonton, Alberta, Canada 12. Health Canada, Ottawa, Ontario, Canada 13. Formerly affiliated with: Public Health Agency of Canada, Ottawa, Ontario, Canada

In mid-August 2014, two paediatric hospitals in the United States (US) reported increases in severe EV-D68-associated respiratory illness [13] that were followed by a nationwide outbreak notable for affecting children with asthma in particular [14].Several countries in Europe also reported EV-D68 activity during the summer and autumn of 2014, including one country, Norway, where EV-D68 was also associated with an increase in hospitalisations for severe respiratory illness [9,[15][16][17][18].In September 2014, the US Centers for Disease Control and Prevention (US CDC) additionally began investigating reports of acute flaccid myelitis of unknown aetiology in children, detecting EV-D68 in some, but not all, of these patients [19][20][21][22]; sporadic cases of EV-D68-associated neurological illness were also reported from France (n=1) and Norway (n=2) [15,23,24].EV-D68 infection is not generally notifiable and laboratory testing for EV-D68 was not widely performed prior to the 2014 outbreak.Consequently, localised clusters of paediatric hospitalisations have largely shaped current understanding of EV-D68 illness, skewing impressions of typical disease severity.Few countries have utilised existing general practice sentinel surveillance schemes for influenza-like illness (ILI) or acute respiratory illness (ARI) to systematically assess outpatient illness due to EV-D68 infection, as conducted in the Netherlands and Germany [8,9,18].
In Canada, pre-existing infrastructure for standardised screening of respiratory specimens was mobilised in response to the alerts from the US.Respiratory specimens collected from ILI patients of all ages attending sentinel outpatient practices in three participating provinces (British Columbia (BC) (population: 4.4 million [25]), Alberta (population: 3.6 million [25]) and Quebec (population: 7.9 million [25])) of the community-based Canadian Sentinel Practitioner Surveillance Network (SPSN) were retrospectively and prospectively screened for EV-D68 before and during the 2014 August-to-December epidemic period.In addition, laboratory-based enhanced passive surveillance was conducted in BC on all detections of EV-D68 from inpatient and outpatient specimens tested at the BC Public Health Microbiology and Reference Laboratory (BC PHMRL) during the 2014 epidemic period.Here we report findings from these two surveillance approaches to inform the epidemiology of EV-D68 including the spectrum of illness, population-based incidence, and potential at-risk groups for severe or unusual sequelae, including neurological or fatal outcomes.

Methods
Epidemiological sampling and laboratory testing protocols are outlined in Table 1 and are summarised below.

Active community-based sentinel surveillance
Community-based sentinel practitioners of the Canadian SPSN collect nasal or nasopharyngeal specimens year-round from patients presenting within seven days of ILI onset defined as acute respiratory illness with fever and cough and at least one of the following symptoms: sore throat, arthralgia, myalgia or prostration [26].Epidemiological information is collected from consenting patients/guardians using a standard questionnaire at the time of specimen collection.Research ethics board (REB) approval is obtained in each participating province.Provincial protocols for EV-D68 detection among EV/ RV-positive specimens are specified in Table 1 [10,[27][28][29][30][31][32].These assays have comparable sensitivities and limits of detection are considered within one log 10 copy/mL, as demonstrated in a national validation study [33].Specific EV-D68 typing was established by partial sequencing of structural capsid viral protein (VP) namely VP1 in BC [10] and Quebec [32] and VP2 in Alberta [28][29][30].

Laboratory-based enhanced passive surveillance in British Columbia
The BC PHMRL is the only site to provide confirmatory diagnosis of EV-D68 in BC.Such testing is usually conducted upon physician request; until 19 September 2014, this was predominantly to confirm diagnosis among inpatients at the province's tertiary paediatric hospital.From 9 September 2014 onwards, this also included routine screening of all respiratory specimens collected from inpatients or outpatients less than 5-years-old.On 19 September 2014, routine EV-D68 screening was extended to all respiratory specimens submitted to the BC PHMRL from all inpatients of any age across the province.Outpatient specimens were also included in routine EV-D68 screening with staggered implementation as shown in Table 1.EV-D68 typing was performed as specified in   a Restricted to non-sentinel cases in British Columbia residents with enhanced surveillance case report forms with valid hospitalisation information collected as at 31 October 2014.Six cases missing information on hospitalisation status were excluded.b Includes 14 infant cases <1 year-old and 20 cases 1-2-years-old (i.e.34 cases <3-years-old).c Includes nine hospitalised infant cases <1 year-old and 16 hospitalised cases 1-2 years-old (i.e. 25 cases <3-years-old).d Includes five non-hospitalised infant cases <1 year-old and four non-hospitalised cases 1-2 years-old (i.e.nine cases <3-years-old).e Five patients remained in hospital at time of reporting and were excluded from length of stay calculations.f Proportions displayed are among those with known information only.g One adult case with neurological illness with specimen collection in November 2014 not included in this table which is restricted to specimens collected from 28 August to 31 October 2014 with full enhanced surveillance case report forms.h In 30 days prior to symptom onset; all five patients reported travel to the United States.i Influenza A(H3N2) (n=1), respiratory syncytial virus (n=1), and Streptococcus pneumoniae (n=8, including one detection in blood and seven in upper respiratory specimen).
into clades corresponding to VP1 designations [12].Phylogenetic trees were constructed in MEGA6, rooted to the 1962 prototype Fermon strain using the maximum-likelihood method with a Jukes-Cantor substitution model and 1,000 bootstrap replicates [35].

Results
Active community-based sentinel surveillance Among 2,078 specimens submitted to the Canadian SPSN, meeting ILI criteria and collected between 1 October 2013 and 31 December 2014, 1,909 (92%) were screened for EV/RV and 221 (12%) tested EV/ RV-positive.There was a mirror-image pattern of EV/ RV versus influenza test-positivity by month, reflecting their differences in seasonality (Figure 1).
During this period, 1,894 of 2,078 (91%) specimens were assessed for EV-D68, of which 18 (1%) tested positive.This includes 1 of 348 (0.3%) collected from October to December 2013 and 11 of 460 (2.4%) from October to December 2014, a significant eight-fold increase in detection rates across two successive years (Chi-square test p=0.01).The remaining EV-D68 detections were from specimens collected in September 2014 (6/37) (Figure 1).There were no EV-D68 co-infections with other respiratory viruses included on the Luminex xTAG RVP panel.

Laboratory-based enhanced passive surveillance in British Columbia
Of 3,716 respiratory specimens collected between 28 August and 31 December 2014 in BC and tested at the BC PHMRL, 239 (6%) were positive for EV-D68, representing 218 unique patients (n=211 excluding seven detected also by the SPSN).The majority (172/211; 82%) had specimens collected between weeks 40 to 45 (early October-early November) (Figure 2) and 72% (139/194) of those with known information were hospitalised.

Case series of EV-D68-associated neurological or fatal outcomes
Five EV-D68 cases (paediatric (n=4); male (n=4)) identified through enhanced passive surveillance in BC were associated with neurological illness including acute flaccid limb weakness (n=3), generalised paralysis (n=1) or head/neck paralysis (n=1) (Table 4) [37].All had preceding respiratory illness and one had concurrent gastrointestinal illness.All were admitted to hospital; three required ventilation support.Case 3 and S. pneumoniae was not isolated from CSF of Cases 3 or 5.As at July 2015 (>9-11 months post onset), all five cases had ongoing neurological deficit.
Three EV-D68 cases in BC with illness onset between August and October 2014 had fatal outcome including: an adult (20-29-years-old) with respiratory failure following acute asthma exacerbation; an adult (≥65-yearsold) with multiple comorbidities and respiratory failure following acute chronic obstructive pulmonary disease exacerbation; and a child (<5-years-old) in whom death was ultimately attributed to Group A streptococcal sepsis.
Of the clade B VP1 sequences, 8 of 179 (4%) had a T146S substitution in the immunogenic D-E loop, and two had adjacent asparagine deletions at positions 144-145 in the D-E loop not found in other published sequences; neither the substitution nor the deletion was found in sequences obtained from neurological or fatal cases.Three VP1 sequences from BC, one SPSN specimen from 2013 and two enhanced passive surveillance specimens from 2014, clustered instead in clade A with recent isolates from France, Italy, and the Netherlands, characterised by an asparagine deletion at position 140 in the D-E loop [7,8].Phylogenetic analysis did not suggest clustering by month, severity, inpatient/outpatient status or asthma history.

Discussion
Canadian investigators used two surveillance approaches to inform risk assessment related to EV-D68-associated illness.These dual surveillance approaches revealed epidemic features of EV-D68 in Canada during the period spanning from August to December 2014.Active sentinel surveillance showed increased EV-D68 detection among outpatient ILI cases affecting all age groups while enhanced passive surveillance showed severe respiratory and neurological disease requiring hospitalisation that occurred at higher incidence, but not exclusively, among children.
The ILI case definition used by the SPSN to standardise outpatient respiratory specimen collection is relatively specific for influenza.It requires fever and cough and at least one other defining symptom such as sore throat [38].By applying this case definition, we will have missed milder illness caused by other respiratory viruses, for which fever may not be a cardinal feature.Accordingly, the actual number of EV-D68 detections by the SPSN was small (n=18).Among outpatient cases of EV-D68 detected through sentinel surveillance in the Netherlands where an ARI case definition was used, fever or cough were each experienced by 13 of 16 cases and sore throat by 8 of 16 [9]; in Germany, which also used an ARI case definition, 16 of 24 EV-D68 cases experienced fever and cough and 11 of 24 experienced fever, cough and sore throat combined [18].With comparable community prevalence, community-based sentinel systems that apply a broader ARI case definition will certainly detect more EV-D68 cases.However, sentinel surveillance systems are not intended for the detection of rare events or for quantifying absolute disease burden [39].They perform best in the detection of highly prevalent conditions and are most valuable for trend analysis.For that purpose, a consistently applied case definition, whether ILI or ARI, is most important.While a single occurrence of a pathogen may reflect chance sporadic detection, multiple case detections across a geographically dispersed network are an indication of widespread community circulation.As such, the eight-fold increase in EV-D68 detection rates by the SPSN in 2014 compared to 2013, even with an ILI case definition, is indicative of epidemic circulation.Nevertheless, it is acknowledged to be an under-estimate of true incidence.
In Canada, adults ≥20-years-old are predominant SPSN participants (>70-75% [26]) and correspondingly comprised two-thirds of outpatient EV-D68 detections.Among specimens obtained from non-elderly adults 20 to 59-years-old and paediatric cases <20-years-old, however, the proportion that tested positive for EV-D68 was equivalent (4%).This suggests that children and adults are at comparable risk for outpatient EV-D68 illness that presents as ILI, although there may be differences based on other presentations.The EV-D68 age distribution we describe as extending to adults is similar to outpatient sentinel surveillance observations reported from the Netherlands and Germany [8,9,18].Although few countries outside of Canada or within Europe have used existing sentinel schemes to explore historic and current EV-D68 patterns [8,9,18], such infrastructure could prove highly informative if invoked elsewhere and may also be efficient for characterising other pathogens displaying sudden, unexpected but widespread activity.
Passive surveillance systems are most sensitive to severe disease, particularly if involving children or clusters.Enhanced passive surveillance conducted in BC also reflected this paediatric hospital-based skew, driven by specimen collection largely initiated at clinician discretion and laboratory-confirmation protocols that initially prioritised specimens collected in hospital and from children.All confirmatory testing for EV-D68 in BC was conducted at the BC PHMRL and this served as the trigger for enhanced data collection in 2014.However, from 19 September 2014, all inpatient respiratory specimens submitted to the BC PHMRL were screened for EV-D68, allowing unique populationbased comparison of hospitalisation incidence by age and sex.Ultimately, about three quarters of EV-D68 detections were hospitalised patients among whom more than 80% were in paediatric age groups, consistent with the age distribution of prior documented outbreaks [5][6][7]11,13,15,17].While this pattern likely reflects a tip-of-the-iceberg hospital-based surveillance phenomenon, children may indeed be at higher risk owing to greater exposure opportunities and lower likelihood of pre-existing cross-reactive immunity, an immuno-epidemiological hypothesis that still requires evaluation.
The majority of EV-D68 cases were detected in early-October to early-November 2014, also consistent with prior documented outbreaks in North America and Europe [3,[8][9][10][11]17] Males were over-represented among paediatric but not adult cases in the 2014 BC enhanced surveillance data, a pattern that has been documented before [3,[6][7][8][9]11,17], but was neither discernible in the small outpatient sentinel series reported here nor in data from Germany [18].Male predominance may reflect increased pre-pubertal prevalence of asthma in boys [40,41].Overall 38% of EV-D68 cases described here reported asthma compared to 7% of the general BC population [40].Asthma is a recognised risk factor for severe EV-D68 illness [1,6,13,15,28], although underlying mechanisms are unclear.Other respiratory viruses, notably RVs with which EV-D68 shares biological features, are associated with acute exacerbation and more severe lower respiratory illness in asthmatic individuals.This effect is thought to be mediated through Th2-skewed pro-inflammatory cytokine production and impaired antiviral responses [42,43].S. pneumoniae was co-detected in upper respiratory specimens from at least 7 of our 111 (6%) hospitalised EV-D68 cases for whom we had complete enhanced surveillance data.While S. pneumoniae is believed to enhance RV-induced disease severity, including asthma exacerbation [44], our finding likely reflects background carriage rates of S. pneumoniae in children.
EV-D68 was associated with fatal outcome in three BC patients, including two adults with underlying comorbidity and one child in whom bacterial co-infection played a role.Enhanced passive surveillance in BC also detected five cases (four paediatric) of EV-D68associated neurological illness, all with preceding respiratory symptoms and prolonged neurological deficit.
The number of EV-D68 cases with neurological manifestations or fatal outcome detected in BC is higher per capita than reported elsewhere during the 2014 epidemic [20,23,24].This likely reflects the centralised, province-wide laboratory screening and enhanced public health follow-up that was undertaken compared to other cluster-driven analyses, although we cannot rule out true regional differences.Other EVs are known causes of neurological disease, particularly poliovirus and EV-A71, but also EV-D94, which is closely related to EV-D68 [45,46].However, only two cases of EV-D68associated neurological illness had been documented globally prior to the 2014 epidemic, both from the US.One involved acute flaccid paralysis in a young adult in 2005 and the second involved a fatal meningomyeloencephalitis in a child in 2008 [3,47].EV-D68 was detected in the CSF of both these cases [3,47].
Clinical features in our neurological case series are consistent with reports elsewhere in 2014.They include acute flaccid limb weakness, bulbar weakness, and/or cranial nerve dysfunction in association with the detection of EV-D68 in respiratory specimens [19][20][21][22][23][24].Also similar to other reports, magnetic resonance imaging findings showed grey matter involvement in multiple spinal levels but mostly affecting the cervical spine [19][20][21][22][23][24].EV-D68 was detected in whole blood from one BC patient, but in the absence of EV-D68 detection in CSF specimens, a causal role for EV-D68 remains unproven in this case series, as elsewhere in 2014 [19][20][21][22][23][24].It should be noted, however, that recovery of poliovirus from the CSF has also only rarely been reported in cases of paralytic poliomyelitis [48].The finding of detectable M. pneumoniae IgM in three of four neurological cases described here is intriguing given similar serologic findings in a small proportion of other recent EV-D68 neurological case reports [20,21] and the independent association between M. pneumoniae and neurological illness [49,50].Like other recent reports [20,21], however, M. pneumoniae was not detected in respiratory or CSF specimens suggesting that antibody findings we report likely reflect prior infection or potentially a false-positive or crossreactive antibody response.
Phylogenetic analysis of EV-D68 viruses from both outpatients and inpatients showed clade B viruses predominated in Canada during the 2014 epidemic period, clustering with recent (2013-2014) sequences from the US and elsewhere but distinct from more historical sequences [9,12].Increased variability in the VP1 region, inclusive of the immunogenic B-C and D-E loops, may have enabled EV-D68 to escape antibody recognition [2,8,9,12,34].We did not identify mutations or clustering by severe outcome, but full genome analysis would be required to assess other neurotropic or virulence markers, such as in the 5'-untranslated region [7,12,51].
There are limitations to our analyses, foremost related to surveillance methods.Heightened awareness through media and other clinician communications likely influenced patient and provider behaviours related to care-seeking, index of suspicion and testing during the 2014 epidemic period.Nevertheless, neither will the surveillance systems have captured all cases nor will the findings reflect true incidence or disease burden.Small numbers limit our power to test statistical associations.A causal versus contributory or coincidental role for EV-D68 in severe illness cannot be concluded; investigations and their timing were mostly at clinician discretion and other aetiologies, including co-infection, may have been under-recognised.We did not assess other types of EV among EV/RV-positive specimens to compare with the EV-D68 experience.Molecular diagnostic testing for EV-D68 was not routinely performed historically and EV-D68 typing assays were developed and validated real-time in response to the evolving 2014 epidemic, also influencing comparisons across space and time.Laboratory protocols showed comparable performance in a national validation study, and all specimens included in the SPSN analysis were collected within seven days of ILI onset (71% within four days); nevertheless, other variation in specimen collection (e.g.type, viral load), handling, transport and processing may have influenced detection rates between participating provinces.These considerations are, however, relevant to all laboratory-based surveillance.Epidemiological data collection and/or reporting were incomplete particularly when drawn from electronic medical records (as per enhanced passive surveillance) rather than from direct patient/clinician interview (as per active sentinel surveillance).Analyses restricted to patients with known information will have underestimated the proportion with some risk factors/conditions.Despite these limitations, the dual surveillance approaches we report suggest generalised increase in EV-D68-associated outpatient illness across a broad age distribution during the 2014 epidemic period.
Severe respiratory and neurological illness requiring hospitalisation predominantly, but not exclusively, affected children, with possible fatal outcome among those with comorbidity or co-infection.Active surveillance, including both outpatient and inpatient settings, is needed from more areas and additional seasons to further inform EV-D68 incidence, spectrum of illness, and potential at-risk groups for severe or unusual outcomes.

Figure 1
Figure 1Epidemic curve of EV/RV, EV-D68 and influenza detections by month of specimen collection, Canadian Sentinel Practitioner Surveillance Network, British Columbia, Alberta and Quebec, 1 October 2013-31 December 2014 (n=1,909) a

Figure 2
Figure 2Epidemic curve by week of specimen collection and hospitalisation status, laboratory-based enhanced passive surveillance, British Columbia, 28 August-31 December 2014 (n=211)

Table 1
Overview of active community-based sentinel surveillance and laboratory-based enhanced passive surveillance used for EV-D68 detection and characterisation, Canada, 2014 EV/RV screening by Luminex xTAG RVP not performed on influenza-positive specimens collected during 2014/15 influenza season (1 October 2014-31 December 2014) in Alberta.
[7]]British Columbia; EV: enterovirus; ILI: influenza-like illness; PHMRL: Public Health Microbiology and Reference Laboratory; RVP:Respiratory Virus Panel; RV: rhinovirus; SPSN: Sentinel Practitioner Surveillance Network.aDefined using specimen collection date.bFullenhancedsurveillancecasereportform completed until 31 October 2014; partial enhanced surveillance data (basic demographic data and hospitalisation status) collected until 31 December 2014.cDefined as acute respiratory illness with fever and cough and at least one of sore throat, arthralgia, myalgia or prostration.dEV-D68cases in out-of-province residents, sentinel patients detected through the Canadian SPSN, and those with unknown/missing geographic information were excluded from enhanced passive surveillance analysis.ePhylogeneticclustersarelabelledcladeA, B, and C as described in Tokarz et al.[12], corresponding to major group 3, 1, and 2 as described in Meijer et al.[9]and lineage 2, sub-lineage 1.2, and sub-lineage 1.1 as described in Lauinger et al.[7], respectively.
229E and OC43, and bocavirus (BC and Quebec only) using versions of the Luminex xTAG Respiratory Virus Panel (RVP) (Luminex Corp., US).During the 2014/15 influenza season, Alberta made the a priori decision to restrict routine EV/RV testing to influenzanegative patients for resource reasons.

Table 1
[2,12,34]etailed case report forms were completed on all EV-D68 detections by the BC PHMRL for specimens collected between 28 August and 31 October 2014, with only basic demographic and hospitalisation status recorded thereafter to 31 December 2014.Forms were completed by local public health practitioners based primarily upon the electronic medical record and reported to the BC Centre for Disease Control (BCCDC) as part of outbreak investigation, exempt from REB approval.Patients presenting with neurological or fatal outcome notified to BCCDC as part of enhanced passive surveillance are further described as a case series based upon information in the electronic medical record and supplemented by direct patient, guardian and/or clinician interview.Phylogenetic analysesPartial VP1 sequences (nucleotides 133-471), including B-C and D-E immunogenic loops, from BC and Quebec were aligned to a subset of representative VP1 sequences in GenBank to establish clade designation[2,12,34]. Since no recombination was observed within VP4, the complete VP4 and partial VP2 (first 215 nucleotides of 5' end) sequences from Alberta were used to align with VP4/VP2 sequences in GenBank and divided

Table 2
Characteristics of patients with specimens collected by the Canadian Sentinel Practitioner Surveillance Network and tested for EV-D68, British Columbia, Alberta and Quebec, 1 August-31 December 2014 (n=506)

Table 3
Characteristics of EV-D68 cases with full case report forms completed, laboratory-based enhanced passive surveillance, British Columbia, 28 August-31 October 2014 (n=146) All values are number (n) and percentage (%) by column where displayed (except in initial header row), unless otherwise indicated.ICU: intensive care unit; NA: not applicable.

Table 4
Summary of clinical and epidemiological findings for EV-D68 cases associated with neurological illness, laboratory-based enhanced passive surveillance, British Columbia, 28 August-31 December 2014 (n=5) fied.Of note, Cases 1, 2 and 5 were Mycoplasma pneumoniae IgM-reactive but, among these, Cases 1 and 2 were M. pneumoniae PCR-negative in respiratory specimens and Cases 2 and 5 were M. pneumoniae PCRnegative in CSF specimens.Streptococcus pneumoniae was detected by PCR in nasopharyngeal specimens from Cases 3 and 5; however, CSF was PCR-negative in CSF: cerebrospinal fluid; EMG: electromyography; EV: enterovirus; GI: gastrointestinal illness; ICU: intensive care unit; ILI: influenza-like illness; IVIG: intravenous immunoglobulin; LUL: left upper limb; NP: nasopharyngeal; RUL: right upper limb.