Epidemiologic and phylogenetic analysis of the 2018 West Nile virus (WNV) outbreak in Israel demonstrates human infection of WNV lineage I

Yaniv Lustig1, Ruslan Gosinov2, Neta Zuckerman1, Yael Glazer2, Laor Orshan3, Danit Sofer1, Eli Schwartz4,7, Gili Schvartz5, Yigal Farnoushi5, Avishai Lublin5, Oran Erster5, Uri Shalom6, Tamar Yeger6, Orna Mor1,7, Emilia Anis2,8, Ella Mendelson1,7 1. Central Virology Laboratory, Ministry of Health, Tel-Hashomer, Israel 2. Division of Epidemiology, Ministry of Health, Jerusalem, Israel 3. Laboratory of Entomology, Ministry of Health, Jerusalem, Israel 4. Institute of Tropical and Travel Medicine, Sheba Medical Center, Tel Hashomer, Ramat-Gan, Israel 5. Kimron Veterinary Institute, Beit Dagan, Israel 6. Ministry of Environmental Protection, Jerusalem, Israel 7. Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv, Israel 8. Braun School of Public Health, Hebrew University and Hadassah, Israel

As at 12 November 2018, an outbreak of West Nile virus (WNV) was responsible for 139 WNV infection cases in Israel.Here, we characterise the epidemiology of the outbreak and demonstrate that only WNV lineage I was circulating in mosquitoes and responsible for WNV infection in humans.This suggests that the concurrence of the outbreak in Israel with WNV outbreaks in several European countries is not due to a common, more virulent WNV genotype.
Sequenced West Nile virus (WNV) strains in Israel typically belong to two distinct clusters within WNV lineage I [1,2] The concurrence of ongoing WNV outbreaks in Israel, as well as in several European countries, during the 2018 transmission season [4,14] prompted us to investigate the epidemiological and phylogenetic characteristics of the outbreak in Israel and assess its relatedness to the outbreak in Europe.

Epidemiological characteristics
WNV infection, a notifiable disease in Israel, is diagnosed by identification of immunoglobulin M (IgM) and G (IgG) antibodies in serum and CSF and WNV RNA in whole blood and urine samples [5].
As at 12 November 2018, 139 cases were diagnosed with WNV infection in Israel, of which 76 involved neurological complications and seven died.Sporadic cases were notified during weeks 9, 11, 15 and 17.A significant increase in the number of weekly cases has been observed since week 24 (Figure 1); the median age of cases was 63 years.
For comparison, in 2016 and 2017, 88 and 47 WNV cases were notified with median ages of 66 and 63 years, f:m ratio was 28:61 and 10:37, respectively.This was within the average number of cases of 79±44 for the years 2001 and 2017 in Israel [6,2].In 2015, there was an outbreak of WNV with 149 notified cases.The 2015 outbreak started later in the WNV season (week 29) than the 2018 outbreak and the median age of cases was 68 years (f:m ratio 64:85).
In 2018, a One Health programme was initiated in Israel, which integrates WNV data obtained from humans, mosquitos, equine and birds.Within this initiative, the public health and veterinary services, Israel Nature and Park Authorities and the ministry of environmental protection report all WNV cases to the division of epidemiology, which, in turn, issues a monthly report during the WNV season [8].
In Israel human WNV infections primarily occur in central Israel, which, in 2018, coincides with confirmed infection in birds, equine and mosquitoes (Figure 2).

Phylogenetic investigation
Seventeen qRT-PCR WNV-RNA-positive samples, obtained from WNV patients (five samples) and mosquito pools (12 samples) during 2018, were amplified by PCR for 890 nucleotides (nt) encoding part of the capsid and membrane (M) proteins and all of the premembrane (prM) protein [7].Raw data was trimmed using Sequencher 5.4 (GeneCodes, Ann Arbor, Michigan, United States (US)) to generate 691 bp sequences of the 17 Israeli samples from 2018 which were aligned with six isolates from 2015 in Israel [2] and 16 WNV lineages I and II reference strains.The phylogenetic tree clearly demonstrates that all five samples obtained from humans clustered with the eastern European subtype of WNV lineage I, cluster 2 (Figure 3).Interestingly, the most closely related WNV reference strain was isolated from the first human case of WNV infection diagnosed in Cyprus in August 2016 [9,10].WNV strains found in mosquitoes belonged to both the Mediterranean and eastern European subtypes of WNV lineage I, cluster 2.

Discussion
In 2000, the largest West Nile fever (WNF) outbreak occurred in Israel with more than 400 WNF and West Nile Neuro-invasive Disease (WNND) cases [11][12][13]; since then, Israel has faced WNV outbreaks every few years [6].While human WNV infections in Europe are prevalent [14], outbreaks are generally less frequent and occur every few years in distinct areas [15].Notably, in 2018, a threefold increase in cases was recorded with simultaneous outbreaks in many countries across Europe, including countries with previously limited WNV circulation [14,15], and WNV lineage II was identified in Greece, Austria and Italy [4,16].Similarly to the outbreak in Europe, the outbreak in Israel started early in the WNV season [14] but is likely unrelated to the European WNV outbreak, as only WNV lineage I was identified in all WNV human cases and WNV-positive mosquito pools.
As an endemic country, WNV morbidity in Israel is high in several areas and especially in the central part of the country [17].Here, we show that outbreaks (including the 2018 outbreak) are not different in their geographical spread but rather in the intensity and attack rate of the disease [6].On the other hand, WNV in Europe appears to have different dynamics with areas with high morbidity in 1 year may have no WNV cases in the following year [14,15].These different patterns could be attributed to environmental conditions and/or seasonal WNV circulation.In this context, the current WNV season is unique due to the high morbidity and geographical spread of WNV in such an extensive area in central Europe and the Mediterranean basin.
Deciphering the underlying causes for WNV outbreaks is essential for prediction and prevention of future WNV infections.As several lineages, clusters and subtypes were identified circulating in Europe and the Mediterranean basin, analysis of their phylogenetic characteristics during an outbreak may determine the spread of the virus and potentially its origin.During the largest WNV outbreak in 2000 in Israel, at least two WNV clusters were infecting the human population [1], while in both 2015 and 2018, only one WNV type was responsible for all sequenced human infections [2].Since 2010, most outbreaks in Europe were attributed to WNV lineage II [3] and two recent studies demonstrated that this WNV lineage II strain has significant genomic homogeneity in Europe, suggesting that it was

Number of cases
Year and week of notification introduced as a result of a locally-amplified single penetration event [18,19].

Conclusion
The year 2018 has been exceptional in its severity with simultaneous and significant WNV outbreaks occurring across many European countries and in Israel.Moreover, infection in humans was detected very early in the transmission season raising the suspicion of emergence of a common, possibly more virulent, WNV strain.Our phylogenetic analysis demonstrates that all WNV sequences detected both in humans and in mosquitoes in Israel were related to WNV lineage I, unlike the outbreak in Europe, thus providing evidence for potential for different introductions of WNV lineages of the virus to Israel and Europe and implicating that indirect causes, e.g.climatic change, could be a potential cause of the simultaneous outbreak in Israel.Phylogenetic analysis was conducted on 40 WNV nt sequences of genes encoding the capsid, premembrane protein and membrane protein, with a total of 691 nt in the final alignment.Sequences included current Israeli WNV outbreak (Israel_2018), previous Israeli WNV outbreak (Israel_2015) and additional references.For sequences from the current 2018 outbreak, their origin is denoted by human/mosquito pictures.
Trees were inferred by maximum likelihood based on the general time reversible model with invariant sites (GTR + I), predicted as the best-fit substitution model by JModelTest.The robustness of branching pattern was tested by 1,000 bootstrap replications and the percentage of successful bootstrap replicates is indicated at the nodes, where only values of > 60% are indicated.

License and copyright
This is an open-access article distributed under the terms of the Creative Commons Attribution (CC BY 4.0) Licence.You may share and adapt the material, but must give appropriate credit to the source, provide a link to the licence, and indicate if changes were made.
This article is copyright of the authors or their affiliated institutions, 2019.

Introduction
Respiratory syncytial virus (RSV) is a frequent cause of respiratory infection and the main cause of bronchiolitis [1], the most common reason for hospital admission in infants in the United Kingdom (UK) [2] and many other developed countries.RSV also leads to over 450,000 primary care consultations in children in the UK every year [3].In childhood, symptoms of RSV infection are more severe at the youngest ages [4].Further, RSV infection in young infants has been linked to wheezing and asthma in later childhood [5,6].
There is a lack of cost-effective preventive strategies for RSV infection.Palivizumab, a humanised monoclonal RSV antibody, has been shown to reduce hospital admissions in children born preterm or with congenital heart disease [7], but is costly [8].It is therefore only recommended in children at high risk of serious complications, who have chronic heart or lung conditions or immunodeficiency [9].However, over 80% of infants who are admitted to hospital with bronchiolitis in the UK are born at term and otherwise healthy [10].
RSV vaccination is likely to be the most effective preventive strategy, and several vaccine candidates are under development [11].Options include targeting pregnant women, infants or older children.Vaccine strategies need to account for the short period of immunity following natural infection [12] and the young age of children with the most serious complications of RSV.Older siblings are an important source of infant infections in low-income country settings [13], however, there have been few studies of the contribution of older siblings to the burden of RSV hospital admissions in high-income countries [14].The design of an effective RSV vaccination programme needs information on risk factors in order to effectively prevent the maximum number of cases and severe complications.A systematic review identified few large-scale studies of multiple risk factors for confirmed RSV infection in economically developed countries [15].Linkage of health administrative databases, including hospital admission data, offer a whole-population cohort design for investigating risk factors for RSV-related hospital admission [10,[16][17][18].These studies have highlighted the importance of clinical risk factors for RSV-related admissions in infancy, such as prematurity and chronic heart and lung disease.However, estimation of risk related to family structure is difficult as such data are limited in electronic hospital databases and diagnostic coding is unreliable for specific infections such as RSV [19,20].
In this study we examine the relative contribution of child, family and healthcare risk factors for RSVassociated hospital admission in the first 3 years of life to inform vaccination and other preventive strategies.

Data sources, study population and period
We developed a birth cohort study of all singleton children born in Scotland to Scottish resident mothers between October 2009 and September 2012, using linkage between the following national administrative health databases: birth and death registration records, maternity records (the Scottish Morbidity Record; SMR-02), Scottish Birth Records (which records neonatal diagnoses), hospital admissions (SMR-01) and the infant vaccination registry (Scottish Immunisation Recall System, SIRS).RSV-confirmed admissions were identified via linkage to the Electronic Communication of Surveillance in Scotland (ECOSS) database, a public health surveillance database held by Health Protection Scotland (HPS), the national infection control agency.We excluded children with a birth weight less than 500 g and children born at less than 24 weeks to ensure exclusion of stillbirths.Deterministic linkage between databases was carried out using the Community Health Index (CHI) number, a unique individual identifier used across the Scottish National Health Service (NHS Scotland) from birth [21].The completeness of CHI number in the datasets that were linked by the electronic Data Research and Innovation Service (eDRIS) for this study was very high, including 99.6% of records in birth registration data, 99.7% in Scottish Birth Records, and 99.8% in SMR-01 (Carole Morris, eDRIS, personal communication, June 2018).Linkage was carried out by eDRIS, and only de-identified data was made available to the research team.
Children were followed from birth until their third birthday, their date of death or the date of moving out of Scotland (as recorded on the national CHI database), whichever occurred first.Outcomes were measured in the period 1 October 2009 to 30 September 2015 to ensure all children were followed up for 3 years from birth (unless they died or emigrated before this).

Child-level risk factors
Gestational age was classified into a four-category variable using an established classification [23]: extreme, severe and moderate preterm (< 34 weeks), near-term (34-36 weeks), term (37-40 weeks) and post-term (≥ 41 weeks).Birth weight for gestational age categories were derived using birthweight centiles for Scottish children [24].Small and large for gestational age was defined as having a birthweight less or greater than the 10th percentile for the particular gestational age, respectively.Children who were neither large nor small for gestational age were categorised as normal for gestational age.We used a previously developed list of ICD-10 codes to identify children with chronic conditions, including chronic heart, lung and neurological conditions, who are at increased risk of RSV-related complications [25].We searched for the relevant codes in Scottish Birth Records and during longitudinal hospital records up to the age of 6 months.Annual quarter of birth was coded into a four-category variable (January-March, April-June, July-September, October-December).Apgar score at 5 minutes was coded as a binary variable (0-7 and ≥ 8).Prolonged postnatal stay after delivery was used as a further indicator of earlylife neurological or respiratory problems with potential long-term sequelae, and was coded into a binary variable: ≤ 14 days or > 14 days.

Family-level risk factors
Number of siblings at birth was derived using the parity variable in SMR-02 maternity records, and coded into a three-category variable: no siblings, one sibling or two or more siblings.Maternal smoking during pregnancy was coded as a binary variable (yes/no).Socioeconomic status was measured using maternal age and arealevel deprivation.Maternal age was coded into a fourcategory variable (< 20, 20-29, 30-39 and ≥ 40 years).The Scottish Index of Multiple Deprivation (SIMD) is an area-level indicator of deprivation derived from several indicators (mainly from the UK Census) [26].SIMD scores are calculated at small-area level, where each area includes 500 to 1,000 persons.SIMD scores were grouped into quintiles and linked to the child via the maternal postcode at delivery.

Health service-related risk factors
We examined vaccination delay as an indicator of access to preventive health services.Children in the birth cohort should have received three doses of pentavalent (diphtheria/tetanus/pertussis/polio/ Haemophilus influenzae type B) vaccine and two doses of pneumococcal conjugate vaccine by 4 months of age.To examine the association between vaccination history and risk of RSV admission we defined a binary variable indicating delayed infant vaccination if a child had not received all required doses by the age of 6 months.

Statistical analyses
All statistical analyses were carried out using Stata version 13 (StataCorp LP, College Station, TX, USA).We determined the number of RSV admissions according to month of age in the first 3 years of life, and estimated RSV admission rates per 1,000 child years by year of age and each risk factor.We estimated the median and interquartile range (IQR) of the length of stay of the RSV admissions according to each risk factor, and the proportion of total bed days (that is, the total number of days in hospital) during RSV admissions by each of the risk factors.Kruskal-Wallis tests were used to compare length of stay distributions.To calculate bed days, children with a length of stay of 0 days were allocated 0.5 bed days.
Only the first RSV admission for each child in was included in the statistical models, and children who were admitted were censored at their admission date.We used Cox proportional hazards regression models to estimate adjusted hazard ratios according to each individual risk factor, adjusted for all others.The proportional hazards assumption was checked using cumulative hazards plots.We included all risk factors in the model a priori.A separate model was fitted to evaluate the effect of delayed infant vaccination (since we defined this variable at age 6 months), with followup started at age 6 months, rather than at birth.Since there was a non-negligible proportion of missing data on key risk factors, we used multiple imputation with 15 imputations to estimate model parameters.A Wald test p value < 0.05 was used to determine whether a particular model parameter was significantly associated with the outcome.Complete case analyses were carried out as sensitivity analyses.Population attributable fractions were calculated for all risk factors which were significantly associated with the outcome using the punafcc command in Stata [27], from the complete case models.For the population attributable fractions, we assumed that the observed associations between each risk factor and RSV admission risk were causal.

Ethical approval
The study was approved by the Public Benefit and Privacy Panel for Health and Social Care, reference number 1516-0405.

Results
The cohort included 169,726 children, who were followed for an average of 2.95 years.The characteristics of children in the cohort are shown in Table S1, and the linkage outcomes in Figure 1.There were 6,158 RSVpositive ECOSS episodes linked to the birth cohort, of which 5,384 (87.4%) were linked to a hospital admission within 14 days of the sample date.1).The overall median length of stay was 2 days but children born at less than 34 weeks or with chronic conditions had a median length of stay of 3 days.Children with known chronic conditions accounted for 11.1% of admissions but 20.4% of bed days.
The overall RSV admission rate in the first, second and third year of life was 21.9, 7.0 and 2.0 per 1,000 child years respectively.Unadjusted RSV admission rates were higher throughout the first 3 years of life for children with chronic conditions, born at less than 34 weeks of gestation, with a postnatal stay of more than 14 days or a 5-minute Apgar score of less than 8 (Table S2).After the first year of life, RSV admission rates were similar by season of birth and size for gestational age.
In the fully adjusted model, premature birth, presence of chronic conditions, birth between July and December and being small for gestational age were the birth characteristics independently associated with an increased risk of RSV admission (  Delayed infant vaccination was associated with an independent and statistically significant increased risk of RSV admission (adjusted HR 1.14, 95% CI: 1.03-1.25;see Table S3).The increased risk associated with presence of older siblings was substantially attenuated for children aged over 6 months (e.g.adjusted HR for presence of one sibling 1.44; 95% CI: 1.31-1.59;Table S3) Complete case analyses were very similar to models based on multiple imputation (Table S4).
Only 6.5% of RSV admissions would be prevented if the excess risk among children with chronic conditions was eliminated (Table 3).If all children had similar risks to post-term babies, 19% of RSV admissions could be prevented.However, removing the risk posed by older siblings or due to lower maternal age would reduce the number of RSV admissions by over 30%.

Discussion
Half of all RSV admissions in the first 35 months of life occur during the first 6 months, and 30% of admissions after age 1 year.Although children born with chronic conditions were at significantly increased risk of RSV admission throughout the first 3 years of life, only 19% of RSV admissions occurred in these high-risk children.Children with older siblings were twice as likely to be admitted with RSV infection and we estimated that reducing this risk would reduce the number of RSV admissions by nearly one third.We also identified a 14% increased risk of RSV admission among children who had a delay in completing the infant vaccination programme.
This is the most comprehensive study of RSV admissions in the UK to date, and one of the few large-scale studies for RSV admission in the literature using laboratory data to confirm RSV infection [14,25,28].A key strength of this study was the use of a national birth cohort including all singleton births in Scotland over 3 years with follow-up through linkage to administrative health databases.This allowed us to quantify the contribution of each risk factor through calculation of population attributable fractions.Our approach also ensured minimal selection bias, loss to follow-up, and sufficient numbers to examine risks in small subgroups of children, including children born at less than 34 weeks' gestation.The well-established national data linkage infrastructure and universal recording of the CHI number on all healthcare interactions in Scotland enabled the examination of both clinical and family risk factors, and the association between vaccination history and risk of RSV admission.Rich administrative data resources such as this are required to measure the impact of a future RSV vaccination programme.
A limitation of using linked administrative health databases is that testing for RSV in children presenting to hospital is likely to favour inclusion of children with more severe symptoms and those in high-risk groups.This is because there is no national swabbing and testing programme for children presenting to hospital with symptoms of respiratory infections in Scotland or elsewhere in the UK.Testing practices are likely to vary both according to hospital and child characteristics, as well as time of year.It is likely that younger children, those with more severe symptoms, children born prematurely or with chronic illness are more likely to be tested.This may have underestimated the number of RSV admissions in older children.Likewise, length of stay would be longer for children who are tested for RSV if the likelihood of testing is related to increased illness severity; hence the estimates presented here may not be representative of all RSV admissions.If the probability of being swabbed and tested is higher among children with underlying health problems, this would lead to an overestimation of the hazard ratios and population attributable fractions for presence of chronic conditions and premature birth.A second limitation is that we could not define some risk factors explored in previous studies, such as early socialisation through group childcare [29], or mode of delivery [30].Finally, we could not account for children who received prophylactic treatment with palivizumab since there is no national individual-level hospital prescribing database in Scotland (or elsewhere in the UK).
Our findings have important implications for the design of a future RSV vaccination programme and for policies to prevent RSV admissions in the pre-vaccine era.As reported in other studies [17,25,28], we identified an increased risk of RSV admission among children born prematurely and with chronic conditions, and RSV admission rates in these children remained higher than among low-risk children across the first 3 years of life.Children in these high-risk groups accounted for a disproportionate number of bed days.Therefore, any future RSV vaccination programme needs to ensure that these children are protected until at least 3 years of age.Maternal vaccination is an attractive strategy since it would protect the youngest infants, however, it would require careful timing during pregnancy to ensure premature babies are protected.
We identified presence of older siblings as an important risk factor for RSV admission.Older siblings are likely to spend substantial amounts of time in closed settings outside the home (nurseries and schools) where infection risk is high [31].A household study in Kenya identified older siblings (the vast majority of whom were attending school) as the source of nearly three quarters of infant infections [13].Children are also more likely than adults to spread infection to infants due to frequent hand-to-mouth contact, and lack of hygienic practices (such as frequent handwashing and sneezing into tissues rather than hands).
Eliminating the risk posed by older siblings could reduce the number of RSV admissions by a third across the first 3 years of life.Vaccinating older siblings could therefore be an important strategy in the UK and other high-income countries in order to prevent RSV admissions among the youngest children [14].As Poletti et al. [32] demonstrate in a low-income country setting, there are different strategies through which protection from infection risk posed by older siblings could be incorporated into an RSV vaccination programme.For example, it could be achieved by a school-based vaccination programme.Such a programme would lead to a reduction in population-level transmission as well as directly protect young siblings of school-aged children from infection.Alternatively, sibling vaccination could complement a newly introduced routine infant vaccination programme for the first few years after introduction; however, the effect would wane with the number of years since introduction.Ultimately, the optimal design of an RSV vaccination programme which effectively prevents infections and hospital admissions among infants and young children will depend on the availability, effectiveness, safety, likely uptake and cost of vaccines for different age groups.However, our results indicate that vaccination of older children should be considered as a potential scenario in future cost-effectiveness models for RSV vaccination programmes in a UK context.Further studies are required to determine the risk of RSV admission according to the age of older siblings.
Our results suggest other measures to reduce the spread from older siblings are also likely to reduce RSV admissions.A systematic review has identified evidence from several clinical trials that handwashing is effective in reducing transmission of respiratory viruses from younger children in particular [33].
Interventions to limit transmission from siblings should focus on infants born between July and December, who had the highest risk of RSV admission in our study.
The risk of RSV admission in this study increased with decreasing maternal age.An increased risk of RSV admission with younger maternal age has also been reported in other studies [28,34].This is likely explained by the strong association between young maternal age and low socioeconomic status.Further studies with more detailed data on living standards are required to explore which aspects of socioeconomic status, such as poor housing quality [35,36], explain the observed increased risk of RSV admission in children with younger mothers.In addition, low socioeconomic status (as indicated by young maternal age) is associated with premature birth and intrauterine growth restriction, which in themselves are risk factors for RSV admission, as we have demonstrated.Thus, the total contribution of young maternal age or maternal smoking to the risk of RSV admission may be underestimated in this study.Future work should also examine the causal pathways through which low socioeconomic status affects the risk of RSV admission.
We observed a small but significant increase in RSV admissions associated with delayed infant vaccination.Our study is the first to examine the role of timely infant vaccination and the risk of RSV admission in a high-income country setting [15].None of the current infant vaccines are expected to provide direct protection from RSV infection (only secondary protection through prevention of bacterial co-infection).Delayed infant vaccination could indicate a lack of access to preventive health services.However, children whose infant vaccinations are delayed or incomplete are more likely to be from poor socioeconomic backgrounds or have long-term chronic illness [37][38][39], which may explain the association observed in this study.
Our study highlights that any future RSV vaccination programme will need to protect children throughout the early life course, and in particular children with chronic conditions, who remained at increased risk throughout the first 3 years of life.Further, protecting young children from infection risk posed by older siblings, including through vaccination of older children, could have a substantial impact on reducing RSV admissions.

*Erratum
A general note under Table 3 was erroneously labelled as note a.This was corrected and the remaining note was reordered on 10 January 2019.

License and copyright
This is an open-access article distributed under the terms of the Creative Commons Attribution (CC BY 4.0) Licence.You may share and adapt the material, but must give appropriate credit to the source, provide a link to the licence, and indicate if changes were made.
This article is copyright of the authors or their affiliated institutions, 2019.

Introduction
Neisseria meningitidis is a major cause of meningitis and septicaemia worldwide.Efforts to control meningococcal disease have been aimed at the development of effective vaccines and subsequent implementation in appropriate immunisation schedules.
Twelve different meningococcal serogroups are recognised and serogroup B (MenB) is currently responsible for most cases of invasive meningococcal disease (IMD) in Europe.Many countries, including the United Kingdom (UK), experienced large outbreaks of serogroup C (MenC) disease in the mid-1990s, due mainly to the ST-11 clonal complex (cc11) [1].In 1999, the UK became the first country to introduce the MenC conjugate (MCC) vaccine in a phased national campaign targeting all those aged less than 18 years over a 12-month period, alongside a routine three-dose infant programme [2].Vaccine eligibility was later extended up to 24 years of age, although this age group was not actively called for vaccination.Routine use of MCC vaccine in many other European countries followed [3].
As MCC vaccines were licensed on immunogenicity studies alone, without direct evidence of clinical efficacy, comprehensive national surveillance was initiated concurrently in order to monitor the vaccines' impact in a population-based setting [2].MCC vaccination was associated with rapid and sustained declines in MenC disease across all age groups through direct and indirect (herd/population) protection [3][4][5].
Invasive disease is rare following nasopharyngeal acquisition where the meningococcus can persist for several months before it is cleared; this asymptomatic carriage, especially in older teenagers, is an important reservoir for infection, onward transmission to susceptible individuals and immunity.In the UK, there was a 66% reduction in MenC carriage among 15-19 year-olds within 1 year of MCC vaccine implementation and this reduction was key to establishing indirect protection across the population [6].
The UK MCC immunisation programme has evolved over time (Box) and, from 2013, has included an adolescent MCC vaccine programme to extend direct protection in teenagers and maintain indirect protection in the wider population.In 2015, this vaccine was replaced with the quadrivalent meningococcal conjugate vaccine (MenACWY) to combat a national MenW outbreak.A multi-component, protein-based vaccine against MenB was also implemented in the infant immunisation schedule at the same time [7].In July 2016, the infant MCC dose was removed because MenC cases were extremely rare in this age group and population protection was likely to be maintained through the adolescent MenACWY programme.

National epidemiology
Public Health England (PHE) conducts enhanced national surveillance of IMD in England.The Meningococcal Reference Unit (MRU) provides a national service for confirming, grouping and characterising invasive meningococcal isolates [11].The MRU also provides free national Polymerase Chain Reaction (PCR)-testing of clinical samples from patients with suspected IMD; Diagnostic laboratories are required by law to notify PHE when they identify Neisseria meningitidis and samples are requested to be sent to the MRU for confirmation and characterisation.PHE routinely reconciles laboratory-confirmed cases to generate a national dataset [12] and follows up each case for demographic data, vaccination history, clinical presentation and outcomes through Health Protection Teams (HPTs), general practitioners and, for younger cases, hospital clinicians.Details documented on HPZone (a web-based software for public health management of infectious diseases used by HPTs throughout England), were also accessed by PHE to obtain additional information on cases since 2010.

Vaccine effectiveness
Using the national dataset, MCC VE was calculated using the screening method [13] for confirmed MenC cases among MCC-eligible individuals (born since 1 September 1981) in England between 01 January 2000 and 30 June 2016.VE was estimated according to time since vaccination (or, if unvaccinated, since the age at last scheduled dose); analysis included unvaccinated children, those who had completed the recommended primary immunisation course for their birth cohort (three doses for those aged 1 year of less before September 2006), or had received one dose after their first birthday at any age.Effectiveness of the primary plus booster dose was based on cases who had received ≥ 1 primary MCC dose followed by the 12-month Hib/MCC booster.The following formula was used: Where PCV is the proportion of vaccinated MenC cases and PPV is the proportion of vaccinated population (i.e.vaccine coverage, matched to each case based on their age and birth cohort and then averaged and 24 months of age.These estimates of partial vaccination were used to adjust the national coverage estimates to give coverage in those who received either no vaccination or the full the schedule.For example, it was estimated that in the three-dose primary cohort, 5.5% received one or two doses by 12 months, so national coverage of 90% for three doses would be adjusted to (90*100)/(100-5.5)= 95.2%,once partial vaccination is excluded.

Molecular characterisation
Typing data and genomes were obtained from the PubMLST Neisseria database and Meningitis Research Foundation (MRF) Meningococcus Genome Library (MGL), to which all MRU isolates from the national dataset have been referred since July 2010 [16,17].A recent phylogenetic network analysis of the known population structure of MenC-associated cc11 sublineages was re-annotated in the context of the current study since cc11 represented the majority of recent invasive MenC isolates in England [18]

Serosurvey
Serum samples collected in 2014 were obtained from the PHE Seroepidemiology Unit; a depository of anonymised residual sera from routine diagnostic testing at participating laboratories.Known immunocompromised individuals are excluded from the collection.The age, sex and year of collection were known for individual samples: immunisation status was not known [20].
Serum bactericidal antibody (SBA) assays were performed against the serogroup C target strain, C11 (phenotype C:16: P1.7-1,1) as previously described [21].The complement source used in the SBA was pooled serum from 3-4 week old rabbits (Pel Freez Biologicals, Arkansas, United States (US)).Titres were expressed as the reciprocal serum dilutions yielding ≥ 50% killing after 60 minutes.The lower limit of detection was a titre of 4. Titres of < 4 were assigned a value of two for geometric mean titre (GMT) analysis.Titres of ≥ 8 were considered protective against MenC disease [22].
Approximately 100 samples were selected from each age-band to fit with different vaccine schedules since MCC introduction and to allow comparison with previous studies.This sample size was selected to achieve reasonable precision around the 95% confidence intervals (CIs) for proportions with SBA titres ≥ 8 and GMTs within the age group of interest.
Since MCC vaccine coverage has been consistently high, the age of the serum donor was used to align the donor with the MenC vaccine schedule they should have been offered and thereby identify which MenC vaccine cohort they belonged to.This then allowed comparison of protection between schedules and over time, using historical data from three previously published similarly designed surveys.These earlier surveys used sera collected from the same Seroepidemiology Unit and tested using the same methodology in the same PHE laboratory during 1996-1999 [8], 2000-2004 [9] and 2009 [10].

Epidemiology
MenC cases in England fell rapidly from 883 in 1998/99 (pre-vaccination year, incidence 1.81/100,000) to 13 cases in 2008/09, continuing at low levels (17-33 cases annually) during 2009/10-2014/15.In 2015/16, there were 42 cases (0.08/100,000), similar to disease levels last observed in 2004/05 (Table 1, Figure 1).Over the last decade, the age distribution has been relatively constant (Figure 1).However, the proportion of cases among those aged 5- Of the 160 cases, 11 (7%) were living in shared accommodation (three university students, four in care homes or assisted living, four in house share or hostel accommodation).There were six cases in known men who have sex with men (MSM) and eight individuals reported using recreational drugs, were injecting drug users or had alcohol dependency.Nineteen cases (12%) had underlying co-morbidities; heart condition (n = 3), diabetes (n = 2), HIV (n = 6), immunosuppression (two on psoriasis treatment, one with chronic lymphocytic leukaemia, one post liver transplant), chronic respiratory condition (n = 3), liver disease (n = 3).There were three clusters of two cases each in primary schools nationally in 2015; all six children were fully-vaccinated (two-dose primary course with booster).

Serosurvey
Results were available for 993 samples collected in 2014 and 323 (33%; 95% CI: 30.1 to 35.9) had SBA titres ≥ 8.The proportion of individuals achieving the seroprotective threshold was remarkably similar across the age groups in 2014 compared with 2009 (Figure 4).In contrast, much higher proportions of 5-19 year-olds were seroprotected during 2000-04, the period immediately after the 1999/2000 national catch-up for 0-18 year-olds.This cohort is now older and responsible for the secondary peak of seroprotection among 15-34 year-olds in subsequent serosurveys.Notably, only 38% of 15-19 year-olds achieved the seroprotective threshold in 2014 compared with 56% in 2009.Only 25% (75/299) children aged 1-14 years were seroprotected against MenC in 2014.

Discussion Epidemiology
MenC disease in England remains at very low levels with an incidence of 0.08/100,000 in 2015/16, an overall reduction of 95.6% compared with 1998/99 before MCC vaccination was introduced.Nearly a third of cases are currently diagnosed in 25-44 year-old adults, many with clinical, social and travel-related risk factors.Overall case fatality remains unacceptably high.Updated VE estimates indicate a rapid decline in protection after 1 year following routine infant immunisation, but this remains high in children immunised at school age (5-18 year-olds) up to at least 8 years after vaccination.
The low MenC disease incidence is reassuring and is consistent with continued herd protection in the wider population and high direct protection in those who have been vaccinated.There have been no cases reported among at-risk individuals, currently defined as those with asplenia, splenic dysfunction or complement deficiency.Those born outside the UK (who are less likely to be immunised with MCC, even if eligible on entry), recently travelled and/or socially mixing with these populations accounted for a quarter of all cases, while a significant proportion of the other cases were living in shared or supported accommodation, which is a known risk factor for IMD [23].An increased risk of IMD in HIV-positive individuals has recently been reported [24], even among those receiving appropriate antiretroviral therapy.Over the past 5 years, six HIVpositive individuals were identified among 160 MenC cases in England, all in the 25-44 year age group, including two who were non-UK born and had recently travelled abroad: Outbreaks of MenC disease in MSM populations have been reported in North America and Europe [25,26].Whole genome sequencing previously showed that MenC case isolates among MSMs in England, while all belonging to lineage 11.2, were interspersed among sporadic community case isolates and did not form a discrete cluster [27].[18].Furthermore, distinct lineage 11.2 populations have independently acquired traits that may facilitate urogenital colonisation [29,30].The acquisition of intact aniA alleles encoding nitrite reductase may enable survival in microanaerobic environments and has been associated with outbreaks of invasive disease among MSM in the US and Europe [30][31][32] and urethritis among heterosexual males in the US [29,30].The urethritis-associated strain has also lost its ability to express a capsule, a further gonococcal trait that may facilitate adherence to the urogenital mucosa.

Serosurvey
Seroprevalence studies have proven valuable in improving our understanding of population immunity, inform mathematical modelling and can complement disease surveillance to help inform national vaccine policy [8][9][10].The 2014 serosurvey confirmed the rapid decline in immunity after infant and toddler immunisation (also observed with the 2009 serosurvey) as well as the low proportion of teenagers protected against MenC disease around the time when the adolescent programme was introduced.Most adolescents in 2014 will have received a threedose infant MCC schedule without a booster.The low proportion of teenagers achieving protective antibody thresholds in the 2014 serosurvey indicated there was poor long-term direct protection in the cohort that was only immunised in infancy and ongoing herd protection across the population could be threatened.This observation, predicted after the 2009 serosurvey, helped support the inclusion of MCC in the adolescent programme in 2013 with the aim of providing high antibody concentrations throughout the peak years of carriage [7].The 2014 serosurvey is already demonstrating higher MCC antibodies among 14-16 year-olds compared with the same age group in the previous serosurvey.The emergency adolescent MenACWY programme in 2015 was implemented to provide direct and indirect protection, not only against MenC but also against the three other serogroups; particularly the emergent hypervirulent MenW strain also belonging to cc11 that was responsible for a quarter of all IMD cases across the UK by 2014/15 [7].[34].A number of countries across Africa, Latin America and elsewhere are also experiencing high MenC disease activity [35,36].Travellers to such countries, including residents visiting family and friends in their home countries, should be appropriately immunised, advised to avoid high-risk behaviours, remain vigilant for symptoms and signs of meningococcal disease and seek immediate medical help if concerned while abroad and when returning to their home countries.

Strengths and limitations
The strength of this study is the consistently high quality of the national surveillance programme with high case ascertainment [12] and active follow-up of cases since 1998/99.This has allowed us to monitor VE in different cohorts over time.Our meningococcal surveillance is complemented with clinical trials, to assess the immunogenicity of different schedules, and serosurveys to monitor population susceptibility.Together, these results have allowed us to adapt the national immunisation programme to provide maximal long-term protection in the most cost-efficient manner.Using the same methodology allows us to monitor population effects over time.A limitation of the serosurvey, however, is that we have no information on the vaccination status of the serum donors or their clinical state.Nonetheless, the large numbers of samples tested, along with the consistently high vaccine coverage nationally has allowed us to analyse the results by birth-cohorts.

Conclusions
We found evidence for rapid waning of immunity following routine infant immunisation, even with a 12-month booster, which was introduced in 2006.Currently, most toddlers (1-4 years) and older children (aged 5-13 years) in England are not seroprotected against MenC disease and are dependent on the indirect protection offered through reduced carriage in young people aged 17-24 years.It is, therefore, reassuring that there is still evidence of herd protection in adults 17 years after mass vaccination of children in 1999.Immunisation of teenagers appears to be key for continued population impact, but also for direct protection before their social behaviour leads to greater exposure, carriage and invasive disease.We have demonstrated high shortterm effectiveness of MCC vaccines in infants and preschool children and the long-term effectiveness of MCC vaccines given to school-aged children and adolescents.These findings have led to several revisions of the UK immunisation schedule to maximise direct and indirect protection.The recent change to emergency MenACWY vaccination for teenagers was brought in to provide more rapid population control for the national MenW outbreak.It should also provide continued direct and population control for MenC and help establish this for MenY disease.
Continued epidemiological and microbiological surveillance will be critical to monitor current programmes which now include newer sub-capsular protein-based vaccines against MenB disease in infants.MenC disease continues to be rare but detailed case-based follow-up has identified certain groups that remain at higher risk for MenC disease despite containment within the wider population, these groups may benefit from targeted vaccination.Stochastic meningococcal outbreaks globally need to be carefully monitored, as well as the recent MenC outbreaks among MSM.High-resolution characterisation is critical to monitor circulating strains and detect known and unknown emerging threats.

Figure 1
Figure 1 Number of West Nile fever cases by epidemiological week of diagnosis, Israel, 2015-18

Figure 2
Figure 2 Spatial distribution of West Nile virus cases in humans, equine, birds and mosquitoes, Israel, 2018

Figure 3
Figure 3 Phylogenetic tree of West Nile virus outbreak strains, Israel, 2015 and 2018

Figure
Figure 2 Distribution of respiratory syncytial virus hospital admissions by month of age in children less than 3 years old, birth cohort study, Scotland, 2009-2015 (n = 5,185)

Figure 2 Figure 3
Figure 2 Clonal complex distribution among meningococcal serogroup C case isolates, England, 2010/11-2015/16 The contribution of child, family and health service factors to respiratory syncytial virus (RSV) hospital admissions in the first 3 years of life: birth cohort study in Scotland, 2009 to 2015 Introduction:Several vaccines for respiratory syncytial virus (RSV) are under development.Designing an effective vaccination programme for RSV requires information about the relative contribution of risk factors for severe RSV symptoms.Aim: To inform preventive strategies in Europe by quantifying the contribution of key child, family and health service risk factors to the burden of RSV hospital admissions in young children.Methods: We constructed a birth cohort study of all singleton children born in Scotland between October 2009 and September 2012 using linkage between birth registration, maternity, vaccination and hospital admission records, with follow-up until the age of 3 years.RSV-confirmed hospital admissions were defined using linkage to national laboratory surveillance data.We estimated hospital admission rates per 1,000 child years and length of stay according to each risk factor.Cox proportional hazard regression models were used to estimate adjusted hazard ratios.

Table 1a
Distribution of respiratory syncytial virus hospital admissions, and median and interquartile range of length of stay in hospital according to risk factor, birth cohort study, Scotland, 2009-2015 (n = 5,185 admissions) IQR: interquartile range; RSV: respiratory syncytial virus.a To calculate bed days, children with a length of stay of 0 days were allocated 0.5 bed days.The 5,185 admissions were in 169,726 children, with total of 15,363.5 bed days.

Table 2
). Being born post-term was associated with a decreased risk.Of the family characteristics, having one older sibling was IQR: interquartile range; RSV: respiratory syncytial virus; SIMD: Scottish Index of Multiple Deprivation.a To calculate bed days, children with a length of stay of 0 days were allocated 0.5 bed days.b Fewer than five children had a missing value.c Including RSV admissions occurring after 6 months of age only.Total number of admissions: 2,645.Total bed days: 6,528.5.The 5,185 admissions were in 169,726 children, with total of 15,363.5 bed days.

Table 1b
Distribution of respiratory syncytial virus hospital admissions, and median and interquartile range of length of stay in hospital according to risk factor, birth cohort study, Scotland, 2009-2015 (n = 5,185 admissions)

Table 2
Crude and adjusted hazard ratios for risk of respiratory syncytial virus hospital admission in children less than 3 years old according to risk factor, birth cohort study, Scotland, 2009-2015 (n = 5,033 admissions) CI: confidence interval; HR: hazard ratio; SIMD: Scottish Index of Multiple Deprivation.aAdjustedfor all other variables in model.bIncludingcongenital heart disease, congenital malformations of the respiratory system, neurological disease and chronic lung disease.The 5,033 admissions were in 169,726 children.

Table 3
Population attributable fraction (as a percentage of admissions prevented) by risk factor and scenario, birth cohort study,Scotland, 2009Scotland,  -2015

Table 2
Meningococcal serogroup C conjugate vaccine effectiveness based on the screening method in immunised cohorts, England, September 2000-June 2016 Unvaccinated cases are based on the date of birth cohort each individual belongs to.b Trend analysis was done by logistic regression for the odds of vaccination in cases against log-time since vaccination with an offset of the log-odds of matched coverage.The exception was for the infant catch-up, where time was modelled as within a year and beyond a year as numbers were too small to model a trend.c Using DoH cover with THIN partial vaccination correction.d Including cases from January 2000.e Including one vaccinated case scheduled for three infant doses but only given a single dose after 12 months of age.f Including cases from April 2006.g Including cases from April 2000. a The current cohort of younger children with poor seroprotection is most likely protected against disease because of this indirect protection.

Table 3
Proportion of samples with serum bactericidal antibody titres ≥ 8 and geometric mean titres by birth cohort and vaccination schedule CI: confidence interval; GMT: geometric mean titre.a Data from Ishola et al., 2009.b These age cohorts would include some individuals targeted for MenC vaccination as part of the adolescent/Fresher MenC vaccination introduced in 2013.