Defining the Epidemiology and Burden of Severe Respiratory Syncytial Virus Infection Among Infants and Children in Western Countries

The primary objective of REGAL was to address seven specific research questions (Table 1). The systematic literature reviews undertaken to answer each of these research questions all used the same broad methodology, which is described in detail below.

Following a study protocol with pre-defined search terms, we conducted a systematic and comprehensive search of the medical literature electronically indexed in MEDLINE (PubMed), Embase and The Cochrane Library. Search strategies were devised for each systematic review and combined free-text search terms with medical subject headings (MeSH). An important part of the protocol was the countries to be included. The overall burden of RSV has been studied in many industrialized countries, and therefore, to ensure a manageable volume of publications, only studies conducted in Western countries were included, which we defined as the United States, Canada, and Europe (including Turkey and the Russian Federation). The full protocol for the systematic reviews is available as part of the online supplement (Supplementary Material 1—REGAL Protocol). Our search included studies conducted in children (defined as ≤18 years) and published between January 1, 1995 and December 31, 2015. The target populations for REGAL included previously healthy term or preterm children [<37 week gestational age (wGA)], those with CLD/BPD, CHD, or other high-risk comorbid conditions (e.g., anatomic pulmonary abnormalities, neuromuscular disorders, Down syndrome, immunodeficiencies and cystic fibrosis) with ‘proven’ or ‘probable’ RSV. Children who had received immunoprophylaxis with palivizumab were included.

No language limits were set on the database searches, with the caveat that English translations of at least the abstract had to be available. Since there is not a universal, standardized definition of severe RSV disease, for the purposes of this article, we have taken this to be ‘RSV infection requiring hospitalization’. Randomized controlled trials, non-randomized controlled trials, crossover trials, single-arm studies, cohort studies (prospective and retrospective), case–control studies (prospective and retrospective), and case series were included, as well as published data from registries and medical databases. We also searched the database ClinicalTrials.gov, which provides information on current ongoing clinical research studies being conducted around the world. In addition, we hand-searched online journals and scanned the reference lists of identified citations and relevant abstracts presented at key meetings to find additional relevant publications for each of the reviews.

Two reviewers (J. Smith and J. Blake) undertook the search adopting a two-phase screening process. In Phase 1, the title (Stage 1) and, then, the abstract (Stage 2) of all identified studies were independently assessed for their relevance in answering the research question (details of the reviewers can be found in the Acknowledgments). The following short-term outcomes were assessed: incidence rates of severe RSV infection requiring medical treatment during the first or subsequent years of life, RSV hospitalization (RSVH) rates, length of stay (LOS) in hospital, RSVH-related outcomes [intensive care unit (ICU) admission, LOS in ICU, requirement for, and duration of, mechanical ventilation, non-invasive ventilation and oxygen], case-fatality rate, and risk factors for severe RSV infection requiring hospitalization. The following long-term outcomes were assessed: subsequent respiratory disease, including recurrent wheezing and asthma up to adulthood (≤18 years) following severe RSV infection in infancy. Other outcomes assessed were the effectiveness of palivizumab in reducing RSVH rates and associated morbidity, long-term sequelae and mortality in different subgroups of children with or without CLD/BPD, and future developments in RSV research including genetic phenotypes and polymorphisms.

All studies marked in Phase 1 as potentially relevant were then independently assessed at the full-text level by the same two reviewers for inclusion (Phase 2). Any disagreements were resolved after discussion with a third reviewer (B. Rodgers-Gray) and X. Carbonell-Estrany. The search results were compiled into a spreadsheet, including documentation of the reason(s) for excluding a study. Data were extracted for all identified articles by one reviewer and quality-checked by the second reviewer, and a written report was produced (Supplementary Material 2—Data Extraction Table). The authors reviewed the search results and report, made any additions and amendments, and all approved the final list of studies for inclusion. For each systematic literature review, we fully documented the inclusion and exclusion processes using a PRISMA flowchart that detailed the number of included and excluded articles.

Included publications were graded according to the Oxford Centre for Evidence-Based Medicine Levels of Evidence [22, 23]: level 1 evidence (local and current random sample surveys [or censuses]); level 2 evidence (systematic review of surveys that allow matching to local circumstances); level 3 evidence (local non-random sample); level 4 evidence (case series). For randomized, controlled trials, a quality assessment for each citation was carried out using the five-point (1 = low quality; 5 = high quality) Jadad Scale [24]. For each study, we also conducted a risk of bias assessment using the RTI Item Bank (score of 1 = very high risk of bias; score of 12 = very low risk of bias) for observational studies [25] and the Cochrane Collaboration’s tool for assessing risk of bias for randomized clinical trials [26].

The analysis in this review article is based on previously published studies and does not involve any new studies of human subjects performed by any of the authors.

A total of 98 publications were included in the final review: 90 identified from the database searches and a further 8 from reference lists/other sources (Fig. 1). Data extraction tables for all 98 studies, including evidence grades and risk of bias assessments, can be found in the online supplement (Supplementary Material 2—Data Extraction Table).

RSV has been associated with 12–63% of all acute respiratory infections (ARIs) [12, 16, 27‐40] and 19–81% of viral ARIs causing hospitalization in infants and children [41‐57] (Table 2). Importantly, while high-risk groups are particularly vulnerable to severe infection and are often the focus of study, the majority (typically >70%) of children hospitalized with RSV-related ARI/bronchiolitis had no underlying medical conditions [14, 16, 44, 56‐58]. To put the burden of RSV into context, in comparison to influenza, retrospective analyses show that RSV causes up to 16 times more hospitalizations and emergency department visits in children aged <5 years [49, 59, 60].

Hospitalization rates for RSV ARIs increase with decreasing age, peaking in the first few months of life [4, 6, 7, 16, 34, 35, 37, 56, 61]. Most retrospective studies have consistently shown RSVH to be the highest in the first year of life [7, 49, 56, 61]. In the prospective study by Hall et al. [16], 1-month-old infants, consistently, were the most likely to be hospitalized, almost twice as often as the next two most at-risk groups: infants <1 month old and infants 2 months old. Incidence rates for confirmed RSVH in the first year of life in healthy term infants may be relatively low at <1% [51], though around 2% of all infants including those with comorbidities are included [61, 62]. More studies in healthy term infants are required. Overall, studies have found that between 75% and 90% of infants hospitalized with RSV were aged ≤12 months, and between 44% and 83% were aged ≤6 months [6, 7, 12, 16, 44, 56, 63‐65]. Only a minority (≤5%) of neonatally hospitalized children are rehospitalized for RSV infection [57, 66, 67].

Annual hospitalization rates for RSV-associated ARIs in the first year of life range from 3.2/1000/year [16] to 42.7/1000/year [37], and decreased with increasing age to 0.6/1000/year [61] to 1.78/1000/year [49] in children aged 1–4 years (Table 2). The reported rates of RSVH, however, vary considerably between studies and across seasons within the same study.

There are very limited data available on the incidence and burden of RSV infection managed on an outpatient basis or in the emergency department [14, 60]. Available studies suggest that the outpatient burden of RSV on healthcare resources has not been fully recognized by healthcare providers [14]. For example, a US study reported that only 3% of outpatients with RSV infection received the diagnosis of RSV-associated illness, as compared with 45% of inpatients (P < 0.001) [14]. Based on study data, RSV infection was estimated to result in one of 334 hospitalizations, one of 38 visits to an emergency department, and one of 13 visits to a primary care office each year in the US [14]. Another US study estimated that, at a national level, more children required an emergency department visit for a RSV infection compared with an influenza infection (21.5 vs. 10.2 visits per 1000 children/year, respectively) [60]. The total estimated number of workdays missed annually by caregivers of children who required emergency department care was 716,404 days for RSV infections [60].

A number of studies have examined the national trends in RSVH rates among infants and children over the past two decades. While some studies conducted in the United States [44, 68] and Canada [69] throughout the 1980s and 1990s and the early 2000s have shown a rise in RSVH rates, other studies, both in the United States [70] and in France [71], conducted between 1998 and 2009, report a decrease in the incidence rate of bronchiolitis and RSVH. Further studies in the United States and Europe report that RSVH rates have remained relatively stable over similar time periods [7, 35, 72]. A firm conclusion on long-term time trends is not possible.

Season-on-season RSVH rates have been shown to vary by up to a factor of two to three in the same geographic region [6, 7, 16]. The length and start date of the RSV season can also vary from year to year [50, 73‐75]. In temperate, northern hemisphere countries the season typically starts in October or November, peaks in December or February, and finishes in March or April [4, 6, 7, 16, 44, 50, 63]. Changes in surveillance programs, testing rates, the RSV tests used, and admission criteria within and between countries, as well as variations in disease severity and levels of circulating virus, all contribute to the varying incidences reported over time and highlight the need for rigorous, uniform, and ongoing data collection.

RSV has a high rate of co-infection with other respiratory viruses with a similar seasonal pattern, such as influenza, rhinovirus (RV), human metapneumovirus (hMPV) and human bocavirus (HBoV) [28, 41, 76, 77], and also with bacterial pathogens [78, 79]. Co-infections, most commonly dual infections, have been reported to occur in up to 50% of RSV-hospitalized children, [76, 78, 79]. In a prospective study of 2525 children aged <14 years by Calvo et al. [76], 599 RSV single infections and 326 RSV multiple infections were detected. The most frequent dual infection was between RSV and RV (120 episodes) and the second between RSV and HBoV (60 episodes). Other dual RSV infections included 3 episodes of RSV and hMPV and 11 episodes of RSV and influenza [76].

Co-infections represent a major confounder in burden of disease analyses of RSV infection [78, 79]. At present, the clinical data available on co-infection, in terms of both the number of viruses involved and the severity of the condition, are variable and even, at times, contradictory [80]. Some studies have reported that multiple infections were associated with more severe disease and longer hospital stays than with single RSV infections [42, 76, 81]. Calvo et al. [76] found that hospitalization was longer for both RSV simple and RSV and HBoV co-infection, lasting about 1 day (4.7 vs. 3.8 days; P < 0.001) longer than in simple HBoV infections. In contrast, Martinez-Roig et al. [80] found an inverse relationship between the number of viruses detected in nasopharyngeal aspirate, the need for oxygen therapy and hospitalization days. In this prospective study of 463 pediatric patients aged between 7 days and 15 years hospitalized with respiratory infections (none with risk factors for respiratory diseases), the most common co-infections were RV and RSV-B (10 cases) and RSV-A and RSV-B (5 cases). Hospitalization decreased with greater number of viruses (P < 0.001). Oxygen therapy was required by 26.75% (1 virus was detected in 55.34% of cases). A larger number of viruses resulted in less need for oxygen (P < 0.001). Ten cases required mechanical ventilation, 4 patients with bacterial co-infection and 5 with viral co-infection (P = 0.69). The most common co-infections were RV and RSV-B (10 cases) and RSV-A and RSV-B (5 cases) [80]. Findings from an earlier Canadian study of 742 children suspected to have a respiratory illness (33% with a comorbidity) support equivalent disease severity between single virus infection and virus co-infection [82]. In multivariable analysis, virus co-infection did not affect admission to hospital after adjustment for age and the presence of underlying conditions [82].

Several important independent risk factors for RSV hospitalization have been identified, including male gender, birth in proximity to RSV season, history of prematurity and tobacco smoke exposure (Table 4). Age, in particular, has been found to be a significant predictor in the severity of infection and RSV hospitalization, with the youngest age groups (<6 months) more frequently hospitalized for a RSV infection [14, 16, 27, 100, 101]. In contrast, some risk factors, such as breastfeeding, have been found to be protective [33, 102]. While the predominance of severe RSV bronchiolitis in boys compared to girls is well known, its mechanism is not yet understood [103]. A study found that the interleukin-9 genetic polymorphism (rs2069885) has an opposite effect on the risk of severe RSV bronchiolitis in boys and girls [103]. Some lesser known risk factors reported in the literature include high altitude above 2500 m [104], maternal atopy [101] and delivery by cesarean section [105]. Black and aboriginal race has also been shown in some studies to be a risk factor for severe disease [14, 65, 90, 101, 106], but not in others [16, 35, 107]. The reasons for these racial disparities, however, remain unclear, but may relate to social factors. Risk factors for RSV-related mortality have not been well described. Insights into risk factors that predispose infants to severe RSV infection and subsequent hospitalization are important to identify strategies to prevent infection, as well as to inform guidelines for palivizumab prophylaxis.

There are a few key limitations when considering this review. With a remit to cover 20 years of study and research into the epidemiology and burden of RSV, it is recognized that improvements and changes in medical practice over time would influence the findings and conclusions drawn. The availability, type, and level of diagnostic testing for RSV undertaken over the review period is particularly relevant. Older studies—if any—predominantly used rapid antigen testing, shell vial culture, and viral culture, which are generally less sensitive than current PCR (polymerase chain reaction)-based assays. As described above, some studies, particularly the older ones, tended to describe the epidemiology and burden of bronchiolitis or ARIs without a totally accurate measure of how much of that burden or any additional burden related to RSV. We consider that there is a really important economical impact of RSV infection on emergency departments and outpatients, on co-infections, and on RSV-specific mortality rates. Unfortunately, there are also limited economical data of outpatient RSV economical burden For this reason, this review may underestimate the true burden of RSV.