We searched MEDLINE, EMBASE, and Google Scholar to identify studies published in English between January 1990 and February 2008. We used the following keywords: Norwalk, norovirus, Norwalk-like virus, human calicivirus, calicivirus, NLV, small round virus, and small round structured virus. We reviewed all abstracts to identify articles that assessed the prevalence of NoV among sporadic cases of diarrhea. To ensure complete capture of all relevant studies, we cross-referenced all articles from the bibliography of the selected articles. After reviewing each article, we selected studies that met the following inclusion criteria: 1) study duration was >1 year, and 2) study used RT-PCR to diagnose caliciviruses (NoV and sapovirus) or NoV in patients with diarrhea. We included studies that tested for caliciviruses, even when they did not differentiate between NoV and sapovirus. For these studies, we multiplied the proportion of caliciviruses detected in each study by the mean proportion of caliciviruses that were NoV among studies that differentiated between NoVs and sapoviruses to yield the estimated NoV prevalence in each study. We excluded studies that did not provide a denominator (i.e., the total number of patients with diarrhea in the study population) or that only conducted molecular analysis using a fraction of the fecal samples (Technical Appendix 1). If the authors presented the data again in another study, only 1 study was included. See Technical Appendix 2, for a list of all references used in the review but not cited in this article.

We stratified studies into 2 settings: community or clinic-based (mild or moderate diarrhea) and hospital-based including emergency department and inpatients (severe diarrhea). We counted cases in which NoV was detected in the presence of >1 other pathogens (i.e., mixed infection) as NoV infection; however, we also present data on mixed infections, when available. Pooled proportions and 95% confidence intervals (CIs) of NoV-positive cases were calculated by using the random effects models (DerSimonnian and Laird method, StatsDirect Ltd, Cheshire, UK). For the studies that included fecal testing on concurrent diarrhea-free controls, the pooled proportions were based on absolute difference in NoV detection rate between cases and controls, thus only including the fraction of cases attributable to NoVs. The Cochran Q statistic and degrees of freedom (df) are presented as a measure of heterogeneity among studies. Analyses were conducted with StatsDirect version 2.5.7 (StatsDirect Ltd).

To calculate the number of outpatient NoV episodes and hospitalizations for children living in industrialized countries (where 23 of 31 studies in our review were conducted), we multiplied the total number of estimated diarrhea episodes in each clinical setting by the pooled proportion attributable to NoV based on the studies we reviewed to yield the number of NoV cases in each setting (9). No data exist on estimates of total diarrhea episodes in industrialized countries. Thus, we divided the estimates of outpatient and inpatient rotavirus episodes for industrialized countries, provided by Parashar et al. (9), by the proportion of diarrhea episodes attributable to rotavirus (23% and 42%, respectively) in the United States (10) and Europe (11) to yield the annual number of total diarrhea episodes in industrialized countries. To estimate the proportion of outpatient (23%) and inpatient (42%) diarrhea episodes attributable to rotavirus, we assumed the midpoint of the proportion of gastroenteritis visits attributable to rotavirus in the United States (19% and 35%, respectively) and 7 European countries (27% and 50% respectively) for each setting, respectively.

To estimate NoV-associated deaths and hospitalizations among children in developing countries, we multiplied global estimates of diarrhea deaths (1) and hospitalizations (9) by the pooled proportion of NoV among children <5 years of age hospitalized with diarrhea. Data from developing countries were sparse on fraction of NoV-associated diarrhea episodes in the outpatient setting.

Overall, we reviewed 235 studies and identified 31 original studies that met our inclusion criteria (Tables 1, 2) (6,12–41). Of these 31 studies, 20 were conducted in high-income countries, 2 were high-middle-income countries, 5 were low-middle income, and 4 were low-income countries, based on World Bank classification of economies. The duration of these studies was 1–5 years. Fourteen studies tested only for NoV (13,14,19,21,22,27,29,30,35,37–41); 17 tested for NoV and sapovirus (6,12,15–18,20,23–26,28,31–34,36). Among 13 of 17 studies that tested for and presented separate detection rates on both caliciviruses, NoV was detected in 84.5% and 88.5% of the community- and hospital-based studies, respectively (6,15,16,20,23–26,31–34,36). In these 13 studies, 69%–90% of the outpatient cases and 61%–100% of the hospital cases with caliciviruses were identified as NoV.

This systematic review of studies that used RT-PCR for detection of NoVs in fecal specimens clearly indicates that these viruses play an important role in the cause of both mild and severe gastroenteritis worldwide. In 1979, Greenberg et al. demonstrated that virtually all children in various countries worldwide acquired antibodies to NoVs by 5–15 years of age (Technical Appendix 2). However, despite strong evidence that NoV infection was ubiquitous, detection rates in children hospitalized with diarrhea were low in early studies that used electron microscopy and antigen assays (4). The advent of conventional RT-PCR for the diagnosis of NoVs has substantially changed our understanding of their epidemiology. Among all reported studies that used conventional RT-PCR, NoVs were detected in ≈12% of children <5 years of age with severe diarrhea, which suggests that these viruses are the second most common cause of severe childhood gastroenteritis, following rotavirus. In addition, although some studies suggest that NoV infections in the community are slightly less severe than rotavirus infections, data also exist to suggest that these childhood infections may be similar in severity, which may particularly apply to hospitalized children (Technical Appendix 2). On the basis of the pooled detection rates of NoV in our review, we would estimate that in the United States alone NoVs may account for >235,000 clinic visits, 91,000 emergency room visits, and 23,000 hospitalizations among children <5 years of age (10). Limited data from developing countries are available to make firm estimates, but NoV disease may cause >1 million hospitalizations and 200,000 deaths each year among children <5 years of age.

Although these figures provide a preliminary indication of the substantial magnitude of illness from NoV disease, they may underestimate the true extent of disease. Evidence suggests that detection of NoVs in fecal specimens by conventional RT-PCRs may be limited by factors such as low virus concentrations in feces, improper specimen storage, inefficient viral RNA extraction, presence of fecal reverse transcriptase inhibitors, and use of different primers (Technical Appendix 2). In addition, NoVs are extremely genetically diverse and none of the reported conventional RT-PCR assays is able to detect all strains (Technical Appendix 2). These hypotheses are supported by the findings of an evaluation of children with gastroenteritis in Peru in which both RT-PCR testing of fecal specimens and serologic assays were used to assess NoV infection, and serologic testing was found to increase the rates of NoV detection from 35% with fecal testing alone to 55% by use of either assay (28). A recent validation study comparing state-of-the-art real-time RT-PCR with conventional RT-PCR found that the sensitivity of real-time RT-PCR was greater than that of the conventional method, especially for samples containing low NoV concentrations or RT-PCR inhibitors (Technical Appendix 2). The broader application of real-time RT-PCR assays to diagnose NoV among children hospitalized with gastroenteritis should provide better estimates of the true prevalence of disease.

Some factors could have led us to overestimate the extent of sporadic NoV disease. Our estimates of sporadic NoV disease prevalence are based on a review of active surveillance data of all diarrhea cases and do not exclude cases originating from an outbreak. In a few studies, NoVs were detected in patients who were co-infected with another pathogen, and only a limited number of studies enrolled concurrent healthy controls, thus making it difficult to determine the fraction of diarrhea cases truly attributable to NoV. Among all studies testing for co-infections, however, the median rate of detecting another pathogen in addition to NoV was low (2%). This finding, combined with the fact that most studies only assessed for NoV among samples that previously tested negative for other bacterial and viral pathogens, suggests that our overall pooled proportion attributable to NoVs is unlikely to be much lower. In addition, for the studies that enrolled diarrhea-free controls, we subtracted control prevalence from the case prevalence of NoV disease when calculating the overall pooled estimate. Lastly, other unmeasured factors underestimating disease prevalence that could not be accounted for, such as inefficient primers, low virus shedding, delays in specimen collection, and lack of a sensitive case-definition are also likely to exist in these studies.

The heterogeneity in the NoV literature is evident and should be considered when interpreting the results of this review. Our systematic approach and strict inclusion criteria likely reduced heterogeneity but do not eliminate biases in the original studies, diversity in study design and population, and publication bias. Nonetheless, the findings of this review suggest that NoVs are a frequent cause of mild and severe sporadic gastroenteritis among children in high- and middle-income countries. In addition, hospital studies in our review only assessed patients admitted with diarrhea. Because of the highly infectious nature of NoVs, substantial additional health and economic effects would also occur from nosocomial disease and outbreaks in healthcare facilities, as previously identified by Lopman et al (7). NoVs are also a frequent cause of severe illness and death from diarrhea among children in developing countries, although firm conclusions cannot be made because of limited data. Systematic evaluations that use broadly reactive, state-of-the-art diagnostic assays, with concurrent evaluation of healthy controls and examination of potential co-infection, are needed to fully understand the role of NoVs in the etiology of sporadic childhood gastroenteritis. These evaluations are especially necessary in developing countries, where diarrhea remains a leading cause of childhood death, causing >1.8 million annual deaths (1).

The increasing evidence documenting the magnitude of the NoV disease prevalence provides support for considering targeted interventions, such as vaccines, for reducing the extent of this illness among young children. However, if one considers that NoV frequently causes both sporadic and epidemic gastroenteritis and can affect all age groups, some other potential targets for vaccination may include elderly persons in nursing homes, who are vulnerable to severe complications, and military recruits, in whom sporadic and epidemic NoV disease is known to incur substantial illness and financial costs from work disruption (7; Technical Appendix 2). The development of vaccines against NoVs will likely be challenging because the immunity to these viruses and the diversity and evolution of circulating strains are incompletely understood (Technical Appendix 2). However, genotype II, cluster 4 NoV strains appeared to be by far the most prevalent strains among the studies we reviewed, and these strains may be the primary targets for vaccine development.

Carefully designed epidemiologic studies that evaluate NoV prevalence in children with diarrhea and a suitable comparison group and that use sensitive molecular assays will help further define target groups that would benefit from vaccines and other interventions. A particularly pressing need exists for better quantifying the extent of severe norovirus disease among children in developing countries and identifying prevention strategies to help reduce the prevalence of deaths from diarrhea in the poorest countries.