Our study illustrates the challenge of identifying the causative agents in children with acute encephalitis. A confirmed or probable etiology was identified in only 41% of enrolled patients which is within the same range as reported in other etiology studies [3], [4], [13], [14], [18]. Worldwide, the causes of encephalitis vary between geographical regions. While JEV was the most common cause of encephalitis in children in Cambodia, enteroviruses and Tick-borne encephalitis virus were the two most common viruses found in young patients in the United States and Sweden, respectively [3], [4], [19]. A study in China revealed enteroviruses, mumps virus and rubella virus as frequent causes of encephalitis in children between 7 months and 13 years of age [20]. Our study identified JEV, enteroviruses and DENV as the most common etiologies in southern Vietnamese children, accounting for 26%, 9.3%, and 4.6% of cases, respectively. The high prevalence of JEV in our patients is in accordance with a previous study in Ho Chi Minh City reporting a prevalence of 45% in children with encephalitis [5]. Similar to previous studies, JEV was associated with high mortality (16%) [5], [21], [22]. Our observations emphasize the need for JE vaccination programs in Vietnam and other regions of Southeast Asia where JEV is endemic [4]. Widespread JE vaccination in developing countries like Vietnam is complicated by high costs and the requirement of multiple vaccine doses. In our study population, only 19.5% of patients had received at least one dose of JE vaccine. There was a trend of lower vaccination rates in JE patients, but this did not reach statistical significance (data not shown).

Enteroviruses are well established causes of aseptic meningitis and encephalitis in young children. While reported case fatality rates of enteroviral central nervous system (CNS) infections are relatively low (0–7%) [3], [13], mortality of confirmed or probable enteroviral encephalitis in our patients was high with 9 of 18 (50%) of patients dying during hospitalisation. However, a definitive diagnosis of enteroviral encephalitis was only established in 4 of these 18 patients, 2 of whom died. In the remaining patients enteroviral encephalitis was not confirmed by virus detection in CSF but suspected based on virus detection in throat and rectum and a clinical syndrome consistent with the diagnosis. Hence, a definitive causative role of enteroviruses in these patients remains unclear, particularly since a convincing link has not yet to be established between detection of enteroviruses in non-CNS sites and encephalitis [23]. In addition, beside enterovirus 71 (EV71), we have not determined whether our patients were infected with enterovirus serotypes of particularly high virulence, such as coxsackievirus B4 and echovirus 11 which have been associated with high case-fatality rates among neonates [24]. Six of 18 (33%) enteroviral infections in our patient group were caused by EV71. In accordance with reported high mortality of EV71 CNS infections, 3 of these patients died [25].

CNS manifestations are rare clinical complications of dengue but are reported with an increasing frequency in endemic areas [26]–[28]. Dengue was found in 4% of 378 pediatric patients with suspected encephalitis in southern Vietnam [27]. Similarly, DENV was identified in nearly 5% of our study patients. The precise pathogenesis of dengue-associated CNS manifestations remains unclear [27], [29]. Patients in this study experienced mild clinical symptoms of dengue without shock or bleeding and all made a full recovery. However, the detection of virus and specific antibodies in CSF suggest that invasion and viral replication in the CNS may play a role in at least a proportion of patients with dengue-associated neurological symptoms.

Me Tri virus is a variant of Semliki Forest virus belonging to the genus Alphavirus isolated from mosquitoes collected in Vietnam [10], [30]. Previous serologic surveillance suggested that this virus might be a frequent cause of encephalitis in young patients in Vietnam with a prevalence of 14% [30]. However, the virus was not detected in any of our patients. Likewise, human parechoviruses are an emerging cause of meningitis and encephalitis in children and infants [31], [32], but were not detected in our study patients. As a previous report suggested that human parechovirus infections exhibit a two-yearly incidence peak [33], the absence of parechovirus infections in our patients may be due to the fact that our study period covered only one year. HSV encephalitis was found in only one patient in our study which is not unexpected considering the young age of our study group.

Bacterial pathogens, including S. pneumoniae and H. influenzae type B, were detected retrospectively by PCR in 12 children (6%), illustrating the difficulties in distinguishing acute viral encephalitis and bacterial meningitis in children based on clinical judgment only. Indeed, the clinical syndromes of these children seemed indistinguishable from the remaining group of patients, although CSF white cell counts were higher as expected (Table 3). Of note, 4 children with bacterial meningitis had evidence of concurrent viral encephalitis based on the detection of virus (EV, CMV) or specific IgM (JEV) in the CSF. These observations suggest that testing for other pathogens should still be considered when a single pathogen has been identified in the CSF. While the clinical relevance of these coinfections remain unclear, it is tempting to speculate about the interplay between viral and bacterial CNS infections, for example by facilitating entry of one or the other into the CNS compartment.

In 59% of the children, no confirmed or probable viral etiology could be identified. Several factors may contribute to this relatively low diagnostic yield.

Firstly, a proportion of children may not have suffered from viral or other infectious CNS illnesses. Other conditions that may have presented in this manner may include pretreated pyogenic meningitis, toxins, atypical bacteria, mycobacterial and other unknown or un-tested bacteria or viruses. As mentioned before, bacterial meningitis was in fact diagnosed retrospectively by PCR in 6% of enrolled children.

A reliable case definition for acute viral encephalitis would be helpful for triage and clinical management of patients, especially in the absence of diagnostic support. When retrospectively using a predefined case definition consisting of a) fever history of less than seven days, b) an alteration or reduction of consciousness (Glasgow coma scale≤14), c) at least one of the following symptoms or signs: seizures (excluding febrile convulsions, defined as a single convulsion lasting less than 15 minutes in patients between 6 months and 6 years of age), focal neurological signs, neck stiffness or cerebrospinal fluid (CSF) pleocytosis, d) no evidence of bacterial infection in CSF specimens, and e) no alternative diagnosis for the clinical syndrome during admission, the viral diagnostic yield only modestly increased from 41 to 45%, whereas definitive/probable laboratory diagnoses of viral CNS infections were established in 20 of 62 patients (32%) who did not meet the criteria of the case definition. Overall clinical outcome and patients characteristics were similar between the two groups even though there were slight differences in terms of patient origins, clinical outcome, history of fever, limb weakness, neck stiffness, percentage of patient with a GCS below 9 and CSF cell count (Table 3). The last five are more likely due biases toward the criteria of predefined case definition.

Diagnostic yield was similar when the criterion of a fever history of less than seven days and febrile convulsions were left out. Diagnostic yield was higher when CSF results were included into this case definition (White cell count <1000 10e6/L, CSF/blood glucose ratio >50%, protein <0.45 g/L, CSF lactate <4 mmol/L): 49% and 37% in patients that did or did not meet the case definition, respectively. However, two thirds of all patients did not meet the case definition. Clearly, further studies are needed to optimize case definitions for viral encephalitis.

The limited detection of viral pathogens in our patients may have been due to clearance of virus from CSF at the time of clinical presentation, as is the case for JEV [34]. This may also explain why, except for dengue virus, flaviviruses were not detected in any of our patients, including those with serologically confirmed JEV infections. The importance of early sampling for reliable diagnostics is suggested by findings in our study: most of the patients were admitted relatively late in the course of illness, possibly in part because of referral delays, and a trend was observed towards shorter illness duration at the time of sampling in patients with positive virus-specific PCRs. The association between earlier sampling time and PCR positivity was statistically significant when analysis was restricted to 132 patients fulfilling our predefined case definition of acute viral encephalitis (3 days versus 4 days respectively, P=0.03).

Alternative Language Abstract S1

Checklist S1