The diagnosis and management of central nervous system (CNS) infections can be challenging in malaria-endemic tropical countries due to limited laboratory facilities and therapeutic options. In addition, restricted availability of health care services that may also be difficult to access can result in late clinical presentations. Viral pathogens are increasingly identified in patients who present with clinical features of CNS infection in this setting, where they may be i) the principal cause of the illness, ii) an alternative post-mortem diagnosis in patients with an established clinical diagnosis of cerebral malaria [1], iii) co-incident pathogens that influence both the CNS manifestations and subsequent mortality of malaria [2], or iv) as in the case of neurotropic human herpes viruses (HHVs) that can establish a latent reservoir through chromosomal integration of DNA, as a bystander reactivated during an acute illness [3],[4].

The types and proportions of viruses that have been identified in CNS infections vary by geo-epidemiological setting. The CNS symptoms caused by cerebral malaria (CM) are a common indication for lumbar puncture (LP) and reactivation of HHVs in this situation is well described [5]. As a result, geographic variability in the incidence, severity and outcome of CM may impact on the viruses detected in children presenting with possible CNS infections [6]. Other possible factors that might influence viral CNS infection include local variability in entero-, adeno- or arbovirus incidence [7], local zoonotic reservoirs that may facilitate the emergence of novel viral pathogens [8], as well as the choice of virus-specific primers that are included in diagnostic multiplex PCR tests.

Few studies have systematically assessed inter-relationships between the presence of viral nucleic acids (NA) in cerebrospinal fluid (CSF) with specific clinical features and outcome. This is especially important for HHVs such as varicella zoster (VZV), herpes simplex viruses-1 (HSV-1), cytomegalovirus (CMV), HHV-6 and HHV-7 as the presence of NA in the CSF may be result of reactivation of viral NA that have been integrated into host cells and may not necessarily represent a primary acute infection. Such studies are, however, challenging because of the potentially large number of candidate viruses and non-viral comparator groups. In a recent study of 513 hospitalized Malawian children, 12 different viruses were isolated from CSF or post mortem brain biopsies, and the presence of any virus increased the risk of death especially in those children with co-incident malaria [2].

Although febrile encephalopathy is common in malaria-endemic coastal Papua New Guinea (PNG), there have been few etiologic studies. In a study of children diagnosed with likely bacterial CNS infections, no causative organism was found in most cases [9]. We hypothesized, therefore, that viruses contribute significantly to the burden of severe CNS illness and its presentation. As part of a detailed prospective observational study of severe childhood infections, we retrospectively analyzed CSF samples from PNG children hospitalized with signs of a CNS infection but without culture- or latex agglutination-proven bacterial or cryptococcal infections for the presence of i) HHVs, ii) picornaviruses, iii) influenza viruses, iv) adenoviruses, v) flaviviruses and vi) bacterial pathogens (Streptococcus pneumoniae (SP), Haemophilus influenzae type b (Hib) and Neisseria meningitidis).

Of 638 children presenting with clinical features suggestive of CNS infection, a LP was performed in 554 (87%). Of this latter group, definite diagnoses of bacterial/fungal meningitis and SSPE were made in 47 (9%) and 22 (4%) children, respectively. There was insufficient CSF in a further 185 children (33%) and 87 of these had a diagnosis of CM. The remaining 300 (54%) with possible CNS infections had CSF available for molecular testing (Figure 1). The median age of this subgroup was 24 (interquartile range [IQR] 18–29) months and 59% were males. One-hundred and eighty-three children (61%) had convulsions, 47 (16%) were comatose at the time of admission, and 42 (14%) had impaired consciousness.

Fifty-five of the 300 children (18%) had at least one type of viral or bacterial NA identified in their CSF, of which 42 had at least one virus identified, including 39 and three with single- and dual infections, respectively. Overlap between viral NA, bacterial NA and malaria was common (Figure 2). Bacterial DNA was identified in 20 children, with 7 of these also with viral NA present. Children with bacterial infection included 14, 5 and 1 with SP, HI and both SP/HI, respectively. Malaria was identified in 93 children (31%). An etiologic agent was not identified in 176 children (59%). The characteristics of children by likely etiologic agent according to malaria status are summarized in Table 1. After adjustment for age, the presence of viral and bacterial NA was associated with malaria infection (odds ratio [OR] 2.05 [95% confidence interval (CI₉₅)] 1.03-4.08, P = 0.04 and 2.62 [1.02-6.8], P = 0.04, respectively).

Any impairment of consciousness was associated with the presence of malaria on blood slide (χ² = 39.6, P <0.001) and a history of multiple convulsions (χ² = 9.8, P = 0.002). The presence of viral or bacterial NA was not associated with impaired consciousness.

Twenty-eight children (10%) had an adverse outcome, including 14 (5%) who died. The remaining 14 (5%) were discharged with neurological disabilities that included children with motor (n = 6), vision (n = 4), speech (n = 3) and sensory (n = 1) deficits. In multivariate analysis, an adverse outcome was independently associated with an abnormal level of consciousness (OR [CI₉₅] 3.5 [1.4-8.8]), raised CSF protein (5.4 [1.8-14.9]) and age (1.014 per month [CI₉₅ 1.002-1.026]). By contrast, malaria was associated with a lower chance of adverse outcome (0.30 [0.09-0.83]). After adjusting for these factors, the presence of viral or bacterial NA was not associated with an adverse outcome.

The present study provides the first comprehensive data relating to the etiology of possible CNS infections in hospitalized children from a malaria-endemic area of PNG. Of 300 children without SSPE or proven bacterial/fungal meningitis, a possible candidate bacterial, viral or parasitic agent, or combination of agents, was identified in only 41%. There were few characteristics that differed between the six diagnostic groups, and thus no reliable clinical differentiation between the effects of bacteria or viruses in the CSF. Many children had more than one possible infection with overlap across different diagnostic groups.

Malaria appeared to have the most important effect on clinical characteristics of children with possible CNS infections. In multivariate analyses, the presence of malaria on blood slide was associated with the detection of bacterial or viral NA. Malaria was also independently associated with impaired consciousness and multiple convulsions, but after adjustment for other factors was inversely associated with an adverse outcome. Neither the presence of viral or bacterial DNA was independently associated with an adverse outcome or impaired consciousness. In post-hoc pairwise comparisons, children with malaria were older than their counterparts without malaria, regardless of the presence of viral or bacterial NA in the CSF.

Human herpes viruses accounted for the majority of the viral infections but there were some clinical differences within this group of related viruses. Children with HHV-6 or HHV-7 were much younger and more likely to have impaired consciousness than those with CMV infections, with an associated trend to a higher rate of complex seizures and adverse outcomes. These observations are consistent with the notion that acute infections associated with HHV6 and −7 might be more likely to present in very young children. The death of the child in whom HHV-7 was the only identifiable etiologic agent highlights the possibility that very severe disease and death may occasionally result from infections with these viruses.

Although HHV-6 and −7 have been shown to cause encephalitis and are associated with febrile convulsions in very young children [16]-[20], demonstrating a causal relationship between the presence of viral NA and clinical disease can be difficult. Like most HHVs, HHV-6 and −7 also are considered neurotropic [19]-[21] with the capacity to establish a latent reservoir. Re-activation of latent infections can cause clinical disease in immune-suppressed individuals [3],[22],[23], but data from developing countries suggest that malaria infection may also predispose to re-activation of HHVs. A small study of 49 Kenyan children with cerebral malaria found that HSV-1 was present in CSF from four children [5]. In a study from Uganda, there was an increase in Epstein-Barr virus in plasma and saliva from children with uncomplicated malaria which cleared with antimalarial treatment [3], suggesting that acute P. falciparum infection may cause HHV reactivation. Our data also support the hypothesis that malaria might cause re-activation of HHVs in some children but without major clinical consequences. Malaria was associated with a 2-fold increase in the chance of detecting a virus when compared to children without malaria. However, although HHVs made up >90% of possible viral pathogens isolated, they did not appear to have a characteristic clinical phenotype.

A recent Malawian study found that malaria-virus co-infections that included infections due to rabies, enteroviruses, mumps and adenoviruses were associated with an increased risk of seizures compared to either infection alone. Mortality in this group was also significantly higher than in children with malaria infection alone [2]. The apparent inconsistency between this finding and the results of the present study may reflect differences in the geographic distribution of encephalopathic viruses as well as their relative propensity to establish a latent reservoir. In addition, regional differences in malaria prevalence, incidence of severe malaria and associated mortality [5], temporal variability in enterovirus, adenovirus and/or arbovirus incidence [7], local zoonotic reservoirs contributing to the emergence of novel pathogens [8], and the choice of virus-specific primers that are included in the multiplex PCR tests employed during each study may all contribute to the distinctive local epidemiology of CNS infections.

Our data also confirm that viral pathogens including dengue and enterovirus that are commonly associated with clinically significant CNS infections occur, albeit rarely, in PNG children from the Madang area. Despite the low incidence of dengue in hospitalized children in the present study (1%), the child with dengue type-2 encephalitis died. This is the first reported case of death due to dengue encephalitis in PNG, highlighting that dengue should be considered an etiologic agent in PNG patients with undiagnosed CNS infections. Interestingly, the clustering of admissions of the three children with dengue type-2 infections indicates that transmission may be seasonal. Only one case of enterovirus infection was documented in our study, but was unable to be typed further.

In contrast to reports from developed [24],[25] and other tropical [2],[26],[27] countries, influenza, adenoviruses and other HHVs were not significant contributors to the burden of encephalopathy in our hospitalized children [7],[28]. Epstein-Barr virus, commonly associated with Burkitt’s lymphoma in malaria-endemic areas including PNG [29],[30], was not identified in the present study and neither was HHV-8, perhaps because PCR was performed on CSF rather than blood or saliva [3],[31].

The present study also raises the possibility of occult culture-negative bacterial infections in children with features of CNS infection. Bacterial infections not identified by blood or CSF culture, or by bacterial antigen testing, were identified in 6.7% of children using molecular diagnostic methods. Although one child with SP/Hib co-infection died, our data do not reveal any clear clinical pattern associated with occult bacterial infections. As with viral infections, the presence of malaria increased the chance of detecting a bacterial pathogen by approximately 2-fold. Although the clinical relevance of ‘incidental’ bacterial infections is not clear from our data, severely ill children with negative CSF or blood cultures might still benefit from empirical use of antibiotics.

Despite detailed investigations covering a large number of candidate etiologic agents, a large proportion of our children with possible CNS infection did not have a convincing final diagnosis. This warrants further studies in this group of vulnerable children, especially as increasingly sensitive molecular and other diagnostic tools continue to be developed [32]. In addition, non-infective causes of encephalopathy, such as environmental toxins, could also be considered where no pathogens are detected.

Our data have some limitations, particularly with the challenges of delineating between acute infections and reactivation of HHVs. Firstly, sophisticated molecular analyses differentiating HHV-6 DNA from actively replicating virus or forms associated with chromosomal integrated DNA were not performed [4]. Secondly, we did not perform tests to define whether children with HHV-6 had HHV-6A or 6B. Thirdly, we did not have neuroimaging diagnostics in our setting that could assist in the diagnosis and management of brain infections. Finally, we did not have plasma available for serological tests such as species-specific immunoglobulin M or G that might assist with making a diagnosis of an acute or latent infection, respectively.

The present study indicates that viral infections such as dengue, enterovirus and perhaps HHV-6 and −7 can cause serious CNS disease in some PNG children. However, there is no apparent effect of the presence of viral NA on clinical features and outcome in this group. By contrast, malaria has a strong positive effect on the likelihood of detection of viral and bacterial NA, is itself associated with clinically important characteristics such as impaired consciousness, and is an independent predictor of a good outcome. This suggests that HHVs may be the result of reactivation from a latent reservoir as a result of malaria without clinical sequelae in most cases. Although the capacity to diagnose significant viral infections in PNG is limited, these data should inform clinicians as to differential diagnoses in an encephalopathic child that should now include dengue and enterovirus infections. Antiviral medications such as aciclovir are routinely not available in PNG. Although antiviral medications have been used successfully in a small number of immunocompetent patients with encephalomyelitis due to HHV-6 [33], the lack of data from well-designed studies, high medication costs and the lack of capacity to make a diagnosis, preclude the incorporation of antiviral recommendations into standard treatment guidelines for countries like PNG. The present data do not argue for a change to these guidelines.