Risk Factors for Nipah Virus Encephalitis in Bangladesh

Patients in Goalando were likely infected by direct contact with fruit bats or their secretions, rather than through contact with an intermediate host.

H enipaviruses (family Paromyxoviridae, genus Henipavirus) are enveloped RNA viruses that cause respiratory illness in pigs and horses and respiratory illness and encephalitis in humans (1)(2)(3)(4)(5)(6). After a 4-to 18-day incubation period, human disease can rapidly progress from mild illness (fever, headache, myalgia) to coma and death within 10 days; the case-fatality ratio is 40%-76% (3,(7)(8)(9)(10). The fi rst recognized human Henipavirus infections occurred in 1994 in Australia, where a respiratory disease among horses was associated with illness in 2 humans (11). The etiologic agent, Hendra virus, was subsequently isolated from asymptomatic fl ying foxes (fruit bats of the family Pteropodidae) (12). Field et al. (2) suggested that horses, identifi ed as the intermediate hosts linked to human illness, may have become infected through indirect contact with fruit bats (e.g., infected fetal bat tissues or fl uids).
An outbreak of encephalitis in Bangladesh was recognized on January 21, 2004; it affected 2 villages of Goalando township, Rajbari District, Dhaka Division, 70 km west of the city of Dhaka (Figure 1). Ten deaths were reported among 12 ill persons with symptoms compatible with NiVE, resulting in a case-fatality ratio of 83% (9,23). Although previous outbreaks of NiVE outside Bangladesh involved primarily men and women >25 years of age (5,16,17,19,26), most (75%) patients in this outbreak were boys <15 years of age. We describe a matched case-control study that was conducted to characterize the epidemiology of NiVE and, specifi cally, to determine if risk for NiVE was associated with contact with animals; an environmental exposure, activity, or behavior; or contact with other NiVE patients during the 2004 NiVE outbreak in Goalando township.

Study Participants, Materials, and Methods
A matched case-control study was conducted in Goalando, Bangladesh (Figure 1), February [18][19][20][21][22]2004. Hypotheses tested in this study, as mentioned above (e.g., increased risk for NiV infection caused by contact with animals, environmental exposure, contact with fruit in season) were based upon factors associated with previous outbreaks of NiVE in Malaysia, Singapore, and Bangladesh.

Case Defi nition
A confi rmed NiVE case-patient was defi ned as any patient with fever and symptoms compatible with encephalitis after December 15, 2003, with NiV-specifi c immunoglobulin M antibodies in cerebrospinal fl uid (CSF) or serum by enzyme immunoassay (EIA). A probable case of NiVE was defi ned as a patient with a diagnosis of encephalitis in whom fever developed and who was living in the same village as a patient with a confi rmed case of NiVE after December 15,2003. Cases remained in the probable category if the patient died and a specimen for laboratory confi rmation could not be obtained.
We conducted a population census of the affected area in February 2004; this census was the basis for selecting controls. We identifi ed 3 controls for each case-patient. The controls were selected randomly from the population and then matched to each case-patient on the basis of gender and age group. All households identifi ed during the census, including houses of case-patients and controls, were mapped by Global Positioning System, and data were uploaded into ERDAS Imagine 8.5 (Leica Geosystems, Atlanta, GA, USA) and merged with a November 2000 IKONOS Geo 1-m satellite image of the outbreak area (Space Imaging, Thornton, CO, USA).
Participation was strictly voluntary, and written informed consent was obtained for all participants; for those <18 years of age, individual and parental consent was obtained. The Bangladesh Ministry of Health and Family Welfare that requested this investigation reviewed and approved all protocols.

Study Population
Probable and confi rmed cases identifi ed in 2 contiguous villages of Goalando township ( Figure 1) were included in this study. Seven of the 12 cases were clustered within 3 households. Of these 7 clustered cases, 3 occurred in 1 household, and the remaining 4 were distributed in 2 separate homes (Figures 1, 2). Therefore, we conducted 2 separate analyses to assess the effect of case clustering on results. The fi rst analysis contained the complete dataset of 12 cases and 36 controls; the subanalysis consisted of 8 cases (we randomly selected 1 case/household) and 24 matched controls. Similar results (proportions, odds ratios [ORs], 95% confi dence intervals [CIs]) were obtained from both analyses. Thus, data presented in this article, including all tables, are derived from the complete dataset.

Specimen Collection and Testing
Serum samples and CSF were tested as previously described (27). When possible, a serum specimen was collected from controls.  Figure 2. Map also shows extreme habitat disturbance; areas under cultivation (for rice, sugar cane) are highlighted with "C," and remaining trees (fruit trees and bamboo stands) with "T." Bottom: Location of outbreak village.

Data Collection and Interviews
After informed consent was obtained, case-patients and controls were interviewed at home by trained interviewers, in their native Bengali language, with a standardized questionnaire. Information such as demographics, types of animal exposures, environmental and occupational exposures, exposure to ill persons, and history of illness was obtained. Proxy interviews of family members and/or friends were conducted for deceased patients. To minimize interview bias, proxy interview methods were also used for all controls that were matched to deceased case-patients.

Statistical Analysis
Exact ORs and 95% CIs were calculated by using a matched univariate logistic regression analysis in SAS version 9.0 (SAS Institute Inc, Cary, NC, USA) (28). Associations were considered statistically signifi cant at p<0.05.

Descriptive Characteristics
Four (33%) cases were confi rmed by EIA; the remaining 8 (67%) case-patients, from whom a diagnostic specimen was not available, were considered probable cases. Among all 13 (36%) controls who consented to blood collection, results of serologic tests for NiV-specifi c antibodies were negative. Furthermore, none of the controls reported having had a perceived fever or symptoms compatible with NiVE from December 15, 2003, through the week the study was conducted (February [18][19][20][21][22]2004). In addition, an antibody prevalence study conducted among persons (n = 300) living in the outbreak site showed no evidence of asymptomatic or mild infection, which suggested that controls entered into the study were likely uninfected (A. Croisier, unpub. data). Proxy interviews were administered to equal proportions of case-patients (83%) and controls ( Table 1). The median age of case-patients included in the study was 11.5 years (range 2-28 years); 9 (75%) were male, and 11 (91%) were <15 years of age ( Table 1). Residences of all case-patients and controls were located within the affected villages, an area with a radius of ≈800 m ( Figure 1).

Animal Exposures
In the matched case-control analysis, a greater percentage of case-patients (60%) than controls (34%) had observed or touched dead animals, although this fi nding was not statistically signifi cant ( Table 2). We observed no differences between case-patients and controls with respect to contact with ill animals ( Table 2), including pigs, ruminants, and fruit bats. Chickens and ducks were often slaughtered for religious purposes or for consumption; however, close contact with these animals and their bodily fl uids (e.g., blood, saliva) during this process was not associated with NiV infection ( Table 2). None of the case-patients or controls had known contact with pigs (healthy or ill) or pig excreta ( Table 2). Four (36%) of 11 case-patients and 7 (19%) of the controls observed fruit bats around their household during the night (OR 4.1, p = 0.49; Table 2). However, some proxy family members and/or friends answering on behalf of patients who had died were unable to answer specifi c questions (e.g., Did you observe fruit bats around your house during the night?).

Environmental and Behavioral Exposures
A greater proportion of case-patients (83%) than controls (51%) reported having climbed trees between December 15, 2003, and February 3, 2004 (OR 8.2, p = 0.025; Table 2). No statistically signifi cant differences were observed between case-patients and controls with respect to outdoor activities such as hunting, fi shing, or playing outdoor games (e.g., hide-and-seek, cricket, soccer). Eating fruit that was locally available (on trees or collected from fruit trees locally) between December and February  (17) was not associated with illness, regardless of how the fruit was collected (from the ground, picked from tree, from the market) (Table 2). Although a greater proportion of casepatients reported environmental exposures (drinking raw date palm sap, harvesting date palm sap, having someone in the household who collects date palm sap, or drinking sap directly from the collection vessel), these differences were not statistically signifi cant (Table 2).

NiVE Case Exposure
There were strong associations between illness and 1) visiting a hospital and/or 2) having had contact with a probable or confi rmed NiVE patient ( Table 2). In one 2-case family cluster, a mother (26 years of age) and her infant son (2 years of age) both became ill and died. The child became symptomatic 2 days before the mother's illness onset (Figure 2; household 4). Among the other affected family clusters, the patients became ill within 3 days of one another (Figure 2; households 1 and 5); all persons in these 2 clusters reported a history of climbing fruit trees. There was no evidence of contact of persons between case households during their illness.

Discussion and Conclusions
In contrast to the patients in the Malaysian and Singapore outbreaks, which occurred primarily among adults, a preponderance of the NiV patients in the January/February 2004 Bangladesh outbreak were young boys. These fi ndings, in the absence of high infection rates among adults or evidence of antibodies to NiV in the general population (investigation team, unpub. data), suggest an association between NiV infection and some childhood activity or specifi c behavior. The odds of NiV infection were significantly elevated among persons who climbed trees, an activity observed almost exclusively among boys <15 years of age. This behavior is quite common among children because they gather fruit from trees. Therefore, these children may have had contact with partially eaten fruit from fruit bats or the secretions/excretions of these animals. Or, the children may have contacted contaminated fruit bat guano or urine in the trees. The percentages of case-patients playing hideand-seek, hunting, and fi shing-all of which were typical behaviorial traits of local boys-were not signifi cantly different than those for controls. These activities generally occur outdoors; however, they do not place a child in direct contact with bat excretions or secretions, as may be true for tree climbing. Therefore, infection was apparently related to a specifi c behavior, tree climbing, rather than age or outdoor activities in general. Furthermore, although other exposures that may have placed persons in closer contact with bat secretions (e.g., collecting fruit or palm sap from trees, drinking palm sap directly from collection vessel) were observed more often among case-patients than controls, these fi ndings were not statistically signifi cant; perhaps because of the small sample size. Nonetheless, our fi ndings can and have been used to help guide NiV outbreak investigations, leading investigators to similar conclusions as ours (29).
Fruit bats forage at night in various trees that are producing ripe fruit and often drink from palm sap collection vessels (30). Fruits are also a major food source for many villagers and, as a result of environmental disturbances (31) in the form of crop development (e.g., jute, rice, and sugar cane), the few remaining fruit trees grow only in close proximity to human dwellings ( Figure 1). This in turn creates a situation in which fruit bats are forced into close proximity with humans, especially while these mammals are foraging and feeding. In addition, date palm sap is routinely collected in rural areas of Bangladesh between December and May. According to villagers, including palm sap harvesters, dead fruit bats are occasionally found in the collection vessels. Local villagers reported that they often observed fruit bats feeding from palm sap collection vessels, and some collectors place cloth over the opening of the vessel to prevent this (investigational team observation). In fact, a greater proportion of case-patients in our study collected palm sap, drank from the palm sap collection vessel, or had a family member who collected palm sap; however, these differences were not statistically signifi cant. The power of our study to detect exposure risks was limited by the outbreak size. Therefore, until additional data are available, remaining cautious of date palm sap collection vessels, especially those visibly contaminated with fruit bat excreta or carcasses, would be prudent.
Numerous investigators have found serologic evidence suggesting that fruit bats of the genus Pteropus are the reservoir hosts for NiV (23,24), and there are reports of NiV isolation from bat urine (20,25) and partially eaten fruit (20). Unpublished laboratory data from the Bangladesh investigation have not supported the presence of an intermediate or primary reservoir host other than P. giganteus. Available data from this study, therefore, suggest direct transmission of NiV to humans through contact with bat secretions or excretions (saliva, urine, guano, partially eaten fruit) during fruit-tree climbing.
Although indirect contact with bats may have been the primary means of infection for this outbreak, Hsu and others (19) demonstrated that contact with ill cows was associated with an increased risk for NiV infection during the 2001 Bangladesh NiV outbreak. Therefore, intermediated hosts should be considered in future NiV outbreaks in Bangladesh.
In contrast to the patients in the Malaysia and Singapore outbreaks (5,16,17,25,26), most of the Bangladesh population (and all of the case-patients included in this study; data not shown) are practicing Muslims who do not consume pork and who avoid contact with pigs. None of the case-patients and controls in our study population reported any contact with pigs or pig excreta, so it is unlikely that these animals played a role in this outbreak.
Clustering of cases within households was a prominent feature of this outbreak ( Figure 2); 1 household contained 3 case-patients, all brothers of ages 7-15 years. However, the longest estimated incubation periods (duration from symptom onset to fi rst known exposure to a NiVE family member) within the clusters reported here were less than the currently recognized 4-day minimum (7). This fi nding suggests that the family clustering may have resulted from a common source of infection (e.g., a specifi c tree they climbed, fruit they consumed, or palm sap collection vessel they were in contact with) rather than person-to-person transmission. Our data also show strong associations between NiV infection and visiting a hospital. However, because the participants were asked if they had visited a hospital within a range of dates (December 15, 2003-February 3, 2004 and not a specifi c date, we were unable to determine if they were ill with NiV before visiting the hospital or whether they acquired their infection there. Some accounts in the literature suggest person-to-person transmission of NiV; therefore, it is plausible that someone could acquire, through contact with a patient's secretions or excretions, an NiV infection while visiting a hospital (6,10,20). Nevertheless, the most probable explanation for the observed association is that NiV encephalitis patients during this outbreak were severely ill, requiring hospitalization.
Although person-to-person transmission may have occurred in this outbreak, the initial infection (index case) may have occurred through contact with bat secretions rather than contact with an intermediate host. A limitation of our study is that we were unable to identify a specifi c mechanism by which person-to-person transmission may have occurred. NiV has been isolated from the respiratory secretions and urine of patients in the Malaysia, Singapore, and current Bangladesh outbreaks (3,8,32,33), which suggests a potential for NiV to be transmitted from person to person. Data based upon chain-of-transmission events and clustering of cases during other 2003 and May 2004 Bangladesh outbreaks led investigators to conclude that human-to-human transmission may have occurred (3,19). Therefore, given the potential for household or nosocomial transmission, we recommend the use of personal protective equipment (i.e., gloves, masks, gowns, and eye protection); strict hand hygiene and surface disinfection during and after contact with an NiVE patient; isolation of patients with confi rmed or suspected NiV infection; and proper disposal of potentially contaminated materials.
In summary, tree climbing, a behavior largely engaged in by young boys, was associated with an increased risk for NiV infection; although the exact mode of transmission is unclear. Our data do not rule out the potential for person-toperson transmission. If person-to-person transmission were extremely effi cient, the conditions and population density of Bangladesh (≈1,000 persons/km 2 ; total population 141 million/144,000/km 2 ) may have resulted in a much larger outbreak. Indeed, a study among health workers in Bangladesh did not fi nd evidence of incidental transmission to persons caring for patients hospitalized with Nipah-related illnesses (34). Bat-to-human was the most probable route of transmission in Goalando; however, some undetermined intermediate or incidental hosts cannot be ruled out. Periodic introductions of NiV to human populations in this region may continue to occur because of the overlapping nature of human and pteropid bat habitats. Moreover, bathuman interactions are likely to increase due to bat habitat loss because the few fruit trees that remain will likely be found in close proximity to human dwellings ( Figure 1).
As a prevention measure, we recommend avoiding contact with fruit bats and their secretions/excretions. We also encourage persons to wash and/or peel fruit, in addition to washing their hands, before preparing meals or consuming fruit. Greater understanding of the relationships between pteropid fruit bats, NiV, and its transmission to humans might offer additional strategies for safe coexistence and disease prevention for Bangladesh and other countries where fruit bats reside. Finally, because the geographic range of this highly lethal pathogen may correspond to the distribution of the genus Pteropus, including parts of China and Australia, most of the Indian subcontinent, and Southeast Asia (12,30), factors that promote transmission from bats to humans need to be defi ned and the role of person-toperson transmission needs to be better characterized.