Bulletin of the World Health Organization

Estimated global incidence of Japanese encephalitis: a systematic review

Grant L Campbell a, Susan L Hills b, Marc Fischer b, Julie A Jacobson c, Charles H Hoke d, Joachim M Hombach e, Anthony A Marfin f, Tom Solomon g, Theodore F Tsai h, Vivien D Tsu i & Amy S Ginsburg i

a. Ross River Consulting, LaPorte, United States of America (USA).
b. Centers for Disease Control and Prevention, Fort Collins, USA.
c. Bill & Melinda Gates Foundation, Seattle, USA.
d. Pharmaceutical Systems Project Management Office, US Army Medical Material Development Activity, Fort Detrick, USA.
e. Initiative for Vaccine Research, World Health Organization, Geneva, Switzerland.
f. Washington State Department of Health, Communicable Disease Epidemiology, Shoreline, USA.
g. Institute of Infection and Global Health, University of Liverpool, Liverpool, England.
h. Scientific Affairs, Novartis Vaccines, Cambridge, USA.
i. Program for Appropriate Technology in Health (PATH), PO Box 900922, Seattle, WA, 98109, USA.

Correspondence to Amy S Ginsburg (e-mail: AGinsburg@path.org).

(Submitted: 09 December 2010 – Revised version received: 17 June 2011 – Accepted: 10 July 2011 – Published online: 03 August 2011.)

Bulletin of the World Health Organization 2011;89:766-774E. doi: 10.2471/BLT.10.085233

Introduction

Japanese encephalitis (JE) is among the most important viral encephalitides in Asia, especially in rural and suburban areas where rice culture and pig farming coexist.13 It has also occurred rarely and sporadically in northern Australia and parts of the Western Pacific.46 JE is due to infection with the JE virus (JEV), a mosquito-borne flavivirus. The main JEV transmission cycle involves Culex tritaeniorhynchus mosquitoes and similar species that lay eggs in rice paddies and other open water sources, with pigs and aquatic birds as principal vertebrate amplifying hosts.1,2,7 Humans are generally thought to be dead-end JEV hosts, i.e. they seldom develop enough viremia to infect feeding mosquitoes. Fewer than 1% of human JEV infections result in JE. Approximately 20–30% of JE cases are fatal and 30–50% of survivors have significant neurologic sequelae.8 JE is primarily a disease of children and most adults in endemic countries have natural immunity after childhood infection, but all age groups are affected. In most temperate areas of Asia, JEV is transmitted mainly during the warm season, when large epidemics can occur. In the tropics and subtropics, transmission can occur year-round but often intensifies during the rainy season.13

The global incidence of JE is unknown because the intensity and quality of JE surveillance and the availability of diagnostic laboratory testing vary throughout the world. Countries that have implemented high-quality childhood JE vaccination programmes have seen a dramatic decline in JE incidence. Although JE is reportable to the World Health Organization (WHO) by its Member States, reporting is highly variable and incomplete. In the late 1980s, Burke and Leake estimated that 50 000 new cases of JE occurred annually among the 2.4 billion people living in the 16 Asian countries considered endemic at the time (approximate overall annual incidence: 2 per 100 000).2 In the intervening two decades, despite major population growth, urbanization, changes in agricultural practices and increased use of the JE vaccine in many countries, this figure has been widely quoted, including very recently.913 In 2000, assuming an annual, age-group-specific incidence of 25 cases per 100 000, Tsai estimated that in the absence of vaccination 175 000 cases of JE would occur annually among Asian children aged 0–14 years living in rural areas.14 The current study used more recent, published, local or national incidence estimates and current population data to produce an updated estimate of the annual global incidence of JE.

Methods

We approximated the JE-affected territory of each of the 24 countries endemic for JE using a recent update15 of an earlier approximation by Tsai16 with some modifications (Table 1, available at: http://www.who.int/bulletin/volumes/89/10/10-085233). Based on these same approximations,15,16 we then stratified the JE-affected territory of some countries (e.g. China excluding Taiwan, India and Nepal) into two or more incidence strata. Because suitable studies of JE incidence were not available for every endemic country or incidence stratum, we sorted JE-endemic countries and incidence strata into 10 incidence groups (A, B, C1, C2 and D through I) based primarily on geographic proximity, ecologic similarity, vaccine programme similarity. Table 1 briefly describes the status of each endemic country’s JE vaccination programme as of 2009, according to recent publications and unpublished sources.8,1720

Incidence data

We identified studies that contained potentially useful data on the incidence of JE in Asia in a manner similar to the one used in a recent study of global typhoid fever incidence.21 Whenever possible, this review followed the relevant guidelines for Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA).22 The review process is described as follows and no protocol is available. We conducted a MedLine search of papers in any language published between 1985 (arbitrarily chosen to emphasize recent data) and April 2010. We used the following search strategy and keywords: Japanese encephalitis and (incidence or prevalence or public health or surveillance or distribution or epidemiology). An initial set of 1374 unique citations was downloaded into a computerized database. The lead author then culled the set to approximately 255 citations by scanning titles and abstracts (if available) and discarding references that did not involve JE surveillance or epidemiology in humans (e.g. studies focused on non-human vertebrates or on entomology, molecular biology or virology). The lead author also obtained a full-text copy of each available reference for review and further culled the list to approximately 75 references by selecting only those that contained explicit human JE incidence data or data that could be used to calculate incidence. Additional references were sought from these papers’ reference lists and from the collections of the authors (including several papers recently submitted or “in press”), and this yielded a total of 80 references of interest. Each of these papers (or the English translation of five Chinese-language papers) was then carefully reviewed by at least two authors to further cull the list, resulting in a final group of 12 studies that provided original, recent population-based and largely laboratory-confirmed incidence data (or hospital-based incidence data in a defined population).13,20,2332 These 12 studies consisted of one study each from Bangladesh, Cambodia, Indonesia, Malaysia and Thailand, two from China (excluding Taiwan) two from Japan and three from Nepal. They represented all but three (B, E and I) of the 10 incidence groups (Table 1). Fig. 1 shows the approximate locations of the study sites in each primary reference. Point estimates of the annual JE incidence in each primary reference are shown as cases per 100 000 persons (Table 2, http://www.who.int/bulletin/volumes/89/10/10-085233). We used an Excel spreadsheet for all calculations in this study.

Fig. 1. Primary reference study area(s)
Fig. 1. <b>Primary reference study area(s)</b>
a Mountain and hill (non-Terai) districts only

a Mountain and hill (non-Terai) districts only

Population data

For each endemic country, we obtained the total population and that of the approximate JE-affected area from the country’s official web site (as identified through www.GeoHive.com). We used the most recent census data (1998–2011) or official inter-census estimates (2000–2010) and dichotomized age as follows: children, aged 0–14 years; adults, aged ≥ 15 years (Table 1). For each country, the estimated total population aged 0–14 years was based on percentages for 2010 obtained from the United Nations33 (for Taiwan, China, the percentage was obtained from The world factbook34). Because a concise source of such percentages for JE-endemic administrative divisions and subdivisions of countries was not available, we assumed that the age distribution within each country was homogeneous.

Assumptions by incidence group

In unvaccinated populations in endemic areas, JE is largely a paediatric disease and most people have acquired active immunity by adulthood. Conversely, in areas with long-standing, high-quality childhood vaccination programmes, JE is usually a rare disease of non-immune adults, especially the elderly. A transitional phase in which the incidence of JE in children and adults is about the same can also occur.3 Because many of our primary references provided point estimates of JE incidence for all age groups combined or for children only (generally 0–14-year-olds), assumptions about age-specific JE incidence were often necessary. We had to make additional assumptions because of the absence of appropriate primary references for incidence groups B, E and I. We assumed that in some countries with a historically non-homogeneous distribution of JE risk (e.g. plain versus highland areas of Sri Lanka and northern versus southern portions of Thailand), the preferential use of vaccine in areas that were formerly higher-risk has homogenized the risk countrywide. Incidence groups are described in Table 1 and additional group-specific assumptions are summarized as follows:

  • Incidence Group A: Overall incidence is 0.003 per 100 000; the child (aged ≤ 14 years) to adult (aged > 14 years) case frequency ratio is 7:1 (based on visual inspection of a figure in the report by Arai et al.23). Because this study23 and the study by Hashimoto et al.24 overlapped in time and because both were based on national surveillance, a weighted average of their respective results could not be used. Therefore, we used the incidence estimate by Arai et al. in our analysis but included the study by Hashimoto et al. in Table 2 for completeness and as general validation of the results obtained by Arai et al.
  • Incidence Group B: JE is rare, with an overall incidence of 0.003 per 100 000, i.e. the authors assumed it to be the same as in historically high-incidence areas with long-standing, high-quality vaccination programmes (i.e. Incidence Group A); the child to adult case frequency ratio is 7:1.23
  • Incidence Group C1: Overall incidence is 3.3 per 100 000 (weighted average of results from Yin et al.20 and Xufang et al.13); the child to adult case frequency ratio is 3:1 (based on the ratio of 156 cases in 0–14-year-olds to 57 cases in persons aged > 14 years among combined residents and non-residents, as reported by Yin et al.20).
  • Incidence Group C2: Overall incidence is 0.01 per 100 000; the child to adult case frequency ratio is 3:1 (as for Incidence Group C1).
  • Incidence Group D: Incidence in children aged 12–14 years is the same as in children aged 0–11 years; incidence in children aged 0–14 years is 10.6 per 100 000 (weighted average of results from Touch et al.,25 Kari et al.26 and Hoke et al.27); the child to adult case frequency ratio is 7:1 (based on the ratio of 43 childhood cases to 6 adult cases detected from July 2001 through 2006 by Wong et al.32).
  • Incidence Group E: Age-specific incidences in this group are assumed by the authors to be half as high as in incidence group D.
  • Incidence Group F: Overall incidence is 2.8 per 100 000 (approximate weighted average of results from two studies from Nepal28,29); the child to adult case frequency ratio is 5:4 (based on the ratio of 532 cases in 0–14-year-olds to 419 cases in persons aged > 14 years, as reported by Wierzba et al.29).
  • Incidence Group G: Overall incidence is 1.0 per 100 000 (approximate weighted average of results from three studies from Nepal28,29,31 and one study from Bangladesh30); the child to adult case frequency ratio is 4:1 (based on the ratio of 73 cases in 0–14-year-olds to 17 cases in persons aged > 14 years, per the table in the report by Bhattachan and colleagues31).
  • Incidence Group H: Overall incidence is 1.5 per 100 000 (based on 49 cases per 600 000 persons over a 5.5-year period in the study by Wong et al.); incidence in children aged 12–14 years is the same as in children aged 0–11 years; the child to adult case frequency ratio is 7:1.32
  • Incidence Group I: Overall incidence is 3.3 per 100 000 (based on the average annual incidence of 6.5 per 100 000 in nearby Taiwan (China) during the immediate pre-vaccination era [1965–1967],35 reduced by 50% following a national mass-vaccination campaign among children in 2009); the child to adult case frequency ratio is 7:1.32

Results

Based on the assumptions noted above, we estimated that approximately 67 900 JE cases typically occur annually in the 24 JE-endemic countries, for an incidence of 1.8 per 100 000 overall (Table 3). Approximately 33 900 (50%) of these cases occur in China (excluding Taiwan) and approximately 55 000 (81%) occur in areas with well established or developing JE vaccination programmes, while approximately 12 900 (19%) occur in areas with minimal or no JE vaccination programmes. Approximately 51 000 (75%) of these cases occur in children aged 0–14 years, which gives an estimated overall annual incidence of 5.4 per 100 000 in this age group.

Discussion

In the pre-JE-vaccination era, tens of thousands of JE cases were often reported annually in Asia. During 1965–1975, more than 1 million cases were reported in China alone.1 In Japan, the Republic of Korea and Taiwan (China) the introduction of routine childhood vaccination programmes against JE beginning 40 to 50 years ago, combined with increased urbanization and evolving agricultural practices, resulted in the virtual elimination of JE, despite continued enzootic JEV transmission.2,14 Other Asian countries were slower to implement childhood JE vaccination programmes largely because of expenses and difficult logistics. Some of these countries now have well developed vaccination programmes, others have less developed programmes, and others still have no programmes at all.8,17,19

During 2006–2009, JE-endemic countries reported 27 059 cases of JE (annual range: 4502–9459; average: 6765) to WHO. Fully 86% of these (23 176 cases; average: 5794 cases per year) were reported from China and India; 16 countries reported a total of 3883 cases (annual average: 971); and 5 countries reported no cases.36 While these data almost certainly include some non-JE cases that were clinically misclassified without the benefit of laboratory confirmation, they almost certainly also represent significant under-reporting of true cases.1,14 The results of the current study suggest that the actual incidence of JE is nearly 10 times higher than reflected in recent reports to WHO. While we estimated that approximately 81% of JE cases presently occur in areas with well established or developing JE vaccination programmes, this probably reflects the epidemiological distribution of JE, since vaccination programmes are more likely to exist in areas with the highest risk of JEV transmission. It may also reflect better surveillance and case-reporting in such areas. Vaccination programmes in many of these areas should be expanded and strengthened. Endemic countries without programmes or with programmes in early development should be fully supported in their efforts to implement and strengthen JE vaccination programmes.

Because we studied JE incidence, not mortality or morbidity (e.g. in terms of disability-adjusted life years [DALY]), we did not attempt to estimate the global burden of JE.37,38 Nevertheless, our results could contribute to efforts to refine estimates of this burden. Based on an earlier estimated frequency of 44 000 new JE cases per year, including 14 000 deaths (32%) and 24 000 cases with sequelae (55%, or 71% of survivors), and an average disability weight of 0.616, Mathers et al. estimated the global burden of JE to be 709 000 DALY (by comparison, the estimated global burden of acute malaria was 46 000 000 DALY).37 If we assume a case fatality rate of 20–30% and long-term neuropsychological morbidity among 30–50% of survivors,8 the results of the current study similarly suggest that approximately 13 600 to 20 400 acutely fatal JE cases occur and that 14 300 to 27 200 JE survivors develop long-term neuropsychological sequelae each year. Although several studies characterizing long-term outcomes in JE survivors have been published,3942 additional studies are needed to allow for refined estimates of the global burden of JE.

The methods we used in the current study resembled the methods that have been employed to estimate the global incidence of other infectious diseases, including typhoid fever, Haemophilus influenzae type b disease and streptococcal disease.21,43,44 Our study is therefore subject to many of the same limitations. First, we derived estimates from sparse data of variable quality, with some notable exceptions (e.g. data from Japan). Furthermore, useful data were lacking from several important countries or parts of countries. The absence of data from India is particularly worrisome because the country has a very large population and several diverse JE-endemic regions Although we did not feel that a formal sensitivity analysis was essential to enable us to appropriately interpret our results, these results would obviously be highly sensitive to relatively small changes in some input data, especially for China and India, which together contain approximately two thirds of the world’s population at risk for JE. This is particularly true for Incidence Group C1 (the higher-incidence stratum of China), which accounted for half of our estimated global total of annual JE cases, and, to a lesser extent, for incidence groups D, F and H (much of India and south-eastern Asia). In contrast, our results would be far less sensitive to changes in input data for incidence groups A, B, C2 and I. Another limitation of our approach is that some of the 12 key references for JE incidence involved hospital-based surveillance in imperfectly defined catchment areas. Such studies tend to underestimate true incidence because they depend on all true cases being hospitalized (as opposed to being treated as outpatients) and within their residential catchment area. Other studies relied on an incomplete network of sentinel hospitals and are similarly subject to an underreporting bias, although such a bias is not readily quantifiable. The following are also potential sources of error that are difficult to quantify: (i) a lack of standardized laboratory testing methods (e.g. the use of various commercially available or in-house test kits for the detection of anti-JEV immunoglobulin M antibody); (ii) incomplete collection of clinical samples (e.g. failure to collect and test both acute- and convalescent-phase samples in all suspected cases, resulting in some degree of underdiagnosis); and (iii) the co-circulation of other cross-reactive flaviviruses (especially dengue viruses) in some JE-endemic areas.

We did not attempt to distinguish between urban and rural incidence because few, if any, of the studies on which our results were based broke down incidence in this way. They involved surveillance in whole countries or in areas that included a mix of urban and rural populations. Nevertheless, the risk of JE continues to be strongly associated with exposure in rural and, to a lesser extent, suburban areas.3 Another limitation of our approach is that JE is an ecologically complex epidemic as well as endemic disease and therefore does not exist in a steady-state, even when incidence is reduced by means of increased vaccine usage. Thus, to state that a set number of new JE cases occurs globally each year (e.g. 50 000, 67 900 or 175 000) is overly simplistic, however convenient it may be for some purposes, because the situation is clearly more dynamic. It was recently discovered, for example, that the JEV is circulating in Tibet (China), formerly believed to be non-endemic.45 More accurate estimates of the global incidence of JE, including its natural fluctuations over time, will await the development of better surveillance and laboratory capability throughout the endemic region, as well as the publication of additional population-based surveillance studies of laboratory-confirmed cases. To this end, agencies funding vaccine programmes should consider concurrently funding well designed, population-based surveillance studies of laboratory-confirmed cases, particularly in JE-endemic countries for which no reliable incidence data currently exist.


Acknowledgements

The authors thank Sybil Barney, James Cheyne, Laure Dumolard, Lauren Gazley, Erin Kester, Megan Le, Jennifer Lehman and Ginger Topel for administrative support; Pritu Dhalaria Raj, Shankar Ghosh and Mong How Ooi for helpful consultations and discussions; Repon Paul and Zundong Yin for sharing “in press” manuscripts; and Tang Yi for graciously providing translations of selected articles published in Chinese.

Funding:

This study was funded by the Program for Appropriate Technology in Health (PATH).

Competing interests:

Theodore Tsai is a full-time employee of Novartis Vaccines. The views expressed in this publication are those of the authors alone and do not necessarily represent the decisions, policy, or views of the Centers for Disease Control and Prevention.

References

Share