The global epidemiology of hepatitis E virus infection: A systematic review and meta‐analysis

Abstract Background and aims Hepatitis E virus (HEV), as an emerging zoonotic pathogen, is a leading cause of acute viral hepatitis worldwide, with a high risk of developing chronic infection in immunocompromised patients. However, the global epidemiology of HEV infection has not been comprehensively assessed. This study aims to map the global prevalence and identify the risk factors of HEV infection by performing a systematic review and meta‐analysis. Methods A systematic searching of articles published in Medline, Embase, Web of science, Cochrane and Google scholar databases till July 2019 was conducted to identify studies with HEV prevalence data. Pooled prevalence among different countries and continents was estimated. HEV IgG seroprevalence of subgroups was compared and risk factors for HEV infection were evaluated using odd ratios (OR). Results We identified 419 related studies which comprised of 1 519 872 individuals. A total of 1 099 717 participants pooled from 287 studies of general population estimated a global anti‐HEV IgG seroprevalence of 12.47% (95% CI 10.42‐14.67; I 2 = 100%). Notably, the use of ELISA kits from different manufacturers has a substantial impact on the global estimation of anti‐HEV IgG seroprevalence. The pooled estimate of anti‐HEV IgM seroprevalence based on 98 studies is 1.47% (95% CI 1.14‐1.85; I 2 = 99%). The overall estimate of HEV viral RNA‐positive rate in general population is 0.20% (95% CI 0.15‐0.25; I 2 = 98%). Consumption of raw meat (P = .0001), exposure to soil (P < .0001), blood transfusion (P = .0138), travelling to endemic areas (P = .0244), contacting with dogs (P = .0416), living in rural areas (P = .0349) and receiving education less than elementary school (P < .0001) were identified as risk factors for anti‐HEV IgG positivity. Conclusions Globally, approximately 939 million corresponding to 1 in 8 individuals have ever experienced HEV infection. 15‐110 million individuals have recent or ongoing HEV infection. Our study highlights the substantial burden of HEV infection and calls for increasing routine screening and preventive measures.


| INTRODUC TI ON
Hepatitis E virus (HEV) as a positive-sense single-stranded RNA virus is a leading cause of acute viral hepatitis worldwide. The infection is usually asymptomatic or self-limiting in the general population. However, acute infection in pregnant women may cause severe clinical outcomes, including fulminant hepatic failure with high mortality rate reaching up to 20%-30%. 1 These patients are mostly from resource-limited regions. In European countries, HEV infection has been frequently reported to bear high risk of developing into chronic hepatitis in immunocompromised individuals, in particular organ transplant patients. 2,3 Thus, HEV is truly imposing a global health burden in both developing and developed countries.
Currently, eight distinct genotypes (GTs) of HEV have been classified. 4 GT 1-4 are known to be the main threat to humans. GT 1 and GT 2 are restricted to human and mainly transmit through contaminated water causing acute hepatitis. GT 3 and GT 4 are zoonotic and have been identified in a wide spectrum of hosts, including human, swine, wild boar, goat, cattle, deer, camel and yak. 5 Both GT 3 and GT 4 can cause chronic infection in organ transplant patients, 2,6 and consumption of raw or undercooked animal meat has been recognized as the main routes of causing sporadic cases in developed countries. 7 In fact, the host range of HEV is ever expanding and the implications of the rare GTs and the newly discovered strains in human health remain largely uncertain. 7 This further complicates the transmission and the risk of HEV infection. In addition to the classical waterborne and foodborne transmission routes, blood transfusion-mediated transmission has been reported in organ transplant patients. 8 Person-to-person transmission has also been proposed. 9 Intriguingly, recent evidence has indicated that pet animals including dogs, cats, rabbits and horses might be accidental hosts for HEV and constitute a potential source for transmitting to human. 10,11 Thus, there is an urgent need to comprehensively understand the risks for HEV infection, in order to device preventive measures.
Globally, it has been roughly estimated that one-third of the population are living in HEV endemic areas. 12 More recently, substantial efforts have been dedicated to systematically evaluate HEV prevalence in different continents (eg the Americas and Europe), 13,14 different countries (eg industrialized countries, China, Iran, Brazil and Somalia) [15][16][17] and special populations or settings (eg blood donors, swine workers and outbreak setting). [18][19][20] Most of these studies are based on seropositivity of anti-HEV IgG antibody. Anti-HEV IgG antibody developed post-infection usually persists for many years, and is thus regarded as a marker for past infection. 21,22 In contrast, anti-HEV IgM antibody is short-lived up to a few months, thus considered as evidence of recent or current infection. Detection of HEV RNA is a bona fide marker for active ongoing infection. In this study, we aimed to systematically estimate the global burden of HEV infection. More specifically, we have mapped the global prevalence of past, recent and ongoing HEV infection and evaluated the key risk factors of infection.

| Data sources and searches
A systematic search was conducted in Medline, Embase, Web of science, Cochrane CENTRAL and Google scholar. Databases were searched for articles in the English language from inception until July 2019. All searches from database were performed by a biomedical information specialist of the medical library, with an exhaustive set of search terms related to hepatitis E virus infection and epidemiology (the full search strategies are provided in the Supporting Information S1). This study is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis. 23 No institutional review board approval was required for this meta-analysis because our study only included data which had been published previously.

| Study selection
Studies were included if they met the following criteria: (a) Studies which contained data about seroprevalence of anti-HEV IgG, anti-HEV IgM or HEV RNA positivity, (b) Studies contained mixed population were excluded unless they clearly and explicitly reported the ongoing HEV infection. Our study highlights the substantial burden of HEV infection and calls for increasing routine screening and preventive measures. Two reviewers (PL and JL) worked independently to determine whether a study met inclusion criteria, abstracted information to assess the methodological validity of each candidate study and extracted data with structured data collection forms. The reviewers resolved discrepancies by jointly reviewing the study in question. If no consensus was reached, a third reviewer (QP), unaware of prior determinations, functioned as an arbiter.

| Data extraction and quality assessment
Eligible studies were further divided into three study populations: identified that reported on the same populations and outcomes, only the most representative and comprehensive study was included for further meta-analysis in order to avoid duplicate data. The quality of studies was assessed using the Joanna Briggs Institute checklist for prevalence studies, which enabled assessment of included studies in relation to risk of bias, rigour and transparency. 24 Studies scoring 1-3 were defined as low quality, 4-6 as average quality and 7-9 as high quality (Table S1). Studies were not excluded on the basis of their quality score to increase transparency and to ensure all available evidence in this area was reported.

| Statistics analysis
After checking for consistency, the Metaprop module in the R-3.5.3 statistical software package was used for meta-analysis. A 95% confidence interval (95% CI) was estimated using Wilson score method, and pooled seroprevalence was calculated with the DerSimonian-Laird random effects model with Freeman-Tukey double arcsine transformation. Heterogeneity across the included studies was assessed using the Cochran Q statistics and I 2 statistics, with I 2 statistics 25%-50%, 50%-75% and >75% considered as mild, moderate and severe heterogeneity respectively. When heterogeneity was higher than 50%, a random effect model will be used. ORs were used to report the risk factors for HEV infection. ORs and their 95% CI were extracted directly from studies when available, with adjusted ORs extracted preferentially over unadjusted ORs. If included studies did not report ORs, crude ORs were calculated from extracted data.
We then pooled the ORs using the DerSimonian and Laird random effect models, with the heterogeneity estimated from the Mantel-Haenszel model. Funnel plots and Egger regression test were used to assess potential publication biases. Additionally, we performed sensitivity analyses using "metainf" in a random model to investigate

| Prevalence of HEV infection in occupational population and special population
Anti-HEV IgG seroprevalence data from veterinarians, swine workers, slaughters and pork sellers were collected to estimate the overall anti-HEV seroprevalence in occupational population. Based on 43 studies with 8776 occupational individuals, the overall seropositivity of anti-HEV IgG is 24.04% (95% CI 18.55-29.99; I 2 = 97%; Figure   S13). In total, data from 126 studies were extracted to analyse the  Table S5).

| Risk factors of HEV
We investigated the potential risk factors for HEV in the general population ( Figure S28). Significant rising effects on anti-HEV IgG seropositivity were observed in consumption of raw meat (OR 1. 45 No statistically significant differences were identified for anti-HEV IgG positivity in respect to different water source (P = .0909), IDU experience (P = .4321), MSM experience (P = .5576) and contacting with cats (P = .4791; Figure S29-S39). Sensitivity analysis detected no study having an obvious effect influence to the pooled estimates of HEV prevalence in the general population (Table S3).

| Anti-HEV IgG detection rate of different ELISA kits
We finally analysed the detection rates of the ELISA kits from six manufacturers. The detection rates of anti-HEV IgG seropositiv-

F I G U R E 3 Hepatitis E virus genotype distribution in our study
Genelabs Diagnostics (6.22%; 95% CI 3.42-9.77; I 2 = 98%) and Abbott Laboratories (5.27%, 95% CI 2.07-9.84; I 2 = 100%) ( Figure   S40; Table S4). from a meta-analysis performed in 2016. 14 A possible explanation for the disparity could be that they collected fewer studies and included small size populations, and thus is prone to cause more bias.

| D ISCUSS I ON
In Americas, we observed a slightly higher seroprevalence rate in North (8.05%) compared to South (7.28%) America, which is consistent with the results from a recent meta-analysis. 13 Of a technical note, it has been well-realized that there are substantial differences in sensitivity and specificity of the anti-HEV IgG ELISA kits from different manufacturers. 29,30 Our results largely agree with the literature that the Wantai assay has the highest sensitivity and has been most widely used. 31  Thus, our estimates may have biases, and we were not able to further sub-analyse regional prevalence, GT-specific burden or clinical outcome, which require future studies in these aspects.
Accumulating knowledge on HEV biology and transmission routes has facilitated the identification of risk factors for the infection. A wide range of domestic or wild animals have been recognized as reservoirs for the zoonotic strains. Consumption of uncooked meat or meat product from swine, wild boar or deer has been widely reported to cause GT 3 HEV infection in European countries. 35,36 As expected, consumption of raw meat is an important risk factor revealed by our meta-analysis. This is in line with previous reports that humans with occupational exposure to pigs are at a high risk of HEV infection. 37,38 In this study, we observed twofold higher anti-HEV seropositivity in occupational population who had frequent contact with pig or pig products compared to the general population.
The host range for HEV is ever expanding and cross-species infections commonly occur. 7 Intriguingly, recent evidence has indicated that companion animals including dogs, cats, rabbits and horses might be accidental hosts for HEV and might constitute a source for HEV transmission to human. 10 Interestingly, when comparing with the general population, veterinarians and dog farm staff who are frequently exposed to dogs have significantly higher anti-HEV antibody positivity. 10 The anti-HEV seroprevalence rates in cats have been reported to be 6.28% in China, 11 8.1% in Korea 45 and 33% in Japan. 46 In this study, we found that people who frequently contact with dogs have higher anti-HEV IgG seropositivity. This was not found in people who contact with cats, but the number of studies is very limited. These results call more attention to address the potential role of pets in HEV zoonotic transmission, although currently it remains unconfirmed whether pets are reservoirs, requiring further investigation.
Previous studies have indicated the differences of HEV seroprevalence between rural and urban areas. [47][48][49] We found that rural compared to urban residents have higher risk of HEV infection. This largely agrees with our findings that high exposure to soil is also a risk factor. In addition, we observed the high risk of HEV infection in individuals with lower education levels, consistent with previous studies. 50,51 Conceivably, this population are more likely living in rural areas with compromised sanitation conditions and more frequent exposure to animals or soil. Although we did not find differences of HEV prevalence with respect to different water source, this does not contradict to the fact that contaminated water is the main source of GT 1 infection, especially during outbreak. In our study, we have excluded studies related to outbreak, and the number of included studies reporting water source is also very limited, which may cause bias.
Of note, there are some limitations of our study. Firstly, the number of available studies, in particular on anti-HEV IgM antibody and viral RNA positivity, is limited. We were also not able to further analyse detailed regional prevalence, GT-specific burden or clinical outcome. Secondly, we have focused on HEV prevalence, but did not estimate the incidence, which is also very relevant for assessing the disease burden. Thirdly, the assays used for HEV detection are heterogeneous in sensitivity and specificity, which may affect the estimates. Fourthly, publication bias existed in our study which was reflected in Funnel and Egger test (Figures S41-S42).
In summary, we found that 1/8 of the global population, cor-

CO N FLI C T O F I NTE R E S T
The authors do not have any disclosures to report.