Hepatitis E virus in the Kathmandu Valley: Insights from a representative longitudinal serosurvey

Background Hepatitis-E virus (HEV), an etiologic agent of acute inflammatory liver disease, is a significant cause of morbidity and mortality in South Asia. HEV is considered endemic in Nepal; but data on population-level infection transmission is sparse. Methods We conducted a longitudinal serosurvey in central Nepal to assess HEV exposure. At each visit, capillary blood samples were collected and analyzed for the presence of anti-HEV IgG antibodies. The study took place between February 2019 and April 2021, with up to 4 visits per participant approximately 6 months apart. Results We collected 2513 samples from 923 participants aged 0–25 years, finding a seroprevalence of 4.8% and a seroincidence rate of 10.9 per 1000 person-years. Young adults and individuals consuming surface water faced the highest incidence of infection. Geospatial analysis identified potential HEV clusters, suggesting a need for targeted interventions. Significance Our findings demonstrate that HEV is endemic in Nepal and that the risk of infection increases with age.


Introduction
Hepatitis E Virus (HEV), first identified in 1983, has emerged as the leading cause of acute clinical hepatitis in South Asia [1].The overall mortality rate associated with HEV is 1-4% [2].Immunocompromised individuals and pregnant women face the highest risk for complications and death, with mortality rates reaching up to 20% among pregnant women in their third trimester [3].The virus is primarily transmitted through food and water contaminated with infected fecal material, disproportionately affecting individuals in locations lacking improved sanitation systems [2,4].
HEV is estimated to cause approximately 20 million new infections annually, leading to around 3.3 million symptomatic cases and 70,000 deaths [5].However, these figures may be underestimated due to limited surveillance capacity and suboptimal access to laboratory diagnostics, which often leave many infections, especially pauci-symptomatic and subclinical cases, undetected and under-reported [6].
Sero-epidemiology provides a valuable tool to augment clinical surveillance of HEV, particularly when diagnostics and reporting systems are limited [7].Sero-epidemiology also offers insights into exposure that are not biased by access to health care and care-seeking behaviors [8].Antibody responses to the four recognized HEV genotypes are similar [2].IgG antibody responses peak 2-6 weeks post-infection [9,10].A study of confirmed HEV patients in Nepal found that IgG responses decayed substantially in the first 6 months but remained elevated above a seropositivity threshold for at least 14 months [11].
Nepal has experienced repeated outbreaks of HEV with sporadic cases of acute hepatitis between outbreaks [12].A study among individuals age 1-93 years seeking orthopedic care in Kathmandu between 2010 and 2012 found an IgG seroprevalence of 47.1% [13].Among healthy blood donors age 18-55 years in Kathmandu in 2014, the age-adjusted seroprevalence of HEV was 3.2% for IgM and 8.3% for IgG [14].A study conducted after the 2015 earthquake among blood donors residing in earthquake-affected areas found a HEV seroprevalence of 3.2% for IgM and 41.9% for IgG [15].Most previous seroprevalence studies were conducted using convenience samples (healthy blood donors or individuals seeking care in a hospital setting), and it is unknown how well these estimates represent the general population.
To meet this gap, we performed a longitudinal HEV serosurvey among a geographically random population-based sample of children and young adults residing in Kathmandu and

Laboratory methods
We eluted the DBS by submerging two filled filter paper protrusions containing 20 ul of dried blood in 133 μL of 1X PBS containing 0.05% Tween buffer and incubating overnight at 4˚C.We then centrifuged the tubes at 10,000xg and aliquoted the supernatant, considering this DBS eluate as a 1:10 dilution of serum.
We analyzed the samples using end-point ELISA with recomWell HEV IgG ELISA kits (Mikrogen Diagnostik), which evaluate immunoglobulin IgG against the recombinant HEV-ORF2 antigen.The Mikrogen HEV ELISA kit was chosen because it had the highest diagnostic performance among commercially available kits, with a published sensitivity of 74% and a specificity of 99% [18].We read the ELISA plates at a wavelength of 450nm (with a reference wavelength of 690nm) using a Bio-Rad iMark ELISA plate reader to produce quantitative optical density (OD) values.

Statistical analysis
We used three complementary approaches to identify the optimal cutoff value (Fig 1).First, we fit 2-component Gaussian finite mixture models to the log-transformed quantitative OD values for all samples at the baseline visit.We plotted the distributions along with cutoffs derived from the mean of the lower mixture component plus 2, 3, 3.5, and 4 standard deviations (SD).The mean + 3.5 standard deviations best separated the two distributions [19] (Fig 1A).Next, we evaluated this cutoff against the distributions of positive, borderline, and negative controls from the ELISA assay (Fig 1B).The mixture-model cutoff perfectly discriminated the negative control values from the borderline and positive controls (Fig 1B).Finally, we plotted the rank order of OD values from the population assay and observed the 3.5 SD mixture model cutoff marked the inflection point in quantitative values (Fig 1C).We conducted a sensitivity analysis for the seroprevalence and seroincidence rates using cutoffs at the mixture model mean plus, 2, 3, and 4 standard deviations.
We calculated seroprevalence as the percentage of samples that were equal to or above the cutoff (seropositive) at each sampling interval.To describe how seropositivity changes over age, we fit generalized additive models with a cubic spline for age [20].We computed simultaneous confidence intervals using a parametric bootstrap of the variance-covariance matrix of the fitted model parameters [21].To investigate variables associated with seropositivity at baseline, we used mixed-effect binomial logit models with a random effect for the community [22].
We defined incident seroconversions or reversions as a change in IgG across the seropositivity cutoff between sampling intervals.We calculated seroincidence, defined as the number of new seroconversions per 1000 person-years, by dividing the number of individuals who seroconverted by the mid-point of person-time at risk between measurements.Similarly, we calculated the seroreversion rate as the number of individuals who seroreverted by the midpoint of person-time at risk between measurements.
We compared the seroconversion rate to the seroincidence rate derived from the cumulative hazard of the age-dependent seroprevalence at the final study visit.We fit a generalized linear model to the seropositivity conditional on age with a complementary log-log link and estimated seroincidence from the model's intercept term [23,24].This approach assumes that antibody responses do not wane after exposure (ie no seroreversion) and that seroincidence is constant over time.
We used negative binomial mixed-effects regression models to explore the influence of various predictors on seroincidence, including area, gender, income level, water source, and frequency of eating outside, while adjusting for age.Each predictor was individually incorporated into the model, which also accounted for repeated measurements on the same individual    across different visits and individual-level variability.An offset was introduced to adjust for person-time.
We conducted a geospatial analysis to visualize the distribution of HEV seropositive individuals and seroconversions across the study area.GPS coordinates were collected in RedCap and plotted using the 'ggmap' in R with stadiamaps for the baselayer.We used kernel density estimation to calculate spatial density by placing a Gaussian kernel over each data point and summing the contributions of all kernels at each location on the map.
Among the 44 individuals seropositive at baseline and 9 the incident seroconversions, there were 15 seroreversion events, yielding a seroreversion rate of 28.3% (15/53).The seroreversion rate over 33.57person-years of observation time was 446.8 (95% CI 250.1-736.9)per 1000 person-years.The seroreversion rate decreased with age and was highest among young children less than 10 years old with 4 seroreversions over 1.87 person-years and a rate 2136 per 1000 person-years.The seroreversion rates by age are presented in the supplemental information (Table C in S1 Text).

Discussion
This representative longitudinal serosurvey reveals ongoing HEV exposure in the Kathmandu Valley of Nepal.With a total of 2513 dried blood samples collected from 923 children and young adults, we observed a seroprevalence of 4.8% and a seroincidence rate of 10.9 seroconversions per 1000 person-years.These findings are consistent with the growing body of evidence suggesting that HEV is endemic in many parts of Nepal [11,12,14,15,25].HEV seroprevalence and seroincidence both increase with age.Among those aged 15-25, 1 in 10 participants had evidence of HEV IgG exposure.This age group is particularly significant as it encompasses women of childbearing age, where the risks of HEV complications and mortality peak [3].These findings align with other studies in Bangladesh [7], Nepal [25] and Laos [26], where seroprevalence increased with age.However, these other studies were limited in their ability to disentangle age and cohort effects.It's possible that seroprevalence among older ages reflected higher periods of exposure in the past.In this study, by measuring incident seroconversions, we were able to characterize a higher risk of exposure with age.These findings align with two cohort from Bangladesh, which also found the seroincidence of HEV increased with age [27,28].The seroreversion rate we estimated here (40.1 per 100 person-years) is higher than what was reported from Bangladesh (15 per 100 person-years) [27].The differences in seroreversion are most likely due to the younger age range of participants in this study.Indeed, the seroreversion rate in young children less than 10 years old in Bangladesh by Dighe et al. (~180 per 100 person-years) [27] was closer to what we observed in this study (210 per 100 personyears).
The clustering of seropositive cases around Banepa in Kavre and the southwestern region of Tribhuvan airport in Kathmandu, suggests possible localized outbreaks or common sources of exposure.The majority of incident seroconversions in Kavre were within five kilometers of Banepa-where a high seroprevalence was observed at baseline.This may suggest propagative transmission from a potential previous outbreak.Such clustering has been observed in other studies [29,30], and suggests a role for geographically-targeted public health interventions.
We identified water source as a potential risk factor for HEV seroincidence but not seroprevalence.Individuals consuming surface water had more than four times the seroincidence rate of HEV compared to those relying on other water sources.The prevalence of consuming surface water among individuals who seroconverted (5/9, 55.6%) was higher than in the general population (179/923, 19.3%) but not in Panauti (105/197, 53.3%)where the majority of seroconversions occurred.However, when restricting just to within Panauti, 3 of 4 individuals who seroconverted consumed surface water, and the other consumed groundwater.Our findings aligns with previous studies that have identified contaminated water as a primary transmission route for HEV [29,30].
While we observed a higher HEV seroprevalence in households with a reported higher income, the seroincidence rate was similar.Possible explanations for the higher seroprevalence among higher-income households include a higher frequency of eating outside the house and living in more urban settings with higher population density.There was also a significant amount of missing data for household income question, likely due to sensitivity around reporting.
Variability in HEV seroprevalence estimates across studies, even within the same region, is well-documented [31].Some of these discrepancies can be attributed to methodologies that overlook the age-dependent nature of seroprevalence and the role of seroreversion.Ignoring the waning of antibodies underestimates seroprevalence and seroincidence when derived from age-dependent seroprevalence estimates [32].In our study, the seroincidence measured from longitudinal seroconversions was three times higher than the seroincidence rate derived from the age-dependent cross-sectional seroprevalence that ignored antibody decay.Dighe et al. also found that the HEV seroincidence rate using longitudinal seroconversions was 5 times that of the age-dependent seroprevalence when ignoring seroreversions [27].Together, these findings underscore the need to factor in seroreversion when reporting HEV seroprevalence and seroincidence rates.Several limitations are important to acknowledge when interpreting these results.First, this study leveraged a cohort designed to characterize enteric fever burden, which did not include individuals above 25 years of age.While this restricted our ability to comment on the HEV burden in older age groups, prior research has indicated that the most significant HEV burden is among young adults [7,25].Future studies are needed to characterize HEV seroincidence in older ages in the Kathmandu Valley region.Second, our use of dried blood samples, though logistically advantageous, might have increased the limit of detection our serological assays.A recent study comparing anti-HEV IgG antibody responses in dried blood spots a sensitivity of 81% and a specificity of 97% comkpared to to fresh serum [33].Given this reduced sensitivity, our findings may underestimate the true HEV seroprevalence and seroincidence by up to 20%.However, another study in Bangladesh found OD values for measuring Anti-HEV IgG response from DBS were equivalent to plasma when stored properly [34].Third, with only 9 seroconversion events, we were underpowered to detect effect sizes less than 2. Fourth, when evaluating the risk associated with water sources, it is likely that individuals consumed water from multiple sources, yet we were only able to capture information about the primary source.Finally, our household sampling strategy could have inadvertently omitted migrant populations and residents of informal settlements.Such populations, often faced with subpar water, sanitation, and hygiene conditions, could potentially have a higher HEV seropositivity rate, potentially biasing our findings towards the null.
In summary, our study demonstrates that HEV is endemic in Nepal and that exposure increases with age.These insights emphasize the need for targeted public health strategies such as vaccination and improved water and sanitation infrastructure.

Fig 1 .
Fig 1. Cutoff identification methods.A) Histogram of quantitative od values among the entire population (N = 923) at the baseline study visit.Each vertical line depicts the cutoff derived from the lower component mixture model fit to the log-transformed OD values plus 2,3,3.5 and 4 standard deviations.B) Histogram of quantitative ELISA OD values for positive controls (red), borderline controls (purple), and negative controls (blue) for each plate.C) The rank order of quantitative od values among the entire population (N = 923) at the baseline study visit, each dot represents and individual result.The black dashed line depicts the cutoff derived from the mixture cutoff of the mean of the lower component plus 3.5 standard deviations.The color depicts whether the sample was seropositive (red) or seronegative (blue) using the 3.5 SD cutoff.https://doi.org/10.1371/journal.pntd.0012375.g001

Fig 2 .
Fig 2. Quantitative anti-HEV IgG antibody responses over time by study site location.The dashed line indicates the mixture model cutoff.Each point represents the quantitative antibody response colored by whether it was positive (red) or negative (blue).Solid gray lines connect individual participant's samples over time.The grey boxes indicate when sampling was paused due to COVID-19 movement restrictions.https://doi.org/10.1371/journal.pntd.0012375.g002

Fig 3 .
Fig 3. Seroprevalence as a function of age at the baseline study visit.Each point represents an individual antibody response that is positive (red) or negative (blue).The age-dependent seroprevalence curve is fit using generalized additive models with a cubic spline for age and simultaneous confidence intervals using a parametric bootstrap of the variance-covariance matrix of the fitted model parameters.https://doi.org/10.1371/journal.pntd.0012375.g003

Fig 4 .
Fig 4. HEV Seroprevalence at baseline and prospective seroconversions across the enrollment areas (Banepa, Panauti, Dhulikhel, Panchkhal: Kavre and Kathmandu) and over the enrollment period of Feb 2019 to Apr 2021.The colored points represent seropositive (red) or seronegative (blue) participants.The black shapes indicate incident seroconversions, with a different shape for each water source.The opacity of the red-shaded area reflects the kernel density estimation.This map was created in R using ggmap.The base layer was created by Stadia Maps (stadiamaps.com)using Stamen design (stamen.com)and OpenStreetMap (openstreetmap.org/copyright).https://doi.org/10.1371/journal.pntd.0012375.g004