Discussion
This study provides the first long-term analysis of JEV seroprevalence in a domestic pig population, and is in a region in India – eastern Uttar Pradesh – in which a high incidence of cases in people occurs annually (A. K. Singh et al., 2020). A key finding was that IgG seroprevalence did not increase with age, indicating that IgG could be transient, and also that JEV infection might not be an immunising infection in pigs.
IgM is an indicator of recent JEV infection in pigs; IgM can be detected 4—7 days post-infection, peaking at approximately 2 weeks and declining over the following 4 weeks (H. Dhanze, Kumar, et al., 2020). If JEV infection was immunising, a reduction in IgM seroprevalence would have been consistently observed with age, even if pigs were not relying on a humoral response to JEV involving IgG. In a review of JEV epidemiology in pigs, Ladreyt et al. (2019) reported that in experimental studies, antibodies to JEV were detected for at least 28 days post-inoculation in two experimental studies, and a study of piglets under field conditions found immunity for up to 3 years (Geevarghese, Shaikh, Jacob, & Bhat, 1994). In this latter study, it was possible that the pigs had been repeatedly infected with JEV and that cross-reaction between antibodies to other flaviviruses (for example, West Nile virus) had occurred. In an experimental case study, IgG was found to decline after 40 days and remained at low levels at one year in a single 100-day old pig experimentally injected with purified recombinant NS1 protein as a JEV associated antigen, but this might not be representative of the immune response following natural JEV infection (H. Dhanze, Kumar, et al., 2020). In a field study of JE in Sri Lanka (Peiris et al., 1993), point estimates of age-stratified prevalence of neutralising antibodies decreased between 7-12 month and 13-24 month pigs at one site (48% [90% CI 29-67] and 19% [90%CI 12-28], respectively), and were similar at others (for example, 21% [90% CI 7-47] and 33% [90%CI 11-65], respectively). It is possible that ages of pigs in the current study were misclassified, but we believe this would be a small number because samples were taken by veterinarians familiar with pig breeds in the region. It is also possible that further stratification within districts would demonstrate increasing IgG seroprevalence with age if JEV transmission has consistent localised patterns within districts. Therefore, we recommend that longitudinal studies of IgG levels in individuals in pig populations at continuous risk of JEV infection should be undertaken to determine whether our hypothesis is correct. There are public health implications where intervention strategies have been based on models in which persistent immunity in pigs was assumed; the force of infection on people from pigs might be higher than estimated if susceptible pigs are not only replenished through births and young pigs losing maternal immunity, but also adult pigs losing acquired immunity.
The district-level spatio-temporal pattern of JEV IgG or IgM seroprevalence did not follow a discernible pattern, which is consistent with endemicity – waves did not reliably occur simultaneously in districts across the region, and neither was a wave of spread apparent from one district to the next. Seroprevalence of IgM was found year-round, indicating a subtropical epidemiological pattern of JEV transmission (Suresh et al., 2022). The seroprevalence throughout the region was consistent with previous, shorter duration studies, with large peaks observed, such as the annual seroprevalence of 60.8% in Sant Kabir Nagar in 2016 and 69.5% in Gorakhpur in 2017 for IgG and IgM, respectively, and some region-wide monthly IgG seroprevalence peaks reaching 100% (H. Dhanze et al., 2014; Kolhe et al., 2015). Decomposition of the monthly time series of seroprevalence for IgG and IgM across all age groups demonstrated that most variance in JEV seroprevalence was due to the unexplained ‘random’ portion of the time-series, followed by the underlying trend. Although annual cyclical waves with 3-4 peaks annually were apparent in the decomposition, the variance of these waves was relatively small. These cyclical waves are also consistent with the expected dynamics of an endemic infectious disease, in which regular epidemic waves occur as the proportion of susceptible pigs in a population increases through births – it does not indicate necessarily that an infection is non-immunising although this can contribute to the effect. Generally, seroprevalence peaks were higher, or arose from a higher baseline, in the monsoon period, indicating a seasonality which has been shown previously (Borah et al., 2013; Kumari & Joshi, 2012; Pantawane et al., 2017; A. K. Singh et al., 2020). In the monsoon season, temperature is warmer and rainfall is higher, providing mosquitoes with more breeding habitat and promoting shorter developmental cycles. The predominant vector of JEV, Culex tritaeniorhynchus , has been shown to be most abundant in the monsoon season in a previous study in Gorakhpur district (Kanojia, Shetty, & Geevarghese, 2003).
Because most variance in the time series decomposition was due to the random (unexplained) portion, autoregressive modelling of JEV IgG seroprevalence in Gorakhpur district was not sufficiently nuanced to be usefully predictive. Interestingly, the underlying trend of IgG seroprevalence in Gorakhpur – higher overall seroprevalence in the first half of the study period, similar to the region-wide trend – was the inverse of the monthly mean relative humidity and total rainfall which were both higher in the second half of the study period. This was reflected in the best fitting model in which mean monthly relative humidity was negatively correlated with seroprevalence. Previous studies have stated that pigs could be used as sentinels for outbreaks of JEV in people (Suresh et al., 2022). Given that seroconversion in pigs can be highly correlated with cases in people 2-4 weeks later (study in Assam; Borah et al. (2013), increased seroprevalence in pig populations could theoretically be used as a predictor of human cases. Whilst the utility of this might be useful for more immediate control options such as fogging in areas detected with high seroprevalence in pigs, the utility of this is low for predictive surveillance due to timeliness. Collection of blood samples, analysis and reporting, as well as implementation of longer-term control vaccination programs in people and the time required to develop immunity would be greater than 2-4 weeks.
A strength of this study included the relatively large dataset over a long time period, enabling comprehensive descriptive analysis of the seroprevalence of pigs in this region. However, limitations included sparse data in some districts, and the lack of longitudinal analysis of pigs. In addition, reported cases in people from the same districts during the study period would enable further evaluation of the value of sero-surveillance in pigs as part of a public health surveillance program.
Cases in people in Uttar Pradesh generally occur from June and peak in September—October (Kumari & Joshi, 2012). This study demonstrates that for sero-surveillance in pigs to be useful to predict the onset and magnitude of human cases at district level so that public health resources for interventions including vaccination programs, mosquito control and bite prevention can be targeted to the highest risk districts, more detailed data about known risk factors is needed at local scales. It is possible that other risk factors such as local availability of mosquito breeding habitat, ground surface temperature, water bird abundance and other landscape factors might be useful in predicting both pig and human cases. Pig sero-surveillance could have a role in confirming predictions and the need for maintenance of interventions in high risk periods for people.