Swine influenza A viruses (IAVs) contribute to risk for pandemic emergence in humans. Emerging livestock systems in low- and middle-income countries (LMICs) have been proposed as hotspots for novel viruses because of the proximity between avian, swine, and human host populations, high densities of smallholder and multispecies farming systems with poor biosecurity, and rapid growth in livestock industries (1-3). However, systematic surveillance of swine IAVs in those settings is nearly nonexistent, limiting our understanding of IAV epidemiology and evolution. We conducted slaughterhouse sampling of pigs over a 2-year period in Cambodia to compare IAV circulation in smallholder versus commercial farms and identify risk factors associated with active IAV infection at slaughterhouses. By performing IAV surveillance in slaughterhouses, we assessed the role of transmission during transport and at slaughterhouses and examined implications for epidemiologic inference of IAV risk along pig value chains, the series of interconnected activities encompassing the production, distribution, and processing of pigs.

We selected 18 slaughterhouses in 4 provinces in Cambodia to encompass pigs from smallholder (<100 pigs) and commercial farms (>100 pigs), after conducting a rapid assessment survey among 52 slaughterhouses to characterize their operations (Appendix). We sampled pigs monthly at each slaughterhouse during March 2020-July 2022 (4). We based sample sizes for each batch (i.e., pigs from the same source tested on the same day at a given slaughterhouse) on 95% probability of detecting >1 positive animal if prevalence within an infected batch was >20% (5). We extracted RNA from nasal swab samples and screened for active IAV shedding using real-time RT-PCR targeting the IAV M gene (6). We screened blood serum samples for IAV nucleoprotein antibodies using ID Screen Influenza A Multi-species ELISA (Innovative Diagnostics, https://www.innovative-diagnostics.com). We collected data on pig breed, age, type, and origin during each sampling visit.

Our study was approved by ethics committees at the London School of Hygiene and Tropical Medicine Institutional Review Board (approval no. 16635) and the Animal Welfare and Ethical Research Board (reference no. 2019-12), National Ethics Committee for Health Research in Cambodia (reference no. 105), Human Research Protection Office (reference no. A-21055), and Animal Care and Use Review Office of the US Army Medical Research and Development Command Office of Research Protections.

To account for chronological and other directional relationships between variables, we developed a directed acyclic graph assuming IAV antibodies are detectable >7 days after exposure (7). ELISA-determined serostatus likely represented IAV exposure on farms because pigs stayed at slaughterhouses only <6 days in this study; virus shedding by pigs starts as early as 1 day after IAV infection and can last >5 days (7). Thus, positive PCR results (i.e., positive infection status) might indicate IAV exposure on the farm shortly before departure to a slaughterhouse, during transport, or at the slaughterhouse.

We developed Bayesian hierarchical logistic regression models to estimate the direct effect of each exposure, adjusted for confounding and batch-clustering effects. We used batch size and duration of stay at a slaughterhouse as continuous variables using fractional polynomial and generalized additive models and categorical variables. We selected functional forms with the largest Bayes factors. We estimated posterior adjusted odds ratios (aOR) using Stan version 2.26.1 (8). We explored spatial trends in seroprevalence based on location of batch origin. We conducted a sensitivity analysis to quantify the potential effects of imperfect diagnostic tests (Appendix).

We sampled 616 batches from 18 slaughterhouses, which provided 4,089 swab and 4,069 serum samples; 340 (55.2%) batches were from commercial and 204 (33.1%) were from smallholder farms in Cambodia, 59 (9.6%) batches were imported from Thailand, and 13 batches were of unknown origin. Estimated transport durations within Cambodia were 0.1-10.1 hours. At slaughterhouses, pigs were penned in groups of 3-30 and kept an average of 3-36 hours before slaughter, depending on the slaughterhouse. Most slaughterhouses reported that pigs were kept 1-6 days. Pens were cleaned daily in 15 slaughterhouses, weekly in 2, and monthly in 1. At least 1 pig tested positive for active infection in 37 (6.0%) batches and for seroconversion in 355 (59.1%) batches (Table).

Seroprevalence among commercial farm pigs was 40.0%, considerably higher than among pigs from smallholders (8.9%). In multivariable analyses, pigs from smallholders were less likely to test seropositive (aOR 0.07; 95% credible interval [CrI] 0.04-0.11) than pigs from commercial farms. Infection prevalence was also lower among smallholder (0.6%) than commercial farm pigs (1.5%), although that association was not statistically significant after adjusting for confounders (Figure 1). Odds of active infection were lower among seropositive pigs (aOR 0.39; 95% CrI 0.18-0.83) and among sows. Several associations provided evidence of transmission at slaughterhouses; specifically, active infection was substantially lower among pigs sampled at slaughterhouses that cleaned pens daily compared with slaughterhouses that cleaned weekly, and increased with duration at the slaughterhouse. We also noted a positive trend between a longer stay at slaughterhouses and seroprevalence (Figure 2 panel B), possibly reflecting risk for exposure shortly before or during transport to the slaughterhouse. The presence of poultry at slaughterhouses did not affect active infection status. Associations were not substantially affected by potential underdetection of infection in a sensitivity analysis (Appendix Figure 4). For commercial but not smallholder farms, seroprevalence averaged across batches varied among districts (Figure 3).