The impact of African swine fever virus on smallholder village pig production: An outbreak investigation in Lao PDR

Abstract African swine fever virus (ASFV) causes a deadly disease of pigs which spread through southeast Asia in 2019. We investigated one of the first outbreaks of ASFV in Lao People's Democratic Republic amongst smallholder villages of Thapangtong District, Savannakhet Province. In this study, two ASFV affected villages were compared to two unaffected villages. Evidence of ASFV‐like clinical signs appeared in pig herds as early as May 2019, with median epidemic days on 1 and 18 June in the two villages, respectively. Using participatory epidemiology mapping techniques, we found statistically significant spatial clustering in both outbreaks (p < 0.001). Villagers reported known risk factors for ASFV transmission – such as free‐ranging management systems and wild boar access – in all four villages. The villagers reported increased pig trader activity from Vietnam before the outbreaks; however, the survey did not determine a single outbreak source. The outbreak caused substantial household financial losses with an average of nine pigs lost to the disease, and Monte Carlo analysis estimated this to be USD 215 per household. ASFV poses a significant threat to food and financial security in smallholder communities such as Thapangtong, where 40.6% of the district's population are affected by poverty. This study shows ASFV management in the region will require increased local government resources, knowledge of informal trader activity and wild boar monitoring alongside education and support to address intra‐village risk factors such as free‐ranging, incorrect waste disposal and swill feeding.


INTRODUCTION
African Swine Fever (ASF) is a disease of domestic pigs and wild suids caused by the African Swine Fever Virus (ASFV). ASFV is a DNA virus present in all secretions, blood and tissues of affected animals (Sánchez-Vizcaíno et al., 2019). It can survive for an extended period in the environment and in refrigerated or frozen meat products. ASFV can spread via direct and indirect contact, with domestic pig/pig, pig/tick and wild boar/environment cycles described in non-African endemic areas (Chenais et al., 2018;Pérez-Sánchez et al., 1994;Sánchez-Vizcaíno et al., 2019).
ASFV is capable of distant spread across landscapes when facilitated by human transportation or management practices (Costard et al., 2009Nurmoja et al., 2018. In naïve pigs and wild boar, clinical signs of ASFV generally follow the peracute or acute disease syndromes (Sánchez-Vizcaíno et al., 2019). The first sign of an ASFV outbreak in a pig herd may be a small number of animals displaying clinical signs of the peracute syndrome, including depression, pyrexia and cutaneous hyperaemia, followed by death 1-4 days later (Sánchez-Vizcaíno et al., 2019). In the acute syndrome, mortality rates can reach 100% within 7 days of clinical signs' appearance (Sánchez-Vizcaíno et al., 2019).
Reports suggest that in 2018, contaminated swill feed carried ASFV to a Chinese pig farm, from where it spread throughout the country (Zhou et al., 2018). The disease affected all production systems, from smallholders to commercial piggeries (FAO, 2020). The disease then spread through South-East Asia, including Vietnam, in early 2019 and was first reported in Lao PDR at the start of June 2019 (FAO, 2020). This outbreak occurred in Toomlan District, Salavane province in southern Lao PDR (FAO, 2020). A month later, in July 2019, neighbouring villages in Thapangtong district, Savannakhet province (Fig-ure 1) first confirmed cases of ASFV (FAO, 2020).
Informal trading, low biosecurity and swill feeding -all common in Lao smallholder pig farming -increase the risk of ASFV spread (Nantima et al., 2015). Smallholder pig-farming practices in Thapangtong are typical of lowland Lao PDR. In a previous survey of Savannakhet smallholder pig keeping practices, performed before the outbreak, median herd size was two pigs per household (Holt et al., 2019). Approximately one-third of pigs ranged freely, and the rest were penned or tethered. Almost all pigs in the villages were either a local breed or crossbreed (94.8%) (Holt et al., 2019).
The Lao government animal disease reporting system begins at the village level: farmers report unusual outbreaks to their Village Veterinary Worker (VVW), a layperson trained in basic animal health management who reports to their District Agriculture and Forestry Office (DAFO). The DAFO communicates with their local Provincial Agriculture and Forestry Office (PAFO), which then informs the Department of Livestock and Fisheries (DLF) and the National Animal Health Laboratory (NAHL) in Vientiane. The DLF handled epidemiology and control measures, while the NAHL performed the laboratory-based diagnosis of ASFV (Samathmanivong, personal communication, 2019). The NAHL used the TaqMan quantitative real-time polymerase chain reaction (rt-PCR) for confirmation of cases (King et al., 2003;Matsumoto et al., 2020).
In the 6 months from July to December 2019, ASFV spread to 17 provinces of Lao PDR, with new case numbers dramatically declining by the end of the year as the available naïve population fell (FAO, 2020).
The case fatality rate averaged 85-100%, often with sudden death and/or elevated mortality as the presenting clinical sign (FAO, 2020).
Lao PDR's 2019 ASFV outbreak stretched the investigation capacity of the local veterinary services as they allocated their limited financial and human resources to national efforts in stamping out affected herds, movement controls and education programs. Globally, information on ASFV ecology and epidemiology among smallholders is sparse, particularly amongst naïve pig populations. The objective of this study was to fill this knowledge gap. As part of our activities, we allocated additional resources and time to investigate the July 2019 ASFV outbreak in Thapangtong district. In this paper we describe the ASFV outbreak, estimate related household financial loss and conduct a preliminary descriptive investigation into risk factors associated with ASFV in the Lao smallholder pig sector using data from Thapangtong district.

Investigating the timeline of Lao government response
The timeline of the local government response and the process for reporting (from village to the province level) was provided by the act-

Household survey
The survey had two phases: a pilot followed by a final question- Subjects were chosen from all the pig-raising households in the selected villages. In Densateung and Phouphanang-Khampia, almost all households were ASFV-affected (

Participatory mapping
After the individual surveys, villagers worked with the investigators to map their village, marking their households' locations, significant landmarks and known areas of wild boar activity. This map was hand-drawn on a large sheet of paper, and each household represented in the survey contributed to the development of the maps.

Data management and analysis
Boars were not included, as only one was reported in the sampled villages.
performed using the lme4 package in RStudio with the glmer() function and the binomial logit method (Bates et al., 2014).

Financial modelling
Household financial losses due to ASFV were estimated by combining

Spatial outbreak modelling
We mapped the outbreak to investigate the spatial component of disease spread in the village. The map data were analyzed with a spacetime permutation (STP) scan statistic (SaTScan; Kulldorff, 2010;Kulldorff et al., 2005). Space-time scan statistics place numerous theoretical circles of different sizes onto a map and calculate the ratio of how many disease cases are observed versus expected within each circle. The circles also extend upwards as cylinders to represent different lengths of time. The height and base are permuted across the map in all possible combinations, and all clusters are recorded (Kulldorff et al., 2005). Unlike many traditional spatial analyses, this study utilized resources from participatory epidemiology approaches. The spatial cluster analyses therefore used the hand-drawn village maps, and the radii of the clusters used the grid (Cartesian) dimensions of the maps created.
For the SaTScan space-time analysis, the maximum cluster size was set to 50% of the study area. The maximum period of the scanning window was set to 10 days based on the average latent period reported in the literature (Guinat et al., 2014). Monte Carlo simulation was used to determine statistical significance by running 999 replications.

Household survey
In the ASFV-affected villages, households owned on average six piglets and two sows. None of these households owned a boar. In the control villages, households owned on average two piglets and one sow. Two of these households owned fattening pigs, and one owned a boar. All pigs in surveyed households were native breeds (Table 1).
Sampled farmers listed pig housing methods with a range of biosecurity levels, from all-day free ranging (n = 38) to full-time enclosures (n = 19), some of the latter being communal rather than private. Of note were the farmers who kept their pigs in enclosures near their rice paddies (n = 6) some distance from the village, which removed their pigs from the village ecosystem (Table 2). Reported contacts between pigs within the villages were numerous (n = 35 villagers confirmed contact), including with neighbours' pigs and feral pigs or wild boar. In Lao PDR, feral pigs and Eurasian wild boar are called muu paa (forest pig), and both closely resemble the domestic village pigs.
Only two farmers (n = 2) reported feeding pork or kitchen swill to their pigs. All surveyed farmers reported feeding a mixture of rice bran and the water used to prepare sticky rice as the pigs' primary diet.
Water sources (other than the rice water) included household water supplies, communal wells and rivers. Of the farmers surveyed, 79 used a communal water source for their pigs and 17 used private water sources.
When asked an open-ended question about how they disposed of their kitchen rubbish, farmers gave various responses, including burying waste. However, the most common method was to burn kitchen waste. Most surveyed households butchered animals inside the house after slaughter, but 14.9% butchered animals outside the house. Many farmers gave the leftover bones to their dogs (50.7%).
Another possible transmission source was using the same syringes and needles to treat multiple sick animals during the outbreak as reported by the VVWs. Several farmers attempted antibiotic therapy, and during a semi-structured interview a VVW explained that they sometimes washed the syringe with soap and water between uses rather than disposing of the syringe.
TA B L E 2 Pig housing methods in Thapangtong district, Lao PDR

Outbreak investigation
We surveyed 49 ASFV-affected households and 50 control households.
Of these 99 households, eight households surveyed in the 'affected villages' did not meet the definition of an 'affected household' (outlined in the Section 2). These households were not included in calculations relating to outbreak characteristics, outbreak losses, spatial modelling or epidemic statistics. However, these eight households were included for the purposes of describing management styles and practices. Across the ASFV-affected households surveyed (n = 41), 330 pigs died with clinical signs of ASFV during the high-risk period. No pigs died in the control villages during the same period.

Outbreak characteristics
During the household surveys, an obvious route of disease entry did not become apparent. Direct contact through the purchase of an infected pig seems unlikely as none of the affected farmers in this survey purchased new pigs in the high-risk period or the 4 weeks prior.
However, all (both affected and unaffected) reported Vietnamese pork traders during the risk period. Farmers were asked about their initial diagnosis, and 21% identified the cause of the deaths as a seasonal disease. However, many were unsure of the cause of the sudden increase in pig deaths (51%). The VVWs were also uncertain about what disease was causing the outbreak. An average of nine pigs died or were culled in affected households surveyed (n = 41). The majority of pigs died and were either buried or burned rather than culled, and only deaths recorded on or after 3 July were culled and buried by authorities.
Of the affected animals (n = 330), the most common early clinical signs were depression (21.5%), fever (15%), inappetence were kept in enclosures adjacent to rice paddies and were therefore not included in the stamping-out measures.

Risk factor analysis
The quasi-complete separation of ASFV outcome by the villages made the data unsuitable for logistic regression (Bates et al., 2014). When including 'village' as a random effect in the logistic regression model, no significant association between the odds of being an ASFV-affected household and housing style, water source, butchering method or pig contact structure was found. The intraclass correlation attributable to the village effect was > 95% for all analyses. Smaller herds of three pigs or less approached statistical significance when taking village into account (p = 0.06). This is likely because smaller herds were significantly associated with the two control villages, Napaxard (p < 0.05) and Xaysomboun (p < 0.001), while the two ASFV-affected villages had more households with larger herds.

Financial loss modelling
Modelling of the financial impact of ASFV in affected villages is presented in Figure 4 where the purple line represents the density of households' losses using the field data. The Monte Carlo simu-lation then drew from a gamma distribution (shape 1.85 and scale 1013712.97) created using the field data in gamma.buster in EpiR. After 10,000 simulations, the mean financial loss estimated in the Monte Carlo analysis was USD215.00, 95% CI (31.19, 569.30) with SEM ± USD26.85 (Figure 4).

Spatial outbreak modelling
Three significant clusters of more than one household and three clus-

DISCUSSION
This study describes the epidemiologic characteristics, including financial losses, associated with ASFV outbreaks in selected villages in The major potential pathways for introducing ASFV discussed here include traders of live pigs/pig products, iatrogenic spread and wild boar. This study did not identify any single, obvious route of ASFV entry into the villages. However, many plausible hypotheses present themselves, and all should be addressed in future disease prevention activities. The study made obvious that conditions within the villages were ideal for the spread of ASFV. A combination of inter-and intra-village control measures will be required in future to prevent the spread and establishment of ASFV in smallholder communities.
A putative source for the ASFV outbreak in southern Lao PDR is the ASFV outbreak in Vietnam that began in early 2019. Both Thapangtong district and the first-affected Toomlan district are on the same major road to Vietnam. Despite a lack of evidence that any ASFV infected live pigs were purchased from traders in the high-risk period or the month prior, the reports of Vietnamese traders suggest increased activity from a region known to have had ASFV in that same period. Whilst the traders did not sell the villagers any pigs, the traders would have been able to contaminate the villages with ASFV contaminated pork meat products, pig wastes from trucks or even by dropping off con- Wild boar and feral pigs are a possible source of the ASFV outbreak described, as in European outbreaks, however current literature suggests this to be unlikely in Lao PDR (Denstedt et al., 2020). Spread of this nature would require prior evidence of ASFV circulating in wild boar populations over enough time for the disease to spread over large distances (Boklund et al., 2020;Schulz et al., 2019). The distance from the Vietnamese border to south-central Laos is large, and wild boar facilitated spread seems unlikely given the above conditions. In this study, two farmers noted that village pigs had contact with wild 'forest pigs' and that the studied villages (ASFV-affected and control) were near forests with forest pig populations. For the disease to spread from Vietnam to Thapangtong, the disease would have to have circulated in wild boar populations over a distance of approximately 168 km without affecting any other villages before Salavane and Thapangtong District. In late 2019, wild boar ASFV outbreaks were noted in the far northern province of Houaphan, meaning wild boar remain a potential future outbreak source in the wild boar-environmental contamination pathway (Denstedt et al., 2020). However, the authors of the wild boar investigation posited that the outbreak was due to a spill over from the domestic population, and not the other way around (Denstedt et al., 2020). A recent scoping review of ASFV transmission suggests that transmission from wild boar to domestic pigs is generally unlikely. The speed of disease spread in 2019 is more suggestive of human involvement in the spread of ASFV (Barrett et al., 2020).
The nature of the outbreak made it so the data were unsuitable for risk factor analysis at the household level as initially planned.
The authors initially designed the study in the assumption that not all households in the villages were going to report being an 'affected household' , however it became apparent very quickly that the biggest risk factor for being an 'affected household' was being in an 'affected village' . Because of the quasi-complete separation of the disease outcome by village, the data was inappropriately structured for logistic regression analysis at the household level. Risk factors for ASFV transmission include free-ranging, swill feeding and poor farm-level biosecurity, many of which were present in both the case and the control villages. While these factors probably impact on ASFV outbreaks in Lao smallholders, it is likely that a whole village risk factor also exists. To estimate risk factors, we believe a village-level analysis must be performed, although we recognize the difficulty of finding enough affected and unaffected villages to perform such a study. In future, a spatial mapping approach using PCR-confirmed villages may provide opportunities to perform such an analysis in the absence of survey data.
Once ASFV entered a village, factors such as wide-spread use of free-ranging and generally higher pig populations allowed for the spread of the virus. Within the affected villages, a combination of direct and indirect transmission pathways facilitated the spread of disease. Sick animals could make contact both within and between herds because two-thirds of pigs were either fully or partially free-range. Sick pigs can spread ASFV via direct contacts, such as a sow to her piglets.
Other pigs may cannibalize a sick or dead pig, and healthy pigs can eat kitchen wastes containing contaminated pork scraps. Their rooting and investigating instincts can lead pigs to uncover shallow-buried contaminated waste or carcasses. As demonstrated in the epidemic curves ( Figure 2 and Figure 3), the disease propagated through the free-ranging and nonfree ranging pig populations once established. Of interest is the considerable difference in IQR for the epidemic days for Densateung (35 days) and Phouphanang-Khampia (5.5 days). It appears that the smallholder village pigs in Densateung operate under a contact structure similar to those in a commercial style farm where the disease spreads slowly before causing serious fatalities. The spread of the disease amongst the pigs of Phouphanang-Khampia more closely resembles that of a single pen of affected animals (Guinat et al., 2014).
Animals with ASFV become infectious when clinical signs develop. The modal period, from clinical signs to death, in this study was one day or less. This short symptomatic period is consistent with reports of ASFV in other Asian and European outbreaks (Guinat et al., 2014;Olesen et al., 2017;Sánchez-Vizcaíno et al., 2019;Tran et al., 2020). Future studies should estimate the R 0 of ASFV transmission at the pig and household level in these villages and compare these estimates with those of commercial piggeries. The results suggest that preventing ASFV entry at the village level is likely the best strategy for protecting whole communities.
The aim of assessing the participatory data for a spatio-temporal relationship between outbreak locations was to quantify how the disease spread through the villages beyond the calculation of an epidemic curve. The statistically significant clustering of disease outbreaks implies that the outbreak sources were not randomly distributed or a universal exposure. In particular, the sequence of localised clusters in Densateung followed a pattern moving across the village in sequentially bigger groups, reflecting the epidemic curve's propagative nature. The STP approach employed in this study, requires only case data, whereas Poisson and Bernoulli spatio-temporal analyses require both case and population at-risk or control data (Gatrell & Durr, 2004). For an outbreak of ASFV, the STP approach is appropriate because all pigs are affected in a village during a short period, and the population can be considered a closed cohort (Ward & Carpenter, 2000). In these low-biosecurity, free-ranging contexts, there are often no control households or animals. In future studies, these outputs could be adjusted by using disease parameters unique to this outbreak, Vientiane FMD outbreak where numerous FMD-affected villages that were presumed to be 'FMD free' by DAFO due to no reports from VVWs, yet retrospective serology determined otherwise (Miller et al., 2018). Here we note the discrepancy in the number of ASFV cases reported by the PAFO to the OIE (n = 80) and the number of animals with ASFV-like clinical signs in the 'high risk period' (n = 330).
The reported clinical signs, whilst typical of acute and peracute ASFV are also typical of classical swine fever, erysipelas, salmonellosis, and highly pathogenic porcine respiratory and reproductive syndrome, all of which are endemic to Lao PDR. Only five animals were sampled and definitively diagnosed as ASFV cases using PCR, highlighting difficulties in a centralized testing system. In the pilot survey, several farmers reported that pig disease was common during June and July (Lao PDR' wet season). They initially thought the deaths were due to this endemic disease syndrome. The familiar clinical signs may have also delayed reporting and action. None of the villagers included ASFV in their initial diagnoses for the pig deaths despite information materials being made available to local authorities by the OIE in early 2019.
The sampling strategy described here allowed us to speak with many household representatives in each village; however, it has its limita- The findings of this study should be utilized in future decisions about the management of ASFV in the region. Trader, VVW and villager behaviour must be managed and control measures put into place for contact between village pigs and wild boar. The resources available to local government authorities must be assessed for them to act promptly in cases of emergency disease outbreaks. When designing control and education strategies, local farming practices, as well as the disease ecology must be considered together in order to develop effective materials to aid in the prevention and management of ASFV outbreaks into the future.

ACKNOWLEDGMENTS
The authors are very grateful for the work of field staff at the Lao [Corrections added on 2 August 2021, after first online publication: The second and third sentences of the Acknowledgments section were corrected.]

CONFLICT OF INTEREST
The authors declare that there were no conflicts of interest in the development of this work.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.