A comparison of three sampling approaches for detecting PRRSV in suckling piglets

Determining whether porcine reproductive and respiratory syndrome virus (PRRSV) is circulating within a breeding herd is a longstanding surveillance challenge. Most commonly, piglets in farrowing rooms are sampled to infer the PRRSV status of the sow herd, with sample size based on the expectation of hypergeometric distribution and piglet selection based on simple random sampling (SRS), i.e., randomly selecting individuals from a population in a manner that all individuals have equal chance of being selected. Conceptually straightforward, the assumptions upon which it is based (homogeneous population and independence of individuals) rarely hold in modern swine facilities. Alternative approaches for sample selection include two-stage stratified sampling (2SS), i.e., randomly selecting litters (first stratum) and randomly selecting piglets (second stratum) within selected litters, and risk-based sampling (RBS), i.e., selecting litters with a higher risk of having viremic piglets, and randomly selecting pigs within those litters. The objectives of this study were to 1) characterize the pattern of distribution of PRRSV-viremic piglets in farrowing rooms and 2) compare the efficiency of SRS, 2SS, and RBS for the detection of PRRSV-viremic piglets. In 12 sow farms, serum samples were collected from all 4510 piglets in 422 litters housed in 23 farrowing rooms and tested for PRRSV RNA. At the population level, the distribution of PRRSV-viremic pigs was analyzed for population homogeneity and spatial clustering. At the litter level, litter size and sow parity were evaluated as risk factors. A non-homogeneous distribution of PRRSV-viremic piglets was observed in nearly all farrowing rooms (15/16), and spatial clustering detected on 11 occasions (11/16). Simulated sampling based on farrowing room data determined that 2SS required 1-to-25 fewer samples than SRS to detect ≥ 1 viremic piglet in 13 of 16 rooms and the same number of samples in 3 rooms. RBS required 1-to-7 fewer samples than 2SS to detect ≥ 1 viremic piglet in 7 of 16 rooms, the same number of samples in 6 rooms, and 1 more sample in 3 rooms. Notably, SRS was less efficient than either 2SS or RBS in detecting PRRSV-viremic piglets in farrowing rooms, regardless of the confidence level. It may be concluded that the core assumptions upon which most current surveillance methods are based do not hold in modern farrowing room facilities. Simulation-based sample size tables for SRS and 2SS are provided.