Weighted pain-related behaviors in pigs undergoing castration based on multilevel logistic regression algorithm

Abstract

Pigs are largely raised for food and translational biomedical research. In several intentional and unintentional situations pigs experience pain. In 2020, the Unesp-Botucatu pig composite acute pain scale (UPAPS) was validated according to the COSMIN guidelines for quick and simple assessment of pain in pigs. Recently, an algorithm has been used to estimate the importance of different behaviors in sheep and horses, however, the statistical weight of each pain-related behavior is still unknown for pigs. Herein, we aimed to investigate whether UPAPS pain-related behaviors in pigs have different statistical weights and to rank their importance. A database from a previous study was used, containing the behavioral records of 45 pigs in the perioperative period of castration. The pigs were filmed between 16 and 24 h before the castration, 3.5 and 4 h after the castration, 1.5–2 h after intramuscular administration of analgesics, and 24 h after the end of the castration. Two experienced observers individually and randomly assessed all videos, without knowing which time-point they were observing (blind analysis). At the end of each video observation, the observers recorded whether they would indicate analgesia, if they considered the pigs were suffering pain, according to their clinical experience, and scored the UPAPS. The intra- and inter-observer reliability were calculated with intraclass correlation coefficient and weighted kappa. A multilevel binomial logistic regression algorithm was constructed to estimate the statistical weights of each UPAPS behavior. The intra- and inter-observer reliability values ranged from 0.71 to 0.88. According to the algorithm, 10 of 19 slope coefficients of the behaviors were significant, evidencing a difference in weight for each pain-related behavior in pigs undergoing castration. The change in posture ‘protecting the affected area’ was the heaviest pain-related behavior based on the algorithm. We conclude that the pain-related behaviors in pigs from the UPAPS have different statistical weights. The algorithm has the potential to indicate the degree of importance of different pain behaviors in pigs undergoing castration and, therefore, to select the most relevant pain behaviors to be targeted when assessing pain in pigs.