Horse acute pain assessment scales and key parameters: A systematic review and meta-analysis

Pain assessment in horses is challenging, as in all non-verbal species, because it relies on the interpretation of indirect behavioural and physiological indicators. Various scales have been developed to recognize and assess pain, but the majority lack validation. Therefore, the present study aimed to systematically review and quantitatively compare the diagnostic performance of equine pain assessment scales across different pain types and clinical contexts. Specifically, we focused on five key metrics: inter-observer reliability, intra-observer reliability, sensitivity, specificity, and internal consistency. The type of pain assessed by the different scales was categorized as colic, orthopaedic, or other. A systematic search of Web of Science, PubMed, and Scopus was conducted for studies published between 2003 and 2024. Out of the 359 records screened, 25 studies met the inclusion criteria: they focused on horses, applied pain assessment scales to live or recorded cases, reported numerical outcomes related to reliability, sensitivity, or specificity, were written in English, and were available in full text. Linear scales showed poor inter-observer reliability (0.71 [95%CI 0.66–0.75]) highlighting their subjectivity, particularly in colic pain assessment. Behaviour only scales demonstrated moderate (0.86 [95%CI 0.83–0.88]) to poor (0.65 [95%CI 0.56–0.73]) inter-observer reliability for the assessment of pain with colic, orthopaedic issues and other pain conditions. Their sensitivity was moderate (0.89 [95%CI 0.86–0.92]) and specificity poor (0.75 [95%CI 0.71–0.79]) compared to the sensitivity and specificity estimates from the meta-analysis for colic pain (0.77 [95%CI 0.65–0.88] and 0.87 [95%CI 0.81–0.92], respectively). Including physiological parameters resulted in excellent inter-observer reliability (0.98 [95%CI 0.97–0.99]) for colic and orthopaedic pain and excellent sensitivity (0.96 [95%CI 0.94–0.97]) for colic pain. The evaluation of facial behaviours had excellent inter-observer reliability for colic (0.93 [95%CI 0.91–0.95]), post-castration (0.92 [95%CI 0.89–0.95]) and head-related pain (0.92 [95%CI 0.87–0.96]). Linear scales are limited by subjectivity and poor reproducibility. Behaviour-only scales show low reliability and diagnostic accuracy, particularly for orthopaedic pain. Incorporating physiological parameters significantly improves both sensitivity and reproducibility, even if these tools may require expert interpretation. Facial expression assessments offer a promising alternative, as several specific facial muscle movements (facial action units, or FAUs) demonstrate high sensitivity and high inter-observer agreement, making them potentially suitable for trained non-experts and applicable in a wider range of clinical settings.