Stability and validation of a bulk tank milk-based model to estimate within-herd bovine leukemia virus prevalence in dairy herds.

Bovine leukemia virus (BLV) causes enzootic bovine leukosis and can lead to significant economic losses in dairy herds. Accurate estimation of within-herd prevalence is an essential step for effective continuous surveillance and control. The objectives of this study were: (1) to build a statistical model predicting within-herd BLV prevalence using Bulk Tank Milk (BTM) ELISA antibody titers, (2) to assess the temporal stability of this model over three years, and (3) to explore the relationship between mean individual cow antibody titers and BTM results. A convenience sample of 30 dairy herds in the Canadian Maritimes was studied. Three years of bi-monthly BTM samples were collected to measure the BTM ELISA antibody titers. Individual ELISA testing of cows contributing to BTM was performed annually for three years, permitting estimation of the true within-herd prevalence. Among the 30 herds, 1453 cows tested BLV-positive in Year 1, 899 in Year 2, and 760 in Year 3. Mean within-herd prevalence were 42 % (SD 24 %), 31 % (SD 20 %), and 29 % (SD 20 %) in Year 1, 2, and 3, respectively. Seven univariable linear regression models for true within-herd prevalence were built, with predictors consisting of single BTM results or averages from two or three rounds of BTM results in proximity to the time of individual ELISA testing. The model using the average of two BTM samples collected two months apart as a predictor prevalence in the herd was selected as the best model based on the highest R² = 0.91 and Predictive R² = 0.90. Model predictions showed strong agreement with observed within-herd prevalence in subsequent years (concordance correlation coefficients 0.94-0.98), indicating temporal stability. Mean individual cow antibody titers were strongly correlated with BTM antibody levels (r = 0.95). This study presents a practical, stable, and cost-effective approach to estimate and monitor BLV within-herd prevalence using BTM ELISA testing. The model can support surveillance programs and help guide herd-level BLV management decisions over time.