Supplementary material for "A large, multi-site investigations into the lipidomics of survival in dairy cows"

Identifying physiological determinants of dairy cow survival, and their potential modulation by parity, may reveal opportunities to improve herd health and longevity. This multi-site, prospective, observational study investigated culling and mortality hazards using lipidomic and standard metabolite assays. Blood samples, stratified by parity, were collected from two cow cohorts (1) dry and (2) peak-milk, from across 29 commercial farms (14 pasture-based, 15 confinement-based). There were 717 non-lactating, late-pregnant, dry cows (~27 d prepartum) and 794 peak-milk cows (~58 DIM) sampled. A total of 186 lipid species (including glycerophospholipids, sphingomyelin, and triacylglycerols), and 11 routinely measured metabolites were evaluated. Cows that were sampled were followed for an average of 693 d and exit reasons recorded. Competing risk survival models were used to estimate the cumulative incidence of culling and mortality by parity and cohort. Blood analytes were autoscaled within cohort and farm, controlling for farm-level effect on metabolites. Survival analysis was performed using an adaptive LASSO Cox full likelihood model that explored associations among blood analytes and hazards of removal, with a shared frailty of farm (accounting for farm-level baseline hazards), and two removal outcomes considered: culling (censoring: death, cull from farm accident, end of follow-up), and mortality (censoring: cull, death from farm accident, end of follow-up). Separate models were used to estimate survival outcome by cohort, and parity groupings (1st, 2nd and 3rd, and > 3rd), or with parity as a categorical covariate. Due to high correlations, analyte data were reduced to 25 clusters using Ward’s hierarchical clustering criterion. Bootstrapping of the LASSO variable selection procedure identified clusters with high selection frequency for use in the final model. The hazards and cumulative incidence of culling and mortality increased with parity. Glycerophospholipids with very long chain omega-3 fatty acids were associated with reduced hazards of culling in parity 1 peak-milk cows [0.39 hazards ratio (HR)], while glycerophospholipids with omega-3 α-linolenic acid reduced hazards of culling in parity > 3 peak-milk cows (0.18 HR). Sphingomyelin with > C18 fatty acyl chains increased the hazards of culling in parity > 3 peak-milk cows (2.02 HR). Clusters containing the routinely evaluated metabolites albumin, globulin, urea, magnesium, glucose, triglycerides, β-hydroxybutyric acid, bilirubin, non-esterified fatty acid were associated with culling and mortality; consistent with their roles in health and reproduction. Lipids collected from dry-cows were poor predictors of survival. Many novel plasma lipid targets for future research into survival of cattle were identified, with pro-inflammatory lipid profiles associated with increased risk of culling and mortality.