Abstract #W21

# W21
Use of circulating metabolites and milk production variables to generate linear regression models for prediction of postpartum liver triglycerides.
C. R. Seely*1, H. T. Holdorf1, R. S. Pralle1, R. C. Oliveira1, J. L. Woolf1, M. R. Moede1, S. J. Erb1, H. M. White1, 1University of Wisconsin-Madison, Madison, WI.

Given the potential impacts of liver triglyceride (lvTG) accumulation on hepatic metabolism, the ability to diagnose fatty liver without a liver biopsy could be advantageous in both the research and applied settings as accumulation of lvTG can only be diagnosed by liver biopsy. Since fatty liver is related to the overall metabolic status of the cow, the objective of this study was to determine if the concentration of lvTG could be predicted from milk production variables and circulating blood metabolites related to energy balance and liver health. Blood and liver samples were taken at −14, +1, and +14 d relative to calving (DRTC) from multiparous Holstein cows (n = 37) enrolled in 2 previously reported studies. Daily milk production and weekly milk composition were collected. Liver TG (% dry matter) was quantified and serum was analyzed for aspartate amino transferase (AST), alanine amino transferase (ALT), albumin (alb), BHB, BUN, and triglyceride (TG). Plasma was analyzed for glucose (glc) and nonesterified fatty acids (NEFA). Through the PROC REG procedures of SAS (9.4), forward stepwise linear regression models utilizing a P < 0.1 and minimum AIC inclusion criterion were fit to predict either +1 or +14 DRTC, or maximum lvTG% from the analyzed blood metabolites and milk variables. Two types of models were explored; 1) a predictive model that used prepartum metabolites to predict maximum lvTG% and 2) a diagnostic model that utilized +1 and +14 DRTC metabolites and milk variables to predict respective lvTG%. Maximum lvTG% was 18.9 ± 1.4% and occurred at +14 ± 1 DRTC. Diagnostic models at +1 DRTC included ALT and AST (R2 = 0.23) and +14 DRTC included blood TG, NEFA, glc, and cumulative milk yield to date (R2 = 0.66). The predictive regression model for maximum lvTG% based on −14 DRTC metabolites included ALT, alb, BUN, and TG (R2 = 0.41). Overall, postpartum diagnostic models were stronger than predictive models; however, additional metabolites should be explored to improve the ability to diagnose or predict lvTG%.

Key Words: transition cow, biomarker, fatty liver