Meta-analysis of the relationship between milk trans-10 C18:1, milk fatty acids <16 C, and milk fat production

The economic value of milk fat and its responsiveness to management strategies provides strong interest in maximizing milk fat production by minimizing occurrence of biohydrogenation-induced milk fat depression (BH-MFD) and maximizing de novo synthesized fatty acids (FA). Tools that allow a timely diagnosis of BH-MFD would improve nutritional management. Specific milk FA or FA categories correlate to milk fat concentration and are of interest for diagnosing the cause of changes in milk fat concentration. The objective of the current study was to characterize the relationship between milk fat concentration and trans-10 C18:1, a proxy for BH-MFD, and FA <16 carbons that originate solely from de novo lipogenesis using a meta-analysis approach that used data from the literature and unpublished Penn State experiments. Prior to the meta-analysis, the effect of FA methylation method on milk FA profile was tested to determine potential bias between papers. There was no difference between sodium methoxide, acid, and acid-base methylation methods on trans-10 C18:1 concentration, but acid methods resulted in loss of short-chain FA. The relationship between trans-10 C18:1 and milk fat percentage was investigated using a 2-component model, where one component described the fraction unresponsive to BH-MFD and the other described a responsive fraction that is exponentially related to trans-10 C18:1. The 2 fractions where characterized utilizing a Bayesian hierarchical model accounting for between-study variability. The model was defined by the function f(x, θ₁, θ₂, θ₃) = θ₁ + θ₂exp(θ₃), where the unresponsive θ₁ fraction was 2.15 ± 0.09%, the responsive θ₂ fraction was 1.55 ± 0.08%, and the exponential term θ₃ was −0.503 ± 0.07 (posterior mean ± posterior standard deviation from the Bayesian hierarchical model). A Lin's concordance correlation coefficient of 0.67 suggested good agreement between observations and predictions from the Bayesian hierarchical model, computed only with the model's mean population parameters. There was a linear relationship between milk fat concentration and FA <16 C as a percentage of total FA (intercept = 2.68 ± 0.237 and slope = 0.043 ± 0.011; coefficient of determination = 0.31). The relationship between milk FA <16 C and milk fat concentration is weaker than what has been published, likely because multiple factors can reduce de novo FA without reducing milk fat and the broad range of diets present in the literature.