Urinary and fecal potassium excretion prediction in dairy cattle: A meta-analytic approach

Abstract

Quantification of potassium (K) excretion in dairy cattle is important to understand the environmental impact of dairy farming. To improve and monitor the environmental impact of dairy cows, there is a need for a simple, inexpensive, and less laborious method to quantify K excretion on dairy farms. The adoption of empirical mathematical models has been shown to be a promising tool to address this issue. Thus, the current study aimed to develop empirical predictive models for K excretion in dairy cattle from urine and feces that can help evaluate efficiency and monitor the environmental impact of milk production. To develop urine K (KUr, g/d) and fecal K (KFa, g/d) excretion prediction models, published literature that involved 45 and 54 treatment means from 10 and 14 studies, respectively, were used. Some studies reported either urinary or fecal K excretion or both, but in total, treatment means used to develop the models were from 17 studies. The linear mixed models were fitted with the fixed effect of K intake, DMI, dietary K content, urine volume, milk yield, and water intake, and the random effect of study weighted according to the number of observations. Leave-one-study out cross-validation was used to evaluate the performance of the proposed models and the best model was based on the lowest root mean square prediction error as a percentage of the observed mean values (RMSPE%) and highest concordance correlation coefficient (CCC). As expected, most daily K excretion was through urine (202.5 ± 92.1 g/d) than through feces (43.5 ± 21.0 g/d), and among the proposed models, the model including dietary K concentration showed poor predictive ability for both KUr and KFa with the lowest CCC values (−0.15 and −0.02, respectively) and systematic bias. The model developed using DMI to predict KFa excretion showed reasonable accuracy, as indicated by RMSPE, CCC, and R2marginal of 46.6%, 0.42, and 48%, respectively. Among the proposed models for KUr and KFa, the model with K intake demonstrated better predictive performance, showing minimal systematic bias and random errors due to data variability of >92%. While these proposed models suggested that reducing K intake can lead to a decrease in K excretion, it is important to ensure that dairy cows receive adequate amounts of this nutrient to maintain optimal health and productivity, especially during periods of heat stress.