A systematic review and meta-analysis of cow-level factors affecting milk urea nitrogen and urinary nitrogen output under pasture-based diets

With dairy cattle farming under pressure to lower its environmental footprint it is important to find effective on-farm proxies for evaluation and monitoring of management practices aimed at reducing the risk of nitrogen (N) losses and optimizing N use efficiency of dairy farm systems. Urinary N (UN) is regarded as the most potent source of N emissions. In contrast to confinement systems, there have been few studies from pasture-based systems associating on-farm animal and nutritional factors with UN output. Thus, the aims of this meta-analysis were to collate a database from pasture-based research to: (a) investigate the associations of management, dietary, and animal variables with MUN concentration, and daily UN output; (b) describe the MUN-UN association; and (c) assess whether animal, management, and dietary factors influence the relationship. We developed a data set consisting of 95 observations representing 919 lactating dairy cattle fed pasture-based diets, which was compiled from 32 unique research publications that reported both MUN and UN output. Multi-level, mixed meta-analysis regression techniques were used to analyze the data. Initially, all variables were assessed as the sole fixed effect in a 2-level random effects model, accounting for within publication heterogeneity. Meta-regression techniques were then used to assess the relationship of all variables with MUN and UN output, respectively, accounting for 3 sources of variability: the sampling error of the individual observation, within publication heterogeneity, and among publication heterogeneity. At the univariable level, despite more than 10 dietary, animal, or management variables being significantly associated with MUN, none explained a large amount of the MUN variation. The variables that explained the greatest amount of variation were dietary crude protein (CP) content and the nitrogen: metabolizable energy content ratio, which explained about 33% and 31% of the variation in MUN concentrations, respectively. Combining factors in multiple regressions improved the model fit, such that the variation within publications explained by dietary CP and N intake increased to 40.0% in the final multiple meta-regression model. For UN output, individual variables explained a greater proportion of variance reported among observations, compared with MUN, whereby diet CP content (pseudo R² = 66.1%), N to metabolizable energy intake ratio (pseudo R² = 64.0%), N intake (pseudo R² = 58.3%), and MUN (pseudo R² = 43.5%) explained the greatest amount of the total variation. Milk urea nitrogen, N intake and dry matter intake were associated with UN output in the final multiple meta-regression model. Substantial heterogeneity existed in both MUN and UN among publications, with among publication heterogeneity accounting for 73.4% of all the variation noted in MUN, and 88.6% of all the variation in UN output. As such, the meta-analyses could not predict MUN and UN to any great extent. It is recommended that a consistent approach to measuring and reporting MUN concentrations and UN output is carried out for all future research in pasture-based systems.