NMR-based metabolomics was applied on urine samples from 32 cows that were fed four levels of crude protein (124, 135, 151, and 166 g/kg DM, respectively) in a crossover design with the aim of identifying urinary metabolites related to nitrogen intake and nitrogen efficiency. Principal component analysis (PCA) on selected regions of the obtained (1)H NMR spectra revealed an effect of crude protein intake on NMR signals in the 0.5-3.0 and 5.0-10.0 ppm regions. Partial least-squares (PLS) regressions confirmed a correlation between the NMR metabolite profile and both nitrogen intake and efficiency. The NMR signals that correlated with nitrogen intake and efficiency included urea, hippurate, phenylacetylglutamine, and p-cresol sulfate, which all contributed to the prediction of nitrogen intake and efficiency. Thus, it was not possible to identify a single metabolite that could be used as a marker to predict nitrogen efficiency, and it can be concluded that a wide-ranging urinary metabolite profile is needed to evaluate nitrogen efficiency in ruminants.