Identifying animals that are superior in terms of feed efficiency may improve the profitability and sustainability of the beef cattle sector. However, measuring feed efficiency is costly and time-consuming. Biomarkers should thus be explored and validated to predict between-animal variation of feed efficiency for both genetic selection and precision feeding. In this work, we aimed to assess and validate two previously identified biomarkers of nitrogen (N) use efficiency in ruminants, plasma urea concentrations and the 15N natural abundance in plasma proteins (plasma δ15N), to predict the between-animal variation in feed efficiency when animals were fed two contrasted diets (high-starch vs high-fibre diets). We used an experimental network design with a total of 588 young bulls tested for feed efficiency through two different traits (feed conversion efficiency [FCE] and residual feed intake [RFI]) during at least 6 months in 12 cohorts (farm × period combination). Animals reared in the same cohort, receiving the same diet and housed in the same pen, were considered as a contemporary group (CG). To analyse between-animal variations and explore relationships between biomarkers and feed efficiency, two statistical approaches, based either on mixed-effect models or regressions from residuals, were conducted to remove the between-CG variability. Between-animal variation of plasma δ15N was significantly correlated with feed efficiency measured through the two criteria traits and regardless of the statistical approach. Conversely, plasma urea was not correlated to FCE and showed only a weak, although significant, correlation with RFI. The response of plasma δ15N to FCE variations was higher when animals were fed a high-starch compared to a high-fibre diet. In addition, we identified two dietary factors, the metabolisable protein to net energy ratio and the rumen protein balance that influenced the relation between plasma δ15N and FCE variations. Concerning the genetic evaluation, and despite the moderate heritability of the two biomarkers (0.28), the size of our experimental setup was insufficient to detect significant genetic correlations between feed efficiency and the biomarkers. However, we validated the potential of plasma δ15N to phenotypically discriminate two animals reared in identical conditions in terms of feed efficiency as long as they differ by at least 0.049 g/g for FCE and 1.67 kg/d for RFI. Altogether, the study showed phenotypic, but non-genetic, relationships between plasma proteins δ15N and feed efficiency that varied according to the efficiency index and the diet utilised.