A robust LC–MS/MS method was developed to quantify a large number of phase I and phase II steroids in urine. The decision limit is for most compounds lower than 1 ng ml−1 with a measurement uncertainty smaller than 30%. The method is fully validated and was applied to assess the influence of administered synthetic steroids and beta-agonists on the steroidogenesis. From three animal experiments, clenbuterol, diethylstilbestrol and stanozolol, the steroid profiles in urine of bovine animals were compared before and after treatment. It was demonstrated that the steroid profiles were altered due to these treatments. A predictive multivariate model was built to identify deviations from normal population steroid profiles. The abuse of synthetic steroids can be detected in urine samples from bovine animals using this model. The samples from the animal experiments were randomly analysed using this method and predictive model. It was shown that these samples were predicted correctly in the exogenous steroids group.