We developed a test that compares sequential measurements of a biomarker against previous readings performed on the same individual. A probability mass function expresses prior information on interindividual variations of intraindividual parameters. Then, the model progressively integrates new readings to more accurately quantify the characteristics of the individual. This Bayesian framework generalizes the two main approaches currently used in forensic toxicology for the detection of abnormal values of a biomarker. The specificity is independent of the number n of previous test results, with a model that gradually evolves from population-derived limits when n = 0 to individual-based cutoff thresholds when n is large. We applied this model to detect abnormal values in an athlete's steroid profile characterized by the testosterone over epitestosterone (T/E) marker. A cross-validation procedure was used for the estimation of prior densities as well as model validation. The heightened sensitivity/specificity relation obtained on a large data set shows that longitudinal monitoring of an athlete's steroid profile may be used efficiently to detect the abuse of testosterone and its precursors in sports. Mild assumptions make the model interesting for other areas of forensic toxicology.