Ten endogenous steroid hormones and metabolites were determined according to the screening procedure for anabolic steroids in spot urine samples from 105 healthy young male athletes (control samples) and 23 males that tested positive for anabolic steroids in the doping control (positive samples). The GC-MS peak areas for each sample were normalized to total area. Multivariate data analysis by Partial Least Square Regression (PLSR) and using a coded Y-variable (positive samples: +1 and control samples -1) allows projection of the most systematic profile structures into a 2D plot revealing a clear distinction between the control and misuser groups. The most important determinants of the location in the loading plot were the ratios of testosterone to epitestosterone and androsterone to etiocholanolone. The ratio between 11-beta-hydroxyandrosterone and 11-beta-hydroxy-etiocholanolone was less important, in accordance with the fact that anabolic-androgenic steroid intake primarily affects the excretion of testosterone from the testis and to a much lesser degree adrenal steroid genesis. We present a preliminary validation of this model (PLS1-DISCRIM) for analysing steroid profiles in doping control samples from several categories of athletes, some of which are suspected for drug misuse, and results from a one dose excretion study in healthy volunteers. Our findings suggest that use of multivariate PLS-regression may give valuable information about anabolic androgenic steroid misuse in sport. When appropriately calibrated, this methodology may delineate drug misusers directly from the screening procedure for anabolic steroids in spot urine tests.