Abstract <p>A steroid profile is an informative tool from the viewpoint of both doping control and clinical diagnostics. Non-targeted analysis makes it possible to identify many compounds that can be potential biomarkers of various diseases. In this work, non-targeted GC–MS analysis of urine samples from 22 patients with cancer of various localizations and 22 healthy volunteers was carried out. Fifty four steroid hormones and some fatty acids were identified in the samples. Statistical processing of the results of non-targeted GC–MS analysis was performed using the Mann–Whitney <i>U</i> test and principal component analysis. In the groups of patients with cancer and healthy controls, 46 steroid hormones and 6 fatty acids showed statistically significant differences. By projecting observations onto a factor plane, we could separate the groups of patients with cancer and healthy people, which shows the promise of using non-targeted GC–MS analysis for the detection of cancer biomarkers.</p>