The development of screening methods for various types of cancer is of utmost importance as the early diagnostics of these diseases significantly increases the chances for patient's successful medical treatment and recovery. In this study we have developed the procedure for chromatographic profiling of urine samples based on solid-phase microextraction and GC–MS. 50 urine samples (20 from the patients with biopsy conformed prostate cancer and 30 from control group) were studied in the optimized experimental conditions. Application of chemometric classification algorithms such as k-nearest neighbors and partial least squares-discriminant analysis allowed construction of predictive models yielding very high sensitivity, specificity and accuracy values all close to 100%. This gives a good promise for further validation of this approach with a broader sample sets.