Background. Bladder cancer (BC) is a common urological cancer, 75 % of which are non-muscle invasive BC. After removal of the primary tumor, the adequate classification of malignancy and the defining of tumor progression risk remains an important issue, since it is associated with frequency of cystoscopy and choice of the BCG- or chemotherapy management.Objective: improve the algorithms of prognosis in intermediate-risk patients with non-muscle-invasive bladder cancer with the consideration of molecular characteristics of the primary tumor.Materials and methods. We studied 125 BC samples; mutations in the FGFR3, PIK3CA, TERT, and TP53 genes were determined by polymerase chain reaction and Sanger sequencing, as well as the expression of the FGFR3, EGFR, ERBB2, FOXA1, and GATA3 genes using realtime polymerase chain reaction.Results. Somatic mutations in the studied loci were detected in 65.6 % of the samples, five new mutations were identified. A decrease of the mutation frequency in the FGFR3 and TERT genes was shown, an increase — TP53 in order (Ta—T1/low G) > (T1/high G) > (>T2/any G). The largest area under the ROC curve (0.807 ± 0.092, р = 0.004) was demonstratedfor the prognostic classifier with the independent variables: mutation in FGFR3 and/or TERT; mutation TP53; overexpression of the FGFR3 gene. The FGFR3, TERT mutations and/or FGFR3 overexpression in the absence of TP53 mutation indicates minimally invasive primary tumor. On the contrary, harboring TP53 mutation indicate the features of muscle-invasive BC at the genetic level. Using this algorithm, we reclassified 21 of T1G3 BC cases as having characteristics associated with non-invasive tumor in 43 %, and invasive BC in 57 % of patients.Conclusion. The aforementioned prognostic model could be used as additional laboratory test in assessing the malignancy and progression risk of non-muscle invasive BC.