Objectives To evaluate the five‐tier Gleason grade group ( GG ) scoring of prostate cancers adopted by the International Society of Urology Pathology ( ISUP ) in 2014, and to propose modifications to optimize its performance. Patients and Methods Data were obtained from PROCURE , a prospective cohort of patients with localized prostate cancer undergoing radical prostatectomy in Québec, 2006–2013. Surgical specimens were evaluated by genitourinary pathologists using 2014 ISUP criteria. Treatment failure was defined as biochemical recurrence and/or initiation of secondary, non‐adjuvant therapy. Analyses were conducted using Kaplan–Meier methods, log‐rank tests, Cox proportional hazards models and Harrell's concordance indices. Results A total of 1 917 patients were included, with a median follow‐up of 69 months. The 5‐year treatment failure rates were 9.6%, 23.5%, 43.1%, 52.6% and 84.3% in GG 1–5, respectively ( P < 0.001 when comparing GG 2 with GG 3). Treatment failure rates for patients in GG 2 and GG 3 with tertiary Gleason 5 pattern were higher than patients in the same group without a tertiary pattern ( P < 0.001), but were similar to rates for patients in GG s 3 or 4 without a tertiary pattern ( P > 0.3). Primary Gleason pattern (4/5) predicted treatment failure in GG 5 (5‐year failure rates 82.3% vs 97.1%, respectively; P = 0.001). The five‐tier GG system had greater accuracy as a prognostic indicator compared with the four‐tier system ( Harrell's concordance index 0.716 vs 0.676). When upgrading patients in GG 2/3 with tertiary Gleason 5 pattern to patients in GG 3/4, and separating patients in GG 5 by primary Gleason pattern, the Harrell's concordance index increased to 0.730. Conclusion The five‐tier GG system increased accuracy for predicting treatment failure compared with the previous grading systems, but can be further improved.