In this population-based register study on the prognosis of Gleason score 8 prostate cancer in needle biopsies, there was a significantly higher mortality in men with Gleason score 5 + 3 cancer compared to those with 4 + 4 cancer and in men with Gleason score 4 + 4 cancer compared to those with 3 + 5 cancer. A majority of these men were treated with androgen deprivation therapy. Those who received treatment with curative intent in general had a very low prostate cancer mortality (0.04–0.15 after 10 years), and significant differences between the grade subgroups were not observed.

Prognostic heterogeneity in Gleason score 8 prostate cancer has been demonstrated in most previous studies [2,3,4,5,6,7]; however, outcomes between Gleason scores 3 + 5, 4 + 4 and 5 + 3 vary between the studies. In a systematic review and meta-analysis, Lu et al. analysed eight reports and concluded that the meta-analysis was hampered by the widely differing methodologies of the studies [4]. In particular, there were differences in the specimens used for the studies with cases consisting of biopsy specimens [2, 3, 5, 7, 8], radical prostatectomy specimens [2] or a combination of the two [6]. Other differences were that some studies relied on single-centre data [3], while others were based on multiple-centre data [2, 8] or registry data [5,6,7]. A problem in several studies was the low number of some of the grade categories. This was in particular true for Gleason score 5 + 3 tumours, and as a consequence, results from these studies need to be interpreted with caution [5, 8]. Selection bias most likely accounts for some observed outcome differences between treatments, such as the low prostate cancer mortality in men who underwent radical prostatectomy or radical radiotherapy in the current study.

Using Surveillance, Epidemiology and End Results (SEER) database registry data, Mahal et al. found that Gleason score 5 + 3 cancers had a significantly worse prognosis than those with a Gleason score of 3 + 5 or 4 + 4 [6]. However, the methodology of grade reporting in SEER was not uniform throughout the study period, and grading was based on the highest Gleason score in either the preoperative needle biopsies or the subsequent radical prostatectomy specimen. As a consequence, the results of this latter study are not comparable with those of our study. By contrast, Gandaglia et al. found a similar outcome in tumours with Gleason scores 4 + 4 and 5 + 3, but their data were based on radical prostatectomy specimens only [2]. Huynh et al. and Rusthoven et al. on the other hand combined cases of Gleason scores 3 + 5 and 5 + 3 and compared them against outcomes of Gleason score 4 + 4 tumours based on the assumption that the presence of a Gleason pattern 5 would be the most critical marker of prognosis [3, 7].

Gleason score 8 cancers on needle biopsy are Gleason score 4 + 4 in the vast majority of cases, and only 9% [3]–21% [8] are being reported as 3 + 5 or 5 + 3. In the present study, 86% were 4 + 4, which is in line with the other large register-based study [6]. Of all Gleason score 8 cancers in earlier series, only ≤ 5% were reported as 5 + 3 [5, 6, 8]. As noted above, some of the previous studies merged Gleason scores 3 + 5 and 5 + 3 due to a paucity of cases and did not report separate results [3, 7].

At the 2014 ISUP consensus conference on Gleason grading, it was recommended that the grading of prostate cancer be based on 5 ISUP grades (also known as grade groups) for reporting purposes. These so-called ISUP grade groups were defined on the basis of Gleason scores, i.e. Grade 1, Gleason score 3 + 3 = 6; Grade 2, Gleason score 3 + 4 = 7; Grade 3, Gleason score 4 + 3 = 7; Grade 4, Gleason scores 3 + 5 = 8, 4 + 4 = 8 or 5 + 3 = 8; and Grade 5 4 + 5 = 9, 5 + 4 = 9 or 5 + 5 = 10. It is apparent that while ISUP grades 1–3 consist of single Gleason scores, this does not apply to ISUP grades 4 and 5, and concerns have been raised that this grouping would have considerable disadvantages by not revealing the exact Gleason grade composition present in individual cases [16]. It has been shown that any percentage of Gleason pattern 5 tumour in a needle biopsy is associated with a worse prognosis when compared to Gleason score 4 + 3 tumours [17]. In a recent study, we analysed the outcome of ISUP grade 5 cancers and found significant differences in mortality between the three grade subgoups (Gleason scores 4 + 5, 5 + 4 and 5 + 5) confirming that a grouping of these Gleason scores would cause loss of granularity of data which, in turn, would diminish any prognostic information derived from grading [12]. In this present study, we found further support for this hypothesis. By restricting our analysis to the subgroups of tumours that constitute ISUP grade 4, we now also demonstrated a prognostic heterogeneity within this ISUP grade. Interestingly, the PCSM for Gleason score 5 + 3 cancers was very similar to that of Gleason score 4 + 5 cancers, reported in our earlier study, with PCSM after 10 years at 0.44 (0.39–0.49) and 0.45 (0.44–0.46), respectively [12].

Thus, while we have demonstrated significant differences in prognosis between the components of ISUP grades 4 and 5 when analysed separately, there was also an overlap between some components of ISUP grade 4 and ISUP grade 5. These data further emphasise that by grouping of Gleason scores to generate ISUP grades, valuable prognostic information is lost.