Independent validation of the prognostic capacity of the ISUP prostate cancer grade grouping system for radiation treated patients with long-term follow-up

Abstract

Background:

There has been a recent proposal to change the grading system of prostate cancer into a five-tier grade grouping system. The prognostic impact of this has been demonstrated in regards only to biochemical recurrence-free survival (bRFS) with short follow-up (3 years).

Methods:

Between 1990 and 2013, 847 consecutive men were treated with definitive external beam radiation therapy at a single academic center. To validate the new grade grouping system, bRFS, distant metastases-free survival (DMFS) and prostate cancer-specific survival (PCSS) were calculated. Adjusted Kaplan–Meier and multivariable Cox regression analyses were performed to assess the independent impact of the new grade grouping system. Discriminatory analyses were performed to compare the commonly used three-tier Gleason score system (6, 7 and 8–10) to the new system.

Results:

The median follow-up of our cohort was 88 months. The 5-grade groups independently validated differing risks of bRFS (group 1 as reference; adjusted hazard ratio (aHR) 1.35, 2.16, 1.79 and 3.84 for groups 2–5, respectively). Furthermore, a clear stratification was demonstrated for DMFS (aHR 2.03, 3.18, 3.62 and 13.77 for groups 2–5, respectively) and PCSS (aHR 3.00, 5.32, 6.02 and 39.02 for groups 2–5, respectively). The 5-grade group system had improved prognostic discrimination for all end points compared with the commonly used three-tiered system (that is, Gleason score 6, 7 and 8–10).

Conclusions:

In a large independent radiotherapy cohort with long-term follow-up, we have validated the bRFS benefit of the proposed five-tier grade grouping system. Furthermore, we have demonstrated that the system is highly prognostic for DMFS and PCSS. Grade group 5 had markedly worse outcomes for all end points, and future work is necessary to improve outcomes in these patients.

Introduction

Current National Comprehensive Cancer Network guidelines risk stratify men with prostate cancer into multiple risk groups based on various clinical and pathologic features.¹ A critical component of this risk stratification scheme is the Gleason score, with current guidelines defining a Gleason score of⩽6 as low-risk, 7 as intermediate-risk and 8–10 as high-risk disease.

Unfortunately, the currently employed three-tiered Gleason grouping system, where a score of 6 of 10 is the lowest sum achievable, is arguably non-intuitive for patients and physicians alike and lacks the necessary resolution to accurately define critical differences in the risk of recurrent disease following treatment for patients within each tier. Numerous institutions have demonstrated that there is a stark contrast in long-term disease control within Gleason score 7 tumors, with a Gleason score of 3+4=7 portending a more favorable prognosis than a score of 4+3=7.^{2, 3, 4, 5, 6, 7} Likewise, the presence of Gleason pattern 5 and a corresponding Gleason score of 9 or 10 is associated with significantly worse long-term disease control than Gleason score 8 tumors.^{8, 9, 10, 11, 12}

A proposal for a novel 5-grade group system was first described in a 2013 study from Johns Hopkins led by Dr Epstein.¹³ The simplified and more contemporary grading system for prostate cancer assigns a grade of 1 through 5, with 1 representing a Gleason score of ⩽6, 2 representing a Gleason score of 3+4=7, 3 representing a Gleason score of 4+3=7, 4 representing a Gleason score of 8 and 5, the highest score, representing Gleason scores of 9 or 10.¹⁴ They then demonstrated that this proposed stratification had increased prognostic accuracy for post-treatment biochemical recurrence-free survival (bRFS) in a large cohort of men treated with either prostatectomy or definitive radiation therapy with a median follow-up of only 3.1 years for the radiotherapy cohort.

Herein we validate these findings in a cohort of men treated with definitive external beam radiation therapy with long-term follow-up (7.3 years), and also assess the prognostic utility of this new grading stratification system for more meaningful patient outcomes including distant metastases-free survival (DMFS) and prostate cancer-specific survival (PCSS). Furthermore, given that the specific treatment utilized for radiotherapy-treated patients (for example, inclusion of seminal vesicles, pelvic lymph nodes and use of androgen deprivation therapy (ADT)) is highly dependent on Gleason score, we have principally assessed the independent impact of the new grade grouping system to diminish confounding clinicopathologic and treatment variables.

Materials and Methods

Patients

Between 1990 and 2013, 847 consecutive men were treated with definitive external bean radiation therapy to an intact prostate at a single academic center after the pathologic diagnosis of prostate cancer. All patients had central pathology review for consistency in Gleason scoring. The analysis of these men was approved through the local institutional review board.

Treatment

All men were treated with definitive external beam radiation with the use of three-dimensional conformal radiation therapy or intensity-modulated radiation therapy. The majority of men (>90%) were treated with dose-escalated radiotherapy to a total dose >75 Gy in 1.8–2.0-Gy fractions. Men were typically treated to the prostate alone, prostate plus the proximal first centimeter of the seminal vesicles or prostate plus the entire seminal vesicles based on the presence of low-, intermediate- or high-risk disease, respectively. Men with high-risk prostate cancer also received treatment to the pelvic lymph nodes at the treating physician’s discretion. Long-term ADT was routinely given to men with high-risk disease, whereas men with intermediate-risk disease received short-term ADT at the treating physician’s discretion.

Outcomes

Primary outcomes analyzed include bRFS, DMFS and PCSS. Biochemical recurrence was defined as the post-treatment PSA nadir+2 ng ml⁻¹ (Phoenix definition).¹⁵ Distant metastases were defined as evidence of metastatic disease clinically, pathologically or radiographically with computed tomography or bone scan. Prostate cancer-related deaths were defined as a death directly attributable to prostate cancer, or death in a man with hormone refractory or progressive metastatic prostate cancer.

Statistical analysis

Using the Cox regression method, adjusted Kaplan–Meier curves were generated to adjust for factors present in the multivariable model.¹⁶ Multivariable Cox regression analyses were performed to assess the independent impact of Gleason grade grouping after adjusting for year of treatment, pretreatment PSA, T-stage and the use of ADT. Year of treatment was dichotomized by 2005 given the change in Gleason grading in 2005 by the International Society of Urological Pathology (ISUP).¹⁷ The discrimination of the three-tiered and newly proposed grading systems were compared using receiver operating characteristic curves and corresponding areas under the curve for both the entire cohort and for patients who did not received ADT. For all statistical analyses, two-tailed P-values of ⩽0.05 were considered statistically significant. Statistical analyses were performed using IBM SPSS version 21.0 (SPSS, Chicago, IL, USA).

Results

Patient characteristics

The median follow-up of the cohort was 88 months. Clinical and pathological patient characteristics for the 847 treated men can be found in Table 1. The median age was 68.7 years. The majority of patients were either intermediate (44.5%) or high risk (31.3%). Similarly, most patients were T1c/T2a (73.0%) and had a baseline PSA <10 ng m⁻¹ (67.9%). A total of 771 patients (91.0%) were treated with dose-escalated radiotherapy to >75 Gy, and 341 men (40.3%) received neoadjuvant and concurrent ADT.

Table 1 Baseline characteristics

Independent validation of bRFS

Adjusted Kaplan–Meier survival curves using a Cox regression model for bRFS (Figure 1) stratified prognosis by the new grade grouping system (P-values compared with group 1; P=0.15, <0.0001, 0.02 and <0.0001 for groups 2–5, respectively). On multivariate Cox regression analysis for biochemical recurrence, the 5-grade group system independently validated a clear stratification of differing risk groups for bRFS (group 1 as reference; adjusted hazard ratio 1.35 (95% confidence interval (CI) 0.90–2.03), 2.16 (95% CI 1.39–3.35), 1.79 (95% CI 1.08–2.97) and 3.84 (95% CI 2.19–6.73) for groups 2–5, respectively). In addition, higher T-stage, PSA and year of treatment were independent predictors of biochemical recurrence (Table 2).

Figure 1

Adjusted Kaplan–Meier analysis of biochemical recurrence-free survival.

Table 2 Multivariable analyses

Distant metastasis-free survival

Adjusted Kaplan–Meier survival curves for DMFS (Figure 2) stratified prognosis by the new grade grouping system (P-values compared with group 1; P=0.11, 0.02, 0.01 and <0.0001 for groups 2–5, respectively). On multivariate Cox regression analysis for distant metastases, the 5-grade group system independently validated a clear stratification of differing risk groups for distant metastases (group 1 as reference; adjusted hazard ratio 2.03 (95% CI 0.85–4.84), 3.18 (95% CI 1.23–8.22), 3.62 (95% CI 1.37–9.54) and 13.77 (95% CI 5.11–37.10) for groups 2–5, respectively). In addition, higher T-stage and PSA were independent predictors of distant metastasis (Table 2).

Figure 2

Adjusted Kaplan–Meier analysis of distant metastasis-free survival.

Prostate cancer-specific survival

Adjusted Kaplan–Meier survival curves for PCSS (Figure 3) stratified prognosis by the new grade grouping system (P-values compared with group 1; P=0.10, 0.02, 0.01 and <0.0001 for groups 2–5, respectively). On multivariate Cox regression analysis for prostate cancer mortality, the 5-grade group system independently validated a clear stratification of differing risk groups for prostate cancer mortality (group 1 as reference; adjusted hazard ratio 3.00 (95% CI 0.80–11.27), 5.32 (95% CI 1.33–21.31), 6.02 (95% CI 1.46–24.82) and 39.02 (95% CI 9.46–160.96) for groups 2–5, respectively). In addition, higher T-stage, PSA and year of treatment were independent predictors of prostate cancer-specific mortality (Table 2).

Figure 3

Adjusted Kaplan–Meier analysis of prostate cancer-specific survival.

Grading group discrimination

The newly proposed 1–5-grade grouping resulted in improved discrimination for biochemical failure, development of distant metastases and prostate cancer-specific mortality as compared with the historical three-tiered low- (Gleason 6), intermediate- (Gleason 7) and high-risk (Gleason 8–10) grouping. For each end point, the new 1–5 grouping system resulted in an increased area under the curve on receiver operating characteristic analysis (Table 3). This held true whether patients did or did not receive ADT as part of their treatment. The discriminatory power with the new system was in fact strongest for prostate cancer-specific mortality in patients who did not received ADT (area under the curve: 0.805). Although the discriminatory power was improved with the new grade grouping system for biochemical recurrence, it remained low for the entire cohort (area under the curve: 0.596).

Table 3 Discriminatory analyses of grade groupings

Discussion

The Gleason histologic grading system, named after Donald Gleason, was initially developed in the 1960’s through prostate cancer research performed by the Veterans Administration Cooperative Urological Research Group.¹⁸ Nine patterns were identified based on unique histologic features, which were later condensed into five distinct histologic patterns.¹⁸ The Gleason score has repeatedly been demonstrated to be a critical prognostic variable for men treated for prostate cancer in the definitive setting with surgery or radiation, as well as in the recurrent setting for men treated with salvage radiation therapy post-prostatectomy.^{8, 9, 10, 11, 19, 20, 21, 22, 23, 24} As its inception, the granularity and resolution offered by the original five-pattern grading system has gradually been diminished, as best evidenced by the current National Comprehensive Cancer Network risk stratification that classifies prostate cancer as low (Gleason score⩽6), intermediate (Gleason score 7) or high risk (Gleason score 8–10).¹ Thus, a scoring system that originally consisted of 25 possible combinations of primary and secondary patterns has been reduced commonly to three generalized risk groups.

Mounting evidence suggests that the current three-tiered risk stratification system (Gleason score ⩽6, 7 and 8–10) is insufficient and over-simplistic to account for critical differences in risk that exist within the intermediate and high-risk groups. Particularly, patients with Gleason score 3+4=7 and 4+3=7 disease have different long-term prognoses.^{2, 3, 4, 5, 6} The landscape of Gleason score 7 prostate cancer is vast, and it is not surprising that men with low-volume Gleason 3+4=7 disease may have a lower risk for disease recurrence than men with high-volume Gleason 4+3=7. The spectrum of prognoses associated with intermediate-risk prostate cancer is best captured by the recently adopted Zumsteg–Spratt risk classification by the National Comprehensive Cancer Network. This classification system divides intermediate-risk prostate cancer into favorable and unfavorable intermediate-risk groups, and demonstrated that patients with favorable intermediate-risk disease have excellent long-term outcomes comparable to men with low-risk disease.⁵ Similar discrepancies exist in long-term prognosis for men with Gleason score 4+4=8 disease compared with men with a primary or secondary Gleason pattern 5. The presence of Gleason pattern 5 results in significantly poorer long-term outcomes.^{9, 11, 25, 26} Last, the current risk stratification schema may be less intuitive for patients with a Gleason score of 6 out of 10, which in fact currently represents the lowest achievable score.

Recently, a contemporary grading system for prostate cancer was proposed by Dr Epstein that is composed of five-prognostic grade groupings.^{13, 14, 27} This grade grouping system addressed several of the highlighted shortcomings of the current three-tiered system and was shown to result in increased prognostic discrimination compared with the currently employed Gleason grading system for the end point of post-treatment PSA recurrence for men treated with either radiation therapy or surgery.¹⁴ The follow-up in this study was short (~3 years for both the surgery and radiotherapy cohorts), and no analyses were presented to understand if the new system provides independent benefit when adjusting for important clinicopathological variables (that is, T-stage, pretreatment PSA and so on). Herein, we provide independent validation of the proposed 5-grade system by Epstein et al., and establish that in addition to improved accuracy for predicting biochemical failure, the new grading system results in improved discrimination for the development of distant metastases and prostate cancer-specific mortality in a cohort of men treated with definitive external beam radiation therapy at a single institution. Our findings thus support the adoption of the new 5-grade prostate cancer staging system, particularly given the improved discrimination for the more meaningful end points of metastases and prostate cancer-specific mortality.

Notable in our analysis is the overall poor prognosis associated for patients within grade group 5. These men have a markedly increased risk for metastatic disease and are much more likely to die as a result of their prostate cancer than any of the other groupings. We have previously demonstrated inferior prognosis associated with the presence of Gleason pattern 5 for men receiving either definitive or post-prostatectomy radiation therapy^{9, 11} and this holds true for men undergoing prostatectomy as well.¹² In our analysis, these poor outcomes were present despite the fact that the vast majority of men in this group were treated with 2 years of ADT and received >75 Gy to the prostate. Clinical trials are needed to assess methods by which to intensify local therapy for these men as are improved methods by which to detect the presence of subclinical metastatic disease.

Although men in group 5 in our analysis had overall poor long-term outcomes, men in groups 3 and 4 were significantly less likely to develop metastatic disease or experience a prostate cancer-specific death. Interestingly, although there was a clear difference with regards to long-term biochemical control for patients in grade grouping 3 (Gleason score 4+3=7) and 4 (Gleason score 4+4=8), these two groups had nearly identical long-term rates of developing metastatic disease and prostate cancer-related death. A possible explanation for this finding could be related to the fact that the majority of men in group 4 were treated with 2 years of ADT, whereas men in group 3 received usually 6 months of ADT or none at all. Long-term ADT has been demonstrated to improve outcomes for men with Gleason 8 disease,²⁸ and there is mechanistic rationale for the benefit associated with combining ADT with radiation therapy in the treatment of prostate cancer through radiation-induced upregulation of the androgen receptor.²⁹ To overcome the confounding of ADT, Epstein et al. demonstrated that in men who received radiotherapy without ADT, the prognostic difference was more apparent than in men who received combined modality therapy.¹⁴

Questions remain with the new grade grouping system. A critical question that remains unanswered is that given the poor prognosis associated with primary or secondary Gleason pattern 5, which grade group is most appropriate for patients with Gleason 5+3=8 or 3+5=8 disease. A recent Surveillance, Epidemiology, and End Results (SEER) analysis comparing outcomes among prostate cancer patients with Gleason 4+4=8, 3+5=8 and 5+3=8 indicated that the presence of primary Gleason pattern 5 resulted in a doubling of the 5-year rate of prostate cancer-specific mortality compared with men with 4+4 or 3+5 disease and similarly a twofold increase in the risk of prostate cancer-specific mortality.²⁶ Given the increased risk for metastatic disease for men with primary or secondary Gleason pattern 5, this begs the question if any patient with primary or secondary Gleason pattern 5 (that is, 3+5=8 or 5+3=8) should be put in group 5 of the new grading system. However, it is important to note that there are limitations to SEER analyses on grade grouping. The SEER database does not control for the consistency in reporting or grading. The data are extracted from pathology reports in hospitals without central re-review, and many SEER sites do not have genitourinary pathology experts. In practice, Gleason 3+5 and 5+3 are exceptionally rare entities.

Our present analysis is not without limitation. Given the retrospective nature of our analysis, we are unable to fully control for confounding factors that likely impacted treatment decisions, such as use and duration of ADT. Furthermore, men were treated over an approximately two-decade period during which time the way individual Gleason patterns have been defined changed. We attempt to control for stage migration by including the year of treatment in our analysis. Last, we were unable to assess the impact of tertiary Gleason pattern 5, as tertiary patterns were not routinely reported at our institution over the study period.

In summary, we provide independent validation of the prognostic benefit for men treated with definitive radiotherapy of the new prostate cancer five-tier grade grouping system with regards to bRFS, and further demonstrate improved prognostic discrimination for the more meaningful outcomes of DMFS and PCSS. Future staging and risk stratification schemes should include this five grade grouping system.