Classifying high-risk versus very high-risk prostate cancer: is it relevant to outcomes of conformal radiotherapy and androgen deprivation?

Physicians and patients, when asked about therapy for localized prostate cancer often look to the National Comprehensive Cancer Network (NCCN) guidelines to provide guidance for selection between different treatment options [1]. Since patients with high-risk (HR) disease have a heterogeneous prognosis, this group has been further subdivided to separate patients thought to have the worst prognosis into the very high-risk (VHR) category [1].

Radiation therapy (RT) has long been considered the primary treatment modality for patients with HR disease and is the only treatment considered by the National Comprehensive Cancer Network (NCCN) to have sufficient evidence to support a Category 1 treatment recommendation [1]. Despite the NCCN treatment recommendations based on improved outcomes for HR patients treated with high dose conformal radiation therapy (CRT) and androgen deprivation therapy (ADT), there is growing interest in the use of radical prostatectomy (RP) for patients with HR disease [2, 3]. Justifications given for considering surgery are high rates of local and systemic failure associated with the use of RT as well as reported good outcomes associated with the use of surgery [2, 3].

Sundi et al. defined a VHR group with adverse prognostic factors predictive for poor outcome following surgery and suggested the need for multimodal therapy to improve outcomes [4, 5]. In consideration of these findings, the 2014 NCCN guidelines were revised and added the presence of primary Gleason grade 5 or ≥5 cores with Gleason score 8–10 as new criteria for inclusion into the VHR group [1]. While relevant for surgical outcomes, the predictive value of the HR/VHR grouping has not been assessed in patients treated with current CRT techniques [6]. Recently, Narang et al. showed inferior outcomes in the VHR versus HR group in a cohort of patients treated with RT and ADT from 1993 through 2006. However, this retrospective study was limited by use of radiation techniques, treatment volumes, dose, and use of ADT that do not reflect current therapeutic approaches [6].

We reviewed treatment outcomes in a cohort of patients with HR disease treated with high dose CRT and ADT to determine if local recurrence (LR) or metastatic disease was predominant. Patients were stratified according to the original and revised NCCN guidelines for HR/VHR groups. The value of this classification system to provide prognostic guidance and improved treatment recommendations for patients with HR disease was assessed.

Failure occurred in 30/203(15%) patients. Median time to failure was 30 m (range: 4 m-76 m). Failures were classified as BR only (n = 4), local (n = 1) or metastatic (n = 25). Table 2 lists the sites of failure. In 8 patients who were initially classified as BR, the use of choline PET-CT imaging showed the location and extent of recurrence. Prostate cancer specific mortality was recorded in 4 patients. Eleven deaths were unrelated to prostate cancer.

Four different NCCN HR/VHR groupings were evaluated. Patients shifted from the HR group to the VHR group as definitions evolved and additional risk features were included (Table 3). The size of the VHR group increased from 18 to 68% of the patients and HR group decreased from 82 to 32%.

KM estimate of 4 year b-DFS for the entire cohort was 87% (95%CI: 82–92%). The 4 year KM survival estimates for b-DFS, CSS and OS were comparable for each of the NCCN subgroups (Table 3) and this was confirmed by Cox regression. On univariate analysis, the NCCN subgroups were not predictive of b-DFS at 4 years. Only DMFS was worse for the VHR group for both pre and post revision NCCN definitions (p = .03 and .01 respectively). This difference was not observed if upstaging using ≥2HR factors was applied and this effect did not persevere on MVA. Cox univariate analysis was also significant for: PSA ≥40 p = 0.001; clinical stages T2c p = .004, T3b p = .02 and > 4 cores with Gleason 8–10 p < .03.

On MVA, PSA ≥ 40 ng/ml was the only significant predictor of b-DFS or DMFS at 4 years with a HR of 3.75 and 3.25, p < 0.005 (Table 4). KM estimates for PSA above and below 40 ng/ml are shown in Figs. 1a and b.

Treatment was well tolerated with significant late ≥ grade 3 GU toxicity of 10% which was predominantly due to reports of nocturia more than 5 times per night. Late ≥ grade 3 GI toxicity of 3.5% was due to rectal bleeding (Table 5).

This study supports the assertion that patients with both HR and VHR prostate cancer treated with high dose CRT, PLNRT and ADT have favorable outcomes with low toxicity. The use of dose escalation, CRT, PLNRT, image guidance and ADT have all been associated with improved outcomes or reduced toxicity [11‐15] and the results of this study are consistent with favorable outcomes reported from studies that used CRT with ADT to treat patients with HR disease [16]. While our results are encouraging, these findings must be interpreted cautiously since longer follow-up time is needed to verify our findings.

Dissimilarities in risk factors and co-morbidities of HR patients treated with RP or CRT make comparison of outcomes between surgical and radiation treatments unreliable. Series reporting treatment outcomes for CRT with HR disease include many patients with advanced disease or comorbidities that would exclude consideration for RP. Despite the inclusion of patients with adverse risk factors, our results compare well to outcomes reported for HR patients treated with RP [2, 17, 18]. Furthermore, we report low rates of acute and late GI and GU toxicity that are consistent with other CRT series [19]. In contrast, HR patients treated with RP often require adjuvant or salvage RT which is associated with increased toxicity when compared to treatment with RP or RT alone [20, 21].

The 2015 revisions to the VHR subgroup were based on the findings of Sundi et al. who reviewed prognostic factors and outcomes from a surgical series of 753 men with HR prostate cancer to create risk factor groupings predictive for metastatic disease and prostate cancer specific mortality [4]. Based on Sundi’s findings, the revised 2015 NCCN guidelines added two additional criteria for inclusion of HR patients into the VHR subgroup [1, 4]. When subdividing our CRT series into HR/VHR subgroupings, we observed that stage migration was substantial. The percentage of patients considered VHR increased from 18 to 62% as the number of criteria considered for inclusion in the VHR subgroup increased. We suggest that redistribution of patients into the VHR group may improve the reliability of comparisons of HR patients treated primarily with surgery versus those treated with radiation.

We found that patients treated with high dose CRT and ADT did well irrespective of current HR/VHR classification and that PSA ≥40 ng/ml was the best discriminator of poor outcomes. In contrast to our findings, Narang et al. showed that when evaluating patients with HR disease who were treated over a 15 year interval from 1993 to 2006, the revised 2015 NCCN HR/VHR subgrouping were predictive for worse outcomes in the VHR subgroup [6]. Although the long follow-up interval is an important strength of Narang’s study, the median follow-up time of our cohort is 50 m (range: 12 m-142 m) and the median time to BF in our reports are similar at 34 m and 30 m. The difference in our findings may be better explained by the differences in treatment received by our respective cohorts. Narang reported the use of a diverse assortment of radiation techniques with a mean dose of 70.2 Gy (range: 64.8–75.6 Gy), and differing ADT protocols. Narang was unable to demonstrate improvement in failure endpoints with dose escalation over 72Gy although use of neo-adjuvant ADT was associated with reduced BF and DMFS. In comparison, the patients in our series received uniform CRT with a minimum dose of 78 Gy (range: 78–82 Gy), PLNRT, and ADT that is reflective of current practice. Increased radiation dose and use of ADT have been shown to be associated with improved b-DFS and DMFS [11, 22, 23]. Pollack et al. in a randomized trial showed that doses <78 Gy versus ≥ 78 Gy were associated with improved b-DFS and DMFS and Denham et al. reported that both dose escalation and increased duration of ADT reduced local progression and BF [22, 23].

The only risk factor predictive for reduced b-DFS or DMFS on MVA in our series was PSA ≥40 ng/ml. This finding is consistent with other studies showing elevated PSA at time of diagnosis to be highly predictive for metastatic disease following either RT or RP [24‐27]. These findings suggest that future modifications to the NCCN guidelines consider PSA level as a criteria for inclusion into the VHR group. We caution that elevated PSA should not be used as an exclusion criteria for definitive treatment since many patients with elevated PSA may benefit from definitive therapy [24].

Patterns of failure analysis shows that distant failure was dominant and isolated initial failure within the lymph nodes, prostate or as BF alone was uncommon. The use of functional imaging has allowed us to identify sites of distant failure early which decreased the number of patients considered LR or BF alone. In contrast to our findings, several studies which used bone scan, CT and prostate biopsy to evaluate patients with BF have reported that the prostate is the most common first site of failure following RT for HR patients [28, 29]. Although prostate biopsy results were not available in our series, choline PET-CT and endorectal-MRI imaging in patients with BF allowed for early detection of metastatic disease in most patients. These findings support the use of CRT with ADT to treat patients with HR disease and suggest that further intensification of local therapy will provide little benefit for HR patients and may only add morbidity. [2, 3, 17, 30, 31].

The NCCN guidelines were revised in 2014 to include choline PET-CT imaging of patients with BF for consideration of focal salvage therapy [32]. Early adaptation of functional imaging with choline PET-CT allowed us to offer patients with local failure or oligo-metastatic disease salvage treatment using targeted radiation therapy and avoid early administration of ADT [8]. Extending the use of functional imaging agents for use during initial staging may further improve outcomes with RP and CRT by identifying and excluding patients with early metastatic disease from receiving definitive therapy.

The absence of a central pathology review is an important study limitation. Since patients were referred from several different institutions, variation between pathologists in assigning Gleason grades to the biopsy specimens may have affected the classification of our patients into HR and VHR subgroups. Although several different CRT treatment techniques were used during the study period and moderate hypo-fractionation and IGRT were instituted only after 2009, univariate and multivariate analysis were unable to demonstrate differences in outcome based on treatment technique. Although most BFs occur within 5 years of treatment [33], longer duration of follow-up is needed to verify our findings.

Prostate cancer patients with HR and VHR disease achieve excellent LC and DMFS with low toxicity when treated with dose escalated CRT, PLNRT and ADT. Reclassification of HR patients into HR/VHR subgroups using original or revised NCCN criteria had no impact on treatment outcomes. Only PSA ≥40 ng/ml was associated with poor prognosis. The use of functional imaging to evaluate BF showed that distant failure was dominant and LR in the prostate rare, challenging the notion that intensification of local therapy will further improve outcomes. Further study and longer follow-up is required to validate these findings.