Phase II trial of combination radiation, hormone, and immunotherapy in grade group 5 prostate cancer

Abstract

Background

Grade group 5 (GG5) prostate cancer (PCa) carries a less favorable prognosis after standard-of-care (SOC) therapy, necessitating novel therapeutic approaches. High-dose rate brachytherapy (HDRBT) and androgen deprivation therapy (ADT) may modulate immune response in GG5 PCa, particularly in tumors with increased immune content. This study evaluated whether the addition of nivolumab to SOC was associated with improved disease control in patients with high-volume GG5 PCa, including those with oligometastatic disease.

Methods

In this non-randomized phase II trial, 31 patients with localized or oligometastatic GG5 PCa and >30% positive biopsy cores were evaluated between September 2018 and April 2021. Patients received four doses of nivolumab (240 mg every 2 weeks) beginning 4 weeks prior to HDRBT, alongside ADT, HDRBT, and external beam radiation. The primary endpoint was to evaluate whether the 2-year freedom from biochemical recurrence (FFBR) rate would exceed a prespecified historical control rate of 75%.

Results

Among the 31 patients, the median follow-up was 38.8 months (IQR 31.0–46.5 months). The addition of nivolumab to SOC RT with ADT was associated with a 2-year FFBR rate of 90.3% (95% CI 74.3% to 98.0%) (median FFBR not reached), exceeding the prespecified historical control rate of 75% (one-sided p value from binomial test=0.024). Definitive and probable nivolumab-related toxicity included 6.3% acute grade 2 and 6.3% acute grade 3 adverse events (AEs), with no grade 4+ AEs observed. A higher Decipher immunosuppression score at diagnosis correlated with early pathologic response (p=0.005) and was independently associated with time to metastatic failure (p=0.044).

Conclusions

Nivolumab combined with SOC was associated with encouraging FFBR in this high-risk GG5 PCa population and may represent a promising therapeutic intensification strategy. The Decipher immunosuppression score may serve as a predictive biomarker for response. These findings warrant further investigation in randomized trials.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Grade group 5 (GG5) prostate cancer (PCa) carries a less favorable prognosis after standard-of-care (SOC) therapy, necessitating novel therapeutic approaches. The current treatment paradigm involves long-course androgen deprivation therapy (ADT) combined with radiation therapy (RT) or surgical resection. RT consists of either external beam radiation therapy (EBRT) monotherapy or EBRT combined with brachytherapy.

WHAT THIS STUDY ADDS

• We designed an approach to leverage both the immunomodulatory effects of ablative RT and ADT in addition to optimizing immune checkpoint inhibitors (ICIs) sequencing to improve PCa disease control. We treated men with an aggressive form of PCa using a combination of hormone therapy, radiation therapy, and an immune-based drug called nivolumab.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study provides prospective clinical evidence supporting the feasibility and potential clinical activity of integrating nivolumab with high-dose rate brachytherapy and ADT followed by EBRT in GG5 PCa, a population with historically poor outcomes. The regimen’s favorable safety profile suggests that ICIs can be feasibly incorporated into existing SOC therapies, warranting further investigation in larger, randomized trials. This study also identified the Ricketts Immunosuppression Score as a potential predictive biomarker, offering a path toward more personalized treatment strategies in aggressive PCa.

Introduction

Men with high-risk and very high-risk prostate cancer (PCa) have poorer oncological outcomes compared with their lower-risk counterparts after standard-of-care (SOC) therapy.^{1 2} High-risk and very high-risk PCa account for approximately two-thirds of deaths for patients with PCa despite being approximately only one-sixth of new diagnoses.³ Treatment intensification with prolonged androgen deprivation therapy (ADT) and brachytherapy (BT) boosts improve disease control for these men.¹ The current treatment paradigm involves long-course ADT combined with radiation therapy (RT) or surgical resection.^{1 2} RT consists of either external beam RT (EBRT) monotherapy or EBRT combined with BT.²

Patients with Gleason grade group 5 (GG5) PCa have significantly poorer outcomes compared with those patients with lower GGs.⁴ Among patients with high-risk and very high-risk PCa, GG5 is an independent risk factor of recurrence (p=0.016).⁵ Patients with localized GG5 PCa treated with high-dose EBRT and long-term ADT have a 5-year biochemical recurrence rate of 27% and a metastatic failure rate of 24%.^{6 7} These men face significantly higher mortality compared with those with less aggressive forms of PCa, highlighting the need for more effective treatment strategies.³

Immune checkpoint inhibitors (ICIs) have demonstrated promising results in various cancer types, however; they have had limited success in PCa.⁸ Previous studies that evaluated the use of RT and ICIs in PCa were largely negative,9,11 with a benefit only seen in subsets of patients and treatment regimens^{10 11} suggesting that sequencing of ICI and ablative-dose RT is critical to generate an effective tumor response. Additionally, recent preclinical data suggested that heterogeneous RT, a dose pattern that naturally arises from BT, enhances adaptive immune priming by stimulating a wide range of immunogenic effects within the tumor microenvironment (TME).¹² These antitumor immunity effects were shown to be enhanced with the addition of ICI therapy.¹² Finally, ADT has been shown to increase T-cell levels and immunogenic responses, potentially boosting the effects of immunotherapy.^{13 14}

Therefore, we designed an approach to leverage both the immunomodulatory effects of ablative RT and ADT in addition to optimizing ICI sequencing to improve PCa disease control. The phase I portion of this trial demonstrated the first clinical data supporting this approach.¹⁵ Here, we present the final analysis of clinical outcomes for this phase II trial evaluating the safety and efficacy of nivolumab in combination with high-dose rate brachytherapy (HDRBT), EBRT, and ADT in patients with high-risk, very high-risk, or oligometastatic GG5 PCa.

Methods

Study design and participants

This was a non-randomized, single-arm, investigator-initiated phase II trial of adults with localized or oligometastatic GG5 prostate adenocarcinoma. Eligible patients were aged ≥18 years with high-volume disease, defined as ≥30% positive biopsy cores with at least 12 cores sampled. Only patients with GG5 disease were eligible; patients meeting high-risk or very high-risk criteria based on prostate-specific antigen (PSA), T stage, or GG4 disease alone were excluded. Oligometastatic disease was defined as ≤3 distant metastatic sites and/or positive pelvic lymph nodes. Patients were required to be candidates for HDRBT and at least short-term ADT, have Eastern Cooperative Oncology Group (ECOG) performance status 0–1, and have adequate hematologic, renal, and hepatic function. Full eligibility criteria are provided in online supplemental table S1. Generalizability of the study population is summarized in online supplemental table S2. The study protocol has been included in online supplemental file 2.

Procedures

Patients received baseline MRI and CT of the abdomen and pelvis as part of staging workup. Whole-body nuclear bone scans were used for systemic imaging workup. Prostate-specific membrane antigen positron emission tomography scans became available at our center in late 2021 and were not available for diagnostic systemic imaging during enrollment but were used for restaging after biochemical recurrence for all but one patient. Lymph nodes measuring ≥1 cm in short-axis diameter were considered clinically involved.

Transperineal 12-core systematic biopsies and blood samples from the time of diagnosis were collected. A Decipher Prostate Cancer Genomic Classifier (Veracyte, South San Francisco, California, USA) was performed on all enrolled patients. Patients were followed every 3 months after completion with PSA surveillance. All patients were also required to undergo multiparametric MRI (mpMRI) for BT planning and 3 months post-EBRT to assess radiologic treatment response.

The trial schema and flow diagrams are presented in figure 1A–C. All trial participants received trimodality therapy including neoadjuvant, concurrent and adjuvant ADT, HDR (two implants; HDR#1 and HDR#2) followed by EBRT. The duration of ADT for all patients on study was a minimum of 12 months as previously described in the Androgen Suppression Combined with Elective Nodal and Dose Escalated Radiation Therapy trial.¹⁶ Specifics on ADT treatment regimen were left to the treating physician’s discretion. EBRT began 2 weeks after the completion of HDR#1 and HDR#2 and was administered at a dose of 45 Gy in 25 fractions at 1.8 Gy per fraction as per the National Comprehensive Cancer Network (NCCN) guidelines.¹⁷ Initiation of ADT had to begin between 4 weeks to 6 months prior to nivolumab. Patients received two 240 mg doses of nivolumab intravenously every 2 weeks for a total of four doses, starting 4 weeks before HDR. HDR consisted of two 1150 cGy implants delivered 2 weeks apart. In patients with seminal vesicle (SV) involvement (T3b), the SVs were included within the HDR implant volume when anatomically feasible and encompassed within the EBRT fields to ensure full therapeutic dosing.

Figure 1. Trial schema and flow diagrams. (A) Trial schema. The study cohort consisted of a quad-modality regimen of ADT, nivolumab, HDRBT, and EBRT. Patients enrolled in the study must have received ADT before RT, but ADT treatment specifics were left to the treating physician’s discretion. Patients received two 240 mg doses of nivolumab intravenously every 2 weeks for a total of four doses, starting 4 weeks before HDRBT. HDRBT consisted of two 1150 cGy implants and were performed 2 weeks apart. Transperineal six-core prostate biopsies were performed immediately before each fraction of HDRBT (specimen collection T2 was within 2 hours after HDR#1). EBRT was administered after HDRBT completion, with a total dose of 4500 cGy delivered in 25 fractions, and daily doses delivered five times per week. (B) Study flow diagram. 114 patients were screened for enrollment, with 34 being ultimately enrolled. All 34 patients received at least two doses of nivolumab. 31 patients completed all therapies and 2 years of follow-up and were included in the final endpoint analysis. (C) Contemporary control flow diagram. Data were retrospectively collected from a contemporary control cohort of 45 patients who met all study enrollment criteria. ADT, androgen deprivation therapy; EBRT, external beam radiation therapy; HDR, high-dose rate; RT, radiation therapy; SOC, standard-of-care.

Fiducial-guided transperineal six-core prostate biopsies were performed at the time of each fraction of HDR (figure 1A). A novel coordinate-tracking system developed at our institution was used to ensure consistent and anatomically reproducible biopsy sampling.¹⁸ This technique used four transrectal ultrasound-guided fiducial markers to enable precise mapping of biopsy locations relative to CT imaging, allowing for accurate correlation with tumor regions.

EBRT was administered after HDR completion, with a total dose of 45 Gy delivered in 25 fractions and daily doses delivered five times per week using intensity-modulated RT with image guidance techniques (figure 1A). Nodal irradiation was only performed if there was imaging evidence of pelvic nodal involvement and when recommended by the treating physician. After completion of EBRT, clinical follow-up was performed every 3 months and included serum PSA and testosterone (TTT) testing for the first 2 years. The only planned post-treatment imaging consisted of an mpMRI at the time of the first follow-up visit. Restaging imaging modality at the time of biochemical recurrence was not prespecified on the protocol and was at the discretion of the treating physician.

In addition to the prospective nivolumab cohort, two institutional comparator cohorts were identified to provide contextual interpretation of outcomes. An institutional contemporary control cohort (hereafter referred to as the contemporary controls) consisted of patients treated at the same institution between January 2013 and November 2021 who met the trial eligibility criteria but received SOC therapy consisting of ADT, HDRBT, and EBRT without nivolumab (figure 1C). These patients were retrospectively identified from the institutional database, and the inclusion date range was chosen to include all patients who were used to calculate the primary endpoint of the study up until nivolumab cohort enrollment closure. A historical control cohort (hereafter referred to as the historical controls), treated prior to study initiation (between January 2013 and December 2017) and meeting the same disease-based eligibility criteria, was also identified and was used in the original study design to estimate the expected primary outcome (online supplemental table S3). Historical controls were treated with trimodality therapy consisting of EBRT, BT boost, and ADT. These comparative cohorts were not prospectively enrolled and were included only for descriptive and contextual comparisons.

Outcomes

The primary outcome in this study was the rate of 2-year freedom from biochemical recurrence (FFBR). 2-year FFBR was measured from the time of enrollment to either the date of biochemical recurrence (BCR) (if occurring within 24 months) or the date of last clinical follow-up, with patients censored at 24 months. FFBR was defined as the absence of BCR. The NCCN definition of BCR was used (PSA rise of 2.0 ng/mL from nadir or evidence of disease progression on imaging during a recurrence evaluation triggered due to a rapidly rising PSA). TTT recovery was defined as >50 ng/dL.

The secondary endpoint was grade 3 or higher acute and late toxicity. Toxicity was assessed according to the Common Terminology Criteria for Adverse Events V.5.0. Assessments were conducted at baseline, during treatment, and at every post-RT follow-up. All toxicity events were reviewed by the investigative team for grading and attribution to each treatment modality. Toxicities deemed possible, probable, and definite in relationship to therapy are reported. Toxicities are grouped into grades 1 and 2, grade 3, and grade 4, and divided into acute and late toxicities, defined as ≤3 months and >3 months from end of nivolumab therapy, respectively.

Exploratory endpoints included the assessment of major pathological response (MPR), defined as ≤1 positive core on transperineal six-core prostate biopsies, as previously described.¹⁵ Fiducial-guided biopsies were performed at two time points: immediately prior to each of the two HDRBT fractions. The first exploratory biopsy (T2) was obtained prior to the first HDRBT implant, after two doses of nivolumab. The second exploratory biopsy (T3) occurred 2 weeks later, prior to the second HDRBT implant and following completion of all four nivolumab doses. These biopsies were evaluated to assess tumor response to treatment and to classify patients as early or late pathological responders. Tumors were defined as early responders if patients achieved MPR at T3 (ie, immediately prior to HDRBT fraction #2, following completion of all four cycles of nivolumab); those that did not achieve MPR at T3 were categorized as late responders. Radiological response was defined as resolution of abnormalities on T₂ and restricted diffusion on diffusion-weighted imaging as determined by a board-certified diagnostic radiologist at 3 months post completion of RT.

As part of the correlative analysis, gene expression profiling was performed using the Human Exon 1.0 ST microarray (Thermo Fisher, Carlsbad, California, USA) at a Clinical Laboratory Improvement Amendments-certified clinical laboratory (Veracyte, San Diego, California, USA). The SCAN algorithm was used for preprocessing and normalization of the transcriptomic data resulting in log₂ gene expression. Complete transcriptomic information from the Decipher GRID registry was obtained for all the samples which contain messenger RNA (mRNA) expression for approximately 46,000 genes.¹⁹ To limit our exploration among known immune-related genomic markers, we used a priori identified immune signature scores available within the GRID database to differentiate between early and late pathological response (online supplemental figure S2).

Statistical analysis

The study was planned with a sample size of 31 evaluable patients to achieve 60% power to test for a 2-year FFBR rate of 90% with the addition of nivolumab against the historical control rate of 75%, using a one-sided binomial test with a type I error rate of 5%. The historical controls provided the benchmark rate used for the trial design. Comparisons with contemporary controls were post hoc and were not included in the original study power calculation. This study assumed a 10% dropout rate; thus, 34 patients were enrolled. The primary endpoint of 2-year FFBR was estimated using the binomial distribution and compared with the historical control rate using a one-sided binomial test. As a secondary analysis, FFBR was also estimated using Kaplan-Meier (KM) methods. 3-year FFBR was measured from the ADT start date to either the date of BCR (if occurring within 36 months) or the date of last clinical follow-up, with patients censored at 36 months. The 3-year time point was chosen to assess sustained differences in FFBR between the nivolumab cohort and comparator cohorts. KM curves were generated for the nivolumab cohort, historical controls, and contemporary controls, and post hoc comparisons between cohorts were performed using log-rank tests. These comparisons were post hoc and the study was not powered for formal statistical comparisons between cohorts.

The area under the receiver operating characteristic (ROC) curve (AUC) was used to measure the ability of the Ricketts Immunosuppression Score (RIS) gene signature biomarker to differentiate early responder versus late responder (pathological), as well as partial responder versus complete responder (radiographic). The AUC was generated by evaluating the ability of the RIS to predict outcome at each observed value of the continuous score. As an AUC of 0.5 would indicate no ability to discriminate, generally, an AUC of 0.7 or higher is considered that the biomarker had good power to predict. To explore the potential protection provided by a higher RIS, a Cox proportional hazard (PH) regression model was fit using time to distant metastasis as the event of interest. For the Cox PH model, the RIS was categorized into high and low based on the median RIS of an independent data set of 319 retrospective GRID biopsy samples with a Gleason grade of 4 or 5. UVA Cox PH models were fit, as well as multivariable models, adjusting for Response Group (early response vs late response), ADT treatment length (> 18 months vs ≤18 months), stage, and Gleason score.

All tests are two-sided, except for the primary endpoint analysis. Statistical significance was defined as p<0.050. Statistical tests were performed using the R programming language and computing environment (R V.4.3.1; R Foundation for Statistical Computing, Vienna, Austria).

Results

Patient and treatment characteristics

Between November 16, 2018, and May 26, 2021, 114 patients with untreated (aside from ADT) high-volume GG5 PCa were screened, of whom 34 were enrolled in the prospective nivolumab cohort (figure 1B). Out of these 34 patients, 3 patients were removed from the study after initiating ICI therapy but before beginning RT due to a colorectal cancer diagnosis, a cardiovascular event-related death, and a head trauma-related death, resulting in a final evaluable cohort of 31 patients. These three patients were included in the toxicity analysis but excluded from the primary endpoint, genomic, and differential clinical outcome analyses. A contemporary control cohort of 45 patients meeting the same eligibility criteria and treated during the same time period with SOC therapy consisting of EBRT, HDRBT boost, and ADT without nivolumab was retrospectively identified for comparison (figure 1C).

Patient, tumor, and treatment details for the nivolumab and contemporary controls are shown in table 1. The nivolumab cohort had a median age of 65.0 years (IQR 57.0–69.0 years) and a median pretreatment PSA of 9.7 ng/mL (IQR 7.4–18.9 ng/mL). Most patients had an ECOG performance status of 0 (93.5%); 74.2% had stage IIIC disease, and 3 (9.7%) patients were oligometastatic. Two of the patients with oligometastatic disease had a single site of distant bone metastasis involving the right iliac wing, and one had two bone metastases located in the right acetabulum and T12 vertebral body. All oligometastatic lesions were treated with metastasis-directed therapy using stereotactic radiosurgery/stereotactic body RT, with dosing regimens including 16 Gy in one fraction for spine metastases, 24 Gy in three fractions, or 30 Gy in five fractions, delivered following completion of EBRT to the primary tumor. The contemporary controls had a median age of 67 years (IQR 60–71 years) and a similar median pretreatment PSA of 8.9 ng/mL (IQR 6.2–23.7 ng/mL). Most patients had an ECOG performance status of 0 (97.8%) and 93.3% had stage IIIC disease. There were no patients with oligometastatic disease within the contemporary control cohort, as trimodality therapy was not part of their care pathway at the study institution. The contemporary controls had a higher rate of nodal irradiation (44.4% vs 16.1%) compared with the nivolumab cohort. Median ADT length was 14.6 months (IQR 14.2–15.0 months) in the nivolumab cohort and 14.4 months (IQR 14.0–14.9 months) in the contemporary controls. ADT consisted primarily of leuprolide (96.9%), and the median time from diagnosis to ADT initiation was similar between cohorts (nivolumab cohort: 1.9 months (IQR 0.6–2.2 months); contemporary controls: 1.9 months (IQR 0.9–2.2 months)). Overall, both cohorts represented clinically aggressive populations, although differences in stage distribution and nodal irradiation should be considered when interpreting between-cohort comparisons.

Table 1. Overall cohort characteristics.

	Nivolumab (n=31)	Contemporary control (n=45)	P value
Age (year)	65 (57–69)	67 (60–71)	0.2
Race			0.3
Asian	1 (3.2)	0
Black/African American	2 (6.5)	6 (13.3)
White	28 (90.3)	38 (84.4)
Unknown	0	1 (2.2)
Ethnicity			0.2
Hispanic	2 (6.4)	1 (2.2)
Non-Hispanic	26 (83.9)	43 (95.6)
Unknown	3 (9.7)	1 (2.2)
ECOG performance status			0.3
0	29 (93.5)	44 (97.8)
1	2 (6.5)	1 (2.2)
PSA (ng/mL)	9.7 (7.4–18.9)	8.9 (6.2–23.7)	0.7
Gleason score			0.9
9 (4+5)	21 (67.7)	31 (68.9)
9 (5+4)	8 (25.8)	11 (24.4)
10 (5+5)	2 (6.5)	3 (6.7)
Percent positive cores			0.4
>30–47.5%	3 (9.7)	5 (11.1)
>47.5–65%	9 (29.0)	12 (26.7)
>65–82.5%	9 (29.0)	7 (15.6)
>82.5–100%	10 (32.3)	21 (46.7)
Prostate volume (mm3)	32 (24.0–42.0)	38.4 (30.0–46.5)	0.08
T3+features on imaging			0.01
None	7 (22.6)	23 (51.1)
T3a	8 (25.8)	13 (28.9)
T3b	14 (45.2)	9 (20.0)
T4	2 (6.4)	0
Prognostic stage group			0.03
IIIC (localized)	23 (74.2)	42 (93.3)
IVA (lymph node involvement)	5 (16.1)	3 (6.7)
IVB (distant metastatic disease)	3 (9.7)	0
ADT start from diagnosis, months	1.9 (0.6–2.2)	1.9 (0.9–2.2)	0.8
ADT length			0.4
≤18 months	21 (67.7)	26 (57.8)
>18 months	10 (32.3)	19 (42.2)
Nodal irradiation field			0.0004
No	26 (83.9)	25 (55.6)
Yes	5 (16.1)	20 (44.4)

Data are median (IQR) or n (%). Baseline was defined as the last non-missing assessment collected at time of enrollment.

Significant (<0.05) p value scores are shown in boldface.

ADT, androgen deprivation therapy; ECOG, Eastern Cooperative Oncology Group; PSA, prostate-specific antigen.

Baseline characteristics of the historical controls are summarized in online supplemental table S3. The median age at diagnosis in the historical controls was 69.5 years (IQR 62–74), and the median pretreatment PSA level was 14.5 ng/mL (IQR 7.2–29.6). The majority of patients were classified as prognostic group IIIC (61.5%), reflecting the predominance of GG5 disease within this cohort. Most patients had an ECOG performance status of 0 (65.4%). Intraprostatic disease burden was also high, with 50.0% of patients having >82.5% positive biopsy cores. Median ADT length was 20 months (IQR 14.5–24.0 months) in the historical controls, and the median time from diagnosis to initiation of ADT was 1.8 months (IQR 0.8–2.2). Overall, the historical cohort represented a clinically aggressive matched population meeting the same GG5 and high-volume disease eligibility criteria used to derive the trial benchmark.

Clinical outcomes

The median follow-up of the nivolumab cohort at data lock (May 28, 2023) was 38.8 months (IQR 31.0–46.5 months). The median time to TTT recovery from first ADT injection was 25.2 months (IQR 23.8–28.9 months) and 24.0 months (IQR 21.4–27.9 months) from time of enrollment. The primary endpoint of 2-year FFBR, assuming a binomial distribution, was 90.3% (95% CI 74.3% to 98.0%), which exceeded the prespecified historical control of 75% (one-sided p value from binomial test=0.024). Using KM methods, the 2-year FFBR rate was 90.3% (95% CI 72.9% to 96.8%), while the median FFBR was not reached (figure 2). For post hoc comparison, KM analyses were also performed for the contemporary and historical controls (online supplemental figure S1A,B). In post hoc comparisons, the contemporary controls (HR for biochemical failure 3.59, 95% CI 1.13 to 12.50; p=0.032) and historical controls (HR 3.58, 95% CI 1.24 to 12.30; p=0.029) had higher hazards of biochemical failure than the nivolumab cohort (online supplemental figure S1A,B). In the nivolumab cohort, no patient experienced a local recurrence at any time. A total of six patients (19.3%) met criteria for biochemical recurrence, with three occurring within the first 2 years and three occurring beyond the 2-year primary endpoint. All six patients who had biochemical recurrence also developed distant metastatic disease, resulting in an overall metastasis rate of 19.3% (6/31 patients). Three patients (9.6%) progressed on ADT. Among the three patients with oligometastatic disease, two experienced disease progression, one at 9.4 months from enrollment (while still on ADT) and another at 37 months. The third person with oligometastatic disease, who had two bone metastases at baseline involving the right acetabulum and T12, had no evidence of disease at 60 months of follow-up.

Figure 2. Kaplan-Meier curve of primary endpoint for the study cohort. Freedom from biochemical recurrence was measured from time of enrollment until date of last follow-up recorded.

16 patients (51.6%) were classified as early responders, while the remaining 15 patients (48.4%) were classified as late responders. Additionally, 15 patients (48.4%) achieved a complete response, while 16 (51.6%) patients achieved a partial response. Early responders showed a substantial tumor clearance at the prespecified systematic biopsies at T3, compared with late responders (figure 3A). Additionally, late responders experienced a rapid progression to disease metastasis compared with early responders (figure 3B). Strikingly, all patients with a late response who experienced treatment failure had both lymph node and bone metastases at the time of initial failure; however, early responders with treatment failure had only a single site of metastasis (either bone or lymph node) at the time of initial failure (figure 3B).

Figure 3. Clinical and pathological patient disease outcomes over time per patient. (A) Scatter plot of tumor burden per patient over time. Tumor burden was defined as the percentage of positive biopsies. Patients were followed longitudinally at multiple time points (T0, T1–T2, (T3). T0 is at biopsy, T1–T2 is at fraction 1 of brachytherapy, and T3 is at fraction 2 of brachytherapy. Early responders (blue) consistently demonstrated a more significant reduction in tumor volumes across all time compared with late responders (yellow). This reduction is sustained or further decreased in early responders, while late responders exhibit more fluctuating or minimal changes in tumor volume, often approaching the baseline or above. (B) Swimmer plot of disease response and clinical events. Each bar represents one subject in the study. Diamond (open) indicates subjects with a single organ metastasis. Diamond (filled) indicates subjects with metastasis in multiple organs (bones and lymph nodes). Early responders (blue triangle) generally exhibited longer periods of sustained responses compared with late responders (red triangle). Although the six failures were evenly distributed between both cohorts, late responders had multisite metastasis at failure, while early responders only had a single site. PSA, prostate-specific antigen.

A subset of known immune-related genes was selected for differential mRNA expression analysis. Using the Wilcoxon rank-sum test, CD8A, STING1, LAG3, TNF, and ANPEP were found to be upregulated in early responders at baseline (T0) (p<0.050; figure 4A). In addition, the log₂ gene expression for IFNG (p<0.001) and VISTA (p=0.012) were found to be upregulated in early responders after two cycles of nivolumab (T1) (figure 4B).

Figure 4. Gene expression and signature differences. (A) STING1, LAG3, CD8A, TNF, and ANPEP all had significantly different (p<0.05) expressions based on pathological response (early response vs late response) at baseline, T0. (B) Genes which were not different at baseline but showed significant differences in gene expression as a treatment effect. (C) RIS was significantly higher among patients classified as pathological early responders and predictive for response. (D) RIS was significantly higher among radiographic complete responders. P value significance is represented with stars on the figure: ***p≤0.001, **p≤0.010, *p≤0.050, *p>0.050. ANPEP, alanyl aminopeptidase; AUC, area under the curve; CD8A, cluster of differentiation 8A; LAG3, lymphocyte-activation gene 3; ns, not significant; RIS, Ricketts Immunosuppression Score; STING1, stimulator of interferon genes 1; T0, baseline timepoint; TNF, tumor necrosis factor.

In our comparison between immune-related signature scores available within the Decipher GRID database, RIS showed significant predictive ability to differentiate between early and late pathological response (OR=3.81, p=0.020) (online supplemental figure S2). The RIS gene signature consisted of 13 genes (CD274, IDO1, FASLG, CTLA4, PDCD1, LAG3, HAVCR2, PDCD1LG2, IL10, TNF, TGFB1, IL12A, PTGS2), first derived and validated in Kardos J and Chai S et al JCI Insights 2016²⁰ and later recapitulated in Ricketts et al 2018.²¹ Therefore, RIS was investigated as a potential biomarker to determine which patients may derive greater benefit from treatment intensification with anti-programmed cell death protein 1 therapy. In our analysis among samples with intermediate to high Decipher score (n=30), the RIS was significantly higher among early responders (p<0.010) with an AUC of 0.80 (95% CI 0.64 to 0.96, sensitivity=0.6, specificity=0.87; figure 4C). RIS was also significantly higher among patients with a complete radiographic response (p<0.01) with an AUC of 0.78 (95% CI 0.61 to 0.96, sensitivity=0.69, specificity=0.86; figure 4D). Multivariable analysis revealed that a low RIS was a significant predictor of a metastatic event (HR 9.256, 95% CI 1.061 to 80.719; p=0.044) when controlled for the response group (early response vs late response), ADT treatment length (>18 months vs ≤18 months), stage, and Gleason score (table 2) for patients treated with SOC plus nivolumab.

Table 2. Time to metastasis using multivariate Cox-regression.

Variable	Unadjusted (UVA)			Adjusted (MVA)
	HR	95% CI	P value	HR	95% CI	P value
Gleason Score
9	1	–	–	1	–	–
10	3.373	0.388 to 29.338	0.27	6.189	0.366 to 104.760	0.2
Stage
IIIC	1	–	–	1	–	–
IVA	4.181	0.339 to 51.518	0.26	5.411	0.295 to 99.348	0.25
IVB	6.856	1.134 to 41.442	0.03	9.702	0.840 to 112.038	0.06
ADT length
≤18 months	1	–	–	1	–	–
>18 months	2.225	0.445 to 11.116	0.33	2.664	0.280 to 25.301	0.39
Responder group
Early	1	–	–	1	–	–
Late	1.651	0.324 to 8.408	0.55	0.391	0.048 to 3.195	0.38
Ricketts risk group
High	1	–	–	–	–	–
Low	4.04	0.783 to 20.844	0.09	9.256	1.061 to 80.719	0.04

P values <0.05 are bold for significance.

ADT, androgen deprivation therapy; MVA, multivariable analysis; UVA, univariable analysis.

Toxicity outcomes

Definitive and probable nivolumab-related toxicity data are shown in online supplemental table S4. Two patients experienced grade 3 dose-limiting toxicities, including autoimmune hepatitis and cardiac QT prolongation. There were no probable or definitive grade 4+ toxicity nor any late toxicity related to nivolumab at any time point. Treatment-related toxicity, including events considered possibly, probably, or definitively related to nivolumab, ADT, HDRBT, or EBRT, is summarized in online supplemental table S5. All patients experienced some form of acute toxicity, with 53.1% experiencing grade 2 events, 18.8% experiencing grade 3 events, and one patient (3.1%) experiencing a grade 4 event (hyperkalemia). No acute grade 5 events were observed. Late toxicities were less common, occurring in 62.5% of patients at grade 1 and 12.5% at grade 2 severity. No grade 3 or higher late events were observed. Most late effects involved the genitourinary, gastrointestinal, or metabolic systems.

Discussion

GG5 PCa is biologically distinct from lower grade groups and is associated with poorer outcomes and greater treatment resistance.^{4 22} Improvements in GG5 outcomes have been achieved with BT boost-based intensification, likely through improved local control and reduced metastatic dissemination.1623,25 We sought to build on this strategy by integrating nivolumab during the immunostimulatory window created by ablative-dose HDRBT and ADT. Preclinical and early clinical data support this approach, and in the present study RIS was associated with both pathologic and radiographic response.^{12 15 21}

Improvements in GG5 PCa outcomes have been seen with the addition of a BT boost to EBRT.^{16 23} Adding a BT boost improves 5-year freedom from distant metastases by approximately 22% (89% vs 67%; p<0.05).²⁴ BT is thought to improve distant metastasis-free survival in localized disease through improving local control and thus limiting the potential for systemic spread.²⁵ We sought to expand on the success of trimodality therapy for men with GG5 PCa by taking advantage of the immunostimulatory window generated by the ablative BT boost to overcome PCa’s ICI resistance. To accomplish this, enrolled patients were given two cycles of nivolumab before their first BT boost fraction to ensure proper immune priming.

Ablative-dose RT helps to modulate the immunosuppressive TME by increasing CD8+ T cell infiltration and reducing myeloid-derived suppressor-cell levels.²⁶ The phase I portion of this study provided some of the earliest clinical data supporting the combination of ablative RT (via HDRBT) with ICI therapy in GG5 PCa.¹⁵ In the present study, RIS was associated with both 1 month post-HDRBT pathologic response and 3-month radiographic response and characterizes a tumor enriched for immune gene signatures while simultaneously expressing immune checkpoint molecules.^{20 21}

Contemporary approaches to treatment intensification increasingly involve the addition of novel androgen-targeted therapies. The Systemic Therapy in Advancing or Metastatic Prostate cancer: Evaluation of Drug Efficacy (STAMPEDE) meta-analysis demonstrated improved outcomes for patients with high risk or node-positive PCa treated with long-term ADT and 2 years of abiraterone in combination with EBRT alone, without BT.²⁷ In that study, the 6-year metastasis-free survival was 82% in the abiraterone arm compared with 69% with ADT alone.²⁷ While STAMPEDE’s regimen avoids procedural intervention, it requires prolonged cytostatic therapy associated with potential long-term metabolic and cardiovascular toxicity, as well as hormonal side effects that may impact quality of life.²⁸ In contrast, the addition of nivolumab during ablative radiation to the primary tumor leverages dose heterogeneity and precise timing to achieve therapeutic synergy through the unique immunostimulatory effects of high-dose distributions within the TME.¹² This approach requires an invasive procedure and more upfront treatment intensification but delivers ablative radiation doses that may enhance local control and stimulate immune responsiveness for effective durable systemic disease control. These two strategies reflect distinct treatment philosophies, with one focused on prolonged systemic suppression and the other on enhanced systemic immune activation with ablative local therapy, and they may ultimately be complementary rather than mutually exclusive. Our findings suggest that immuno-radiotherapeutic intensification may represent a viable alternative in selected patients and underscore the need for future trials.

The identification of patients with PCa who derive the most benefit from therapeutic escalation remains a critical challenge. Decipher has been widely used as a prognostic classifier for biochemical recurrence and metastasis risk,^{29 30} and the Post-Operative Radiation Therapy Outcomes Score (PORTOS) has demonstrated predictive value in postoperative and definitive radiation settings.³¹ Patients with high PORTOS scores in the SAKK 09/10 and RTOG 0126 trials experienced significantly improved biochemical control with dose escalation, whereas those with low scores did not, underscoring the need for individualized radiation dosing strategies.³¹ RIS builds on this framework by incorporating immune-related gene expression, distinguishing itself as a predictor of treatment response in trimodality SOC with the addition of ICI therapy.

In this study, high RIS was independently associated with early pathological and radiographic response. Patients with low RIS scores were also found to be at greater risk of metastatic disease progression with ICI added to SOC trimodality therapy. The paradox of immune-enriched tumors exhibiting concurrent immune escape mechanisms suggests that RIS could help identify patients that would have a meaningful improvement in disease control with checkpoint inhibition to overcome resistance. Future prospective studies should validate RIS in independent cohorts and explore its integration with existing genomic and clinical predictors to optimize patient selection for novel therapeutic strategies in high-risk PCa.

Interpretation of these findings should also account for the unusually adverse-risk population enrolled in this trial. In addition to GG5 histology, protocol eligibility required high-volume biopsy involvement (≥30% positive cores), and the study permitted stage IVA and selected oligometastatic stage IVB disease. Positive biopsy core burden independently predicts inferior biochemical control, metastatic failure, and PCa-specific mortality after definitive radiotherapy.32,34 Half of our historical cohort had >82.5% positive biopsy cores which may help explain their high rate of 2-year FFBR. Consistent with this, a recent multi-institutional GG5-only series treated with contemporary high-dose EBRT and long-term ADT reported a 5-year biochemical recurrence free survival (bRFS) of 73.1% despite modern treatment intensification.³⁵ These features help contextualize why matched institutional comparator cohorts may be more informative for this study than broader high-risk trial populations.

Despite the promising findings of this phase II trial, several limitations must be acknowledged. First, the study was non-randomized and single-arm, which limits the ability to directly compare the efficacy of nivolumab in combination with SOC therapy against SOC alone. Without a randomized control group, it is difficult to ascertain whether the observed improvement in FFBR was entirely attributable to the addition of nivolumab or whether other confounding factors contributed to the outcomes. Although comparisons with contemporary and historical controls were performed to provide context, these analyses were post hoc and cannot substitute for randomized comparisons. Additionally, the study was powered at 60%, which is below conventional standards; however, this was an intentional design choice given the exploratory nature of the trial, the anticipated large effect size, and the ethical considerations of introducing a novel immune-radiotherapy regimen. The aim was to detect a meaningful signal of clinical activity with a modest sample size while minimizing patient risk. Importantly, the observed 2-year FFBR rate of 90.3% exceeded the prespecified threshold of 75% and reached statistical significance (p=0.024), supporting potential clinical activity in this setting. Given this result, the study provides useful data to inform future randomized investigations. That said, the relatively small sample size (n=31 evaluable patients) limits the generalizability of the findings, and the study population, drawn from a single tertiary cancer center, may introduce referral bias. The relatively short median follow-up of 38.8 months, while sufficient to evaluate 2-year FFBR, does not provide long-term data on overall survival, late toxicities, or the durability of treatment response. Although RIS emerged as a potential predictive biomarker, these findings should be interpreted with caution given the exploratory nature of the analysis, small sample size, and lack of both internal and external validation. As such, the reported ROC metrics likely reflect apparent performance and may overestimate true predictive accuracy. Further validation in larger, independent studies is necessary to confirm the predictive utility of the RIS and its potential integration into treatment decision-making. Importantly, the trial was designed and initiated prior to the emergence of data from STAMPEDE²⁸ and Prostate-Only Versus Whole-Pelvic Radiation Therapy in High-Risk and Very High-Risk Prostate Cancer³⁶, which have since reshaped treatment standards for high-risk and node-positive disease by supporting androgen receptor pathway inhibitor intensification, longer durations of ADT, and elective nodal irradiation in selected patients. As such, our approach does not incorporate these evolving standards, and the role of ICI therapy within the current treatment landscape remains to be further explored.

Despite its limitations, this study provides prospective clinical evidence supporting the feasibility and potential clinical activity of integrating nivolumab with HDRBT and ADT followed by EBRT in GG5 PCa, a population with historically poor outcomes. The regimen’s favorable safety profile suggests that ICIs can be feasibly incorporated into existing SOC therapies, warranting further investigation in larger, randomized trials. This study also identified the RIS as a potential predictive biomarker, offering a path toward more personalized treatment strategies in aggressive PCa.

Conclusion

Adding nivolumab to SOC for GG5 PCa was associated with encouraging FFBR outcomes without significant added toxicity. Additionally, RIS was identified as a potential biomarker that may help predict favorable treatment response, contributing to ongoing efforts to optimize patient selection for ICI-based approaches. Our findings support further investigation in a larger randomized phase II/III study appropriately powered to compare SOC trimodality therapy with and without nivolumab in GG5 PCa.

Data availability statement

No data are available.