A Comprehensive Look at Genomic Classifier Performance in Intermediate Risk Prostate Cancer: Insights from the NRG Oncology/RTOG 01-26 Trial, Journal Club - Rashid Sayyid & Zachary Klaassen

August 22, 2023

Rashid Sayyid and Zach Klaassen discuss the publication on genomic classifier performance in intermediate-risk prostate cancer, focusing on the results of the NRG Oncology/RTOG 01-26 randomized Phase 3 trial. The study explores the use of the Decipher 22-gene genomic classifier in risk stratification, outperforming other common risk classification systems for prostatectomy and pre-treatment biopsy samples. The study validates the Decipher test in a Phase 3 trial using pre-treatment biopsy samples in men with intermediate-risk disease planned for radiation without ADT. The results demonstrated that the genomic classifier was independently prognostic for several oncological endpoints, even after accounting for factors such as Gleason Score. In conclusion, the Decipher genomic classifier is presented as the only gene expression biomarker with randomized Phase 3 validation in men with intermediate-risk prostate cancer, aiding in personalized decision-making.


Rashid Sayyid, MD, MSc, Urologic Oncology Fellow, Division of Urology, University of Toronto, Toronto, Ontario

Zachary Klaassen, MD, MSc, Urologic Oncologist, Assistant Professor Surgery/Urology at the Medical College of Georgia at Augusta University, Georgia Cancer Center, Augusta, GA

Read the Full Video Transcript

Rashid Sayyid: Hello, everyone. This is Rashid Sayyid. I'm a urologic oncology fellow at the University of Toronto and along with Zach Klaassen, Associate Professor and Program Director at Augusta University, we'll be discussing the recent publication looking at genomic classifier performance in intermediate risk prostate cancer, specifically results from the recent NRG Oncology/RTOG 01-26 randomized Phase 3 trial. This study was recently published by Spratt et al. in the International Journal of Radiation Oncology, Biology, and Physics.

We know that intermediate risk prostate cancer remains a heterogeneous patient population, and the currently available clinical pathologic variables have modest discriminatory ability. This has important implications for treatment selection, particularly in the radiation therapy cohort, whether we give ADT or not. So there's been a lot of efforts in trying to risk stratify this population. In 2018, Spratt et al. also demonstrated that a combinatory 22-gene genomic classifier, Decipher. They performed a clinical risk stratification system using this test and demonstrated that it outperforms each of the NCCN and the CAPRA risk classification systems using tissue of prostatectomy and pre-treatment biopsy samples.

Clearly this was the first signal that this test can perform commonly available tools to risk stratify these patients. Next Berlin et al. demonstrated that Decipher can also be used to risk stratify intermediate risk patients undergoing radiation therapy using a prospective registry. They demonstrated that a higher genomic classifier score or a higher Decipher score, but not the NCCN risk group, was associated with increased biochemical failure rates, increased rates of distant metastases. When you combined the genomic classifier, Decipher with the NCCN model, it outperformed each of the Decipher-only or the NCCN-only models significantly for predicting these outcomes. As such, there's been a push to combine these genomic tests with currently available tools such as the NCCN or the CAPRA risk stratification system to risk stratify these patients. But however, the Decipher test in this setting has not yet been validated in a Phase 3 trial.

For this reason, the purpose of this study was to validate Decipher in a Phase 3 trial using pre-treatment biopsy samples in men with intermediate-risk disease planned for radiation without ADT. For the purposes of this study, the authors accessed archival diagnostic biopsy specimens from men treated in the NRG/RTOG 01-26, which is a Phase 3 trial of intermediate-risk patients randomized to receive 3D conformal or intensity modulated external beam radiotherapy without ADT at a dose of 70.2 versus 79.2 Gy. This study was initially reported in 2018 with an eight-year median follow-up, but has been since extended to 12 years for the intents of this analysis. The investigators performed a central pathology review, biopsy samples, and then the biopsy specimen with the highest grade tumor focus was selected for microdissection and then subsequent RNA extraction, complementary DNA amplification, oligonucleotide microarray hybridization, and microarray quality control.

The Decipher scores were calculated based on a locked model, and the scores generated were from a scale of zero to one. The genomic classifier scores were categorized as follows: into low risk if it was less than 0.27 and then intermediate 0.27-0.4, and higher risk greater than 0.4. So you may wonder why were lower cutoffs used than what is typically used for Decipher? Typically, with Decipher as opposed to 0.27 and 0.4, we use 0.45 and 0.6 to define the different risk groups. But with a longer age, the older the tissue gets, inherently the lower the score gets. To account for the fact that tissue from the sample is 13, 20 years old, the authors adjusted for this with the lower cut-off scores that we see in this analysis.

The primary objective of this analysis was to evaluate the independent association of the Decipher score with oncologic outcomes on a multi-variable analysis, whereas the secondary objective was to explore whether there was any association and interactions between the genomic classifier and the radiation treatment being given. Again, it was 70.2 versus 79.2 in this trial. The primary endpoint was time to disease progression, which was defined as either biochemical failure, local failure, distant metastasis, prostate cancer related mortality or receipt of salvage therapy. Then they also assessed secondary endpoints, which included time to biochemical failure or time to distant metastasis, time to prostate cancer mortality, time to receipt of salvage therapy, metastasis-free survival, overall survival. The primary endpoint was a composite endpoint, whereas the secondary endpoint was the individual endpoints alone.

Analysis by treatment arm was performed using the intent-to-treat principle, and continuous and categorical variables were compared using the Wilcoxon ranks on the Chi square tests. The authors performed survival analysis given this was a time-to-event analysis. So they looked at Kaplan-Meier curves with Cox modeling, and then they also performed competing risks analysis, which differs from the standard survival analysis in that it accounts for competing risks from other causes such as other-cause death. Whereas other-cause death is censored in the Kaplan-Meier curves with the Fine and Gray models and cumulative incidence function curves, you account for this competing risk which precluded your event of interest from occurring.

The multi-variable models, regardless of whether it was Cox or it was the Fine and Gray model competing risk, were adjusted for the clinical risk groups using Gleason Score and PSA. The treatment modality received, 3D conformal radiotherapy versus intensity modulated or randomization arm, meaning the dose of radiotherapy received. Then the interaction between the randomization arm and the GC risk group was also analyzed with the randomization arm, the GC risk group combined to estimate a 10-year survival rate in this study. At this point, I'll turn it over to Zach to go over the results and discussion.

Zach Klaassen: Thanks so much, Rashid, for that comprehensive introduction. This is the CONSORT diagram for this study. As you can see here, patients that received allocated treatment in this trial included 1,449. There were 847 unique patient samples available, subsequently 449 unique patient samples with genomic classifier scores available, and finally, 215 unique patient samples with genomic classifier scores that passed quality control. We can see in table one, this is the delineation between these patients. There were 107 patients that received 70.2 Gy compared to 108 patients that received 79.2 Gy radiation.

We'll go through the highlights of the demographics and clinical characteristics in this analytical cohort. Median age of these patients was 70, roughly 88% were white, and 7% were African-American. In terms of performance status, most of these patients had excellent performance status, 92% with a Zubrod performance status of zero. The median PSA at study entry was 7.2. The most common Gleason Score was Gleason 3+4, 61%. Second most common was Gleason 4+3 at 24%, and 14% of patients had Gleason 3+3 disease. In terms of T stage, a 50-50 split between T1 and T2, number of NCCN intermediate risk features, most commonly one at 73%. Roughly two-thirds of these patients received 3D conformal radiotherapy and one-third IMRT. The median follow-up for these patients for overall survival was quite long at 12.8 years median follow-up.

This is the distribution of genomic classifier by randomization arm. We can see the median genomic classifier score in the entire cohort was 0.26 with no difference in genomic classifier score by randomization arm. This is the delineation by T stage. We do see some heterogeneity in genomic classifier score, especially when we look at T2a, T2b versus T1b and T1c, with higher scores in the higher T stage tumors.

This is the delineation by Gleason pattern. There's a slight increase as we increase Gleason Score for the genomic classifier score, but this was not statistically significant in this study. This is by PSA at study entry. Some heterogeneity here but not statistically significant with regards to different PSA study levels and the corresponding genomic classifier scores.

This is the main results page here showing prognostic performance of the genomic classifier score in multi-variable Fine-Gray competing risks analysis or Cox proportional hazard models for all of these endpoints that the authors assessed. We see here in the box that every single one of these outcomes was significantly worse with increasing genomic classifier score, so this is per 0.1 unit increase in score, we see the corresponding hazard ratios. For biochemical failure, either by the Phoenix or ASTRO definition, we see worse outcomes with increasing genomic classifier score; receipt of salvage therapy hazard ratio, 1.17 with increasing score; disease progression hazard ratio, 1.12; distant metastasis, 1.28. The highest predictor prostate cancer specific mortality, 1.45 hazard ratio, although only 11 events in this analysis. We also see increasing genomic classifier score associated with worse metastasis-free and overall survival. These were all statistically significant outcomes.

This looks at the interaction plot for genomic classifier risk group and randomization arm for the metastasis-free survival endpoint. If we look at the right here, this is genomic classifier, intermediate and high, and we see that these patients' dose escalations, so 79.2 versus 70.2 Gy showed a greater absolute benefit with a 10-year metastasis-free survival of 75% for those getting escalated compared to 54% for standard dose. This tells us that for these genomic classifier intermediate/high patients, they did have a benefit with regards to MFS if they received 79.2 Gy of radiation.

This study represents the first validation of any tissue-based expression biomarker in the context of a Phase 3 RCT from pre-treatment biopsy tissue in intermediate-risk localized prostate cancer. The Decipher genomic classifier was independently prognostic on multi-variable analysis for all oncological endpoints assessed, even after accounting for highly prognostic factors such as Gleason Score. The study showed that patients with low genomic classifier scores had very low rates of death from prostate cancer at 4% with extended follow-up.

On the other hand, patients with high genomic classifier scores had unacceptably high rates of disease progression at more than 60% and distant metastasis at more than 15% with radiotherapy alone. Previous work from Dr. Spratt's group and the MARCAP Consortium meta-analysis showed there's a 40% relative reduction in 10-year distant metastasis from the addition of short-term ADT to radiotherapy, with a hazard ratio of 0.60. Thus, the 10-year rate of distant metastasis of patients enrolled in RTOG 01-26 was 4% for genomic classifier low patients, and a relative reduction of 40% would translate into a 1.6% absolute reduction in 10-year distant metastasis rate.

In conclusion, the Decipher 22-gene genomic classifier represents the only gene expression biomarker with randomized Phase 3 validation in men with intermediate risk prostate cancer. The genomic classifier is independently prognostic for disease progression, biochemical recurrence, distant metastasis, metastasis-free survival, and prostate cancer specific mortality on multi-variable analysis treated with radiation therapy alone in the NRG/RTOG 01-26 trial.

Finally, men with low genomic classifier scores had low rates of metastatic progression or death from prostate cancer with long-term follow-up, and the use of the genomic classifier can assist with more accurate estimates of absolute benefit from treatment intensification for better personalized shared decision-making. We thank you very much for your attention. We hope you enjoyed this UroToday Journal Club discussion.