Introduction to mHSPC – What Are Relevant Prognostic/Predictive Factors for the Management of Patients? "Presentation" - Matthew Smith
November 15, 2024
At the 2024 Advanced Prostate Cancer Consensus Conference (APCCC), Matthew Smith presents an overview of prognostic and predictive biomarkers in metastatic castration-sensitive prostate cancer, examining established factors like disease patterns and volume while highlighting the Decipher score's prognostic value.
Biographies:
Matthew Smith, MD, PhD, Oncologist, Professor of Medicine, Harvard Medical School, Hematology/Oncology, Massachusetts General Hospital, Boston, MA
Biographies:
Matthew Smith, MD, PhD, Oncologist, Professor of Medicine, Harvard Medical School, Hematology/Oncology, Massachusetts General Hospital, Boston, MA
Read the Full Video Transcript
Matthew Smith: Thank you, Silke and Aurelius, as always for organizing an absolutely outstanding meeting. Well done. I'd also like to thank Chris Sweeney and Gert Attard for some very useful conversations as I prepared my presentation today. So here are my aims. A few caveats. I'll be focusing on mCSPC as instructed and overall survival. Other disease states and clinical outcomes may be considered in future sessions. Further, I was asked to provide an introduction. This is not a comprehensive review. So if I don't mention your favorite reference gene or signature, please don't be offended.
So let me begin with prognostic markers. This is a qualitative example of a prognostic biomarker on the cartoon on the left. A definition of a prognostic biomarker requires a significant association between a biomarker and an outcome, in this case overall survival, and there are many established prognostic biomarkers in mCSPC. This is a landmark study from Nick James showing the outcomes with ADT alone in the STAMPEDE trial. As you all know, poor clinical outcomes with the median overall survival of only 42 months, with survival significantly associated with anatomic sites of disease, patients with soft tissue metastases doing best, those with bone only and bone and soft tissue doing significantly worse.
Overall, survival is also strongly associated with disease volume and disease presentation. This is very nice work from Francini and Chris Sweeney, published in 2018, showing markedly different outcomes by volume of disease and manner of presentation. Patients with de novo high volume metastases, for example, having a median overall survival of only 43 months with ADT alone. Those with recurrent low volume disease having median overall survival of twice that at 92 months. So this is all kind of familiar review, but as part of the requested introduction I'm describing this.
The mutational landscape of mCSPC is complex. And mutations in cancer-relevant genes are correlated directly across the spectrum of castration-sensitive prostate cancer. The prevalence rates for various commonly altered genes vary by manner of presentation and extent of disease, as shown here, for example, in de novo metastatic disease, mutation rates for p53 and PTEN are 38 and 34%—more than twice the observed rates in biochemical-only recurrent disease.
In the same study, the investigators reported that TP53 mutations were significantly associated with overall survival by disease state. Those with p53 mutations having significantly worse overall survival whether they had oligometastatic disease or multiple sites of metastases at presentation. This is work by Hamid and Chris Sweeney. Again, these are genomic alterations in tumor suppressor genes, showing that genomic alterations in TP53, PTEN, and/or RB were associated with significantly worse overall survival, and that survival was particularly worse for patients who had alterations in two or more of those tumor suppressor genes.
This is really extraordinary work that was mentioned earlier. This is a transcriptome-wide analysis from the abiraterone arms of the STAMPEDE study. This is in preprint form reported by Gerhardt Attard. If you haven't looked at it, I encourage you to do so. It's really a remarkable body of work here showing the results of testing from 57 pre-specified signatures in their transcriptome-wide analysis, those shown in the pink and red dots, those being significantly associated with survival. Decipher scores performed particularly well with significant prognostic association in metastatic disease and significant association with metastasis-free survival for patients with localized high-risk disease.
Here's data from the same analysis showing the marked separation of patients by decipher score. Those with scores greater than the median having a particularly poor prognosis, with a median overall survival of less than 36 months compared to patients with low decipher scores having survival twice that long. Now, while there are many prognostic markers, it's quite a different story for predictive biomarkers. And in a way, it was really a trick assignment I was given by Silke and Aurelius. This is shown on the left as an example of a predictive biomarker.
Definition of a predictive biomarker requires data from a randomized controlled trial and a significant interaction between biomarker and a treatment effect. And cut to the chase, at the present time, there are no rigorously defined predictive biomarkers in mCSPC. But what I will describe are some provocative early studies looking at what appeared to be meaningful associations.
So this is really a tour de force work looking at a meta-analysis of the various randomized controlled trials in mCSPC, forest plots for the various, in this case, volumes of disease in different doublet treatment combinations, making the contrast in the treatment effect observed by high and low volume disease for ARPI in docetaxel. So as you all know, most of the treatment benefit with docetaxel appears to be seen in patients with high volume disease. Here you see a hazard ratio of 0.73, falls off quite sharply to 0.91 in patients with low volume metastases in this meta-analysis.
Very different story with ARPIs, where you see a consistent treatment effect for both high volume and low volume disease, with hazard ratios of 0.66 and 0.58, respectively. A similar story is told when you look at the manner of presentation. Again, ARPI treatment effect appears to be preserved across de novo and recurrent metastatic disease, with hazard ratios of 0.65 and 0.61, respectively. In contrast, with docetaxel, most of the benefit appears to be conferred to patients with de novo metastatic disease, with a hazard ratio of 0.78 compared to that of 0.90 for those with recurrent metastatic disease.
This is data from the CHAARTED study. Very nice analysis by Hamid et al reported looking at different signatures and the treatment effect with docetaxel. You see there's a difference between basal signature and luminal B signature, with much greater treatment effect seen in luminal B, although this did not meet criteria for statistical significance as a predictive biomarker. Similar provocative results were seen with a decipher score analysis showing that most of the treatment effect is conferred in patients with higher decipher scores, the higher quartiles here on the two panels on the right.
This is a figure you saw earlier today by Dr. Spratt, showing a difference in overall survival by addition of abiraterone to ADT in the STAMPEDE abiraterone arms by decipher score, showing that most of the benefit over time is seen in patients with high decipher scores. Although again, this just missed reaching statistical significance, so it can't formally be called a predictive biomarker.
So in summary, there are a number of well-established prognostic biomarkers in prostate cancer. Disease presentation, anatomic pattern of metastases, and disease volume are all associated with overall survival in mCSPC. It should come as no surprise to anyone in this audience. There are a number of single gene alterations, including PTEN and p53, that are associated with shorter overall survival, and a number of signatures, including higher decipher scores, are also associated with shorter survival. At the present time, there are no robustly defined predictive biomarkers in mCSPC. Thank you.
Matthew Smith: Thank you, Silke and Aurelius, as always for organizing an absolutely outstanding meeting. Well done. I'd also like to thank Chris Sweeney and Gert Attard for some very useful conversations as I prepared my presentation today. So here are my aims. A few caveats. I'll be focusing on mCSPC as instructed and overall survival. Other disease states and clinical outcomes may be considered in future sessions. Further, I was asked to provide an introduction. This is not a comprehensive review. So if I don't mention your favorite reference gene or signature, please don't be offended.
So let me begin with prognostic markers. This is a qualitative example of a prognostic biomarker on the cartoon on the left. A definition of a prognostic biomarker requires a significant association between a biomarker and an outcome, in this case overall survival, and there are many established prognostic biomarkers in mCSPC. This is a landmark study from Nick James showing the outcomes with ADT alone in the STAMPEDE trial. As you all know, poor clinical outcomes with the median overall survival of only 42 months, with survival significantly associated with anatomic sites of disease, patients with soft tissue metastases doing best, those with bone only and bone and soft tissue doing significantly worse.
Overall, survival is also strongly associated with disease volume and disease presentation. This is very nice work from Francini and Chris Sweeney, published in 2018, showing markedly different outcomes by volume of disease and manner of presentation. Patients with de novo high volume metastases, for example, having a median overall survival of only 43 months with ADT alone. Those with recurrent low volume disease having median overall survival of twice that at 92 months. So this is all kind of familiar review, but as part of the requested introduction I'm describing this.
The mutational landscape of mCSPC is complex. And mutations in cancer-relevant genes are correlated directly across the spectrum of castration-sensitive prostate cancer. The prevalence rates for various commonly altered genes vary by manner of presentation and extent of disease, as shown here, for example, in de novo metastatic disease, mutation rates for p53 and PTEN are 38 and 34%—more than twice the observed rates in biochemical-only recurrent disease.
In the same study, the investigators reported that TP53 mutations were significantly associated with overall survival by disease state. Those with p53 mutations having significantly worse overall survival whether they had oligometastatic disease or multiple sites of metastases at presentation. This is work by Hamid and Chris Sweeney. Again, these are genomic alterations in tumor suppressor genes, showing that genomic alterations in TP53, PTEN, and/or RB were associated with significantly worse overall survival, and that survival was particularly worse for patients who had alterations in two or more of those tumor suppressor genes.
This is really extraordinary work that was mentioned earlier. This is a transcriptome-wide analysis from the abiraterone arms of the STAMPEDE study. This is in preprint form reported by Gerhardt Attard. If you haven't looked at it, I encourage you to do so. It's really a remarkable body of work here showing the results of testing from 57 pre-specified signatures in their transcriptome-wide analysis, those shown in the pink and red dots, those being significantly associated with survival. Decipher scores performed particularly well with significant prognostic association in metastatic disease and significant association with metastasis-free survival for patients with localized high-risk disease.
Here's data from the same analysis showing the marked separation of patients by decipher score. Those with scores greater than the median having a particularly poor prognosis, with a median overall survival of less than 36 months compared to patients with low decipher scores having survival twice that long. Now, while there are many prognostic markers, it's quite a different story for predictive biomarkers. And in a way, it was really a trick assignment I was given by Silke and Aurelius. This is shown on the left as an example of a predictive biomarker.
Definition of a predictive biomarker requires data from a randomized controlled trial and a significant interaction between biomarker and a treatment effect. And cut to the chase, at the present time, there are no rigorously defined predictive biomarkers in mCSPC. But what I will describe are some provocative early studies looking at what appeared to be meaningful associations.
So this is really a tour de force work looking at a meta-analysis of the various randomized controlled trials in mCSPC, forest plots for the various, in this case, volumes of disease in different doublet treatment combinations, making the contrast in the treatment effect observed by high and low volume disease for ARPI in docetaxel. So as you all know, most of the treatment benefit with docetaxel appears to be seen in patients with high volume disease. Here you see a hazard ratio of 0.73, falls off quite sharply to 0.91 in patients with low volume metastases in this meta-analysis.
Very different story with ARPIs, where you see a consistent treatment effect for both high volume and low volume disease, with hazard ratios of 0.66 and 0.58, respectively. A similar story is told when you look at the manner of presentation. Again, ARPI treatment effect appears to be preserved across de novo and recurrent metastatic disease, with hazard ratios of 0.65 and 0.61, respectively. In contrast, with docetaxel, most of the benefit appears to be conferred to patients with de novo metastatic disease, with a hazard ratio of 0.78 compared to that of 0.90 for those with recurrent metastatic disease.
This is data from the CHAARTED study. Very nice analysis by Hamid et al reported looking at different signatures and the treatment effect with docetaxel. You see there's a difference between basal signature and luminal B signature, with much greater treatment effect seen in luminal B, although this did not meet criteria for statistical significance as a predictive biomarker. Similar provocative results were seen with a decipher score analysis showing that most of the treatment effect is conferred in patients with higher decipher scores, the higher quartiles here on the two panels on the right.
This is a figure you saw earlier today by Dr. Spratt, showing a difference in overall survival by addition of abiraterone to ADT in the STAMPEDE abiraterone arms by decipher score, showing that most of the benefit over time is seen in patients with high decipher scores. Although again, this just missed reaching statistical significance, so it can't formally be called a predictive biomarker.
So in summary, there are a number of well-established prognostic biomarkers in prostate cancer. Disease presentation, anatomic pattern of metastases, and disease volume are all associated with overall survival in mCSPC. It should come as no surprise to anyone in this audience. There are a number of single gene alterations, including PTEN and p53, that are associated with shorter overall survival, and a number of signatures, including higher decipher scores, are also associated with shorter survival. At the present time, there are no robustly defined predictive biomarkers in mCSPC. Thank you.