Video
Improving Management of Prostate Cancer: The Decipher Prostate Genomic Classifier
Author(s):
Focused discussion on the Decipher Prostate Genomic Classifier and real-world data from the American Urological Association (AUA) Annual Meeting in the setting of prostate cancer management.
Transcript:
Michael S. Leapman, MD, MHS: The Decipher Prostate Genomic Classifier is a tool that can be used to help risk-stratify patients with prostate cancer. It is offered at 2 different junctures. One is for patients who’ve had a prostate biopsy, and we’re trying to learn more about the characteristics of their cancer and the potential risk. The other is among patients who’ve been treated with radical prostatectomy, and we’re also trying to understand their prognosis and potential outcome. I would say that it runs the spectrum. I typically use genomic classifiers for patients who are considering active surveillance or observational treatment at first and want to know if they may be a candidate or if they may harbor more aggressive characteristics in their cancer. For patients after radical therapy, we will use tools like Decipher to help make decisions about how patients should be monitored, or whether early salvage radiation should be offered for patients who have unfavorable pathology. The Decipher classifier assesses 22 features associated with prostate cancer aggressiveness.
In this project, we sought to understand the clinical decision-making and the clinical outcomes that occur for patients who were tested with a genomic classifier in the real-world setting. Something important to understand about the Decipher classifier and most risk-stratification tools is that they are typically developed with patients who’ve already been treated for prostate cancer in archival tissue, tissue that’s been accessed from banks and libraries where we follow the outcome over time and try to understand what may happen. Now that these tools have entered the real world, a very important question is understanding what happens and how do these tools perform in this real-world setting. The goal of this project was to, first and foremost, build a real-world data linkage to the genomic classifier. What this entails is taking the comprehensive records for all patients who received Decipher testing in the commercial setting, in real-life practice, and try to link them with data records to understand what happened to their long-term outcomes after treatment or after testing. This linkage was accomplished through a secure platform where data was securely exchanged between 2 parties, and we ultimately were able to link over 92,000 patients who had Decipher testing and had some clinical follow-up to help understand what happened afterward. Through this project, we validated that we were able to identify key events relevant to prostate cancer, including dates of diagnosis, dates of biopsy, dates of treatment, and importantly, a recurrence if it did happen, or the development of metastasis of prostate cancer if that occurred.
What we found in this study is that the Decipher classifier was independently associated with risks of biochemical recurrence and metastasis among patients treated with radical prostatectomy. This relationship remains significant when adjusting for relevant clinical and pathologic features, including a patient’s Gleason score, surgical margin status, stage, and nodal status as well.
I do think that the findings in my clinical practice, in my experience, do ring true with what we found in this study, that there is a wide distribution of genomic classifier results in the real-world setting. We frequently make changes to our management based on these findings, usually the intensity of how we monitor patients, as well as the intensity with which we will deliver treatment. I do think that these findings line up and make sense. It’s also reassuring to see that these prognostic associations from which we base most of our confidence in this test also hold up in the real-world setting, where patients and providers are aware of these results and modify their treatment based on that.
I think these data are an important confirmation that we can dig deeper and provide a more refined risk assessment beyond some of the conventional clinical variables that we identify in a pathology report. It also is a very important leap forward in matching clinical and genomic data on a large scale. I think this will be an exciting area for future work.
Transcript edited for clarity.
