Investigating the Relationship Between MRI Findings, Genomic Markers, and Gleason Score - Eric Kim
May 10, 2024
Eric Kim discusses the variation in prostate cancer genomic subtypes related to MRI PI-RADS scores and racial differences. He focused on the correlations between mpMRI findings and the Decipher genomic classifier in a racially diverse cohort of men. The study revealed that MRI and Decipher scores generally correlated well, especially in non-African Americans, but this correlation was weaker among African American men. High PI-RADS scores were associated with more aggressive prostate cancer subtypes, a trend less pronounced in African Americans. The study, emphasizing the need for personalized treatment, also addresses the underrepresentation of African-American men in prostate cancer research, suggesting that existing models are based predominantly on data from Caucasian men.
Biographies:
Eric Kim, MD, Urologist, The University of Nevada, Reno School of Medicine, Reno, NV
Preston Sprenkle, MD, Associate Professor of Urology, Director of the Urologic Oncology Fellowship and Research, Yale School of Medicine, New Haven, CT
Biographies:
Eric Kim, MD, Urologist, The University of Nevada, Reno School of Medicine, Reno, NV
Preston Sprenkle, MD, Associate Professor of Urology, Director of the Urologic Oncology Fellowship and Research, Yale School of Medicine, New Haven, CT
Read the Full Video Transcript
Preston Sprenkle: Good morning. I'm Preston Sprenkle. I'm a urologic oncologist at Yale University. We're here at the AUA. I'm speaking with Dr. Eric Kim, a urologic oncologist just recently from Wash U and soon to be at the University of Nevada at Reno.
Eric Kim: Thanks for having us and happy to discuss our project.
Preston Sprenkle: Great. Tell me a little bit more about it. I mean, it seems like we're looking at the correlation of a bunch of things that are very active and hot topics in prostate cancer right now.
Eric Kim: Absolutely.
Preston Sprenkle: Tell me a little bit more about how this came about and some of the results.
Eric Kim: Yeah, so we initially started with a smaller cohort, and we just noticed that PI-RADS score, as you know, will track to Gleason. Higher PI-RADS, higher Gleason. Decipher tracks to Gleason. Higher Decipher GC, higher Gleason. But for some reason, MRI and Decipher were not correlating with each other, and I think that's something you've put out before. We said, "Let's look in our entire cohort," which I think ended up being about 750 patients. It should be, I think we've updated it again, it may be about 900 just to try to find exactly where that incongruity is coming from. What's the driver because obviously MRI is telling you something biological or else why would it track to Gleason, which we obviously know links to outcome and Decipher we know links to biology maybe even better than Gleason. If you look at AUCs for Decipher for metastasis, it's like 0.8 something versus Gleason's, I think 0.6 something. So it just really came from that and just motivated residents and fellows that want to do the chart times and all the hard work that comes with that. As you know.
Preston Sprenkle: Yeah, that's great. So tell me a little bit more about where MRI... You mentioned you didn't include Gleason. Were you able to control for Gleason score as part of a subanalysis in this? I didn't see that in the abstract.
Eric Kim: Yeah, so we did do a multivariate analysis. Not everything was included in the abstract, obviously. But yeah, even controlling for Gleason, there seems to be some correlation with MRI appearance and the subtypes. Some of the deeper dives into Decipher GRID, the PAM50 subtypes or the PSC subtypes, there seems to be a correlation to subtype, especially in the Caucasian men.
A real limitation of our field, limitation everywhere, is the lack of African-American men in these cohorts. I think we had about 140, 150 or so African-American men, which is a pretty nice-sized cohort. But if you think about African-American men being at two to three times the likelihood of having prostate cancer, knowing that in St. Louis, probably similar to New Haven, the population is what, 30, 40% of our city's population, it's kind of sad that we don't have better representation. But with that being said, that relationship that we found between Decipher GC, but then also the Decipher GRID subtypes and MRI did not really hold up in African-American men.
Personally, I think it's just all these tests have been designed off of Caucasian men because that's who's had the MRIs, that's who's had the Decipher scores to build these models.
Preston Sprenkle: It did seem that there was some signal, wasn't statistically significant, but do you think that that still could just be because of the small population size?
Eric Kim: Potentially. What was interesting was just that the subtypes didn't hold up. I know that we pulled out, I think PTEN and P53 as specific genes of interest, and PTEN obviously actionable with PARP inhibitors or whatnot. But we were really hoping that the subtypes would hold up so that in an ideal world, we would find out that MRI is telling us about the tumor in its microenvironment. It's telling you more about the biology of the disease, not just tumor, no tumor, spot, no spot. There should be some more, I would hope, deeper information coming from it. And that's where it all started.
Preston Sprenkle: Right. So it seems like if we had consistency, then you could have reproducible treatments for everyone, but it seems that we may need to be personalizing those therapies more.
Eric Kim: For sure. I mean, I think as it stands in, at least the data that we've pulled out of our experience, it makes you think that you got to sprinkle in MRI, you got to sprinkle in Decipher, you got to sprinkle in the clinical information, you got to put it all together, like you said, to really personalize the treatment.
Again, if you look at independently or individually what's going to be the strongest predictor, it probably is Decipher. Really, we tried to put out something... I don't know if you guys have for MRI and long-term outcomes or PI-RADS and predicting metastasis and maybe PI-RADS 5. There's a small signal there, but Gleason score doesn't hold up as well. The clinical, if you look at CAPRA, it's not that much different than Gleason score.
Preston Sprenkle: Good morning. I'm Preston Sprenkle. I'm a urologic oncologist at Yale University. We're here at the AUA. I'm speaking with Dr. Eric Kim, a urologic oncologist just recently from Wash U and soon to be at the University of Nevada at Reno.
Eric Kim: Thanks for having us and happy to discuss our project.
Preston Sprenkle: Great. Tell me a little bit more about it. I mean, it seems like we're looking at the correlation of a bunch of things that are very active and hot topics in prostate cancer right now.
Eric Kim: Absolutely.
Preston Sprenkle: Tell me a little bit more about how this came about and some of the results.
Eric Kim: Yeah, so we initially started with a smaller cohort, and we just noticed that PI-RADS score, as you know, will track to Gleason. Higher PI-RADS, higher Gleason. Decipher tracks to Gleason. Higher Decipher GC, higher Gleason. But for some reason, MRI and Decipher were not correlating with each other, and I think that's something you've put out before. We said, "Let's look in our entire cohort," which I think ended up being about 750 patients. It should be, I think we've updated it again, it may be about 900 just to try to find exactly where that incongruity is coming from. What's the driver because obviously MRI is telling you something biological or else why would it track to Gleason, which we obviously know links to outcome and Decipher we know links to biology maybe even better than Gleason. If you look at AUCs for Decipher for metastasis, it's like 0.8 something versus Gleason's, I think 0.6 something. So it just really came from that and just motivated residents and fellows that want to do the chart times and all the hard work that comes with that. As you know.
Preston Sprenkle: Yeah, that's great. So tell me a little bit more about where MRI... You mentioned you didn't include Gleason. Were you able to control for Gleason score as part of a subanalysis in this? I didn't see that in the abstract.
Eric Kim: Yeah, so we did do a multivariate analysis. Not everything was included in the abstract, obviously. But yeah, even controlling for Gleason, there seems to be some correlation with MRI appearance and the subtypes. Some of the deeper dives into Decipher GRID, the PAM50 subtypes or the PSC subtypes, there seems to be a correlation to subtype, especially in the Caucasian men.
A real limitation of our field, limitation everywhere, is the lack of African-American men in these cohorts. I think we had about 140, 150 or so African-American men, which is a pretty nice-sized cohort. But if you think about African-American men being at two to three times the likelihood of having prostate cancer, knowing that in St. Louis, probably similar to New Haven, the population is what, 30, 40% of our city's population, it's kind of sad that we don't have better representation. But with that being said, that relationship that we found between Decipher GC, but then also the Decipher GRID subtypes and MRI did not really hold up in African-American men.
Personally, I think it's just all these tests have been designed off of Caucasian men because that's who's had the MRIs, that's who's had the Decipher scores to build these models.
Preston Sprenkle: It did seem that there was some signal, wasn't statistically significant, but do you think that that still could just be because of the small population size?
Eric Kim: Potentially. What was interesting was just that the subtypes didn't hold up. I know that we pulled out, I think PTEN and P53 as specific genes of interest, and PTEN obviously actionable with PARP inhibitors or whatnot. But we were really hoping that the subtypes would hold up so that in an ideal world, we would find out that MRI is telling us about the tumor in its microenvironment. It's telling you more about the biology of the disease, not just tumor, no tumor, spot, no spot. There should be some more, I would hope, deeper information coming from it. And that's where it all started.
Preston Sprenkle: Right. So it seems like if we had consistency, then you could have reproducible treatments for everyone, but it seems that we may need to be personalizing those therapies more.
Eric Kim: For sure. I mean, I think as it stands in, at least the data that we've pulled out of our experience, it makes you think that you got to sprinkle in MRI, you got to sprinkle in Decipher, you got to sprinkle in the clinical information, you got to put it all together, like you said, to really personalize the treatment.
Again, if you look at independently or individually what's going to be the strongest predictor, it probably is Decipher. Really, we tried to put out something... I don't know if you guys have for MRI and long-term outcomes or PI-RADS and predicting metastasis and maybe PI-RADS 5. There's a small signal there, but Gleason score doesn't hold up as well. The clinical, if you look at CAPRA, it's not that much different than Gleason score.