Zachary Klaassen: Hi, my name is Zach Klaassen, urologic oncologist at the Georgia Cancer Center in Augusta, Georgia. I'm delighted to be joined on UroToday, as always, by Dr. Andrea Necchi, who is a medical oncologist at San Raffaele in Milan, Italy. Andrea, thanks for joining us on UroToday.

Andrea Necchi: Thank you, Zach. It's a pleasure.

Zachary Klaassen: Really exciting day that you presented at ASCO. Looking at the impact of MRI radiomics from the PURE-01 study and really kind of figuring out if we can identify safely patients that have a pCR and maybe not necessarily have to do a cystectomy in that muscle-invasive bladder cancer space. And that's what we'll be talking about today.

Before we get into that data, maybe just bring our listeners up to speed on PURE-01. You've done some great work with it, and just the trial design, the outcomes at a high level.

Andrea Necchi: Yeah, PURE-01 study is a long-dated journey, you know?

Zachary Klaassen: Yeah.

Andrea Necchi: We started in 2016, 2017 in administering 3/4 of a very short course of three cycles of neoadjuvant pembrolizumab monotherapy before cystectomy in clinical T2, T4, N0, M0 muscle-invasive bladder cancer patients. The results were very, very intriguing at the end, and they've been confirmed in the long run. We present this year at the meeting, in a poster session, the five-year median follow-up outcomes showing that the pCRs are maintained over time in most of the patients.

Zachary Klaassen: Right.

Andrea Necchi: So overall, quite successful results. And then the possibility to enrich a study like this with a lot of translational work, including imaging work, that is part of the presentation that we made today with the MRI.

Zachary Klaassen: And tell us about the schedule of MRI. So you're getting it before treatment, after treatment. Where's that data coming from for the radiomics analysis?

Andrea Necchi: Yeah. We made MRI just before starting pembrolizumab. So after the TURBT the patients had already performed for diagnosis and for staging, and at the end of the neoadjuvant treatment. So after the three courses of pembrolizumab, just before radical cystectomy. And we matched the pre- and post-therapy imaging in association with the primary outcome of the study, that was the pathologic complete response – pathological response at the cystectomy level.

Zachary Klaassen: Excellent. And just tell us about the high-level results from the analysis you presented at ASCO.

Andrea Necchi: Yeah. It was strong work made in collaboration with the radiologists at MSK. Leo Schwartz and the radiology department at MSK did a fantastic job in collaboration with us at San Raffaele. So a radiologist at San Raffaele identified the lesions of each patient. We deal with 112 patients overall, meaning that more than 200 MRI images in pre- and post-therapy.

They identified the lesion. They segmented the lesion. And then in New York, at the MSK, they performed their radiomic analysis by extracting the main features of the images in terms of texture, in terms of image, in terms of a lot of other features, something like 230, or more than 230 features that overall define the radiomic features that can be extracted by a single lesion, by a single image.

Zachary Klaassen: Wow.

Andrea Necchi: And then we performed a logistic regression analysis, multiple logistic regression, putting all these features in a unique model, including clinical stage as a kind of clinical adjustment variable. And we made this analysis towards pathological complete response – it was the primary outcome of the study – and pathological major response, meaning downstaging to non-muscle-invasive disease.

Zachary Klaassen: I see.

Andrea Necchi: We did so for pre-therapy exams, for post-therapy exams, and for the delta post- and pre-therapy. What we saw is that, in particular for post-therapy MRI, the features were able to reach an AUC, so a reliability and accuracy for predicting, in particular, the pathological major response – residual non-muscle-invasive disease – of more than 0.90.

Zachary Klaassen: Wow.

Andrea Necchi: So very, very accurate.

Zachary Klaassen: Yeah.

Andrea Necchi: The accuracy was a little lower, like 0.8, for the pathological complete response outcome. And it was even lower for pre-therapy examination. So the point is that when looking at the post-therapy images, we are pretty well set in the possibility of predicting the non-muscle-invasive residual disease. That's important for potential future application of this kind of tool that is also something like an automatic tool.

It's one of the first times that we apply, and that we see, a machine-learning tool applied to MRI, applied to muscle-invasive bladder cancer. So the point is the possibility to make these things accurate, regardless of the expertise of the radiologist, that is quite impacting when determining the various course, when determining – when judging MRI features, you need to have an experienced radiologist. But in this way, you need to have just a computer analyzing the images, so we can apply it in the community, the results.

Zachary Klaassen: Yeah. There's so many ways we could take this information. And I think that's what's exciting about it. We think about it, if we're able to safely predict from MRI radiomics, whether that's with ctDNA, biomarkers, et cetera – obviously we're focusing on radiomics – we could predict patients that had a pathological complete response. Could we then spare them a cystectomy, trimodal therapy? Lots of implications there too.

Just talk about how, as we go forward, you mentioned some of the great points about implementation in the community. How could this take us to where we could do safely bladder sparing with muscle-invasive bladder cancer?

Andrea Necchi: Well, a great point. So MRI and advanced imaging in general is being considered – or is being included – in the newer definition of clinical complete response to any neoadjuvant therapy. It is now the primary focus of any kind of perioperative approach to these patients, with the aim of just focusing on a different internal endpoint as compared to the pathological complete response that could be equally accurate in predicting the outcome of the patients, with the advantage of, as you said, potentially skipping the need to undergo radical cystectomy for most of the patients, or at least for patients who are considered as major responders or complete responders.

Zachary Klaassen: Sure.

Andrea Necchi: So the negative imaging findings are still a matter of controversy.

Zachary Klaassen: Right.

Andrea Necchi: So we likely need support from an AI and machine-learning tool, for example, with the aim of improving our accuracy in predicting the residual complete response or absence of any residual disease or minimal infiltrating disease. Having said that, there is still an issue with regards to how could we consider the non-muscle-invasive disease after any kind of systemic therapy.

Zachary Klaassen: Right.

Andrea Necchi: So is it sufficient to move towards maintenance systemic therapy, to perform a ReTURBT, or perform intravesical neural therapy to save the bladder and to save the life of the patients? Or is the residual disease, the pTa, pT1 high-grade disease, still putting the patients at high risk of developing disease progression and metastasis?

Zachary Klaassen: Yeah.

Andrea Necchi: Emerging data from a Retain study or from the Hoosier Oncology study, pointed to the fact that there is a proportion of these patients with pT1 high-grade disease that are at risk of developing systemic metastasis.

Zachary Klaassen: Right.

Andrea Necchi: So the point is on how to manage the best, for the safety measure for the patient. This kind of residual disease is still a matter of debate, provided that we are sufficiently able to predict, or not, the presence of this kind of disease after therapy.

Zachary Klaassen: Yeah, great answer. I think there's so much excitement in the neoadjuvant/adjuvant space. And certainly, this data is really exciting – just building blocks for future analyses. And always just the goal of avoiding cystectomy when we can, but also still selecting patients where they're going to benefit. I think that's what it comes down to. Andrea, always a great discussion. Thanks so much for joining us on UroToday.

Andrea Necchi: Thank you and UroToday for inviting me. Thanks.

Zachary Klaassen: Of course.