Baseline 68Ga-PSMA-11 PET/CT parameters can predict response to 177Lu-PSMA-617

In this interview, Phillip H. Kuo, MD, PhD, FACR, shares key findings and takeaways from the study, “Quantitative ⁶⁸Ga-PSMA-11 PET and clinical outcomes in metastatic castration-resistant prostate cancer following ¹⁷⁷Lu-PSMA-617 (VISION trial),”¹ for which he served as the lead author. Kuo is a professor of radiology, section chief of nuclear medicine, and director of theranostics for the department of radiology at City of Hope National Medical Center in Duarte, California.

This transcript was AI generated and edited by human editors for clarity.

Phillip H. Kuo, MD, PhD, FACR

Could you describe the background/rationale for this study?

When I advise young faculty, I tell them when you're trying to get published for promotion, make sure you have something short-term, medium-term, and long-term. This was definitely in the “long-term” category. This research was published in one of our top journals, Radiology. The company Endocyte was the one who started off the VISION trial (NCT03511664) in 2017. They acquired rights to ¹⁷⁷Lu-PSMA-617, which at that time, of course, wasn't even known as Pluvicto. Endocyte quickly developed the protocol for the FDA the phase 3 trial, which is now known as the VISION trial.

When we met with the FDA in early 2018—the Endocyte team brought me along as a consultant—it was a very interesting and stressful time, because this was a very unconventional development of a radiopharmaceutical, much less a theranostic. At that time, there were not very many examples of approved theranostic systems. So, this was very cutting edge, but also a little nerve-wracking, because we didn't have a template to follow for regulatory approval of a theranostic system like we have now.

There [also] hadn't been any conventional industry sponsored phase 1/2 trials. A lot of the data that were being relied on were from the great work out of many sites out of Europe–– including of course, in Germany, where this molecule originated from, and an amazing group out of Australia and New Zealand. But that was it. There weren’t the standard phase 1/2 multi-institutional trials that could help us understand site heterogeneity. [We didn’t have data] with good control groups. That left many gaps the FDA wanted filled.

One of the major things that the FDA always wants to be sure about is that your drug is engaging the target. The other thing that was of concern was that there weren’t the typical dose escalation studies that you would have seen in earlier trials and that the FDA would have already vetted. ¹⁷⁷Lu-PSMA is given at 200 mCi in a flat dose. You can dose reduce to 160 mCi if there are [adverse events] or bone marrow suppression, or something of that nature. You might wonder, "What's the added benefit of 200 mCi vs, for example, 150 mCi?” Are we giving more toxicity without much benefit? We weren't able to hand them trials that they'd already approved as phase 1/2. That was actually the genesis of this idea way back in 2018. It took a long time to get this study published.

Rather than send us back to do a dose escalation study or some confirmatory target studies, the FDA accepted that if we're keeping the dose the same, at 200 mCi, what we can then vary to correlate with survival is how much radiation the tumors are getting. We can do that by assessing the target expression. Hence, the genesis of this idea where we took all the baseline PSMA-PET scans that these patients had for eligibility, segmented all the tumors, and derived parameters on an organ system and whole-body level. All the scans had already been deemed eligible using the VISION imaging eligibility criteria. The FDA also wanted to be sure that these novel VISION imaging eligibility criteria could be applied reproducibly and thus another FDA-mandated sub-study was interreader and intrareader agreement with the VISION eligibility rules for PSMA-PET and diagnostic CT. The idea was that, ostensibly, about 15% of the patients that would have done the worst were excluded from randomization.

In the VISION trial, those eligible patients then went on to get randomized to either standard of care alone or standard of care plus ¹⁷⁷Lu-PSMA-617. We took that randomized patient population and quantified all their baseline PSMA-PET scans. That was challenging, because many have very high burden of disease. We used an artificial intelligence [AI] algorithm to segment out the disease on the scans and provided validated data to the FDA for approval of a breakthrough drug. To my knowledge, that's the first time AI was used in this way for regulatory drug approval.

Invicro’s AI algorithm—and there's now many more software algorithms—was used for bone metastases. If you combine the CT and PET data, the AI needed a relatively small number of training set data to perform accurate segmentation of the bone disease, and 91% of patients had bone disease. All that bone disease was segmented by AI, and I checked all of that and found it to perform well. But of course, there's always be some issues with AI.

Then I trained a small army of image analysts from the imaging CRO Invicro to semi-automate, with software, all the soft tissue disease, so all the lymph nodes, liver metastases, lung nodes, and adrenal metastases. Then I modified, as needed, every single one of those. These were over 10,000 volumes of interest [VOIs]. We sent all the parameters for statistical analysis and correlated them with survival, which was the goal of the substudy. The primary objective was radiographic progression-free survival. The original plan was for a quartile analysis.

What were the key findings from this study?

It showed exactly what we hoped. We categorized the segmented tumors by their organ system; for example, bone, liver, or lymph node. If you take all those VOIs in the patient, whether it's from bone, lymph node, or adrenal, and add up all those VOIs and take the average of all those voxels, you get the whole-body SUV_mean. Additionally, we used other parameters like SUV_max, tumor volume, and the product of tumor volume and SUV_mean. The most important parameter was the whole-body SUV_mean; that had the most statistical significance in this complex matrix of radiographic progression-free survival (rPFS) and overall survival (OS), against all these various PSMA-PET parameters.

When we divided the patient population into quartiles by whole-body SUV_mean, with each quartile you go up, there was an improvement in patient survival and rPFS. If you compare the treatment arm of ¹⁷⁷Lu-PSMA-617 plus standard of care vs standard of care alone, the rPFS benefit was only 1.9 months in the lowest quartile, but 10.2 months in the highest quartile. So, a huge difference. We also showed that even in this lowest quartile, there was a benefit, as we showed in the VISION trial overall. It came with very tolerable side effects and improvement in quality of life compared to standard of care alone.

We also did the analysis for OS. That was a secondary end point, and we showed that compared with standard of care alone, there was an improvement in median OS. Even in the lowest quartile of whole-body SUV_mean, [patients achieved] a 3.2-month improvement in OS compared to standard of care alone. When you go to the highest quartile for whole body SUV_mean, the OS difference between the treatment arm and the standard of care arm alone was 6.4 months, so actually a doubling between the lowest and highest quartile.

If you do it on a continuous basis [rather than by quartile], for every 1 unit increase in whole body tumor SUV_mean, there was a 12% decrease in the risk of an rPFS event and a 10% decrease in the risk of death. We showed that it was quite continuous; there is no optimal whole-body SUV_mean. People have thrown out the number of whole-body SUV_mean of 10 as an optimal cut-off, but that just happens to be the bottom limit for the top quartile. If we did a different patient population and did the quartiles, it wouldn't be 10. It just happens that the threshold for the top quartile was 10. But when you do the continuous analysis, it's quite linear. There is no cut-off or optimal value. Even though that was not the objective in order to gain approval and get the substudy performed, my colleagues and I knew that this would potentially be a great clinical value in the future.

What is the main take-home message based on these findings?

The main take-home point is there is so much that we have yet to learn and optimize, but we already can start doing it. We already are doing it. This is a true theranostic paradigm; we image first, then we treat. And now, something that we did not do in the VISION trial but is already being done across many sites and has rapidly become the absolute standard of care, is imaging patients when they're on the drug. There are already a lot of data to support the advantage of image, treat, and then continue with on-treatment imaging to optimize therapy with your imaging biomarker. We can readily image patients while they're on treatment and get a real-time assessment of how they're doing and provide this information to the medical oncologists or urologists, if they’re treating them in earlier disease states.

We can truly do theranostics the way it was meant to be, where we combine all these biomarkers. Serum PSA is a very good biomarker, but, particularly in the advanced disease states, it's far from perfect. But we [can] combine that with imaging biomarkers to optimize treatments for patients and inform not just the treating team, but the patients themselves. Many patients are well informed and want to know about their plan of care, [especially] the risk and benefits. We can inform them even more so now. I really think that's where it's going. We started and showed with this trial that we can do it from the very onset with the baseline PSMA-PET. Now, many excellent groups across the world are already showing the power of the imaging biomarker while they're on treatment.

I just want to end with 1 final caveat. Although we see research such as this work as being very important in the management of patients and educating not only us as clinicians, but also the patients about the risk vs benefit, there's still a lot of work to do. We have to validate artificial intelligence algorithms for real world clinical care. The entire VISION study was done with ⁶⁸Ga-PSMA-11. So, [we need to] take great caution in trying to extrapolate any of these SUV values to any other tracer. It goes to show that we have real potential just beyond excluding, for example, the 10% to 20% of patients that are going to do the worst with ¹⁷⁷Lu-PSMA-617. [There was also] very exciting data that just got published about the PSMAfore trial, which is often known as the pre-taxane study for ¹⁷⁷Lu-PSMA-617. As we move this therapy earlier in the disease state, there are going to be more treatment options for patients. Knowing how patients are going to benefit is going to play an even more important role as we move this earlier in the disease state for metastatic castration-resistant prostate cancer.

Is there anything else that you’d like to add?

I do want to thank my many colleagues and co-authors, particularly Ken Herrmann, MD, and Andrew J. Armstrong, MD, MSc, from the University Hospital Essen and Duke [University], respectively, as well as many other people. It really takes a village to do this kind of work.

REFERENCE

1. Kuo PH, Morris MJ, Hesterman J, et al. Quantitative ⁶⁸Ga-PSMA-11 PET and clinical outcomes in metastatic castration-resistant prostate cancer following ¹⁷⁷Lu-PSMA-617 (VISION trial). Radiology. 2024;312(2):e233460. doi:10.1148/radiol.233460