Moderator: The next paper is by none other than Dr. Andrew Scott from the University of Melbourne. The title is "Reproducibility and Repeatability of Quantitative 18F-FDHT Uptake Metrics in Castration-Resistant Prostate Cancer Metastasis: A Prospective Multicenter Study". Andrew.

Andrew Scott: Okay. I'm presenting this on behalf of colleagues from a number of institutions as part of a multicenter, multinational trial looking at a fluorodihydrotestosterone, FDHT, in patients with castration-resistant prostate cancer. The background is that we know that androgen-receptor expression in castrate-resistant prostate cancer is an important target for potential therapeutic options. Although the ability of the androgen receptor to up and downregulate the androgen splice variants which can be introduced or alternate signaling pathways can obviously impact the way in which patients respond to anti-androgen therapy. The Memorial Sloan Kettering group had pioneered the development of FDHT as a marker of looking at the actual expression of the androgen receptor itself, and this particular trial was designed to look at repeatability and reproducibility of FDHT in patients with metastatic castrate-resistant prostate cancer.

The intent was that we would look at, initially, whether or not this was a test which could be reliably introduced in individual patients at multiple centers in multiple continents, and whether or not there was any relationship between FDHT and perfusion/diffusion changes as assessed on whole-body MR, and a subsequent cohort analysis that will be looked at patients that are actually on abiraterone treatment.

This particular presentation is merely to demonstrate the repeatability and reproducibility of the technique because it had only been evaluated in patient cohorts at one site, and so for its general applicability, this was part of the initial validation cohort. The inclusion and exclusion criteria are as listed. Patients with progressive metastatic castrate-resistant prostate cancer with imaging or biochemical evidence of progression, but visible lesions on either CT, bone scan or MRI and informed consent, and with the exclusion of patients with a paddock or inland sufficiencies or concurrent androgen receptor inhibitor use.

We had an ambitious initial study design in which it was intended to look at whether or not you could repeat FDHT patients within the same patient and get reproducible results. The concept was that, initially, you'd look at FDHT PET scans being performed on subsequent days and then to think about doing them a week apart, and then three weeks apart with the underlying hypothesis that if we, in the first cohort of patients, saw a difference in lesion variability of greater than 15% we would go on to looking at the subsequent cohorts. As it turned out, and as I will show you, there was less than a 15% lesional variability identified, and so only the first cohort was evaluated.

All patients underwent two separate FDHT PET/CT scans on two consecutive days. They had low dose CT beforehand followed by dynamic PET/CT so that the input function could be calculated. There was blood drawn from which whole-body activity and plasma concentration of FDHT and metabolites was analyzed in all patients. Then after a short break, a whole-body scan was performed approximately 45 minutes post-injection in all patients.

You'd all be familiar on the left-hand side with the FDHT distribution. It does have a reasonable blood pool activity and excretion via the hepatobiliary system. On the right-hand side is the same patient with an FDG PET and, again, there'll be no surprise that there's often a difference between in this case a bone in metastatic lesions in the context of androgen receptor expression versus FDG uptake reflective of the fact that AR expression can be quite variable in patients with metastatic disease.

Now, in order to do this study properly, we actually engaged with the SNMMI collaborative trials next week, CTN for centralized data management, and we used a Keosys imaging system to upload all the images so that they could all be analyzed and interpreted in a common manner. All of the PET/CT scans were analyzed by two experienced observers, nuclear medicine practitioners, one, a resident, but one radiologist nuclear medicine physician using the Keosys DICOM viewing platform. Lesions were considered suspicious for metastatic disease independent from the history of the patient when there was uptake, which was higher than blood pool or background tissue, and not corresponding to a known site of physiological excretion of the tracer.

The actual lesion analysis was performed according to this schema, a fairly standard one for bones, lymph nodes, and soft tissue as anyone who's been involved with these types of imaging studies in prostate cancer patients would be very familiar. The actual analysis itself, as I mentioned, was done centrally for this particular analysis, although we also performed center by center analysis and those results will be reported separately. There were regions of interest as well as a qualitative assessment of sites of disease identified. The volumes of interest, we're using a 50% isocontour of the SUVmax, and then from these calculations for SUVmax, SUVpeak and SUVmean corrected for bodyweight were all derived. There was also an analysis of the plasma activity so some corrections based on our background blood activity of the parental compound could be performed. In addition, there were tumor volumes either through metabolically active tumor volume from 50% threshold of SUVmax corrected for background or total lesion uptake, which was defined as SUVmax times the MATV was calculated.

Now, this is where the analysis became quite complicated, and if anyone's got a question, I recommend that you talk to the first author of this paper, which is not me, because when you're looking at reproducibility coefficients, which is 1.96 times the standard deviation of the absolute difference in single or multiple lesions, and I'm glad I remembered that because I couldn't be able to work it out if I tried... So that was looked at, so that's called an RC. Then there were also qualitative comparisons and statistical analyses based on interclass correlated coefficients and Bland-Altman plots.

There were 27 patients, 140 FDHT avid lesions were included. Now, there was no difference in patient characteristics between groups at all of the centers, and there was no relationship of patient characteristics, age, Gleason score with any of the parameters which were assessed. The important take-home point was that SUV values were found to be that which had the greatest repeatability and without significant differences in reproducibility between them. The test and retest scans all showed very strong correlations. Now, the repeatability coefficients ranged from 21% to 24%, and for the volume metric analysis, because there's slightly more variability in terms of the definition of volumes, there was also a good correlation but it didn't match quite as well as we saw with SUVs.

For those who were interested in the detail, this shows the overall and then center by center results for each of the SUV parameters, as well as the volumetric parameters. You can see that the P values running differently in terms of repeatability for the volume metric parameters. We also looked at the lesion level and patient level for each of the values, including normalization for the plasma input curve. There was a better correlation when the plasma input curves were included, but this was only relatively small, but it was, in fact, observed.

In conclusion, FDHT imaging was remarkably repeatable and reproducible across multiple patients, in multiple centers, in multiple continents. SUV was found to have the best repeatability of the measurements that were assessed, and you'll be seeing in subsequent meetings further results on additional analyses from these cohorts, as well as the results of an ongoing accrual cohort in abiraterone-treated patients. We'd like to acknowledge the Movember Foundation which funded this study and the results of this have just come out in the Journal of Nuclear Medicine. Thank you.

Moderator: Thank you very much, Professor Scott. Any questions or comments? Can you please go to the microphone?

Speaker 1: Congratulations on the paper though, it's really interesting. I just have one question regarding the SUVmax thresholds that were used. There are papers, they refer to 40%, 45%, there was use of 50%, and that gives a huge variation of the volume. Why was chosen 50% of, just to have an idea, we are struggling with this right now, so we're trying to get the best information we can on that.

Andrew Scott: Well, it was a consensus decision. It wasn't based ... there is literature out there on what the threshold needs to be, and it was essentially consensus that was decided a priori before we started. That's the only reason.

Speaker 1: Okay.

Speaker 2: I think, at the end, it's not that important to know the threshold but if you ... 

Moderator: Okay, thank you, Dr. Wall?

Richard Wall: I wanted to ask you about that one equation. The volume, I think it was 46% repeatability coefficient for the volumes, that is a big difference. The thresholding, obviously, I know you know this, but it means that that isn't really so usable in individual patients as the SUVmax, and I think it does point to some of the challenges with getting volumes as a percentage of the SUVmax, which inherently has variability as well. I wasn't sure how you did the background correction but I just would say from this, when you say the repeatability is good, the SUVmax repeatability was pretty good but the volumes are really not good. I think that maybe a bottom-up approach, a threshold-based from some parameter coming up like with FDG we've done for PERCIST with the liver as a background, may be more repeatable. I don't know what that threshold should be or if you have an opinion on that.

Andrew Scott: Sure. That has not been analyzed as yet, so I agree. This was based, as I said, on the a priori methods that we decided on, so that was what we saw. I agree.

Moderator: I have one quick question. How does an FDHT fit into the management of your patients? Do you use it clinically?

Andrew Scott: No.

Moderator: Why did you do this study then?

Andrew Scott: It's purely in the context of trying to have a biomarker for AR, which may or may not have an impact on treatment options. It certainly had an applicability in the context of the Memorial group for looking at therapeutics, which work specifically on the androgen receptor itself. I think it has a potential role in drug development, it may have a role in the prediction of response, but that's still to be determined, and that's the next phase of the trial that we're doing.

Moderator: Okay. Thank you so much. All right.