(Length of Presentation: 8 min)
This presentation represents a study performed to assess if [18F]FDHT PET/CT could be a valuable imaging biomarker in patients with prostate cancer.
Gem Kramer, Radiology and Nuclear Medicine VU University Medical Center Amsterdam the Netherlands
Read the Full Video Transcript
Gem Kramer: Thank you very much for the introduction. First, this study was financially supported by the Movember Foundation.
So as we all know, the androgen receptor is critical for the development of prostate cancer and is one of the motors in early stage prostate cancer. However, even if the prostate cancer had progressed into castration-resistant prostate cancer, it remains to play a role in either through ligand-dependent activation, like overexpression or de novo androgenesis the tumor itself or through ligand-independent activation.
So, recently, in recent years, all kinds of new therapies have been developed like enzalutamide or abiraterone targeting these androgen receptors, however, 50% of the patients receiving these treatments remain non-responders. And to avoid unnecessary treatment and costs it will be very nice to have a prognostic imaging biomarker, in this case, imaging, to evaluate response and to indicate which patients are going to respond to therapy and which not.
This is where the FDHT comes in. The F18 fluorodihydrotestosterone, so actually it's fluorine labeled to dihydrotestosterone, as we see on the picture to the right and this enables you to noninvasively image androgen receptors. So if you want to use this tracer as a biomarker you can do it using visualization, but if you want to adequately assess response one would need quantification as well.
The golden standard, of course, is a non-linear regression. Only this is not generally applicable in clinical practice, so, therefore, we need more simplified methods. The aim of this study was, therefore, to assess whether the simplified methods for quantification of FDHT uptake in patients with mCRPC are adequate.
So, what did we do in this study? First, we tried to determine which was the optimal non-linear model describing the pharmacokinetics of FDHT. Second, we looked at continuous arterial sampling versus image-derived input function, and calibrate for arterial and venous samples and see where it was possible to leave the continuous arterial sampling out. Second, we looked at linearization models like Patlak and Logan plots and compared them to non-linear regression. And last, but not least, we looked also at different static methods of simplification. So, the most simple of body weight but also tissue to blood ratio and sort of corrected for the parent plasma. As you see here, this line is the FDHT so the parent plasma of FDHT. And you see it rapidly metabolizes, so after 30 minutes there's almost nothing left. So we also correct for this rapid metabolization and through correcting for parent plasma and also for the area on the curve plasma.
We included eight patients for this study, both test and retest data. Four scans were available with continuous arterial sampling, arterial samples, and venous sampling. And 12 were available using only venous samples. And only patients with mCRPC were included, so they do have testosterone levels lower than 1.7 nanomoles per liter. And they had to have progressive disease according to PSA rise or new lesions according to imaging through RECIST or two new lesions on a bone scan. And they were excluded if they were already treated with anti-androgens, like enzalutamide, or at a low HB so you couldn't do the arterial sampling.
So the scan protocol was as follows. First, there was a dynamic FDHT-PET scan for 30 minutes using continuous arterial sampling. And also we did five arterial manual samples to correct for metabolites later on, and three venous to compare those. After this first dynamic scan, we took a short break of 15 minutes, where the patient had to go to the toilet and when he came back we put him on the table again and we did a whole body FDHT-PET scan at 45 minutes post-injection.
Using the scans, we obtained than an image-derived input function by placing a writ of interest in the ascending aorta and calibrating those with the venous and arterial samples. And we did the same, we also obtained a writ of interest for the tumor TAC by using a 50 percent --.
Using this data, we performed a nonlinear aggression to obtain the optimal model like I already explained and we found that the two tissue 3K model with bloodborne infection was a model ultimately describing the pharmacokinetics. This is an irreversible model, we are going to be looking at Ki for now.
We compared the Ki of the continuous sampling with the Kiof the image derived input function with corrected for venous sampling and we saw a near perfect correlation of 0.89 of 98th. So wherefore assumed that we can replace the arterial sampling by need of at least corrected for three venous samples.
We looked further. We went to the Patlak and Logan analysis and compared those and what we found was that the Patlak always gave normal varies where Logan had a lot of misfits, which we suspected cause Logan actually is more likely to, irreversible traces where this was more irreversible. And here if you compare the Ki on nonlinear aggression with the Patlak Ki, also see a good correlation.
The only thing of this method, the nonlinear regression is you need the extra scan, for cleaning would be nice to use the extra scan so we looked at SUV body weight and what we saw here was that the correlation went down to the S squared was only 0.77 but there seems to be kind of split in the data so we were looking further into this. And this data is actually of one patient while the other 12 patients are on the other line shown a really good correlation of still above 0.90. So looked into this individual patient previous, really early in the disease and also fast progressive so this might be a different kind of metabolism is not, faster metabolizer compared to the patient who is already further in their disease. This was one with no treatment, no dose, no nothing before this.
So we tried to correct for this and actually was using the tissue to blood ratio and we found that actually here the surf corrected for the area of the curve of current plasma completely reversed the effect, no outline anymore. Suggesting also this is the metabolism of the tracer is different in this one patient and if you look for the other ways to simplify it you see that it doesn't really change. So if you correct serve only for the parent plasma or blood concentration the R squared stays more or less the same.
In conclusion, quantification of FDHG update performed using the simplified method, it's possible. We can replace the continuous arterial sampling by -- if you use the surf correctly the area of the curve of parent plasma or the Patlak you get a near perfect correlation. However, you need an extra scan also to obtain the area on the curve of parent plasma and when you further simple case the methods this results in a decreased accuracy of the scans and the quantification. These are the acknowledgments, I would like to thank you for your attention.