0:00

The, uh, title of my, uh, talk is, is perhaps a little bit different than,

0:04

than what may have been on the flyer. Uh, but I think it, uh,

0:06

I think it's gonna cover everything that, that we'll talk about. So,

0:09

there's certain things about PSMA PET that we definitely know. Uh, we have, uh,

0:14

pivotal phase three clinical trials that make us as certain as we possibly can

0:17

be in clinical medicine,

0:18

that PSMA PET does certain things and works as advertised in certain ways. Uh,

0:23

there are some things that we might know, um, various interpretive pitfalls,

0:26

which we'll talk about, uh, non prostate cancers that have PSMA uptake,

0:31

things along those lines.

0:32

And then there are things that we need to know that we, that we don't know yet,

0:36

and that would include, um, things along the lines of response assessment,

0:40

uh, radios, uh, artificial intelligence, and we'll,

0:44

we'll touch briefly on those things as well. I,

0:46

I am just recently arriving in North Carolina, so I haven't had a chance to,

0:49

to change all my slides over yet. So, uh,

0:51

you'll see it says Johns Hopkins and all of my slides have this kind of, uh,

0:55

backdrop of the Hopkins dome on them. So, uh, apologies for that. Uh, uh,

0:59

the next time I, I join you for anything, uh,

1:01

I promise there'll be University of North Carolina Slides. Alright,

1:05

just real briefly, PSA, it's a transmembrane carboxy peptidase.

1:08

It's highly expressed on prostate cancer cells.

1:10

About 95% of primary tumors at least have PSMA expression. Uh,

1:14

we know that that can change over time, and as patient's tumors, uh,

1:18

evolve through various treatment, uh, modalities and, uh, and change over time,

1:22

that that, that PSMA expression may not always be maintained. But again,

1:26

at least at the, at the start point, when patients have primary tumors,

1:29

at least 95% of those will have PSMA expression.

1:31

And there does appear to be a correlation between expression levels of tumor

1:35

aggressiveness. That's a histopathologic finding.

1:37

I don't know that we always perceive that correl,

1:42

that that correlation or that association, uh, on the scan level, uh,

1:45

because of things like partial volume effects and perhaps other things that are,

1:49

that are at work at a macroscopic level. Uh, but again, if you look at, uh, uh,

1:53

if you look at just expression levels versus tumor aggressiveness,

1:56

there does seem to be a relationship there.

1:58

And there are now some emerging ideas that, uh,

2:02

that simple metrics like SUV MAX on A-P-S-M-A scan may actually have

2:06

prognostic significance,

2:07

and it probably comes back to this fundamental relationship. P ssm A as,

2:12

as we said, is a, is a membrane protein.

2:14

It has a large extracellular domain that has several sites on it that we can

2:18

leverage for binding of small molecules or, uh,

2:20

antibodies or even other kinds of molecules, uh,

2:23

for both imaging and for therapy. And what we know about p SSM a pet so far,

2:27

and again, these, these things, uh,

2:29

everything on this slide is at least backed up by prospective phase two data.

2:33

Uh,

2:34

and most of it is backed up by prospective phase three pivotal trials,

2:39

uh, across, uh, again, every PSMA PET agent, uh, they,

2:43

they all use sort of relatively similar trial designs. In particular, uh, the,

2:47

the two more recent and Zuma, uh, used very similar trial designs, uh,

2:52

and then gallium PSMA 11, uh,

2:54

which is sold under the trade names of ELIX and locs. Um, the, uh,

2:59

the data there at least closely parallel the data from the, the other two, uh,

3:03

agents, even though the clinical trial designs may have been slightly different.

3:08

Uh, we do know that psm a pet has a moderate sensitivity,

3:10

only a moderate sensitivity,

3:12

but a very high specificity for preoperative nodal staging.

3:15

We know that they have a high detection. We know that psm,

3:17

a PET has a high detection deficiency for sites of biochemical recurrence.

3:20

And we know that pssm a PET is effective for guiding metastasis directed therapy

3:24

for patients with limited volume or oligometastatic disease.

3:27

And also for selecting patients for endo radiotherapy, uh, which are generally,

3:32

uh, lutetium labeled ligands, at least in North America.

3:34

They're so far lutetium labeled ligands, uh, primarily Pluto,

3:37

which is a regulatory approved agent, or LUTETIUM 1 77 PSMA 6 1 7,

3:42

as it's more broadly known. Alright, so here's,

3:47

here's an example of P ssm, A PET in a, uh, in a couple of patients with, uh,

3:50

high risk prostate cancer. So, and, and again, to, uh,

3:54

as we'll see over the next few slides, this is sort of a, uh,

3:57

an area where for nodal involvement, the sensitivity is moderate, uh,

4:01

but the specificity is extremely high.

4:04

Most patients that you image with, uh, unfavorable, intermediate,

4:08

high or very high risk prostate cancer,

4:09

which is typically with a guidelines suggest, uh,

4:11

that we should be imaging those populations.

4:15

Those patients are gonna look like the patient on the left.

4:16

They'll have very high uptake in a primary tumor in the prostate,

4:20

but they won't have any evidence of nodal involvement. Uh, we,

4:23

we know from nomograms,

4:24

these patients could have potentially up to a 25% chance of,

4:27

of nodal involvement. But we also know that our sensitivity is, is moderate.

4:30

So our, uh, most patients, the significant majority of patients,

4:35

even those at risk for nodal involvement, uh,

4:37

will not have evidence of nodal involvement on the scan. Uh,

4:41

my urology colleagues like to note that, that this indicates that, uh,

4:45

nodal staging of prostate cancer patients is still ul,

4:47

still ultimately comes down to surgery. It's ultimately surgical staging.

4:51

We can do the best we can with PSMA pet, uh, but, uh,

4:55

the surgeon is going to find disease that, that we did not perceive on,

4:58

on the PSM a scan occasionally, and I would say this is uncommon,

5:02

patients will present more like the patient on the right,

5:04

where they have unsuspected systemic disease.

5:06

This patient has nodal involvement in the pelvis and their retroperitoneum.

5:09

And in the less supraclavicular space, um, this, uh, these,

5:13

these nodes are all sub sub centimeters. So prior to the pssm, a PET scan,

5:17

this patient would've been classified as clinically localized. Uh,

5:20

but in this particular case, the patient actually has systemic disease.

5:23

This happened a couple of times in our phase two single center trial in a few

5:27

times in, uh, in the pivotal, uh, phase three trial. That was,

5:31

that was done with the same agent. Uh, but this is a relatively rare event.

5:34

However, for the patient,

5:35

it's life changing and that they switch from what would be a curative, uh,

5:39

attempted curative therapy, uh, to a more palliative, uh,

5:44

paradigm where they'll be receiving, uh, presumably lifetime, uh,

5:47

systemic therapy. Alright, let's, uh,

5:51

let's quickly take a look at the, uh, the Osprey trial, which, uh, was,

5:55

was done with the agent that sound known as, as purify. Uh, however,

5:58

very similar trial design and, uh, very similar results with, uh, with, uh,

6:03

the other agents that are, that are out there. Uh, this, uh,

6:07

particular trial design was sort of a follow on to the, um,

6:11

to the trial that was,

6:13

that involved both of the patients I showed you on the previous slide,

6:16

which was, again, a single center, phase two. Uh,

6:18

the OPR trial was a phase three. It actually had two cohorts,

6:21

but we're only gonna focus on one cohort, uh, which was, uh, a total of, uh,

6:26

um, 268, uh, uh, patients who, uh, underwent,

6:31

uh, a, a, uh,

6:32

D-C-F-P-Y-L PET CT or PS MA targeted PET ct.

6:36

And of those 252 patients underwent a radical prostatectomy with an extended

6:41

pelvic lymph node dissection. The, um,

6:44

co-primary endpoints of the trial were sensitivity and specificity for pelvic

6:48

nodal involvement. And the results of that trial, which again,

6:52

had been recapitulated with, with other agents, uh, in very,

6:55

in very similar clinical settings, was again, a sensitivity of, of only 40%,

7:00

which was surprising. And that primary endpoint was missed, uh, by, by, in,

7:04

in this particular case. Uh, interestingly, the co-primary endpoint of, uh,

7:08

pluma, uh, and again,

7:10

a very similar clinical trial setting that included some patients that had, uh,

7:14

uh, unfavorable intermediate risk. These were all higher,

7:16

very high risk patients, uh, also met, met, also failed to meet its, um,

7:23

sensitivity co-primary endpoint.

7:25

But the unifying theme of these agents is that the specificity remains

7:28

incredibly high. And no matter who you exclude,

7:31

no matter what sort of sub-analyses you do post hoc, uh, uh,

7:36

playing with the numbers,

7:37

you'll generally find that no matter what you sort of drive your sensitivity to

7:41

the specificity remains rock solid. This,

7:44

this has important implications for how I think we should approach reading these

7:47

scans. If you have a patient who's presenting for initial staging,

7:52

uh, and presumably those patients are doing gonna be unfavorable, intermediate,

7:56

high, or very high risk prostate cancer,

7:59

if the specificity is truly approaches a hundred percent,

8:03

and that specificity is robust,

8:05

then in places where prostate cancer would make sense to spread to proximal

8:10

external iliacs, ator, sate, internal iliacs, um, presacral,

8:15

peri rectal, and then if those areas are involved,

8:18

we have to start worrying about the common iliacs.

8:20

And if those areas are involved,

8:21

we have to start worrying about the retroperitoneum.

8:23

But if we see disease that that sort of fits that paradigm of,

8:27

of where, uh,

8:29

prostate cancer spreads to our specificity for that uptake is going to be

8:33

incredibly high and we should confidently call even very subtle findings. Uh,

8:38

that's my approach.

8:38

If I see anything above blood pool in those areas that we just described, um,

8:42

and I can correlate that to a lymph node, even if it's very small,

8:47

I'm highly suspicious that that lymph node has evolved. I'd say the,

8:50

the caveat to this is you will sometimes encounter patients who have systemic

8:54

inflammatory processes,

8:55

and they'll have relatively diffuse adenopathy with low level uptake.

8:59

And I think those patients are an important pitfall to at least be aware that,

9:03

uh, uh, that someone with diffuse adenopathy with uptake, uh,

9:07

may be fooling you. And maybe in that couple of percent of patients where,

9:11

where the specificity, uh, doesn't quite hit a hundred percent.

9:15

Alright, so we also said that we know that, uh,

9:18

the PSMA PET has a high detection efficiency for,

9:20

for finding sites of biochemical recurrence. Uh,

9:23

here you can see a local recurrence that was a cult on, uh, col on ct. Uh,

9:27

but there's an enhancing nodule visible there on MRI.

9:31

Even more commonly than that, patients will have one or perhaps two, uh,

9:35

lymph nodes in the pelvis that, that are,

9:37

that indicate their signs of recurrence. Uh,

9:39

the pitfall here is trying to avoid, uh,

9:41

mistaking ureteral excretion for uptake in the lymph node that can be

9:46

challenging.

9:46

And it's really incumbent upon us to look very carefully and make sure that

9:50

we're not over calling, uh, mureal excretion as a, as a pelvic lymph node.

9:56

The, uh, CHONDRO trial was,

9:58

was now one of several clinical trials that had been done in this space with

10:00

various PSMA agents. Uh, spotlight was the, uh, uh,

10:04

was the PLUMA equivalent, and then there was also a, uh,

10:07

a dual center study between UCSF and UCLA, uh, looking at this, uh,

10:11

same patient population or a similar patient population, uh,

10:14

with Gallium PSMA 11.

10:19

The results of the condor trial, I think also sort of informed how,

10:22

how we should approach the scan, or not just the condor trial, but,

10:25

but all of these trials have, uh, have generally found similar, similar results.

10:30

The big result is, uh, is actually on the next slide, but I do want to, uh,

10:34

I do wanna focus on the slide for just a second.

10:36

One thing the FDA has done is PSMA agents have been approved is start to, uh,

10:42

is start to have all these new metrics, uh, correct localization rate,

10:45

correct detection rate.

10:46

These generally come down to being some variation on a theme of positive

10:51

predictive value, and usually at the lesion level,

10:54

although sometimes not quite at the lesion level,

10:56

sometimes at more of a patient level. Uh, and it's, uh, honestly, the,

11:00

the FDA hasn't done us any favors here. They've,

11:02

they've just kind of muddied the waters and made things confusing. However, uh,

11:06

for, uh, uh, for this particular trial, the, uh,

11:11

the primary endpoint was something called the correct localization rate,

11:14

which was a, uh, a positive predictive value at the lesion level,

11:19

is at least how we should think about it.

11:21

And we won't talk about the standard of truth,

11:22

but there was sort of a hierarchical standard of truth.

11:25

The correct localization rate does not change much with PSA level. So that's,

11:30

uh, it changes a little bit.

11:31

There's a trend towards higher correct localization rate at, at higher PSAs,

11:35

but it's not nearly as strong as the, uh, as the relationship between,

11:41

uh, detection efficiency and PSA level. Again,

11:44

what I think this means for how we should approach PSMA scans in a biochemical

11:48

recurrence setting is that we know that at very low PSA levels that many of

11:51

those scans are going to to be negative. However,

11:55

if our correct localization rate or positive predictive value at the lesion

11:58

level is relatively robust across PSA levels,

12:02

the same finding may mean the same thing regardless of the patient's PSA level.

12:06

And what I mean by that is a small obterator fossil lymph node in a,

12:10

in a biochemical recurrence patient with a PSA of 0.3 or a PSA of three,

12:15

I would treat that finding in the same way and be highly suspicious that it

12:19

represents a true positive side of disease,

12:21

because that positive predictive value at the lesion level is,

12:25

is so high regardless of PSA. So that's at least my approach to, uh, to the, uh,

12:29

the scans here. Uh, but again, I,

12:31

maybe the most important take home message is that PSMA isn't a magic bullet.

12:35

And at low PSA levels in the recurrence setting,

12:37

it is not uncommon for a scan to be normal. Now, if we,

12:42

if we do assume that, uh, that these, uh,

12:45

that we have a high detection efficiency at, uh,

12:50

for most biochemical recurrence patients,

12:51

and that we can see things that we can't see with conventional imaging,

12:55

the next step down that line is to say, well,

12:56

if we see a limited volume of disease,

12:58

is that patient eligible for metastasis directed therapy? Uh,

13:02

perhaps avoiding or at least kicking the can down the road on, uh,

13:06

on systemic therapy like androgen deprivation, which, uh,

13:10

in addition to things like, uh, quality of life and sexual side effects,

13:13

also puts men at risk of, um, heart attacks and strokes and other,

13:17

other sort of bad complications. So, uh, so we'd like to avoid that if we can.

13:22

Of course, most men aren't going to be able to avoid that,

13:24

but occasionally we can at least, uh,

13:26

perhaps buy a couple of years off of systemic therapy. Here is a patient who,

13:30

after prostatectomy, had what appears to be a,

13:32

a presacral or peri recal lymph node,

13:35

that was the only visible site of disease on AP SM A scan. Uh, the patient, uh,

13:39

underwent, uh,

13:40

stereotactic body radiation therapy to that site of disease did not start

13:44

systemic therapy. His PSA became undetectable.

13:47

And over the two years of follow up of, of this particular study, uh,

13:50

his PSA never became detectable again,

13:53

I have absolutely no idea if this patient's cured. I,

13:55

I would be hesitant to use cure and anyone who has disease outside of the

13:58

prostate, but at least this patient's been at least two years without requiring

14:02

deprivation therapy. We also see things at the other end of the spectrum,

14:07

where, again, the lack of sensitivity of conventional imaging may fool us. Uh,

14:10

here was a bone scan that had maybe one suspicious spot on it, uh, but,

14:15

uh, in fact, this patient had hundreds of marrow based lesions that were, uh,

14:19

invisible on both CT and bone scan. So this patient,

14:22

obviously not eligible for metastasis directed therapy. Uh,

14:25

this patient's volume of disease requires that, uh,

14:28

that he undergoes systemic therapy.

14:30

And there are patients that sort of fall in the middle.

14:32

They have low volume disease, but perhaps not as low volume as we'd like. Uh,

14:36

this patient, for instance, has both pelvic and retroperitoneal adenopathy.

14:40

And when SBRT without starting, uh, without starting ADT, uh,

14:44

and returned a few months later for a follow-up PSMA PET scan,

14:47

and it was found to have disseminated bone metastases.

14:50

So those bone metastases were almost certainly there prior to him undergoing

14:54

radiation therapy,

14:55

but they were perhaps below the detection efficiency of the scan. So, again,

14:58

pssm a not a magic bullet. Uh,

14:59

it will serve some patients well in selecting them for metastasis therapy. Uh,

15:03

in other patients, uh, it may not, uh,

15:05

it may not serve so well because they may just have such low volume disease,

15:08

if you will, the true micrometastatic disease, uh,

15:11

that we're just not able to treat all of their sites of disease.

15:16

However, if a,

15:17

if a man has sort of progressed to the point that he has more widespread

15:20

metastatic disease,

15:21

and in particular in the us this would mean post chemotherapy,

15:24

metastatic castration resistant disease, uh,

15:26

to be eligible for treatment with lutetium M-P-S-M-A, or again,

15:30

in the US right now, that's Pluto. Uh, these, uh, uh,

15:35

we, we select patients, uh, for Pluto with, uh, uh,

15:39

with A-P-S-M-A scan upfront, uh,

15:41

so that we can confirm that they have both sites of disease that, that have, uh,

15:45

that are expressing the target,

15:46

and also that they don't have sites of disease that aren't expressing that

15:49

target. And that won't benefit from, from PSA targeted into radiotherapy.

15:54

Uh,

15:54

somewhere around 40% of patients that are treated are going to have an objective

15:58

biochemical response with a, uh, greater than 50% drop in their PSA.

16:03

There'll be about 70% of patients that have any drop in their PSA. Uh,

16:07

and that efficacy all comes in, in a, uh,

16:11

at a relatively small cost in terms of toxicities. Uh,

16:15

there can be, uh, nephropathy, but it's generally low grade and self-limited.

16:20

There can be Xerostomia, again, generally low grade and self-limited.

16:25

And we now have sort of the pivotal trial results with the teaching.

16:28

And PSMA six 17, uh,

16:29

this was the Vision trials published in the New England Journal a couple of

16:32

years ago. Uh, and it showed that in a prospective multicenter setting that, uh,

16:37

there was, uh,

16:38

improvement in and progression-free survival freedom from skeletal events.

16:42

And then the big one, of course, overall survival in those patients who, uh,

16:46

received, uh,

16:47

Pluto on top of best standard of care therapy relative to those men who only

16:52

received best standard of care therapy. Let me, uh,

16:57

let me shift gears a little bit from what we know to what we think we might

16:59

know. And, uh, I think there's, uh, there's a drive right now. Uh,

17:04

but this is one of my, my academic interests. And, uh,

17:07

but it's also academic interests of other folks that, um, we,

17:12

we need to, we need to do a better job of, of sort of bringing in, in a,

17:16

into alignment, uh, everyone's interpretations of the same findings. Uh,

17:21

you'll find a lot of inter reader, uh,

17:23

a lot of inter reader variability right now in terms of how people approach

17:27

findings on, on scans. A solitary rib meta,

17:29

a solitary rib lesion may be a metastasis to one reader,

17:32

and it may be fibrous dysplasia to another reader. And of course,

17:35

all of this is gonna eventually, uh, hopefully shake out as,

17:38

as we collectively become more experienced and all read hundreds or even

17:42

thousands of these scans over the years. But we, uh, uh,

17:47

we do need some way of homogenizing our interpretations.

17:49

We need to communicate findings to our clinical colleagues, uh,

17:52

and other fields have already done this. And,

17:55

and Pi Rads is probably the big one. Uh, I,

17:58

I sort of came into being as a radiologist when Pi Rads already existed,

18:01

but all the stories I hear about, uh, sort of the pre py rad era was that, uh,

18:06

uh, you know, no one, uh, uh, everyone was reading prostate MRRI differently.

18:10

Everyone was interpreting what those reads meant differently,

18:13

and it was hard to sort of get a nice, coherent signal outta of that noise.

18:17

And Py Rad sort of brought everything into alignment,

18:19

where now the expectation is everyone's reading, uh, with Py Rads,

18:22

there may still be some inter reader variability, of course,

18:25

but at least PI Rad has provided sort of a structure that, that, uh, that can,

18:29

uh, uh, on which everyone can hang their hat. Now,

18:32

there are a number of competing approaches, but I'm only gonna talk about one,

18:36

and that's PSMA rads, uh, since it's kind of like birads or at least, uh, uh,

18:40

is potentially like by RAD or PI rads. And then it's a, a Likert scale of, of,

18:45

uh, how, uh, likely prostate cancer is in given patients.

18:50

But let me first show you some pitfalls as to why this would be an important

18:54

thing to do. Uh, obviously, if everything out is,

18:56

that's hot on the scan is cancer,

18:58

and everything that isn't hot on the scan isn't cancer, uh,

19:01

our lives would be a lot easier. Uh, but of course,

19:03

there are pitfalls in both directions. For example, a, uh,

19:07

a scan that may have negative or variable uptake, uh, as in this patient, uh,

19:12

may indicate that a patient has neuroendocrine prostate cancer,

19:14

neuroendocrine prostate cancer tends to lose PSMA expression.

19:18

And so you can get what are, uh,

19:21

absolutely biopsy proven positive liver metastases in a patient that doesn't

19:25

wind up having any PSMA uptake. Uh, and this is just the,

19:30

the biopsy evidence of that. So, uh, adenocarcinoma, uh,

19:34

has PSMA expression, but as you can see in those central couple of panels,

19:37

all that PSA expression goes away in the sort of truly neuroendocrine,

19:42

differentiated prostate cancer that was present in the liver of this patient.

19:47

Uh,

19:47

there are other things that are probably neo vascularized or inflamed that can

19:52

have PSA uptake. Uh, we don't necessarily always understand the biology of this,

19:56

but here's one example. This is a patient with a vertebral compression fracture.

20:00

This, uh, uh, this has nice linear uptake. It looks a lot like it would on, say,

20:04

an FDG PET scan, uh, but no vocality here to suggest that this is a, uh,

20:09

a, a pathologic compression fracture.

20:14

Maybe one of the more famous pitfalls in, uh,

20:17

NPSA imaging is the presence of, uh, peripheral ganglia.

20:22

So, dorsal root ganglia, cervicothoracic,

20:25

or stella ganglia and celiac ganglia are all variably

20:30

positive in patients. And the celiac ganglion is, is really the hardest one.

20:33

So that's the panel in the upper right there.

20:35

A couple of panels in the upper right,

20:37

it looks like a retroperitoneal lymph node.

20:39

It looks for all the world like a retroperitoneal lymph node,

20:41

and in some patients, it's quite prominent. In some patients, it's quite avid.

20:44

So it really,

20:45

it's incumbent upon us to know the anatomy well and not mistake these lesions

20:49

for, uh,

20:50

or not mistake these normal structures with variable uptake for sites of

20:54

prostate cancer. Uh, here's just another example of that. Uh, and here,

20:59

here's that, that celiac ganglion. So again, it,

21:01

it really does look like a lymph node. It's kind of kidney being shaped. Uh,

21:04

and this patient's relatively prominent and they have relatively prominent

21:07

uptake. Uh, but again, we just have to be very careful not to, uh,

21:11

not to mistakenly call those sites of prostate cancer. Uh, and then here's, uh,

21:15

here's an example of the bone. The bones are,

21:17

are full of hazards for those of us interpreting these scans. Uh,

21:20

here's a patient who had badgett disease in the sacrum, which, uh,

21:24

which had uptake throughout the sacrum. Uh,

21:27

you can see the bone scan there also had uptake throughout the sacrum. Uh, but,

21:30

uh, uh, but the MRI shows, uh,

21:33

shows that there's preservation of the intramedullary fat.

21:36

So this is just a nice example of Paget's disease. And as noted,

21:41

inflammatory things can, can have uptake. So here's a,

21:43

a patient with cortical lanar necrosis after getting radiation therapy to the

21:47

brain. Uh, you can see that there's uptake in the, uh, areas of enhancement,

21:52

uh, on the, uh, on the MRI. So this, uh, uh, this uptake is, uh,

21:57

is potentially confusing to us.

21:59

You'll see this in other parts of the body as well.

22:01

Patients who have gotten prior SBRT, uh, to say, a rib lesion,

22:05

the underlying lung, the post-radiation changes there, uh,

22:07

will often have low level diffuse uptake.

22:11

And then there's PSMA expression in non prostate cancer.

22:13

So this is primarily in the tumor, neo vasculature, the neovascular endothelium.

22:18

Uh, but, uh,

22:19

what we can potentially leverage this in these other cancers for diagnostic

22:22

purposes.

22:23

We also have to really be aware that if a pattern of apparent disease does not,

22:28

uh, comport with what we believe prostate cancer should look like,

22:32

we may have to suggest that another prostate or then another cancer type at

22:35

present.

22:35

Lemme show you a couple of examples of what used to be called conventional renal

22:39

cell carcinoma, we now call clear cell renal cell carcinoma. Uh,

22:43

he adheres a patient with relatively widespread metastases,

22:46

perhaps the clue on his whole body map images that he's missing a kidney. Uh,

22:50

but these were all, uh, these were all metastatic renal cell carcinoma, uh,

22:54

and specifically clear cell renal cell carcinoma. And you can see even very,

22:58

very small lesions there. There's quite avid uptake. But I would, uh,

23:02

I would posit that something like a, uh,

23:04

subcutaneous perineal metastasis would be highly unusual in prostate cancer.

23:09

Something like an intramuscular metastasis would be highly unusual in prostate

23:13

cancer. But those things may indicate the presence of,

23:15

of something like a renal cell carcinoma that has more unusual patterns of

23:19

disease spread. Uh,

23:21

brain parenchymal metastases would also be very unusual in prostate cancer. Uh,

23:25

here's one from, uh, from, again, a clear cell renal cell carcinoma.

23:30

And then relative to FDGI, I tend to think FDG gets kind of a bad rap for RCC.

23:34

It's not bad in the metastatic setting, uh, but in a patient, uh,

23:37

contemporaneously, imaged with, uh, with F DG and PSMA, uh,

23:41

the PSMA not only had higher uptake in known lesions, but also, uh, uh,

23:46

but also, uh, wound up, uh,

23:47

picking up a couple of lesions that would've been missed on the F DG scan.

23:52

And then one last example of a clear cell RCC. This is a patient who was imaged,

23:56

uh, right before he passed away, unfortunately,

23:58

and then underwent a rapid autopsy. And so, uh,

24:02

sites of disease that would've potentially been very hard to biopsy,

24:04

such as intramuscular metastases, uh, were, were available for,

24:08

for histopathologic inspection. So, not only does, uh,

24:12

PSMA do a sensitive job of picking up clear cell renal cell carcinoma mets, uh,

24:17

it also is very specific for, for those same metastases. Uh,

24:21

I would say it is, uh, uh, again,

24:23

impor perhaps most important for those of us reading this, this, uh,

24:27

modality for prostate cancer,

24:28

just to be aware that very strange patterns of spread from prostate,

24:32

what look like strange patterns of spread spread for prostate cancer may be

24:36

something else. And again, we, we have to be aware,

24:38

we do have to be aware of that. Uh, lemme show you, uh, a couple of,

24:42

uh, these are sort of high-grade glioma examples. I,

24:46

I don't think this is a specific finding.

24:47

As you saw post-radiation changes can also have uptake. So the, uh, uh,

24:52

so while not specific, uh, this, uh,

24:54

areas of blood brainin barrier breakdown will be highlighted by,

24:57

by P SMM A scans,

25:01

there's at least some uptake of PSMA and, uh,

25:06

in thyroid cancer. Uh, I've overall been, however,

25:09

disappointed with our experience with trying to image these patients. Uh,

25:12

there seems to be both disease that's iodine avid psm, a negative, uh,

25:17

disease that is iodine avid, PSMA avid, and disease. That is, uh,

25:22

I, uh, uh, let's see. I forget even what I said. Uh, that's, uh,

25:26

iodine negative, PSM a avid. So all those things,

25:29

all those permutations can exist. So I don't know that there's a great role for,

25:32

for PS APE moving forward in thyroid cancer.

25:36

But all of those brings us to the idea that, again,

25:38

we have to have ways of either, uh,

25:41

structuring our interpretation of what we think these various things are,

25:45

these various pitfalls or non prostate cancers.

25:48

And we have to communicate that to our clinical colleagues, uh,

25:52

and help them sort of make decisions as to, uh, what new workup, uh,

25:56

what new workup steps might be for the patient if it's not so straightforward.

26:00

It's just they have prostate cancer in the site, in this, in that site. Uh,

26:04

in our clinical, uh, in our clinical, uh, workflow, uh, we use PSMA rads,

26:09

uh, I suppose to some degree, this is our bias. We, we, uh,

26:13

sort of developed and validated this system. So,

26:15

so it does tend to be something that, uh, that we're very comfortable with. Uh,

26:19

there, there are other systems out there, uh, promise. Uh,

26:23

there's an original sort of EA and m uh, Delphi consensus model.

26:28

And then there's something called E-P-S-M-A,

26:30

which tried to subsume all of those things together. Uh, but, uh, wound up, uh,

26:34

perhaps becoming a little over complicated. And then, uh, there are new efforts.

26:38

There's now a PSMA RADS 2.0. There's also Promise 2.0, uh, and all,

26:43

but all these things have the same goal of trying to standardize some aspect of

26:47

our approach to the scan. Uh, I'll very, very briefly show you sort of, uh,

26:52

what we, uh, what we've done with, uh, with sort of PS MA RADS version one,

26:56

uh, which was really just a, uh, a way to, again,

27:01

create a five point scale of likelihood of prostate cancer.

27:04

One is definitively not prostate cancer, uh,

27:07

either from there not being any findings or from things that do have uptake,

27:12

uh, being either biopsy proven or pathognomonic shown on some other imaging to,

27:17

uh, not be sites of prostate cancer. PS PSMA rads.

27:20

Two things that are very unlikely to be prostate cancer, uh,

27:23

but perhaps a low level uptake in something that could be obscuring aside of

27:27

prostate cancer. PSMA RADS three has, has subcategories.

27:32

Uh, and those include, uh, RADS three A for, uh,

27:37

indeterminate soft tissue RADS three B for indeterminate bone. Uh,

27:41

just because of the workup. Uh,

27:43

the sort of workup steps for each of those might be a little different.

27:45

That's why we, we subcategorize them. And then things that, uh, uh,

27:50

PSMA RAD three, three C or things might be other cancers, uh,

27:53

but that have uptake. Uh,

27:55

here you can see an example of a potential lung cancer. Uh,

27:57

this was an isolated finding in this patient. We will always, within reason,

28:02

try to get tissue on isolated findings in the lungs. Uh,

28:05

although they often do turn out to be prostate cancer,

28:07

that is still a relatively rare pattern of,

28:09

of disease spread for prostate cancer. So our,

28:12

our preference is to make sure that it isn't an underlying lung cancer. Uh,

28:15

you can see there are also things that might be cancer,

28:17

such as a relatively large pulmonary nodule that lack uptake of,

28:21

of psm a targeted agents. And those we would categorize as a psm a rats three D.

28:26

And then the fours and fives are easy. Fours are things that are, uh,

28:30

almost certainly prostate cancer, but don't necessarily have a CT correlate.

28:33

Might be a submeter lymph node,

28:35

it might be a bone lesion that's marrow based and doesn't have a,

28:38

an associated sclerotic reaction. And then those things that, uh,

28:42

are either things like enlarged lymph nodes or sclerotic bone lesions that

28:46

have uptake, uh, those go into the five category. So it's,

28:50

it's a relatively simple approach, uh,

28:52

but it does make you come down hard on sort of giving some kind of number or

28:56

decision to things that, uh, that may, uh, uh,

28:59

where it may have to make a decision.

29:00

Does this need additional workup or is this definitively benign? Or is it, uh,

29:04

is it not definitively benign? Alright, and this is just an example. Uh,

29:09

I won't won't go into the details here. Um, this, uh, this approach has a, uh,

29:13

has a higher high inner observer reliability, uh,

29:17

both on sort of an overall scan score, highest score for,

29:20

for an individual lesion, as well as individual lesion scores.

29:24

These aren't perfect, uh, and experience does matter.

29:27

Experience readers will tend to cluster together a little bit more,

29:31

but relative to something like py rads, the,

29:33

the repeatability here is actually substantially higher.

29:37

And those things that are indeterminate, unfortunately,

29:39

really are indeterminate. So if it's an indeterminate lymph node, uh,

29:43

when you follow it up over time,

29:44

maybe about 75% of those will manifest as as having been true positive,

29:49

and maybe 25% won't. Uh, and, uh, bone lesions, it's actually worse.

29:54

So with bone lesions, the, uh, uh,

29:56

only about 20% of indeterminate bone lesions wind up, uh, wind up being, uh, uh,

30:00

being true positive on follow up.

30:03

And advanced reconstruction algorithms can help a little bit.

30:06

Here's just a patient who is imaged with and without, uh, a, uh,

30:10

point spread function reconstruction. And you can see, uh,

30:12

a lymph node that's may be a little wishy-washy, although probably callable, uh,

30:16

the top couple of panels, uh, is, uh, is more definitively callable in the,

30:21

uh, in the bottom panels. Alright, so let me,

30:25

let me move on to some emerging ideas. So these are things we, we don't know,

30:28

but, but hopefully will at some point. And there are, uh,

30:32

there are things here I'll probably go by pretty quickly, uh, make sure we,

30:35

we have time for, for questions. 'cause we probably covered sort of the,

30:38

the really practical aspects of things up, up to this point. And now we're,

30:42

we're going to, uh, do some thought exercises about maybe where the field of,

30:45

of PSMA PET is headed. So let me, uh, but let me launch, uh, right into things.

30:50

Uh, so very briefly, image quantitation.

30:52

So I'm a big believer in image quantitation from perspective of biomarker

30:56

development.

30:57

I don't think there are meaningful P-S-A-P-S-M-A-S-U-V

31:02

cutoffs that we can really apply, but I can tell you that, uh,

31:06

liver uptake is highly reproducible, uh, across various radio tracers. Uh,

31:11

they'll all have different liver uptakes,

31:12

but the liver uptake is sort of the most repeatable organ in, in PSMA pet.

31:17

Uh, there's also, there's an agent over in Europe called psm,

31:20

a 1007 that has very high liver uptake because it's hepa biliary excreted. Uh,

31:25

so that, I actually don't know if the,

31:27

if the quantitation is most repeatable in the liver.

31:29

But for the agents that are FDA approved in the us this, uh,

31:32

this would hold true. Uh,

31:34

there probably is some sort of tumor sync effect when you see really extreme

31:39

degrees of, of tumor infiltration, particularly in the skeleton.

31:43

There's definitely a drop in uptake in other organs,

31:46

but it doesn't really come into play at sort of the, uh,

31:50

the levels of disease that we often see here.

31:53

You can see three patients who have varying levels of, of disease involvement,

31:57

but it really looks like their,

31:58

their organs are all very similar and degrees of uptake. Again,

32:02

there's a little bit of a tumor sink effect, but it is, uh,

32:04

there are other things that affect PSMA uptake that,

32:07

that would also play a role in sort of the day-to-day variability.

32:11

So it won't be all tumor syn effect, uh,

32:13

or just differences in sort of tumor volume between patients.

32:18

Uh, let's see. I think I'm gonna skip that. Uh, but here, uh, uh,

32:23

but here I'm gonna, I'm gonna maybe slow down just a little bit because I,

32:26

I think here we have a couple of important ideas. So, uh,

32:30

we know that, uh, we know that, uh, test retest studies are not easy to do. Uh,

32:36

patients, uh, getting imaged, you know, at one time, and then again, you know,

32:39

within a week without any intercurrent therapy or, or initiation of new therapy,

32:44

uh, it's hard to sort of line all that up. It's expensive to do that.

32:47

It's lots of scanner time. But, uh, we, uh,

32:50

we're lucky enough to have the Prostate Cancer Foundation, uh,

32:53

fund us to do a test retest trial, uh,

32:55

with PA in patients with metastatic disease, uh, with, uh,

32:59

with A-P-S-M-A PET scan. And turns out that, uh,

33:03

that the scans are highly repeatable. So, uh,

33:06

even in low volume disease or in high volume disease, uh, the SUVs are,

33:11

are robust. Uh, and, uh,

33:13

here you can see some bland alt plots and correlations. Uh, the,

33:18

I will say that the p the SUVs are more robust at higher SUVs.

33:23

So as you get these patients who have metastatic castration resistance,

33:26

extensive volumes of disease, and very high SUVs,

33:29

those SUVs will be very repeatable. So the idea that, uh, a,

33:34

a solitary PSMA scan at, at,

33:36

at a sort of a time point prior to initiation of something like pto, uh,

33:40

is going to tell you something about the dosimetry of those patients and how

33:44

much dose their tumors going to get.

33:47

I think there's a lot of validity to that idea.

33:49

I don't know that we're leveraging that very well,

33:50

but hopefully we will at some point. But, uh, but again, yeah,

33:53

higher SUVs are more robustly repeatable, uh,

33:57

and I think that's in and of itself an interesting finding.

34:00

The SUVs and FDG PET just don't have the dynamic range to actually have told us

34:04

that, although it's relatively intuitive, perhaps. Uh, but we,

34:07

we didn't know that prior to the advent of, of pssm a pet, uh,

34:11

where we would have SUVs from down at, uh, you know,

34:14

0.8 or 0.9 sometimes in early subtle sites of disease up to over a hundred and,

34:19

uh, and really avid sites of disease. Alright,

34:23

this is just more about that. Okay. So, radios has, has gotten a lot of play,

34:27

and it would be great if radios could perhaps, uh,

34:30

provide us some insights or predictive biomarkers, uh,

34:33

that might tell us something about PSMA pet. Unfortunately,

34:38

I don't tend to be a huge believer in radios for PET in general. Uh,

34:42

there are broad categories of things that tend to be, uh,

34:46

sort of low frequency features of the images,

34:48

things like entropy or homogeneity, uh,

34:51

that do have some repeatability or reproducibility, uh, at the, uh,

34:55

at the radio level. Uh, I'll also say though, that, uh, um,

35:00

that most radio features are not reproducible, and that has to do with,

35:04

with aspects of the, the pet reconstruction matrix, where there are, uh,

35:10

inestimable parameters that are high frequency features of the images that are

35:14

basically going to never be repeatable,

35:16

because we're just throwing random numbers in that, uh,

35:18

in that reconstruction matrix. I, uh, probably need an update this,

35:22

if you would like to read more. Uh, this, uh, manuscripts now been published in,

35:26

uh, uh, in, uh, the prostate, uh, Rudy Veer, who's done a lot of,

35:31

uh, a lot of this work. Uh, uh, who's over at Warburg, uh,

35:34

is the first author on that paper. It, it's worth a read,

35:37

it's kind of an interesting fundamental pet, uh,

35:41

fundamental pet finding that that just happens to be applied in this case to

35:44

PSMA and artificial intelligence. You know,

35:48

I think every radiologist worries that at some point, uh,

35:50

these algorithms are gonna get so good at looking at images that we're gonna be

35:52

out of jobs. Uh, I, I think our,

35:55

our jobs are hard enough that that isn't like right on the horizon, but, uh,

36:00

but it is certainly going to come to pass that AI is gonna replace a lot of what

36:04

we do day to day.

36:04

So we do have to at least have that in the back of our minds right now. AI,

36:08

I think, is gonna be more of a friend to us.

36:10

It's gonna help us with lesion classification.

36:12

We can do whole body tumor burden assessments, uh,

36:15

and prognostication and decision-making will,

36:17

will hopefully be things that we can start to lean on AI for a little bit.

36:23

Uh, with psm, a pet AI is already pretty good. I,

36:26

I think it's better at hotspot detection than it really is at classification.

36:30

Uh, the, the FDA approved product that's out there, that's,

36:34

that's a sort of PSMA AI tool, uh,

36:36

will pick up solitary rib lesions that are fibrous dysplasia and things like

36:40

that. It's basically picking up things that are outside of the normal bio

36:43

distribution. It does a very good job at that.

36:45

It provides a quantitative readout, uh,

36:47

but it isn't at the point where it's helping us make decisions about what

36:52

the lesions really are or what the site's uptake really are and how seriously we

36:56

need to take them.

36:57

That's really where we have to go for AI to ultimately be useful with,

37:01

for us in this, this context. Uh, but, uh, but it, uh, isn't quite there yet.

37:05

I think it's, it's a question of data and a follow up and, uh,

37:10

and really good data. So data that is based on, uh, uh,

37:15

that's based on either long-term follow-up or, or histopathology correlations.

37:19

And, and that kind of data is sort of hit and miss in the field.

37:25

I, uh, I always wonder if we can supercharge some of these sort of, uh, um, the,

37:30

these very image oriented algorithms with even better

37:35

images. Uh, this is just an example of cinematic rendering of, of a, of a psm,

37:39

a PET ct, which is, uh, uh, which is a, a technique that was pioneered by,

37:44

by Siemens. Uh,

37:45

it's available on their VB 40 and above workstations. I think.

37:50

Uh, and, uh, this part isn't necessarily out there and ready for prime time yet.

37:54

Uh, to differentiate the PET data from the CT data,

37:57

you have to internally light the PET data. Uh, but nonetheless, I'd,

38:00

I'd really be curious if, uh, if, uh,

38:03

GPUs can sort of derive more out of these, uh,

38:08

interesting and color-coded, uh, and sort of, uh,

38:13

1, 1 1 scan and one image kind of image, uh,

38:17

than they can out of sort of just the two D images being, being fed,

38:20

generally fed to them. We, uh,

38:25

in addition to, uh, in addition to thinking about radios and ai, uh,

38:29

we need to also think just more broadly about imaging biomarker development. I,

38:34

I think this is, this is where a lot of the field of PSA is headed. Uh, again,

38:37

some of that'll be driven by, by AI and, uh, and,

38:40

and perhaps by radios to some degree. But,

38:43

but we could do simpler things than that. We can, we can look at, uh,

38:48

say in this particular, it was very nice study from, uh,

38:51

from the Netherlands and Australia. They looked at patients who were

38:56

pssm a positive, uh, on the scan, uh, histopathology negative,

39:01

or I'm sorry, histopathology positive.

39:03

And then patients who were negative on the scan and histopathology positive.

39:07

So essentially false negative scans. Those false negative scans,

39:10

those patients did a lot better.

39:12

So the false negative scan is a prognostic biomarker of how the patient's going

39:16

to do with surgery and how long their biochemical, uh,

39:19

recurrence-free survival will be. Uh, another example,

39:23

perhaps of an imaging biomarker is that if we, uh,

39:26

if we look at the phase two oral trial, this is a post-hoc analysis,

39:29

but nonetheless, uh, a nice post-hoc analysis. It shows that if you, uh,

39:33

if you treat everything that's PSA avid, uh,

39:37

and visible on conventional imaging versus everything that's only visible on

39:41

conventional imaging,

39:43

the patients who got everything that was PSMA AVID treated,

39:45

did better than those patients who may have had PSA avid disease that wasn't

39:50

visible on conventional imaging. Not a surprise perhaps, but shows us that,

39:55

uh, that we have to be using sensitive imaging if we're going to do this.

39:58

And then if we use that sensitive imaging,

40:00

we know that those patients are generally gonna do very well.

40:05

Uh, now, uh, maybe switching gears a little bit, but,

40:09

but keeping in the sort of imaging biomarker space, uh,

40:14

initial experience with, with PSMA in response to ADT.

40:18

So there's a, uh,

40:19

poorly understood biology of increasing PSMA expression, uh,

40:24

with a decrease in androgen signaling.

40:28

That's at least what happens in the lab is whether that really happens in

40:31

patients is hard to say. Uh,

40:33

here is a patient in whom that did seem to hold true,

40:35

where the initiation of ADT uh,

40:37

brought out new lesions and made other lesions more avid.

40:40

It seemed like this was going to be a way that we could, um,

40:43

all of a sudden now give Pluto after a shot of ADT, uh,

40:47

and the patient's gonna respond really well.

40:49

Or if we're trying to do metastasis directed therapy,

40:51

maybe we find one new site of disease and, and that changes, uh,

40:55

how we treat the patient. But, uh,

40:57

but it turns out it's actually much uglier than this. Uh, and so if,

41:00

if you look at sort of a different context of cutting off androgen signaling was

41:04

abiraterone or enzalutamide, uh, patient scans do all sorts of things.

41:08

Some get hotter, some get colder, some have new disease, some don't. Uh,

41:12

but you can sort out on kind of a whole body level what the,

41:15

what the imaging biomarkers are, and they have associations with, uh,

41:19

time to therapy change and overall survival.

41:22

So it's exciting to think that our changes and uptake are going to, uh,

41:27

sort or sort of a dynamic readout of changes and uptake can tell us how a

41:31

patient is doing, when they're gonna need to change therapy,

41:33

how long they're gonna live. Uh, AI is certainly gonna help us with this. We,

41:38

we have to really mine, mine out these, uh, uh,

41:41

these studies and do larger studies that are really gonna help us with this

41:45

and talk about that. Uh,

41:49

and PSA PET has the, the opportunity to help us in places where,

41:55

uh, where something like APSA may fall apart.

41:58

So patients will eventually develop what may be PSA non screening tumors.

42:01

They may still have PSMA expression,

42:03

and there are difficult situations like bipolar androgen therapy where patients

42:07

are androgen deprived, but every six weeks gets super physiologic testosterone.

42:11

Well, their PSAs, of course, spike when they get that testosterone,

42:14

but does that mean they're progressing or is,

42:16

or are we just maintaining sort of the sensitivity of their, their disease for,

42:20

for, for androgen uh, signaling, uh, for,

42:24

for things that target androgen signaling. And turns out,

42:27

if you have any lesions you are progressing,

42:29

then eventually that'll manifest on CT and bone scan.

42:32

But you can tell it earlier on P SSM A, so that early time point P SM a,

42:36

again, may be sort of a, a real biomarker for,

42:40

for whether the patient's doing well on the disease, uh, well on that,

42:43

that given therapy or not, and whether they need to change therapy. With that,

42:48

I'll, I'll kind of wrap up and I, I think I'm about on track to, to have, uh,

42:51

some good time here to answer some questions. But, um, I would just like to,

42:55

to wrap up by saying there are already multiple indications for diagnostic PSMA

42:58

based imaging. We've talked about those,

43:00

but are interpreted footfalls that suggest a need for a structured reporting

43:03

approach. Doesn't have to be PSMA rads, and that's my bias, but, uh, uh,

43:07

P smma RADS Promise E-P-S-M-A. Uh, I think, uh,

43:11

all of us should at least think about these, uh,

43:12

think about these possibilities. And then, uh,

43:15

we're just starting to understand PSMA targeted PET findings as imaging

43:17

biomarkers. And we have a lot of work left to do, uh, uh, to really, uh,

43:22

to really make that happen.