Quantitative Morphology in Kidney Research
February 13-14, 2012 Conference Videos

Quantitative Morphology to Generate Surrogate Outcomes: Morphologic Changes as Surrogate Outcomes – How Are These Defined and Validated? FDA Perspective
Aliza Thompson and Lynne Yao, U.S. Food and Drug Administration (FDA)

Video Transcript

00:00:00,000 --> 00:00:09,266
JEFFREY KOPP: So, the last session was about starting to think about taking these tests to the clinic and the next speaker from the FDA is going

00:00:09,266 --> 00:00:20,566
to help us think at least some of these studies we’re looking at potentially might be brought to the clinical practice at some point and therefore are

00:00:20,566 --> 00:00:27,132
going to have to pass through the FDA if they’re going to be actually used in clinical medicine as predictors to help us choose therapies and so

00:00:27,133 --> 00:00:35,366
forth. So, what is the road map—to use that commonly-used term—that we need to be thinking about now as we develop these various

00:00:35,366 --> 00:00:45,366
tests? Our speaker is Dr. Aliza Thompson, a nephrologist who is a team leader with CDER, Center for Drug Evaluation and Research, at the

00:00:45,366 --> 00:00:55,132
FDA and she is the perennial nephrology group go-to person when we want somebody from the FDA to come help us think through what the

00:00:55,133 --> 00:01:07,933
process is; in this case, biomarker and particularly, renal pathology biomarkers. So Aliza, thank you very much.

00:01:07,933 --> 00:01:17,633
ALIZA THOMPSON: Hi. Can everyone hear me? Maybe I’m too loud. Thanks for this invitation to speak and Happy Valentine’s Day to everyone.

00:01:17,633 --> 00:01:28,633
Clearly, the work that’s being discussed at this meeting has many important applications and I’m going to focus on one application, which is its

00:01:28,633 --> 00:01:41,199
use as a tool in drug development. If you look at drug development, certainly histology has been used in the preclinical setting in animal studies as

00:01:41,200 --> 00:01:50,533
a way to inform our understanding of a drug’s potential for nephrotoxicity in the clinical setting, but to a much lesser extent it’s been used,

00:01:50,533 --> 00:02:00,366
actually, within the context of clinical trials as a way to facilitate the development of therapies that could potentially improve the lives of people

00:02:00,366 --> 00:02:12,666
with kidney diseases, and so I’m very hopeful that this will be one of many discussions we will have about the use of this as a tool within the

00:02:12,666 --> 00:02:29,199
context of drug development. Let’s make sure I can work this. Okay. And so, the topics I’m going to touch upon in this talk are a few. One is I’m

00:02:29,200 --> 00:02:35,666
going to give you a general sense of what the efficacy requirement is, then I’m going to talk a little bit about how we evaluate potential

00:02:35,666 --> 00:02:44,232
surrogate endpoints, talk about the use of surrogates and their implications for assessing a drug’s safety profile, and then from there talk

00:02:44,233 --> 00:02:51,833
about morphologic changes in the kidney as a potential surrogate endpoint for the purpose of drug development, keeping in mind that, of

00:02:51,833 --> 00:02:58,833
course, there are a lot of ways we could use changes in kidney morphology within the context of drug development, but I’m just going to focus

00:02:58,833 --> 00:03:12,699
here. Of course, the disclaimer. What you want to hear is what the FDA thinks, but regrettably I can’t tell you what the FDA thinks. So, the regulatory

00:03:12,700 --> 00:03:21,600
framework. Before a drug can be marketed in the United States it has to be shown to be safe and effective for its intended use. This requirement

00:03:21,600 --> 00:03:31,366
really arose just within the last 50 years and came about in the setting of a safety crisis. Probably as many of you know, thalidomide had

00:03:31,366 --> 00:03:40,099
been marketed outside the United States both as a sedative and as a treatment for morning sickness in pregnant women, and unfortunately it

00:03:40,100 --> 00:03:51,900
was later discovered that this was a horrible [---] and led to the birth of many children with marked abnormalities, specifically in limb formation. In the

00:03:51,900 --> 00:04:01,666
U.S. the drug wasn’t approved and the crisis was narrowly averted but nonetheless, as a result of this crisis, it led to the passage of laws

00:04:01,666 --> 00:04:11,732
that dramatically expanded both the responsibility of the FDA and also its power and essentially led to what many would call the Efficacy

00:04:11,733 --> 00:04:22,533
Requirement. If you look at actually what the law said it really spoke more and speaks more to the quantity and the quality of data that you need in

00:04:22,533 --> 00:04:31,499
terms of showing that your drug works. So, what the law refers to is “substantial evidence” of efficacy and that this evidence is supposed to

00:04:31,500 --> 00:04:39,833
come from adequate and well-controlled trials. In practice what that’s generally been taken to mean is that you need two studies, you need P values

00:04:39,833 --> 00:04:50,333
less than .05 in those studies, and that those studies should be randomized studies with either an active or a placebo control. But if you ask sort

00:04:50,333 --> 00:04:56,966
of what endpoints you need to show your drug effects, that was something that really came about later and something that the courts really

00:04:56,966 --> 00:05:10,266
addressed, and what the courts concluded later on was that that endpoint that you had to affect actually had to be clinically meaningful. So again,

00:05:10,266 --> 00:05:16,732
drug approval can be based on clinically meaningful endpoints which the agency has come to interpret as an endpoint that reflects

00:05:16,733 --> 00:05:24,366
how a patient feels, functions, or survives; or we’ve also allowed the use of surrogate endpoints and from a regulatory standpoint,

00:05:24,366 --> 00:05:31,466
surrogates have a very specific meaning. I know when we refer to surrogates often we just mean that they’re substitutes, but from a regulatory

00:05:31,466 --> 00:05:39,699
standpoint a surrogate is supposed to be some…can be a laboratory measure or a physical sign, but it’s used in therapeutic trials as a substitute or

00:05:39,700 --> 00:05:48,500
in lieu of the actual clinical outcome that you care about. And so really, surrogate endpoints are expected to predict the clinical benefit of your

00:05:48,500 --> 00:05:59,900
therapy on that outcome of interest. From a regulatory perspective we view two types or we have two types of surrogates. One is an

00:05:59,900 --> 00:06:07,700
accepted surrogate, something like blood pressure which, if you show an effect on that endpoint, it’s a basis for outright approval. You’re

00:06:07,700 --> 00:06:15,100
done; we believe that your therapy is providing a benefit to patients. There’s another type of surrogate, however, and that’s a surrogate

00:06:15,100 --> 00:06:24,866
where there’s a little bit less certainty about what it’s predicting and those types of surrogates are used in the setting of accelerated approval for

00:06:24,866 --> 00:06:34,466
diseases that are serious or life-threatening, and this type of surrogate or approval using this type of pathway comes with a requirement after

00:06:34,466 --> 00:06:46,166
approval to actually complete studies verifying the benefit in the post-marketing setting. It’s obvious to all, I think, there is a hierarchy of

00:06:46,166 --> 00:06:55,632
endpoints, so showing your treatment effects and mortality is incredibly compelling evidence of a drug’s efficacy. Showing that your treatment

00:06:55,633 --> 00:07:07,166
affects a surrogate can provide evidence of efficacy, but it’s just less compelling. I just also want to make the point on this slide that ultimately

00:07:07,166 --> 00:07:16,799
when we look at drugs and we try to understand what they do we often think about them in terms of what their efficacy is and then, on the other

00:07:16,800 --> 00:07:26,666
hand, what their safety is, but ultimately what we need to get some understanding of is whether or not they provide net benefit. So, I just also want

00:07:26,666 --> 00:07:36,732
to make the point that if you’re having a mortality benefit or you’re affecting cancer, that the tolerance for risk is higher; you can have a

00:07:36,733 --> 00:07:45,733
chemotherapy that has nephrotoxicity. On the other hand, if your effect is just on blood pressure, your tolerance for risk is much less

00:07:45,733 --> 00:07:57,399
and we’re probably not going to approve an antihypertensive that has marked nephrotoxicity. So, why use surrogates? Well, I think this is

00:07:57,400 --> 00:08:07,333
probably obvious to all of you. The appeal of surrogates is that they can be used in clinical trials to establish effects in smaller studies and

00:08:07,333 --> 00:08:14,333
allow us to conduct studies of shorter duration, and in terms of at least nephrology drug development, something that often rises is that

00:08:14,333 --> 00:08:24,133
the endpoints that we really care about may be rare events or they may be events that occur very late in the disease course, and so to a large

00:08:24,133 --> 00:08:32,133
extent, surrogates are seen as a way to enable drug development where it might otherwise not be possible, in essence, for early stages of

00:08:32,133 --> 00:08:43,299
disease or in settings where the disease is less severe. So, how do we look at surrogate endpoints? How do we evaluate them? I’ll tell you

00:08:43,300 --> 00:08:51,700
about the approach in a minute but I just want to open this up by pointing out that the approach that we use in our evidentiary standard is

00:08:51,700 --> 00:08:59,100
probably higher than that that’s used by the medical community in everyday practice or even what’s endorsed or the approach taken in

00:08:59,100 --> 00:09:10,400
guidelines. I think there are two reasons for this. One is that how conservative we are in our thinking depends in part on who we are making

00:09:10,400 --> 00:09:22,900
the decisions for, but it’s also because it depends upon how definitive our recommendation must be, and I want to illustrate this with two examples. I’m

00:09:22,900 --> 00:09:33,266
told, for example, that I have cancer. I may be very willing to try a therapy where there’s not a lot of data to support it; it’s my body and I make a

00:09:33,266 --> 00:09:44,366
decision. However, practicing as a physician, I may put a little bit more caveats around it and may not recommend it to a patient, and certainly in

00:09:44,366 --> 00:09:54,866
writing a practice guideline, I may not endorse it at all. So, that’s one example. Then in terms of the definitiveness of the recommendation, I want to

00:09:54,866 --> 00:10:02,266
make the point that in everyday practice when you see patients—those of you who see patients and make recommendations for therapies—you

00:10:02,266 --> 00:10:12,099
can make statements like, “Well, it hasn’t been shown but there are some data out there that suggest it might be beneficial,” when you all or

00:10:12,100 --> 00:10:18,766
those of you who are involved in writing the guidelines write guidelines, you can grade them. You can say, “Well, this is not based on strong

00:10:18,766 --> 00:10:30,399
data but we, as experts, think it’s maybe a good idea.” Unfortunately or fortunately, we as an agency don’t get to qualify our statement that the

00:10:30,400 --> 00:10:39,366
drug is effective or going to provide net benefit. The assumption is once it’s approved by the agency that at least in the population that’s

00:10:39,366 --> 00:10:46,566
described in the label, it should work. So again, what you’re about to hear is maybe a little bit more conservative but I think there’s a reason for

00:10:46,566 --> 00:10:55,466
that. So, what are the types of data that we often look to when we’re evaluating surrogate endpoints? I think a key thing is whether or not

00:10:55,466 --> 00:11:03,966
there’s a strong reason to believe that the biomarker is actually on the causal pathway to the disease. Other types of data that we look at

00:11:03,966 --> 00:11:11,532
include observational data, which essentially show a relationship between the marker and the outcome of interest; and finally—and this is sort

00:11:11,533 --> 00:11:19,799
of the highest level of evidence—is the extent to which there are actually data from interventional trials showing that your treatment’s effect on the

00:11:19,800 --> 00:11:32,600
marker predicts the treatment’s effect on the clinical outcome that we’re interested in. So, the agency has over the years adopted some

00:11:32,600 --> 00:11:39,466
surrogates, not a lot, and I think these examples are going to show you some of the different approaches that have been used or the types of

00:11:39,466 --> 00:11:49,132
data to support surrogate endpoints. So, let’s take the example of blood pressure as a surrogate. Why do we buy it? Well, over the years there

00:11:49,133 --> 00:11:59,233
have been numerous intervention trials using drugs from distinct pharmacologic classes and that experience has told us that treatment effects

00:11:59,233 --> 00:12:08,666
using these drugs on blood pressure ultimately predicts treatment effects on cardiovascular outcomes, so that’s one example. However,

00:12:20,766 --> 00:12:22,966
there’s another realm of surrogates and I would say something like: electrolytes fall in that realm. Certainly you know there are drugs that are

00:12:22,966 --> 00:12:31,032
approved to lower potassium levels. We also have now a whole class of drugs that been approved to raise serum sodium levels, and I

00:12:31,033 --> 00:12:35,166
would argue that those drugs or the use of sodium or potassium or phosphate as a surrogate endpoint, what didn’t lie in the same type of data

00:12:35,166 --> 00:12:43,466
that we have for blood pressure, but it more lies in the realm of biologic plausibility. We know that when your potassium goes very high you end up

00:12:43,466 --> 00:12:50,466
in a cardiac arrhythmia and so there’s a lot of plausibility that potassium is causing that, and we know that when you lower it you can see the

00:12:50,466 --> 00:12:58,999
changes in the ECG, and similarly it would argue for something like sodium, you know, that as levels get very low you see mental status

00:12:59,000 --> 00:13:05,200
changes, you can see cerebral edema, herniation, and we know that bringing the sodium back up can reverse some of these things. So

00:13:05,200 --> 00:13:15,200
again, two ways of approaching a surrogate are validating it and I again want to stress that when there’s biological plausibility that it’s really on the

00:13:15,200 --> 00:13:24,133
causal pathway; in some ways it’s an easier route. I would also like to make the point that it’s through all these data and intervention trials using

00:13:24,133 --> 00:13:32,799
these different agents that showed that effects on blood pressure predicted outcomes. I’m going to argue that. That’s, in part, why we think high

00:13:32,800 --> 00:13:43,900
blood pressure actually causes strokes and heart attacks in those types of things. So, surrogate endpoints. Certainly it is not unreasonable, or in

00:13:43,900 --> 00:13:51,800
fact, it’s very sensible to think that a marker that could identify patients that are at risk of poor outcomes is also going to predict a treatment’s

00:13:51,800 --> 00:14:02,133
benefit. However, this isn’t always the case and I just want to point to some examples of where markers that have performed very well at

00:14:02,133 --> 00:14:11,099
predicting a patient’s risk haven’t done such a good job at predicting a treatment’s effect on clinical outcomes, and I’ll pull one from the field of

00:14:11,100 --> 00:14:21,400
cardiology and one familiar to perhaps all of you from the field of nephrology. So, let’s talk about something like ventricular premature

00:14:21,400 --> 00:14:30,566
depolarizations. They were a good risk marker for sudden and non-sudden cardiac death after MIs. Unfortunately, when some antiarrhythmics

00:14:30,566 --> 00:14:39,832
were used in efforts to suppress these arrhythmias, you actually saw more deaths in the arm that got these treatments. Switching over to

00:14:39,833 --> 00:14:50,999
the field of nephrology and hemoglobin levels, low hemoglobin levels do a good job at identifying patients in populations at greater risk for cardiac

00:14:51,000 --> 00:15:00,266
disease and mortality and certainly do so in the area of chronic kidney disease. However, in studies where erythropoiesis stimulating agents

00:15:00,266 --> 00:15:11,399
were given in efforts to improve hemoglobin levels, actually using these drugs to target higher versus lower levels led to worse outcomes. So

00:15:11,400 --> 00:15:20,300
again, you can be a good marker of risk but not actually predict your treatment’s benefit. So, why do surrogates fail? Well, sometimes it’s simply

00:15:20,300 --> 00:15:29,366
because we got the relationship wrong but a lot of people believe that what’s happening in a lot of settings is that the drug has unexpected and off-

00:15:29,366 --> 00:15:39,132
target effects and that these are essentially negating the benefit, or worse, actually resulting in harm to patients. I just want to point out that

00:15:39,133 --> 00:15:45,033
this isn’t an issue just for surrogate endpoints but anytime you use some sort of intermediate endpoint which isn’t really the ultimate endpoint

00:15:45,033 --> 00:15:57,266
that you care about. So, I just want to make a few quick comments about safety because I think the use of surrogate endpoints has important

00:15:57,266 --> 00:16:04,099
implications for assessing a drug’s safety. Compared to how we look at effectiveness, safety is certainly a less-well or less-assessed

00:16:04,100 --> 00:16:15,133
in clinical trials, and as a compensation for this, we view safety with a very different lens than we view efficacy. So for example, to get a drug

00:16:15,133 --> 00:16:22,333
approved you may want a P value less than .05. You don’t need to see that to be worried that there’s a safety signal, it can be sufficient to see

00:16:22,333 --> 00:16:32,333
an adverse trend, and it’s not necessary that this biomarker be fully validated as a marker or in its ability to predict a treatment effect. It can be

00:16:32,333 --> 00:16:40,533
enough to use a marker where there’s just some association between the marker and bad outcome to raise concern. The other point just to

00:16:40,533 --> 00:16:47,499
make, it’s important to consider when you look at safety, it’s not just what’s actually seen within the development program but also what risks and

00:16:47,500 --> 00:16:56,266
level of risks the development program can exclude and how well the studies were designed to actually capture those risks. A case in point,

00:16:56,266 --> 00:17:05,599
again, to look to the erythropoiesis stimulating agents, it really took studies that were designed to show that these treatments were going to help

00:17:05,600 --> 00:17:16,100
people in terms of CV benefits to actually demonstrate their risks. Again, you need larger studies to actually see the safety signals. So,

00:17:16,100 --> 00:17:24,333
how does this relate to surrogate endpoints? Why am I telling you any of this? Well, keep in mind that the goal of therapy is to provide net

00:17:24,333 --> 00:17:33,699
benefit and also keep in mind that the concept of safety is not, again, that a drug is “safe” but it’s safe within the context of the benefit that it’s

00:17:33,700 --> 00:17:42,966
providing. Again, compared to an outcome like mortality or some clinical outcome that the patient directly perceives the nature of a benefit

00:17:42,966 --> 00:17:52,932
provided by a surrogate endpoint is just less obvious, and the implications for that is two-fold. One is that you’re going to have less tolerance

00:17:52,933 --> 00:18:04,533
for risk if you’ve established your efficacy by showing effects on a surrogate, but the other implication is that the size of your development

00:18:04,533 --> 00:18:12,866
program—how much safety data you’re going to need—is going to be dictated, to a larger extent, by safety concerns if you employ a surrogate

00:18:12,866 --> 00:18:22,666
endpoint in the development program. With that I wanted to move on to talk about the topic of today’s meeting or topic related to today’s meeting

00:18:22,666 --> 00:18:31,932
which is morphologic changes in the kidney as a surrogate endpoint. If you look at drug development, or at least drug development in the

00:18:31,933 --> 00:18:41,699
kidney arena, histology has been used to a very limited extent, I would argue, to support drug approval and there are really two situations that

00:18:41,700 --> 00:18:50,700
I’m aware of. One is it’s been used in the setting of therapies that are being developed to treat acute organ rejection and kidney transplant, and

00:18:50,700 --> 00:18:59,833
here biopsy proven acute rejection has been used as an endpoint; and then the other setting is in Fabry Disease where they looked at effects on

00:18:59,833 --> 00:19:09,133
a molecule and its deposition in certain cell types, including cells in the kidney, as a way to establish the efficacy of a therapy. In this latter setting it

00:19:09,133 --> 00:19:19,399
was one of the…the first setting is the concept of a surrogate endpoint that leads to outright approval, the second setting was a surrogate

00:19:19,400 --> 00:19:27,833
where there was still some uncertainty, and so it came with it a requirement to do a study and the post-marketing setting verifying the treatment

00:19:27,833 --> 00:19:37,566
benefit. So, what are considerations in using morphologic changes as a surrogate endpoint? I think they fall into a number of categories. One, of

00:19:37,566 --> 00:19:45,432
course, is: what’s the clinical significance of the finding? Another issue that gets raised is, even if you decided the finding is probably clinically

00:19:45,433 --> 00:19:54,866
significant, you have to figure out what size and effect on the finding is likely to matter, is likely to translate into a treatment benefit, and then I think

00:19:54,866 --> 00:20:00,932
there’s always the question, “Will use actually enable more efficient drug development?” Then, there are finally practical and ethical

00:20:00,933 --> 00:20:08,033
considerations. So, what’s the clinical significance? Again, this goes back to the types of data you’d like to see supporting a potential

00:20:08,033 --> 00:20:18,399
surrogate. What’s the nature of the data linking to a particular finding in a specific disease to adverse renal outcomes? Again, it’s not sufficient

00:20:18,400 --> 00:20:23,266
to know that the marker does a good job at diagnosing disease or identifying patients at high risk of progression; what you really want to

00:20:23,266 --> 00:20:32,232
know or what’s very helpful to know is that reasonably the finding is actually mediating the progressive loss of renal function. Then comes

00:20:32,233 --> 00:20:42,299
the question of effect size and I’m almost embarrassed to bring this up in an audience filled with people who think about quantifying things,

00:20:42,300 --> 00:20:49,666
but here we think about effect size, too, or measurement. I want to point out, again, the clinical significance of any treatment-induced

00:20:49,666 --> 00:20:59,766
change on a morphologic finding rests really in its ability to translate into a treatment-induced effect on renal outcomes and one can certainly have,

00:20:59,766 --> 00:21:07,866
when you see small treatment effects on a marker, is that it may not translate at all or at least over the course of the patient’s lifetime into

00:21:07,866 --> 00:21:15,532
treatment effects on the outcome. The other issue is that, with small treatment effects on the marker, you get worried about small treatment

00:21:15,533 --> 00:21:22,766
effects on your outcome of interest, and again, here there is greater concern that the effect on the surrogate could be weighed by some off-

00:21:22,766 --> 00:21:32,132
target or adverse treatment effect. What about more efficient drug development? Why do we want to use surrogates? Well, we want to use

00:21:32,133 --> 00:21:39,733
them because they’re supposed to translate into smaller studies and studies of shorter duration, right? The problem is that if there’s measurement

00:21:39,733 --> 00:21:46,299
error it’s going to limit the ability to detect a treatment effect using the surrogate. It’s also going to make it hard to validate your surrogate

00:21:46,300 --> 00:21:57,933
and measurement error. I think issues both with accuracy and precision have been a major topic of this meeting. I’m just going to point out two

00:21:57,933 --> 00:22:05,599
obvious sources. One is, when you take a kidney and when you do a kidney biopsy, you’re getting just one sample of what’s going on in a much

00:22:05,600 --> 00:22:16,833
larger organ, and so if that sample’s not representative of the disease process going on in the organ, it’s not going to be helpful and I think

00:22:16,833 --> 00:22:25,299
this has an important application, of course, in that for the use of histology as an endpoint, it’s probably going to be better and more suited for

00:22:25,300 --> 00:22:34,133
using diseases where the process is more diffuse throughout the organ. And then the other issue, which has been brought up by many

00:22:34,133 --> 00:22:45,799
speakers and important clearly for the use of morphologic changes in any setting: you’re going to need to have, really, a common and

00:22:45,800 --> 00:22:55,533
standardized approach for defining and quantifying the changes that you’re seeing. So, what about practical and ethical considerations

00:22:55,533 --> 00:23:04,433
that are going to make it hard to use this in drug development? Well, biopsies aren’t risk-free and at least in certain disease settings it’s going to be

00:23:04,433 --> 00:23:15,699
difficult to obtain biopsies or justify the risk. I also want to point out that there’s a limited ability to get repeat biopsy samples because of the risks

00:23:15,700 --> 00:23:29,900
associated with it, so it’s another constraint or limitation with the use of histology, at least, as an endpoint. In thinking about the rule of changes or

00:23:29,900 --> 00:23:41,966
morphologic changes as a surrogate endpoint, I think there’s perhaps one setting that it may be the most useful or particularly well-suited and I

00:23:41,966 --> 00:23:51,932
think the scenario is this: you have a slowly progressive disease that, at least in the vast majority of patients, is going to progress or lead

00:23:51,933 --> 00:23:59,766
them to end-stage renal disease over the course of their lifetime. In that setting, using histology to demonstrate a treatment’s benefit in these

00:23:59,766 --> 00:24:10,932
patients at a time when their renal function is still preserved could be a very important use of this type of endpoint, but then the first issue you have

00:24:10,933 --> 00:24:21,633
to get at, which is the real difficult one which is validating the surrogate. So with that, I’m going to close. I just want to again highlight that many

00:24:32,800 --> 00:24:35,000
important rules for the work that you’re doing in drug development, I just focused on one which is as a surrogate or a basis for drug approval. I also

00:24:35,000 --> 00:24:48,533
want to point out that there are now two nephrologists in CDRH, which is the center that regulates devices and also diagnostics and I’m

00:24:48,533 --> 00:24:57,733
very happy to share their names and pass on information if you have questions related to sort of validating those types of instruments and tools

00:24:57,733 --> 00:25:03,766
and tests and getting them out for clinical practice. Thanks.

00:25:03,766 --> 00:25:10,532
ROBERT STAR: Rob Star, NIH. That was beautiful, as always. I actually want to go to the…could you go back one slide to the last thing

00:25:10,533 --> 00:25:15,633
you talked about which was how to, you said, validate this, probably qualify to be the surrogate?

00:25:15,633 --> 00:25:18,233
ALIZA THOMPSON: Yeah, it’s always a struggle what to figure to put on the slide.

00:25:18,233 --> 00:25:27,099
ROBERT STAR: But I just want the people in the room to understand what that actually means and how difficult that might be. So in the case of

00:25:27,100 --> 00:25:35,600
polycystic kidney disease, again, we’re talking about an imaging surrogate instead of a morphologic one, but I think that the same

00:25:35,600 --> 00:25:44,433
considerations probably apply from the way you look at it. What the FDA appears to be asking for is an observational study that shows that the

00:25:44,433 --> 00:25:54,133
kidney growth occurs before the renal function goes away and then two randomized controlled trials, preferably using different drugs that show

00:25:54,133 --> 00:26:04,466
the changes in the surrogate of total kidney volume, correlate with later changes in renal function. So it’s a huge amount of information to

00:26:04,466 --> 00:26:09,766
validate or to qualify the surrogate. Is that approximately correct?

00:26:09,766 --> 00:26:20,532
ALIZA THOMPSON: Yeah, I think that the example of polycystic kidney disease is a great example. We don’t or at this point we’re not approving

00:26:20,533 --> 00:26:28,499
therapies based on changes in kidney volume and one of the concerns that’s been raised…you know, there’s a lot of plausibility. You take a

00:26:28,500 --> 00:26:34,400
kidney and you make it huge, that you’ve caused damage, but what we’re talking about here in terms of the endpoint, if people are interested, is

00:26:34,400 --> 00:26:45,733
really small changes in kidney volume and trying to understand their clinical significance. The sense or how we’ve approached this is that we

00:26:45,733 --> 00:26:53,366
haven’t been prepared to accept small changes in kidney volume as an endpoint but would, in fact, want to see data from intervention trials. Now,

00:26:53,366 --> 00:26:58,899
whether or not we need two in a whole range of distinct pharmacologic interventions, I don’t know. It may be that you just get one intervention trial in

00:26:58,900 --> 00:27:07,766
and people start to feel better about it. It may be that you needed more than that, I don’t know. But again, that is the concept, that you do need data

00:27:07,766 --> 00:27:17,299
from intervention trials to support some of these. It’s a hard path.

00:27:17,300 --> 00:27:29,533
MALE: I was wondering if you could expand on that point. I guess my question is: at what point and what kind of data do you need to actually

00:27:29,533 --> 00:27:38,199
reach a tipping point in your decision-making and how many studies does it take, generally, and how many different types of interventions do you

00:27:38,200 --> 00:27:45,000
need to see before it reaches the level of FDA proof of concept?

00:27:45,000 --> 00:27:51,466
ALIZA THOMPSON: Just like there’s not a standardized approach to measuring a lot of these things you guys measure, there’s no formal

00:27:51,466 --> 00:27:59,999
or standardized approach to validating a surrogate. I think, again, a lot of it rests on the extent to which you can make the case that your

00:28:00,000 --> 00:28:08,233
marker is on the causal pathway and again, looking back to electrolytes, you know, the vaptans—conivaptan and tolvaptan—were

00:28:08,233 --> 00:28:14,866
approved within recent history based on a surrogate, and again, you didn’t need large intervention trials showing that effects on serum

00:28:14,866 --> 00:28:23,732
sodium predicted your effect on outcome. If you have a finding like scar in the kidney you show that glomeruli are sclerosed and people have

00:28:23,733 --> 00:28:30,699
done a pretty good job of quantifying the relationship, so telling me what some given change in the kidney is going to predict in terms

00:28:30,700 --> 00:28:39,866
of rate of progression to dialysis, I think that’s a setting where it’s not clear to me you’d ever need an intervention trial showing that your treatment

00:28:39,866 --> 00:28:48,699
effect on the surrogate predicts a treatment effect on the outcome. In other settings, some other finding, you know, a little bit of deposition

00:28:48,700 --> 00:28:58,966
here in the mesangium, I think there you’re going to need the intervention trials and how many, again, it’s going to be subjective and it’s going to

00:28:58,966 --> 00:29:11,966
be based, I think, on essentially the totality of the data from all those classes. Not helpful. You should come in. One of the goals of coming here

00:29:11,966 --> 00:29:20,399
today or one of the reasons I was very excited to get the opportunity to speak is this is an issue that is occasionally raised at meetings or in

00:29:20,400 --> 00:29:31,033
e-mails but never sort of explore it within the context of a specific disease and a specific histology or morphologic change, and I think that

00:29:31,033 --> 00:29:38,999
you really can’t have this conversation unless you’re exploring it within a very specific context.

00:29:39,000 --> 00:29:50,900
MICHAEL MAUER: Thank you for your talk, Aliza. I think I understand, at least in part, the FDA conundrum and maybe Fabry Disease is a good

00:29:50,900 --> 00:30:03,766
example of it. So, the indication was primarily provided on the clearance of endothelial cells with the Fabrizine meta-dose at the milligram per

00:30:03,766 --> 00:30:15,032
kilo every two weeks intravenously. Now, it worked beautifully for endothelial cells. There were other cells like podocytes, vascular cells,

00:30:15,033 --> 00:30:28,133
tubular cells that clear much more slowly. Now if we pick another drug where the dose is .4 milligrams per kilo it beautifully clears endothelial

00:30:28,133 --> 00:30:45,166
cells. It doesn’t clear podocytes as well as the higher dose but the endothelial cell clearance, which is very easy, is now the standard—the

00:30:45,166 --> 00:30:58,399
level of the playing field—and it may take years and years of long-term follow-up study to find out that the .4 milligram per kilo dose is simply

00:30:58,400 --> 00:31:07,200
inadequate. So, this is a real struggle and Fabry Disease, I think, is a good indication of that, of how that can go wrong.

00:31:07,200 --> 00:31:14,300
ALIZA THOMPSON: Right, and I just want to point out that, unfortunately Lynne Yao who’s a nephrologist—pediatric nephrologist—by training

00:31:14,300 --> 00:31:22,866
and works in inborn errors and metabolism, was supposed to come today and talk a little bit about Fabry Disease and the experience in Fabry

00:31:22,866 --> 00:31:33,532
disease. Unfortunately, she couldn’t make it, but I very much agree that that’s very challenging and raises a whole bunch of issues and I think

00:31:33,533 --> 00:31:40,199
highlights one important issue when you look at morphology findings is you’re going to highlight some finding and say, “Look here,” but what do

00:31:40,200 --> 00:31:46,200
you make of the other findings that are occurring in the kidney and what that’s going to mean in terms of your treatment providing a benefit? Very

00:31:46,200 --> 00:31:58,333
difficult, but clearly the work of people in this area in terms of elucidating what all these findings mean will be very important.

00:31:58,333 --> 00:32:09,366
MICHAEL MAUER: I have a very specific question; let me see if I can frame it as a specific question. So if we have a structural parameter—I’

00:32:09,366 --> 00:32:20,599
m just going to pick one that I like---mesangial fractional volume and let’s say that it explains 50% of the variability and GFR loss, which is

00:32:20,600 --> 00:32:29,366
pretty good for a single parameter and we haven’t got glomerular number, so there’s going to be noise and there’s noise in the measurement

00:32:29,366 --> 00:32:40,266
and there’s noise in the sampling, but let’s say it explains 50% of the variability in GFR loss as a very slowly progressive kidney disease. So, if

00:32:40,266 --> 00:32:51,232
we do a baseline and a five-year biopsy, the change is going to be quite small that you’re going to measure and in and of itself the change would

00:32:51,233 --> 00:33:07,566
be insufficient to affect GFR. So now let’s say you have a 50% reduction in the change of that parameter that explains 50% of GFR change in

00:33:07,566 --> 00:33:16,299
more advanced cases. How would that be looked at from your perspective?

00:33:16,300 --> 00:33:24,700
ALIZA THOMPSON: Well, I think it raises a few issues. One I just want to touch on. So for finding, like the one you’re describing in terms of

00:33:24,700 --> 00:33:32,133
what you’re probably going to ultimately need, is you’re probably going to need data from intervention trials unless you can find some other

00:33:32,133 --> 00:33:40,333
way from animal studies or something else that actually shows you that what you’re seeing is actually the cause or causing a loss of renal

00:33:40,333 --> 00:33:51,599
function somehow, directly leading to, you know, death of tissue. Absent that, you’re probably going to need some intervention trials. But I think it

00:33:51,600 --> 00:34:01,566
gets back to some of the earlier points I made about effect size, which is there’s always this concern when the effect size is small, about

00:34:01,566 --> 00:34:09,166
whether or not it’s going to translate into your outcome of interest, which yes, on the one hand it’s GFR decline but ultimately it’s the progression

00:34:09,166 --> 00:34:18,066
to dialysis because that’s what people perceive, right? People are asymptomatic from kidney diseases for a long period of time, but also with

00:34:18,066 --> 00:34:27,299
small treatment effects we also have a greater concern about the fact that they could be outweighed by the toxicity of the therapy. But the

00:34:27,300 --> 00:34:38,366
work you’re describing is important work and I think that it’s going to require us to sit down and have more careful discussions about, if it were to

00:34:38,366 --> 00:34:45,466
be validated, how one would approach its validation or qualification and what types of data would be important, because I think, you know

00:34:45,466 --> 00:34:55,599
again, going back to the scenario where I think histology or morphologic changes could provide the most benefit is certainly at very early stages

00:34:55,600 --> 00:35:05,233
of progressive diseases where you’re not going to really be able to show an effect on people who have well-preserved renal function, so

00:35:05,233 --> 00:35:13,699
using creatinine is not an endpoint. But again, the concept of the type of disease isn’t just that it’s slowly progressive, but the patients are, over the

00:35:13,700 --> 00:35:24,200
course of their lifetime, very likely to go onto ESRD and really perceive something from their disease. I look forward to further discussions

00:35:24,200 --> 00:35:32,133
with you.
MARILYN HAILPERIN: Hi. I’m Marilyn Hailperin and

00:35:32,133 --> 00:35:42,599
I’m from the NephCure Foundation. Our mission is to support research to find the cause, cure, and better treatments for primary nephrotic syndrome

00:35:42,600 --> 00:35:55,733
and particularly FSGS, so it may be a question to you and also to the larger community here. How do we facilitate engaging the right constituents to

00:35:55,733 --> 00:36:04,833
sit down, sort of, with the FDA and have discussions around what are the appropriate clinical pathways to validate, to investigate,

00:36:04,833 --> 00:36:15,099
potential surrogate indicators? Our intention, of course…surrogate endpoints..our intention, of course, is to get pharmaceutical companies to

00:36:15,100 --> 00:36:24,500
pay attention to our rare disease and they won’t pay attention to your rare disease unless there’s some mechanism to suggest that there will be

00:36:24,500 --> 00:36:40,800
FDA approval at the end of the commitment for creating those drugs. Are there case studies, examples of best practices where a constituency

00:36:40,800 --> 00:36:50,500
group made up of maybe patient advocacy and other physicians, clinicians, scientists, who’ve been able to do this successfully that we could

00:36:50,500 --> 00:36:55,200
apply to this nephrology community?
ALIZA THOMPSON: How the approach has gone

00:36:55,200 --> 00:37:04,366
thus far, and again I’m very appreciative to the NIH for inviting me to these types of meetings. These meetings have been an opportunity for at

00:37:04,366 --> 00:37:13,999
least our division to sort of present some of our approach to the validation of surrogates and then one thing that happens after these meetings and

00:37:14,000 --> 00:37:21,800
we’ve encouraged is for the communication. So if you send me an e-mail, we’ve been connecting with more recently the members from the ASN

00:37:21,800 --> 00:37:35,866
about endpoints and specifically to use proteinuria as an endpoint in some diseases associated with nephrotic syndrome. So, how a

00:37:35,866 --> 00:37:44,199
lot of this sort of qualification goes about or for the discussions about these endpoints goes about is by reaching out after these meetings and

00:37:44,200 --> 00:37:55,466
we set up calls or a path for it to try and gather more data that would address our concerns about the validity of some of these endpoints.

00:37:55,466 --> 00:38:03,866
Because again, you guys know the data; we, however, know what the requirements are or at least how we’ve interpreted their requirements

00:38:03,866 --> 00:38:10,266
and it’s really the meeting of these two things that’s going to define the path forward. So, feel free to e-mail me. I think I put my e-mail up.

00:38:10,266 --> 00:38:15,699
MARILYN HAILPERIN: You’re on speed dial.

00:38:15,700 --> 00:38:22,933
PAUL KIMMEL: Thank you, Aliza, very nice presentation. I’m just wondering, since we have so many morphologists and pathologists around,

00:38:22,933 --> 00:38:33,633
you focused on, I guess, clinical trial endpoints. Is it worth talking about the PSTC and how morphometric data, histologic data, were used in

00:38:33,633 --> 00:38:45,833
screening and how the FDA qualified those as sort of a structure for people to understand how histologic material may be used in working with

00:38:45,833 --> 00:38:48,733
the FDA?
ALIZA THOMPSON: Yeah, I think that’s excellent

00:38:48,733 --> 00:38:56,266
and Paul, you’re going to actually stay at the microphone so you can correct or fill in any blanks. I focused my talk on surrogate endpoints

00:38:56,266 --> 00:39:00,966
but maybe it’s because when I saw the title I thought “surrogate endpoints” and to me surrogate endpoints mean something very

00:39:00,966 --> 00:39:12,866
specific. But I want to say that, you know, drug toxicity, specifically renal toxicity, is a huge issue that prevents the development of therapies and

00:39:12,866 --> 00:39:21,732
one of the big issues with drug toxicity and nephrotoxicity is, of course, creatinine is not really an indicator of acute kidney injury, it’s an

00:39:21,733 --> 00:39:30,099
indicator and a late indicator of effects on glomerular function, and so there’s been a lot of interest in finding markers that can do a good job

00:39:30,100 --> 00:39:41,300
of allowing or that are sensitive in allowing an earlier indication of a nephrotoxicity to the kidney. So, a group in industry working through CPATH,

00:39:41,300 --> 00:39:53,633
which is a nonprofit institute, came together and tried to put together data to qualify the use of several markers or biomarkers of kidney toxicity

00:39:53,633 --> 00:40:08,099
and data specifically in the rat showing the relationship between these biomarkers and toxicity on histology. I’m trying to think of what

00:40:08,100 --> 00:40:16,066
else to say. In sitting here today and hearing some of the discussion I think they would have been aided and moreover think that, even within

00:40:16,066 --> 00:40:24,566
the context of animal studies, we could use a lot of your expertise in interpreting some of these findings and also using it to understand the

00:40:24,566 --> 00:40:34,966
toxicities that therapies are having, but Paul is going to chime in and say what I didn’t touch upon that I should of.

00:40:34,966 --> 00:40:39,299
PAUL KIMMEL: God forbid. I mean, it was a beautiful explanation but I just think that pathologic specimens can be used in the screening process

00:40:39,300 --> 00:40:45,966
for drugs. They can also be used, probably, as entry criteria and if there were more refined measurements, more reproducible

00:40:45,966 --> 00:40:52,099
measurements, I think there are many pathways for using pathologic material in clinical trials.

00:40:52,100 --> 00:41:01,033
ALIZA THOMPSON: Yeah, and I just want to just further that. I can give you one example where we use histology currently as a tool in clinical

00:41:01,033 --> 00:41:08,566
trials, which is something like with this nephritis whereas you know a lot of therapies are directed at certain types or classes of lupus nephritis and

00:41:08,566 --> 00:41:15,932
it’s been used as a way to identify patients with focal or diffuse proliferative or membranous lupus nephritis who were thought to have more

00:41:15,933 --> 00:41:23,599
severe disease and required treatment, but I don’t think we’ve optimized our use there. People use it for the diagnosis but probably should be looking

00:41:23,600 --> 00:41:32,333
at a whole bunch of characteristics on histology as a way to understand what patients are more likely to benefit from a particular therapy,

00:41:32,333 --> 00:41:41,866
particularly given some understanding of how they think the therapy is going to actually affect with this nephritis and its disease.

00:41:41,866 --> 00:41:52,166
JOHN BASGEN: This talk about surrogates reminded me of another possible surrogate other than podocyte number and the theoretical

00:41:52,166 --> 00:42:01,432
mathematicians tell us that this problem of counting number can’t be done in a two-dimensional plane because it’s a zero-dimensional

00:42:01,433 --> 00:42:10,499
parameter. So, this is where the two sections, and a volume that you’re sampling, makes it work. But, aren’t we using podocyte number as a

00:42:10,500 --> 00:42:25,133
surrogate of how well the kidney is filtering? Is there another surrogate or measurement of filtration, glomerular filtration? Well, there is—a

00:42:25,133 --> 00:42:35,833
filtration surface of the glomeruli—which its surface is a two-dimensional parameter which then can be easily measured in thin sections. You

00:42:35,833 --> 00:42:44,933
don’t have to identify whether they’re sick podocytes or disappearing podocytes. In the sclerosed area there isn’t any filtration surface

00:42:44,933 --> 00:42:57,966
anymore. So I’m wondering…you need to do electron microscopy to measure this surface of the glomerular capillaries; you’d like to know the

00:42:57,966 --> 00:43:05,432
volume of the glomeruli, so we’re back to the problem of getting a good measurement of glomerular volume; and if you knew how many

00:43:05,433 --> 00:43:20,833
glomeruli there were in the kidney you could multiply those three parameters together to get this total filtration surface per kidney. Now, Mike

00:43:20,833 --> 00:43:34,433
Mauer did part of this study a couple of decades ago on correlating this filtration surface with a measure of kidney function and I’m just saying

00:43:34,433 --> 00:43:45,666
that, because it’s easy for me to do electron microscopy, it’s a lot easier for me to measure this filtration—well, the surface density of the

00:43:45,666 --> 00:43:55,832
filtration—than to count podocytes. So, is there a better surrogate for functions than podocyte numbers?

00:43:55,833 --> 00:44:01,899
ALIZA THOMPSON: There may be and that’s really…you all are the subject matter experts. Come up with what you think is going to be good.

00:44:01,900 --> 00:44:08,700
ASHRAF EL-MEANAWY: Ashraf El-Meanawy from Milwaukee. Can these morphometric parameter accessed by data for entry of

00:44:08,700 --> 00:44:16,733
selection criteria for patients, say, when we get to the point that we have glomerular volume or number and we go and pick only diabetic patients

00:44:16,733 --> 00:44:22,133
who have big glomeruli as a treatment group in a treatment study, would that be acceptable?

00:44:22,133 --> 00:44:30,266
ALIZA THOMPSON: Yeah, and I think that that is an excellent use of these types of markers, is to identify patients who are likely to have poor

00:44:30,266 --> 00:44:39,766
outcomes, which is considered or the concept is that you’re doing a prognostic enrichment. Currently, we do that in trials of diabetic

00:44:39,766 --> 00:44:47,132
nephropathy by picking people who have a lot of proteinuria or you pick people with impaired renal function, but it could turn out that your marker,

00:44:47,133 --> 00:44:55,899
your histology actually, does a much better job at predicting people who are going to progress. The other way we do it is something called predictive

00:44:55,900 --> 00:45:05,266
enrichment which is this concept that patients with certain characteristics are going to be more likely to respond to your treatment, and again, it

00:45:05,266 --> 00:45:14,832
could be that because your drug causes B cell depletion and based on your morphologic changes or histology are seeing that in certain

00:45:14,833 --> 00:45:22,133
subsets you think these cells are somehow more…it’s the same disease but it’s actually maybe a different disease in some patients and

00:45:22,133 --> 00:45:28,366
you think there’s a reason to believe that the B cell is playing more of a role because of some finding, you could also use that to selectively

00:45:28,366 --> 00:45:40,266
enroll patients. I think what you’re outlining is a wonderful use, an early use, of the changes in morphology as a way to facilitate drug

00:45:40,266 --> 00:45:55,666
development, and you don’t need to have them validated, at least this formal process that I’ve described or qualified. Excellent. Thank you.

Date Last Updated: 10/4/2012

General Inquiries may be addressed to:
Office of Communications and Public Liaison
Building 31, Rm 9A06
31 Center Drive, MSC 2560
Bethesda, MD 20892-2560
Phone: 301.496.3583