Posted by Jeff Berk; BOLT International
Today FDA refused to approve Amyvid™ for the detection of beta-amyloid plaque in the brains of living patients. The complete response was primarily focused on the need to establish a reader training program for market implementation that helps to ensure reader accuracy and consistency of interpretations of existing Amyvid scans. In our note, below, we want to expound on a few aspects related to amyloid imaging. The first is that when we last polled them, our Thought Leaders actually thought that even more interesting than florbetapir is [18F]-AZD4694; which is being co-developed by AstraZeneca and Banner Alzheimer’s Institute (Phoenix, AZ). AZ’s diagnostic is reported to have better sensitivity due to a significantly lower white matter background signal. If 18F PIB sets the bar for amyloid imaging at 70%-75%, then AV-45 is imaging in the 60%-65% range. AZ is claiming that their detection is “non-inferior” to 18F PIB.
“I saw some stuff from AstraZeneca that looks really good. I don’t know. And it is going to take a while for all that to shake out”.
“I think that AstraZeneca has a very interesting compound. I think F18 PIB could be okay, although I think GE doesn’t seem to be moving forward too expeditiously and Bayer is the fourth”.
“AV-45 doesn’t look bad. It looks as good as any F18 agent with the possible exception if the stuff that AstraZeneca is saying, and again talk is cheap, but if what AstraZeneca is saying about their amyloid imaging agents having significant lower white matter background they may be the best that exists. But again talk is cheap and until you line it up with some common measure, you know what is your yardstick? You don’t know. Because in the end there are going to be several similar agents. I firmly believe that we are imaging amyloid with PIB. We have lined up PIB with the GE F18 compound, which is an F18 PIB that we developed here. So we know pretty closely how those two stack up and we know that F18 compound isn’t as good as PIB because there is less of a signal to noise. It gets up to about 70-75% of PIB. The best we can tell from AV45 or the Bayer compound is they are about the same. They are in the 60-70% of PIB range signal to noise. None of them look as good as PIB. I am not sure about AstraZeneca yet. If you hear what they say they would have you believe it is as good as PIB but until you do it you don’t know. But I think the Bayer compound, Avid’s AV45, and from what I have seen of the AstraZeneca compounds I believe they are all actually imaging amyloid”.
Point number 2 is that the real value of amyloid imaging agents will be in amnestic MCI (mild cognitive impairment) patients:
Ø There is strong consensus with the premise that the best target for effective disease modification is the asymptomatic population. This is also the population that is hardest to study. Surrogate endpoints, none of which are validated today, are going to facilitate a shift of the AD treatment paradigm from today’s late symptomatic approach to early intervention in rationally selected patients. The combination of 11C or 18F imaging and spinal fluid assessment measuring the ratio of A-beta 42 and tau is critical to predicting which patients will progress from MCI to AD, and detecting drug activity in the early phase of clinical development. What was shown at AAN is that using this ratio as a biomarker can predict the development of dementia in non-demented patients within 18-24 months. Amnestic MCI patients convert to AD at about 15% per year. Amnestic MCI has a high conversion rate to AD which is 15% per year. The data shows that the PIB-positives are converting to AD at 40% per year whereas the PIB-negative are converting to AD at almost 0% per year, and they are converting to non-AD dementias at about 10% per year. These biomarkers used in combination thus should minimize the commercial risk at a time long before a change in cognition can be detected.
“I believe that the best target population for effective disease modification is the asymptomatic population. I believe that we can identify people who have Alzheimer’s disease, but have no symptoms using biomarkers. I believe that we will be able to validate surrogate endpoints so that we can study drugs in this population but just to make it clear that we are not quite there yet”.
Ø For Pharma the key regulatory question is: what criteria will FDA/ EMEA accept for a biomarker for a prevention trial? At this time because the unmet need is so great, the regulatory agencies will accept something less than a fully validated volumetric (MRI) or FDG-PET endpoint for a prevention trial. The criteria set up by the FDA, that would also be acceptable to EMEA, would be that your drug effect on your surrogate endpoint is “reasonably” likely to predict later clinical benefit. If for example bapineuzumab shows in the current phase III program that there is a correlation between a cognitive effect and an MRI volumetric effect, that would be enough to make it reasonably likely to expect that a bapineuzumab effect on just the volumetric measure would predict a later clinical benefit. If bapineuzumab has a modest effect related to a volumetric effect in AD, that in itself would be sufficient to allow them to conduct an asymptomatic AD trial using simply a volumetric outcome.
“We are really entering difficult ground for FDA to agree ahead of time to what the endpoints of a study are going to be in these preventive trials. Right. What is the state of agreement? What should the endpoints be in these kind of prevention trials? Again, I think the best endpoint would be a volumetric one, although one could also propose an FDG PET endpoint. And I think that the agencies are open to this. They are not ready to accept it but they are open to it. The path towards validation – the bar is not as high as some people think it is. You don’t need a fully validated surrogate because the need is so great. The criteria set up by the FDA that would also be acceptable to EMEA, would be that your drug effect on your surrogate endpoint is “reasonably” likely to predict later clinical benefit. I believe we are not that far from “reasonable likelihood.” And how we might get there would be, if for example bapineuzumab shows in the current phase III program that there is a correlation between a cognitive effect and an MRI volumetric effect as they try to show in phase II, although I don’t think they did. If they show that in phase III I think that would be enough to make it reasonably likely to expect that a bapineuzumab effect on just the volumetric measure would predict a later clinical benefit. So they might have a modest effect related to a volumetric effect in AD and that itself would be sufficient to allow them to conduct an asymptomatic AD trial using simply a volumetric outcome”.
Ø The societal cost of AD therapy is going to be very high and there is a need for predictive biomarkers to match therapeutics with the patients who can benefit from them. These predictive biomarkers will be able to address the question of whether or not a patient truly has AD, but without therapies available to alter the disease, it is not clear that the general practitioner should be able to prescribe such a screen. Payers may limit reimbursement to cases where neurologists have prescreened a patient and believes that there is a reason to confirm the presence of Alzheimer’s.
“I think amyloid imaging will be available, I think it will surpass FDG-PET. Then the question will be what will be restrictions on how it is used or applied. That will be good to tell if a person has Alzheimer’s or not, it will be a dichotomous type scan where it either says you have Alzheimer’s or it doesn’t say you have Alzheimer’s. But the question will be is it appropriate for a primary care physician who is not sure how to diagnose Alzheimer’s disease appropriate to order amyloid imaging because that is expensive? Or would it be that you set up kind of a tiered approach well only neurologists can order them and this is after they do other tests or something like that”.
“If you are talking about an asymptomatic person, and granted you have identified this plaque entanglement so there is reason to put this patient into a high risk category, but otherwise asymptomatic I am just wondering how willing patients are going to be to be compliant. So actually I don’t agree with that concern. That has been mentioned by many people, in particularly the FDA, that if you are going to go to an asymptomatic population you had better have more sensitivity to all safety and tolerability issues. But I don’t agree with that. I think that, in fact, if anything I might say the opposite that the people who are most eager to manage their disease and are willing to take expensive and risk and tolerability issues are the people who are still asymptomatic. I think they are desperate for treatment and they would be willing to put up with a lot”.
Finally some Thought Leader perspective on PIB:
Ø One of the potential strengths of 11C-6-OH-BTA-1 (PIB; GE Healthcare) is that it may selectively bind to the specific tertiary structures of A-beta that are associated with the more toxic disease-related deposits. All Alzheimer’s patients, diagnosed by ADAS-COG, display PIB binding, whereas cases from normal aging or from temporal dementia where there have been small amounts of amyloid in the brain after death, had produced PIB scans that were completely negative.
“As we have gone through the years and the animal models with PIB, and then in the human studies, and these things haven’t even yet begun to come out in papers, but they will over the next year. There are some types of A-beta that PIB binds to and some types of A-beta deposits that PIB doesn’t bind to. The first ones we knew about were in the transgenic mice and we and others have published papers on that. We have seen the same thing in other mammals and now the study in primates that all A-beta deposits aren’t the same. One way we looked at that is gee PIB isn’t as good as we thought it was. It doesn’t detect all these deposits. But if you want to be an optimist you could even look at this, and this is probably a mixture, but it is better than we thought it was. Not only is it picking out A-beta deposits in AD brain, but it may be actually picking out the most disease-related deposits. We know that a little diffusive amyloid can occur in the brain without any particular reaction around it in several brain areas. These primates don’t develop Alzheimer’s disease. There are no primate models of Alzheimer’s disease. They age so they are probably a pretty decent model of normal aging. But we know that just the presence of amyloid alone isn’t enough to make Alzheimer’s disease. It probably needs to be in a specific tertiary structure. And PIB may actually be one way to measure that structure. I know there are people using it in vitro. You can make a variety of polymorphic forms of A-beta aggregates. Some of them are much more toxic than others. People are trying to see if PIB can pick out one of these forms from the other. I think that is the interesting thing that it is not just A-beta deposits but certain types of A-beta deposits that seem to be associated with Alzheimer’s disease. We have never seen an Alzheimer’s patient that has been diagnosed here at Pittsburgh that didn’t have a bunch of PIB binding. But we have had some patients that have had other dementias, or we know of cases from normal aging or from temporal dementia where there have been small amounts of amyloid in the brain after death, but the PIB scan was completely negative. It didn’t pick it up at all. I think it may be because that really wasn’t what was going on in that person. For one person they were cognitively normal. The other two had different sorts of dementias and they had those pathologies that were probably driving it and the little bit of amyloid was probably incidental”.
Ø The imaging by PIB of reduced brain amyloid should therefore be more closely tied to a physiological effect of an anti-amyloid therapy than a change in CSF a-beta. A change in a-beta, measured by PIB, isn’t predictive. It’s telling you that a physiological change has occurred; one that you would expect to see if an anti-amyloid strategy was working. There are small 5-10% increases in A-beta in a certain subset of controls MCI and ADs over a year. In patients given an active immunotherapy or an effective beta- or gamma-secretase inhibitor you would expect to see at least a stabilization, or a decrease in a-beta over time.
“It is going to be most valuable on therapies that have a change, a decrease, so the immunotherapies. We are presuming that we are starting the secretase inhibitors; we are in a clinical situation where we are starting it based on risk as a preventative therapy. So there wouldn’t be a baseline level. Now that is probably not the way it is going to be derived in the trials. There are probably going to be amyloid to begin with. So in all of these therapies while you could probably use CSF A-beta to screen in if you need a chain score you need to have amyloid imaging before and after the treatment. So if you need them both places, if you are going to have amyloid imaging before the treatment to me there is no reason to do CSF. If you only need the one screen at the beginning then CSF and A-beta are interchangeable. That is only part of the picture. The second part of the picture is at this stage, certainly with immunotherapies probably in trials of beta, gamma and any kind of anti-amyloid therapies, it is going to be in people who have amyloid already in their brains. You are going to be looking for either a cessation of increase A-beta, which now over the past year many people are showing that there are small 5-10% increases in A-beta in a certain subset of controls MCI and ADs over a year. So with the beta and gamma secretase inhibitors you could look for a stabilization of that, although I believe that if you are really effectively treating, inhibiting beta and gamma secretase there is enough of a dynamic process in the turnover of A-beta that the brain mechanisms will cause a decrease over time. I think you can detect and would be able to detect a decrease in amyloid burden with really effective secretase inhibition if we ever get there. Certainly with the immunotherapies you would feel like in order to be effective those therapies should cause a decrease in A-beta. So that is pretreatment to define a baseline and to screen in but if you just want to screen in you can use CSF, but to define the baseline, and then depending on the design post treatment one or more follow ups to see if you have decreased A-beta or at least stabilized prevented an increase”.