Trials and Tribulations: Does ADAPT Have to Adapt?

Updated 4 October 2002

Q&A with John Breitner
Comment by Eddie Koo
Q&A with Todd Golde
Q&A with Greg Cole
Q&A with Karen Hsiao Ashe
Comment by Marcia Gordon
Letter by Monique Breteler to John Breitner
Comment by Paul Aisen
Comment by Leon Thal
Q&A with Giulio Pasinetti
Comment by Pat McGeer
Comment by Curtis Meinert

Q&A with John Breitner, Principal Investigator of ADAPT. Questions by Gabrielle Strobel.—Posted 25 September 2002

Q: What happened at your site after the letter appeared, in terms of reactions from the press and patients?
A: We were very concerned about it. Luckily, the fallout thus far has been small. Of the enrolled people two have called up but were readily reassured. The public is smarter than we sometimes give it credit for. The trial is continuing at all sites without disturbance. We had four inquiries from eastern-based media but by the time we could respond it was past their deadline, and our statements were not cited in the stories the next day. CNN ran a segment interviewing Sidney Wolfe of Public citizen but no scientist from the ADAPT trial. Reuters, the Wall Street Journal, and other newspapers carried stories.

Q: Have you had any official reaction from NIA/NIH?
A: Of course. They take it as seriously as we do. They have prepared an official response. I have seen it but it is not public yet.

Q: Do you see scientific merit to Public Citizen's arguments?
A: The statements are scientifically invalid, and they are biased and distorted. There is a new hypothesis out there, which is intriguing. In cultured cells and animals, some but not all of these NSAIDs have the capability to reduce Ab production, probably by affecting g-secretase. It is pretty clear that celecoxib and naproxen don't. But no one knows if the drugs have the same effect in humans, not even in the laboratory. And no one knows if Ab is the real mechanism by which NSAIDS affect AD risk. The only relevant human data is what Monique Breteler's group presented in Stockholm. But those data are from an exquisitely small number of people who have taken NSAIDs and gotten AD. This data has too low a confidence level to be anything near conclusive at this point in time.
ADAPT will test whether at least two of these drugs prevent AD. There are plenty of epidemiological data suggesting naproxen will do so. We are watching these basic research efforts very closely and I am not disparaging them at all. Perhaps they are right. But we need a lot more information before we have reason to abort a massive clinical effort such as ADAPT. We are two years into ADAPT and it would be a shame to stop it now.

Q: The letter claims that the drugs were chosen because the companies agreed to donate them. Accurate?
A: This is a distorted statement. When we wrote the grant in 1998, we indicated that we would buy the drug. Then it became apparent that we would face serious problems in formulating drug and placebo suitable for a trial. ADAPT is a masked trial involving two different drugs and placebo. That means we needed to have study drug and placebo pills that are indistinguishable, and we would have to formulate and manufacture those ourselves. This is a very substantial practical concern for an academic-based trial. We went to great lengths to get the pharmaceutical companies making ibuprofen to make these preparations for us. They said they would donate ibuprofen in a gel capsule form but would not provide a placebo. We would have had to make the placebo ourselves. Moreover, the overencapsulated capsules would have been more than inch long and our elderly, frail study subjects would have had to swallow four of those every day for years. We were concerned that would cause many dropouts. Nonetheless we went down that road until Leon Thal of the Alzheimer's Disease Cooperative Study (ADCS) told us that Bayer had made Aleve, i.e. naproxen, available with matching placebo for some of his clinical trials. We approached Bayer and they donated identical drug and placebo. Mind you, at that time there was not a shred of evidence indicating there was a difference in any beneficial effect of ibuprofen and naproxen. So that is how the choice between ibuprofen and naproxen was made.
I should add, had we gone with the overencapsulated forms of ibuprofen and shown efficacy in the trial, we would then have needed to show bioavailability and bioequivalence of our doctored, encapsulated medicines with the commercially available form of ibuprofen. So you really want a placebo that looks exactly like the commercially available form of the drug you are testing, plus unlabeled pills of that drug.

Q: How about celecoxib?
A: With celecoxib the story is more bizarre. We tried to get Merck or Pharmacia to make available Vioxx or Celebrex and a matching placebo. Merck already had a dubious reputation in the academic trial scene because they promised drug to the ADCS and then reneged shortly before the trial was about to start. They also refused our request. It took us a year to persuade Pharmacia to provide unmarked drug and placebo.
Let me emphasize also that we promised these drug companies nothing. They do not have seats on our steering committee; they do not have access to our trial data; they cannot advertise that NIA is testing their drugs for AD. Naproxen and ibuprofen have been off-patent for many years, so the only possible gain to the companies is another indication after 7 years, but that market will be available to every company. Even the celecoxib patent will have expired by the time it could be approved for AD, so there will be generics on the market by then. Clearly, the companies have no financial interest in this.

Q: I see the practical considerations. What was the scientific rationale behind the drug choice?
A: We initially proposed a trial with a single agent. The NIH study section, aware that COX-2 inhibitors were coming on the market and apparently had a better safety profile, urged us to use a selective COX-2 inhibitor. We initially were circumspect about this suggestion because all of the epidemiological evidence was on the conventional NSAIDs. But for a prevention trial you must go the safest route, so we decided to use a parallel design testing one drug from each class. Beyond that, there was no scientific reason at the time that one NSAID would have an edge over the other; it was like "choosing between twiddledum or twiddledee."

Q: Could you include another arm to test ibuprofen? That way you could compare two NSAIDs supported by epidemiology, one of which lowers Ab (ibuprofen) and the other does not (naproxen)?
A: That is a very appealing idea in theory, but difficult to implement in practice. I discussed this with Curtis Meinert, who chairs the ADAPT steering committee. For example, for the new arm you could not use the members of the placebo group who have been recruited over the past two years. You'd have to add new placebo subjects to go along with everybody in the ibuprofen arm. Adding arms mid-stream has been tried before in other trials and it has always caused great problems.
No, I'd like to see another trial for ibuprofen. You have to recruit anew anyway, so why not do a new trial? And with the increasing interest in AD by the pharmaceutical industry, maybe the formulation problems we had two years ago would be easier to solve nowadays.

Q: How about indomethacin or flurbiprofen?
A: Both have been off-patent for years. Indomethacin is the most toxic as regards gastrointestinal bleeding. All the epidemiological data, which was so impressive, derived principally from ibuprofen and naproxen. The other ones represent single-digit percentages of NSAID usage. They might have worked but ,again, for a large prevention trial you need more prior human evidence and a better safety profile than these two have.

Q: What would you like to see happen now?
A: It is high time that people study the mechanisms for NSAID action in AD more closely. People have assumed because of pioneering work from Pat McGeer and others that they act through suppressing inflammation, and indeed they may. But then why do prednisone (an anti-inflammatory steroid drug) and hydroxychloroquine (an anti-malarial that has strong anti-inflammatory activity) not work? There are real questions about the inflammation hypothesis.
I am an epidemiologist. There is a large body of evidence showing an inverse relationship between use of these NSAIDs and incidence of AD. I don't know why, but would very much welcome clarification of the mechanism. As yet, the new research coming along does not justify stopping the trial of an agent that is so commonly used and has such strong epidemiological support, nor does it justify the choice of other agents that have a worse safety record. We will only get the answer if we do the trial.

Q: How about the letter's other allegation, namely that the consent form used when enrolling people into the study was inadequate?
A: The package insert data they cite to support these claims are based on a higher dose of naproxen, not on the over-the-counter dose used in the trial. More importantly, those side effect rates cited in the package insert are intended to warn physicians of all known complications of the drugs, and not just those that would pose a practical risk for ADAPT participants. Finally, the package insert data are based on data from other population who were not selected to avoid these risks. We select subjects specifically to avoid these risks. That is why you have inclusion and exclusion criteria in a trial protocol. The trial protocol and the consent form have been reviewed by seven different IRBs who, together, I submit, have more expertise in protecting clinical trial participants than does Public Citizen. This claim is irresponsible.

Let me add that had the letter been successful in derailing the trial, it would have done immeasurable damage to the cause of finding better treatments for Alzheimer's disease. Is that in the public interest? If they are seriously interested in the public good, they might have talked with us first to find out that many of their concerns are unjustified. Instead, they say they have "reviewed the literature and discovered" that the trial is ill-conceived. This letter is naïve and threatens needless harm to the cause of research against Alzheimer's disease. —John Breitner, University of Washington School of Medicine, Seattle.

Comment by Eddie Koo:—Posted 25 September 2002

Many of Sidney Wolfe's accusations of ADAPT are unfounded although I appreciate their caution that giving these drugs to elderly individuals can be potentially dangerous. I will focus only on those aspects that concern our work as I think our results have been misrepresented or misinterpreted.

First, NSAIDs are not -secretase inhibitors, as claimed in the letter. They only modulated g-secretase activity by reducing Ab42 in favor of shorter Ab peptides. g-secretase inhibitors have their limitations as potential therapeutics, as well, through inhibiting notch cleavage, etc.

Second, the dosage of NSAIDs we had to use to achieve this effect is very high, a key issue we have to address scientifically. What's FDA approved and what were given in the Rotterdam study are likely much lower than what we gave to the mice. Therefore, whether very high doses are necessary and achievable in humans in order to see the Ab42 effect remains to be proven. Consequently, making a strong conclusion that our results imply that it is the Ab42 effect that underlies the proposed beneficial effects of NSAIDs in humans is a stretch, to say the least.

Third, we have never discounted the inflammatory responses. We never claimed the inflammatory responses are a result of AD. Even if the inflammatory responses are secondary doesn't mean that modulating these responses cannot alter the risk for AD or the disease course. It is likely that AD treatment must be approached via different angles.

Finally, reanalysis of the Rotterdam study did not take into account our results of all the FDA-approved NSAIDs. The authors assumed that all compounds not mentioned as positive must therefore be negative, which is incorrect (though not their fault, we just didn't have all the data when our Nature paper came out). Even if true, Wolfe ignored the fact that the confidence level is a lot lower in the reanalyzed data. —Eddie Koo, University of California, San Diego.

Q&A with Todd Golde—Posted 25 September 2002

Q: What do you think could be done to resolve the question of which NSAIDs to test in a large, expensive, long-term trial?
A: The issue of which NSAIDs to test is quite uncertain at this time. A definitive answer as to which drug or drugs to test is unlikely to be forthcoming in the near future. There is epidemiologic support for the use of naproxen in the ADAPT. However, the epidemiologic data for ibuprofen is arguably stronger. Moreover, because ibuprofen lowers Ab42 and has been shown to retard amyloid deposition in a mouse model, one could argue that ibuprofen should be one of the drugs tested. From a hypothesis-testing point, a trial of naproxen and ibuprofen could generate valuable information. Such a trial would enable comparison of two drugs with similar anti-inflammatory properties, only one of which is active on Ab42. The counter argument to this is that ibuprofen's anti-amyloid effects have only been shown in animal models, and only at high doses. Thus, ibuprofen will not necessarily lower Ab42 in humans.

Because NSAIDs that are non-selective COX inhibitors are likely to have significant side-effects, the inclusion of celecoxib as an NSAID with a reduced risk for side effects is certainly reasonable. If COX2 inhibition were a key factor in the apparent protective effect of NSAIDs in AD, then adminstration of a COX2-selective NSAID would have huge advantages. Again, the counter argument here is that there is little hard data to support the use of celecoxib. It is not included in the epidemiologic studies, and there is no data from any animal models to support its use.

There is no immediate simple answer to how to figure out which NSAID to use. The best I can give is to do the trial, perhaps modifying it to include a third arm testing ibuprofen. Although this is costly, in relative terms the millions of dollars spent on such a trial are basically insignificant compared to the billions of dollars that AD costs society each year.

As a field, we have yet to learn how to do fully powered AD prevention studies with clear-cut endpoints. Even if the ADAPT trial fails, it will provide extremely important information regarding the anti-inflammatory hypothesis and how to run the next AD prevention trial. Of course, for the sake of those with AD we hope that no AD trial fails.

Q: Can you update us on last year's Nature paper?
A: Yes. We presented data at the international conference in Stockholm on the effects of most FDA-approved NSAIDs with respect to Ab42 lowering. These showed that in acute dosing studies in Tg2576 animals, flurbiprofen was the most effective NSAID at lowering Ab42. For these studies we used 50mg/kg/day of each drug. In humans, the maximum dose of flurbiprofen is something like 200 mg/day, whereas ibuprofen can be administered at 1,600-3,200 mg/day for short periods of time. Thus, it is unclear whether approved doses of flurbipofen would be better then approved doses of ibuprofen for lowering Ab42. However, flurbipofen is adminstered as a racemic mix; we have found hat R-flurbipofen is equally effective at lowering Ab42, but lacks activity against cyclooxygenase. Thus, higher doses of R-flurbiprofen are likely to have fewer side efects than adminstration of traditional NSAIDs, and may enable us to determine whether R-flurbiprofen can lower Ab42 in humans. Myriad Genetics, who has licensed the use of R-flurbiprofen form Encore pharmaceuticals has recently filed an IND to test R-flurbiprofen in AD (see http://www.corporate-ir.net/ireye/ir_site.zhtml?ticker=mygn&script=410&l ayout=9&item_id=327790). This planned phase 1 trial will be co-sponsered by the NIA.

Personally, I am disturbed by the actions of Public Citizen in that they use our data to support their argument, but never once spoke to either Eddie or myself for our take on it. This is irresponsible. — Todd Golde, Mayo Clinic, Jacksonville, Florida.

Q & A with Greg Cole. Questions by Gabrielle Strobel. —Posted 25 September 2002

Q: Do you see any merit in the letter's scientific argument? And what do your own mouse studies of NSAIDs show?
A: I think the scientific argument has some merit but is flawed for the following reasons.

a) Amyloid and Ab42 are very important endpoints, but not the only important endpoints likely to impact AD prevention. For example, regulating responses to amyloid/Ab are also potentially important pathways to control, and these may involve both excitotoxic and inflammatory pathways. Cox inhibitors, including naproxen and selective cox-2 inhibitors, can be protective in these pathways. Therefore, the fact that a drug doesn't share the Ab42 lowering effect of the subset of NSAIDs identified by Weggen et al doesn't imply a lack of efficacy by this drug. And the authors of that paper certainly would not claim that, either. It is entirely reasonable to believe that reducing excitotoxicity and at least some aspects of inflammation should be beneficial and may be a common basis of benefit from many different NSAIDs. In that case, reducing Ab42 would be a potential added benefit if effective NSAID doses can be reached without dose-limiting toxicity.

b) Perhaps it is just my ignorance but I don't understand how the Rotterdam study results have the power to safely conclude that only NSAIDs that lower Ab42 have contributed to protective NSAID effects. The most widely used drug in their study was Diclofenac, which accounted for 43 percent of the prescriptions and 36 percent of cumulative duration. Naproxen was 17.64 percent of prescriptions and 16.46 percent of duration. In fact, more than 2/3 of the NSAIDs exposure was from NSAIDs that fail to lower Ab42 in vitro, and no more than 30 percent of the NSAIDs used lowered Ab42. Is there enough power in the study to draw conclusions that differentiated NSAIDs? For the overall NSAID group with more than 2 years exposure, there was 80 percent reduction in risk, but less than 33 percent of the NSAIDs used lowered Ab in vitro.

How can they be responsible for the effect? How is this possible? Sulindac and indomethacin account for fewer than 5 percent of the total NSAID prescriptions, so little can be said about the efficacy of these drugs from this study. Does this mean that the group with longer exposure and efficient protection is heavily weighted toward ibuprofen, which accounts for only 21 percent of the total cases but roughly two thirds of the Ab42-lowering drugs? Does that mean that ibuprofen for more than a few years works very well? It would be useful to know the actual figures for different drugs and see the statistics. In conclusion, I suspect that the Rotterdam study data, like the Cache County and Baltimore Longitudinal studies that had roughly 50 percent ibuprofen users in the NSAID group, show that ibuprofen may work very well to prevent AD. For any single NSAID, it has the best epidemiological support.

In our hands, ibuprofen had a very significant effect in an AbPP-transgenic mouse model. The actual plasma levels achieved with an effective dose in chow in AbPP transgenics have been measured by Doug and Sally Boudinot in our mice and are in the range obtained in humans with analgesic dosing, for example 400 to 800 mg per day. Given that more than 26 epidemiological studies and Breitner's twin-sibling study show apparent prevention effects, plus acute and chronic animal studies show efficacy, there is a compelling rationale to go forward with prevention trials.

c) The argument that because specific NSAIDs like naproxen or cox-2 inhibitors have failed to treat AD they won't work for prevention, either, is at best weak and probably irrelevant. The NSAID epidemiology strongly supports a need for intervention at an early pre-symptomatic stage, which is the whole point of a primary prevention trial. Pathogenesis in a complicated cascade like AD is very likely stage-dependent, just as it is for atherosclerosis or cancers. One would not expect that taking aspirin or stopping smoking would work at late, symptomatic stages of these diseases. This is easy to imagine with AD. For example, if the formation of seeds for AbPP or tau amyloid deposition is a rate-limiting event and levels are 10 percent above a threshhold level, treatments that reduce amyloid by a modest 20 percent level may be very effective in markedly delaying the disease. They would be much less effective if the level of monomer was 30 or 40 percent above that threshold, as in some transgenic models. Amyloid formation is seeded and autocatalytic. Interventions influencing amyloid might not be very effective if the level of amyloid accumulated was already sufficient to drive pathogenesis. For example, if intracellular tau aggregation is already seeded and well under way, lowering extracellular amyloid formation may be no longer be relevant. Not getting hit in the head any more doesn't stop dementia pugilistica and tangle formation from progressing in retired boxers. The intervention had to be early.

d) Finally, NSAIDS that don't lower AbPP42 production may influence amyloid deposition by regulating pro-amyloidogenic co-factors such as a1 antichymotrypsin or ApoE. The argument for ACT is very clear and supported by data presented by our group and others.

Q: In your opinion, should the trial continue as is?
A: Yes. There is a basis for testing naproxen and a reasonable rationale for COX including COX-2 inhibition as a target. Although I would be happier with ibuprofen as a drug choice, there are many likely endpoints shared with naproxen and other NSAIDS that provide a strong rationale for a trial. These would be related to COX inhibition and likely to Control inflammatory cascades, alpha 1 ACT and excitotoxicity, and possibly produce beneficial cognitive or vascular effects. In fact, there is an interesting paper from Glenda Halliday and collaborators in Australia (Arch Neurol 2000 58(3):517-9 ) that examined a group of AD patients who were regular longterm NSAID users and found that they had improved neuropsychological test scores but no difference in pathology-including plaques and tangles. The majority of these human NSAID users were taking naproxen at the same dose used in the ADAPT Study, suggesting possible AD benefits unrelated to pathology. We don't know for sure how NSAIDs work to prevent AD, and naproxen may well work. It is important that we keep NSAID prevention trials on track.

Q: How about indomethacin or flurbiprofen?
A: Indomethacin and flurbiprofen are too toxic for prevention. R-flurbiprofen or the NO-fluribprofen derivative have potential (see related news item ). However, they need to go through phase I trials in elderly and probably AD patients to find doses with biomarker efficacy and very clear safety, and probably should be tested on AD patients for efficacy. —Q & A with Greg Cole. Questions by Gabrielle Strobel. 2002.

Q&A with Karen Hsiao Ashe—Posted 25 September 2002

Q: Do you see any merit in the letter's scientific argument?
A: First off, I support Alzforum's effort to present a more balanced picture than the letter did. I will answer your questions as best I can without divulging unpublished data. I do see some merit in the scientific argument. However, this argument is by no means firm or complete. For instance, data examining the effects of NSAIDs on memory in transgenic mice could potentially strengthen or weaken it, depending upon the results.

Q: I believe you are testing certain NSAIDs in Tg2576 mice. Did all have similar effects on behavioral endpoints or did you see differences? Are the NSAIDs that affect Ab cleavage showing better behavioral results than the ones that don't?
A: Although I have done these experiments comparing ibuprofen, naprosyn, and rofecoxib, I prefer to wait until they are published before sharing the results.

Q: Did you test flurbiprofen?
A: We did not test flurbiprofen for effects on memory.

Q: In your opinion, is there enough in vitro and animal science on Ab-altering NSAIDs to change ADAPT now?
A: No. All the work on NSAIDs using in-vitro systems and animal subjects have focussed upon A b as an outcome measure. We have recently shown a clear but complex relationship between Ab and memory (Westerman et al., J Neurosci, 2002). The implication of our work is that currently available measures of Ab are not sufficient to predict effects on memory. It is therefore very important to extend observations made with NSAIDs in animal subjects to examining effects on memory before recommending changes to ADAPT.

Q: Are there other potential mechanisms for NSAIDs in aging neurons, besides Cox and A bPP processing that go into the equation?
A: There are potentially other mechanisms by which NSAIDs work that are not mentioned in the letter. One interesting putative mechanism, recently published by Boutand and colleagues (J. Neurochemistry, 2002, Vol 82, 1003-1006) involves prostaglandin H2-mediated Ab crosslinking. There may be others that remain to be discovered. — Karen Hsiao Ashe, University of Minnesota, Minneapolis.

Comment by Marcia Gordon— Posted 25 September 2002

The demand to stop the trial strikes me as similar to Monday-morning quarterbacking. It is possible that the trial might be different if we'd had all of the data we have today 5 years ago, when the work to carry out an NSAID prevention trial began to build consensus.

I agree with Public Citizen that the inflammatory response probably occurs as a response to amyloid build-up. The genetics of the disease clearly place deposition of amyloid as the primary and most necessary event in the development of AD. However, they have left out a key piece of the puzzle—I find the data arguing that it is the brain inflammatory response that is responsible for killing neurons quite convincing and compelling. The transgenic mice and some human patients referred to as "high-plaque normals" clearly demonstrate that accumulation of massive concentrations of amyloid is not the only requirement for developing neurotoxicity. It is this death of neurons that causes the tragic symptoms of AD, not the accumulation of amyloid, which naturally occurs in all of us as we age. I find it quite plausible (yes, still) that NSAIDs may provide benefit in AD through mechanisms unrelated to their ability to inhibit g-secretase. The only way we will know for sure is to complete clinical testing, using a time frame long enough to accurately portray the clinical picture of AD.

I find it ironic that Public Citizen stresses the potential adverse consequences of NSAIDs, but the traditional NSAID chosen for the ADAPT trial, naprosyn, is widely recognized as having a better safety profile than most other drugs of this class, including 2 of the 3 other agents specifically mentioned by Public Citizen, indomethacin and sulindac. The information conveyed to patients in clinical trials is reviewed by IRB committees and carefully worded. It is required by law that information be conveyed in a way that is understandable to the lay population.

We teach medical students from the first day they enter medical school the official position of the American Medical Association: prudent practice dictates medications should be used in the lowest possible dose necessary to produce clinical benefit for the shortest possible duration in every situation. This statement in the consumer information cannot be taken as a mandate that NSAIDs should never be taken for chronic conditions. Many Americans do take NSAIDs for long periods of time for various types of arthritis and other conditions. Naprosyn and celecoxib are available over-the-counter, which means many Americans take these agents at their discretion without direct medical supervision. It is true that NSAIDs do have potential for adverse consequences, and some patients must withdraw from chronic use because of these adverse events. Basic scientists and also the pharmaceutical industry are working diligently to develop newer NSAIDs with fewer adverse side effects.

The arguments raised by Public Citizen document the necessity of performing additional clinical trials specifically to test the g-secretase inhibition hypothesis. Rather than stopping the ADAPT, it should be expanded to include a third treatment arm incorporating an NSAID with strong g-secretase-inhibiting activity. This would allow comparison with the prototypcial NSAIDs already included in the ADAPT trial. For all clinical trials, there is a review board to supervise the study. It might be appropriate for this review board to recommend that the language of the informed consent concerning adverse events be strengthened. Patients might also be screened for occult blood in the stool at routine visits to more closely monitor potential GI adverse events.

We as basic scientists are working on other NSAID agents. Flurbiprofen is used a lot in Europe, but not too much here. I'm not sure why, possibly it is a patent position issue. I don't believe it has any special merit for inclusion in a clinical trial. We are also working with newer agents that might have better safety profiles (see related news item ). And we are comparing various NSAIDs to determine which causes the most amyloid reduction in transgenic mice. Also, specifically note that although my university is one of the 6 ADAPT sites, I have no personal involvement or interaction with the trial. — Marcia Gordon, University of South Florida College of Medicine, Tampa.

Letter by Monique Breteler to John Breitner—Posted 25 September 2002

Dear John,

I have read the public letter sent by the US advocacy group Public Citizen to the Department of Health and Human Services, asking its secretary Tommy Thompson to suspend the ADAPT trial. I appreciate the important watchdog function of lay and patient organizations, yet if they make scientific judgments these should be thoroughly founded. I profoundly disagree with the interpretation of the existing evidence and the scientific argument in mentioned letter. The authors of the letter partly base their opinion on our research findings from the Rotterdam Study, which is why I would like to comment on their arguments and give you my current views on whether anti-inflammatory drugs might be beneficial in the prevention of Alzheimer's disease.

Public Citizen postulates that there is no longer any biological basis for the ADAPT trial. Their arguments include that 1) the inflammatory hypothesis of AD no longer holds and Ab accumulation is central to Alzheimer's disease; 2) there is convincing evidence now that some NSAIDs do lower Abwhereas others do not; and 3) the NSAIDs that are being used in the ADAPT trial were not effective in preventing disease progression.

1. The inflammatory hypothesis no longer holds…

There is ample evidence that inflammation processes do play a role in Alzheimer's disease. Whether inflammation is causal, contributes to the progression of the disease, or merely marks the ongoing pathologic process is unclear. The fact that amyloid deposition in the brain triggers a local inflammatory response does not preclude that inflammation may contribute to disease initiation or progression. Our observation in the Rotterdam Study that persons with higher high plasma levels of inflammatory proteins were at an increased risk of dementia and Alzheimer's disease is compatible with inflammation contributing to the pathogenesis of dementia.

It is highly likely that Ab accumulation in the brain does play a central role in Alzheimer's disease. However, what causes this accumulation in the majority of cases is largely unclear. Alzheimer's disease is a multifactorial and heterogeneous disorder. This implies that there is no single cascade of events that ultimately leads to the clinical syndrome. Moreover, it implies that there may be different mechanisms on which one could intervene to prevent or delay onset of disease.

Most basic research focuses on specific Alzheimer pathology. However, there is now convincing evidence that the larger proportion of elderly dementia patients actually has a mixture of degenerative as well as vascular pathologies in their brains that may all have contributed to the clinical syndrome. Prevention of vascular pathology may be an effective strategy to postpone onset of clinical Alzheimer's disease. Since inflammation is involved in the occurrence and progression of atherosclerosis, anti-inflammatory strategies might prove effective in the prevention of Alzheimer's disease through an effect on vascular pathology.

2. Some NSAIDs lower Ab, others do not.

Observations that some specific NSAIDs do have an effect in vitro or mice on amyloid processing are extremely interesting. However, they do not rule out that those or other NSAIDs may also have effects on different yet relevant mechanisms. Also, it is unknown whether the NSAIDs that lowered Ab have an effect on amyloid processing in humans.

Triggered by the cell culture and mouse observations, we re-analyzed the Rotterdam Study data. Our findings in that re-analysis are compatible with the view that different NSAIDs may have different effects on risk of Alzheimer's disease. However, our findings should be interpreted carefully. As in any observational study, there may have been residual confounding. Also, we had limited power. The lab and epidemiological data strongly suggests that this is a research area that merits further investigation.

As yet there is no proof that lowering Ab levels reduces the risk of Alzheimer's disease, nor that that might be achieved through specific NSAIDs.

The NSAIDs being used in ADAPT were ineffective in preventing disease progression. The fact that a drug is not effective in stopping progression of overt disease does not imply that the drug may not be effective at a completely different stage. Indeed, since different processes are likely to be involved at different stages of the disease, it is to be expected that drugs that prevent onset of disease have no effect on disease progression and vice versa. — Monique Breteler, Erasmus University, Rotterdam, The Netherlands.

Comment by Paul Aisen—Posted 25 September 2002

I have read the letter and was quite distressed by it. I am not involved in the ADAPT, though I have had a number of discussions with John Breitner about it, and have shared data from the ADCS NSAID study with the ADAPT investigators. Also, I was involved with the NIA review of the ADAPT application.

I would describe the letter as an inflammatory tirade rather than a scientific argument. While it does review some relevant scientific observations, the discussion is one-sided. I do not agree with the conclusions and recommendations of the letter. Celecoxib and naproxen were chosen based on a number of considerations, including epidemiologic data suggesting that NSAIDs reduce the risk of AD, and basic research suggesting that a number of inflammatory processes, including cyclooxygenase activity, may contribute to neurodegeneration in AD, as well as tolerability issues.

The recent evidence suggesting that a subset of NSAIDs favorably influence AbPP processing is quite interesting, and may be important. But it represents one recent addition to a huge body of data on inflammation and anti-inflammatory drugs in AD. I believe that this new line of evidence warrants further investigation, perhaps including clinical studies.

It is quite difficult, though not necessarily impossible, to alter a large clinical trial that is under way. One major issue is expense: it would require a huge additional investment to add an ibuprofen arm. Indomethacin would probably not be a good choice, because of toxicity concerns. Ibuprofen or sulindac would be preferable. I believe that discussion of additional trials that would test the hypothesis that ibuprofen or sulindac can favorably alter the course of AD is appropriate at this time. Paul Aisen, Georgetown University, Washington, D.C.

Comment by Leon Thal—Posted 25 September 2002

The addition of a third arm with ibuprophen is a consideration. It would be logistically difficult but possible to do. The real question is does ibuprofen make any more sense than naproxen since tolerated doses over 5 years may well be too low to block formation of b42. —Leon Thal, University of California, San Diego and Alzheimer's Disease Cooperative Study.

Q&A with Giulio Pasinetti—Posted 25 September 2002

Q: Do you see merit in the letter's scientific argument?
A: No. However, I definitely understand the concern of this group with regard to the iatrogenic consequences associated with NSAIDs (gastrointestinal toxicity), and do acknowledge that it is an ambitious trial. Current data clearly indicate a protective effect for NSAIDs in AD, and the best proof of concept is a prophylactic trial. I do not get the impression that a scientific argument has been presented by the letter in question. Epidemiological evidence show that NSAIDs protect against AD, one of the most significant studies supporting this was published in the New England Journal of Medicine late November (see related news story). This study, which tracked patients for about 8 years found that NSAIDs, regardless of their type, protected against AD. Ibuprofen accounted for about 21 percent of NSAID use and naproxen about 18 percent, selective COX-2 inhibitors were not evaluated most likely due to their recent introduction to the market. Based upon the stronger anti-inflammatory effects associated with naproxen, this drug may be a good candidate. With regard to selective COX-2 inhibitors, a large body of data shows upregulation of this enzyme in the brain during the earliest phases of AD dementia , and thus begs the question as to whether these drugs (which are more tolerable than traditional NSAIDs; e,g. ibuprofen) will be effective in preventing AD dementia (neuropathology) rather than treating it.

Q: In your opinion, should the trial continue as is?
A: Yes. The trial is a much-needed step in the critical evaluation of what epidemiology suggests. Given that there is currently no effective treatment for AD, and that with proper monitoring NSAIDs are safe in healthy individuals, this trial provides a great deal of promise for those people who are at significantly higher risk for AD. However adjunctive treatment with gastro-protective compounds (although adding to the cost of this trial) may be helpful, and further control of Helicobacter pylori (a culprit in more than 90 peptide of peptic ulcers) is essential to successful and safe management of the ADAPT.

Q: Is there a constructive way to deal with the underlying issue that some emerging research points to NSAIDs other than celecoxib and naproxen as perhaps more promising? Add an ibuprofen arm?
A: That we have finally overcome the hurdle of acknowledging the role of inflammation in AD is encouraging for future studies, however, cost will remain a limiting factor. Further problems lie in recruiting large enough patient populations to test multiple drugs, and selecting the specific drug to administer. Conducting a trial that employs a selective COX-2 inhibitor and a non-selective COX inhibitor (as does ADAPT) is probably the best place to begin dissecting the role of NSAID compounds. At the same time, experimental studies must continue to clarify all the mechanisms of NSAIDs so that we can further deduce the potential role of these compounds in AD, and better classify the drugs themselves. An ibuprofen arm may be useful, although COX-1 inhibitors may present serious concerns for tolerability in a "preventive" treatment.
Interestingly, in a tolerability study conducted at Mount Sinai School of Medicine we found that the preferential COX-2 inhibitors nimesulide was outstandingly well tolerated in the AD population, in contrast to other drugs, for periods exceeding 2 years ( see related news story and comments). We also found that post-hoc analysis of individual items suggests possible benefit on clinical assessment of orientation and judgment, and possible mood elevation in response to short-term therapy with nimesulide, compared to placebo.

Q: How about indomethacin or flurbiprofen?
A: The major roadblock of anti-inflammatory drugs studies is tolerability. For example, there was reported a 59 percent drop out rate in Joe Rogers's indomethacin trial (Abstract), and a recent review confirmed that indomethacin is unsuitable for AD (Tabet and Fldman, 2002). Moreover, indomethacin represented only 4.6 percent and flurbiprofen 0.23 percent of the NSAIDs in the recent Rotterdam study (NEJM), thus it is currently difficult to support a protective effect for these compounds. The chronic use of NO-releasing drugs may pose serious hazards even among current users, especially within the aging population that is at high risk for multiple neurodegenerative disorders. Further, given that treatment with certain NSAIDs is currently the number one cause of iatrogenic illness, we need to proceed cautiously. We strongly urge the systematic and careful monitoring of current chronic users of NO-releasing drugs for signs of neurological and possibly gastro-intestinal problems. In view of the mechanisms evoked by NO-releasing drugs including activation of microglia in the brain, I would be extremely cautious in applying these drugs in a frail population. — Giulio Pasinetti, Mount Sinai School of Medicine, New York.

Comment by Pat McGeer—24 September 2002

There is some merit in the Public Citizen letter since it draws attention to the failures of COX-2 inhibitors and naproxen in AD clinical trials and suggests that the informed should be more thorough in warning of the dangers of NSAID use. However, the letter loses credibility in describing the ADAPT trial as unethical and calling for its termination. The authors overlook that naproxen, along with ibuprofen and some other traditional NSAIDs, are considered safe enough to be over-the counter agents. They also overlook that NSAIDs, prescription and non-prescription, are the most widely used of all classes of drugs. Physicians do not hesitate to prescribe NSAIDs for arthritic conditions that are not life-threatening. In contrast, AD is an extremely dangerous condition. There is no effective treatment, and there is inexorable progression to death. This needs to be a fundamental consideration in all clinical trials. The outlook for AD patients is hopeless unless the agent being tested can eliminate or slow the neuronal death. Epidemiological and pathological data indicate that NSAIDs may offer some hope. The inflammation in AD brain is more severe than in osteoarthritic joints, but it is silent because the brain has no pain fibers.

Having said that, the problem with the ADAPT trial is that the drugs chosen have already been shown to be ineffective. There needs to be a re-evaluation, not a cancellation. Three trials of COX-2 inhibitors have failed, two with rofecoxib and one with celecoxib. Since COX-2 is concentrated in neurons, not microglia, and normally facilitates neurotransmitter action, these failures are hardly surprising. Further trials of COX-2 inhibitors would appear to be unnecessary, so this arm of the ADAPT trial could be dropped. Naproxen is more logical since it is one of the traditional NSAIDs upon which the overall epidemioligical data of AD sparing are based. But a trial of low-dose naproxen (400mg/day) also failed, and the same low dose has been selected for the ADAPT trial. So far no data have been published on how well naproxen reaches the brain, and the epidemiological data are not robust enough to assess naproxen in comparison with more widely utilized NSAIDs. So a better choice of a traditional NSAID might be in order. Since the trial is in its early stages, a reorganization could easily take place.

Which candidates have the best chance of success? The best epidemiological data exist for ibuprofen. It is the most widely utilized NSAID, probably because it is considered to be the safest. In common with others of the profen class, it readily crosses the blood-brain barrier. Indomethacin is a more powerful NSAID which also has easy access to the brain. It has been shown in a pilot clinical trial to inhibit AD progression. Diclofenac is a third NSAID that reaches the brain. It has also been tried in a small clinical trial. Although statistical significance was not reached at the p<.05 level, the diclofenac patients scored better than the placebo patients on each of the measures. Much has been made of the high dropout in the indomethacin and diclofenac trials due to GI side effects. But the right approach is not to abandon these drugs in favor of ones that do not work, but to find out how best to use them. That could include adjusting dose levels, or using pump inhibitors or H2 antagonists in sensitive individuals. Another more promising approach is to use representatives of a new class of NSAIDs, which are modified by the addition of a nitrate ester. These so-called NO-NSAIDs have little GI toxicity. They are cleaved by the time they circulate so they retain the properties of their parent NSAID. These NO-NSAIDs are in early clinical trials for other indications, and their application to AD could be accelerated. — Pat McGeer, University of British Columbia, Vancouver.

Comment by Curtis Meinert, Chairman, ADAPT steering committee, —24 September 2002

Let me comment with some historical perspective gained over the course of previous clinical trials.

What is the only drug ever removed from the market by the Secretary of Health and Human Services? Answer: Phenformin. So what, you may ask? Well, at about the point we are now at in ADAPT during the University Group Diabetes Program (UGDP), there came a new oral agent for adult-onset diabetes—phenformin. Like all new ideas it was the best, and the hypothesis underlying its use was said to almost certainly be the correct one. Some in the UGDP argued that we were studying the wrong treatments and that the trial would be useless unless we added phenformin. Those who argued against such a change lost and, therefore, the treatment was added. Six or seven years later the drug was history because it was dangerous.

In the Coronary Drug Project (CDP), like ADAPT a secondary prevention trial, we came under attack midway through it because results from another study suggested clofibrate was the cat's meow. Critics argued that these results made the CDP unethical because clofibrate was now the treatment of choice and, therefore, the CDP should be stopped. What happened? We stayed the course and clofibrate turned out to be as effective as placebo.

I am quietly amused by the way in-vitro and animal data are evaluated when they fit a belief system as opposed to when they do not. Everyone, it seems, supports the Ab hypothesis based largely on one study (not even replicated so far as I know). But what happened to animal studies reported during the UGDP suggesting a mechanism of effect as to why tolbutamide (one of the treatments stopped in the UGDP) might, in fact, be harmful? The results were dismissed as flawed and ignored. They did not fit the belief system at the time. I am not enough of a student of the Rotterdam results, but I wonder if those who say the ratio has changed use data based on small numbers from an ad-hoc subgroup analysis and choose to ignore the aggregate results of that study? Why? Could it be that the new results fit a belief system?

We may not have the right treatments, and if we produce a null result there will be those who will say "We told you so". Further, the chance of finding a benefit, even if one exists, is like looking for a needle in a haystack. We have limited power now. Water it down by spreading ourselves even more thinly and the trial approaches being unethical because we have no chance of showing benefit, even if one exists.

If every new result or criticism from some group with their own agenda requires a change to the consent form, we need to change it frequently and let the people decide how to evaluate that new information. I am not sure that would be helpful to anyone. —Curtis Meinert, Johns Hopkins University, Baltimore, Maryland.