April 15, 2003

Letter on Ethical Concerns with the Alzheimer’s Disease Anti-Inflammatory Prevention Trial (ADAPT)

Ruth Kirschstein, M.D.,
Deputy Director
National Institutes of Health
Bethesda, MD 20892 

Dear Dr. Kirchstein:

We appreciate the lengthy discussion of the points in our September 4, 2002 letter to Secretary of Health and Human Services Tommy Thompson concerning the ADAPT trial on Alzheimer’s disease prevention.  Here, we reply to the arguments made in your response to our letter.

RESPONSE TO POINTS CONCERNING STUDY HYPOTHESIS

At this point, unfortunately, it is not really possible to know whether a particular drug will work in preventing Alzheimer’s disease (AD) since so little is known about the genesis of this complex pathological state. As a result, it is possible to make a case for many different mechanisms of disease and the use of many different drugs. However, in deciding which drugs to test in a specific prevention trial, one must use a higher safety standard than usual because one is treating basically healthy (and in this case, elderly) people on a long-term basis. The goal should be to test the safest drugs with the best scientific rationale for success, based on both theoretical and experimental data.

It seems clear from available experimental data that some non-steroidal anti-inflammatory drugs (NSAIDs) are more likely to succeed than others and may be both safer and cheaper. Such a drug, logically, would be the one to test; this is why we supported the use of ibuprofen.

DRUG DOSES AND EXPOSURE

Questions have been raised as to how ibuprofen’s success in lowering A$42 levels (A$42 is the form of amyloid protein that is thought to be responsible for neural toxicity in AD). in animal studies relates to its potential usefulness in humans. A landmark study found that ibuprofen significantly lowered A$42 levels in mice[1] The authors tended to discount the results as they applied to people since “enormous doses” were used in mice.[2] However, in reality, and very importantly, mouse exposure was low compared to humans. This is because, in order to take into account the differences in metabolic rates between species and provide a more reliable basis for comparison, toxicologists routinely compare the amount of drug to which animals and humans are exposed on the basis of surface area, not dose. The Food and Drug Administration (FDA) routinely uses surface area in drug labeling when actual plasma drug levels are not available.

In order to compare drug exposures across species, one first transforms dose (mg/kg) to surface area (mg/m2). Each species has its own surface area conversion factor: those used by the FDA  are 3 kg/m2 for mice and 37 kg/m2 for a 60 kg human.[3] The calculation for comparing ibuprofen exposures is shown below (using the 50 mg/kg/day dose given to mice[1] and an 800 mg/day dose administered to people, assuming a 60 kg weight person):

Mouse:  50  mg/kg  X 3 kg/m2 = 150 mg/m2/day
(exposure to ibuprofen in Weggen et al)[1]

Human: 800 mg/60kg =13 mg/kg  X 37 kg/m2 = 493 mg/m2/day

Thus, the drug exposure at which there was significant lowering of A$42 in the mouse was only 30% of that expected to occur in humans at an 800 mg/day ibuprofen dose. These data (rather than reflecting “enormous doses” in animals), actually give increased weight to ibuprofen’s ability to influence A$42 levels in humans.

NIH states that the celecoxib dose is “the highest dose compatible with safety concerns,” yet the 400 mg dose of celecoxib used in the study is the maximal dose approved for rheumatoid arthritis and twice the maximal dose for osteoarthritis in the FDA-approved label (in trials of 3 to 6 months duration rather than the seven years in the ADAPT trial). Furthermore, celecoxib has been approved only since December 1998; thus its adverse effects are still being unraveled.

The NIH letter also states that, “Safety concerns dictated the use of a lower dose of naproxen.” However, this isn’t a “lower dose:” on a mg/day basis, it is essentially the same as that for celecoxib (the celecoxib dose is 440 mg/day and the dose for naproxen is 400 mg/day). Furthermore, patients over 65 years of age are instructed in the naproxen non-prescription drug label not to exceed the 220 mg bid dose (italics added). So for those over 65 years in age, 220 mg bid (taken in ADAPT) is the high dose, thereby increasing the likelihood of adverse effects in this elderly population.

RESPONSE TO POINTS CONCERNING ADVERSE EFFECTS AND THE IMPLICATIONS FOR INFORMED CONSENT

The NIH letter states that safety is “paramount in all NIH studies” and that “the participant must be informed in writing of all the potential risks and benefits…” (italics added). The latter clearly has not been done in ADAPT as one can see by comparing the FDA-approved label with the consent form, as we did in our September 4, 2002 letter.[4] And although the NIH letter says that there is an additional safety back-up with FDA medical officers reviewing NIH-generated protocols, these medical officers are under so much pressure to keep current with commercial applications (Investigational New Drug and New Drug Applications) that they have little time for either protocols or reports of adverse effects in trials initiated by NIH,[5] perhaps assuming that these trials already have the needed oversight.  The ethical issues we are raising in this trial include not only whether any adverse effects encountered will be acted upon, but whether, given the current state of scientific knowledge, the patients are given full information at the onset of the trial.

The NIH letter states that the consent form for ADAPT is extensive, “consisting of 20 pages.” This presumably includes the consent forms for eligibility, cognitive testing, genetic testing, and DNA banking, as well as enrollment. The enrollment consent form example that we have is 6 pages, only 1.5 pages of which are concerned with “Risks/Discomforts.” We haven’t seen more recent versions (we were told that these are not available, that each IRB had their own but are not required to submit copies to NIA).[6]

In our letter, we described the many ways in which the ADAPT consent form falls short of the information required to be in the FDA-approved labels for the drugs being tested. However, we are not advocating providing patients with the full FDA-approved labels for celecoxib and naproxen; rather, we think something like the “Information for Patients” section present in the celecoxib label would be useful, coupled with a list of specific physical signs of adverse effects.[7] Further support for our position comes from the celecoxib label that emphatically states “Patients should be informed about the signs and/or symptoms of serious GI toxicity and the steps to take if they occur,” yet no information on signs or symptoms is provided to patients in the ADAPT consent form we reviewed.

The NIH letter continues, “Trial participants are encouraged to ask questions,” but patients, lacking background information, don’t know what to ask. The choice shouldn’t be between the complete PDR labels and the skimpy statements in the ADAPT informed consent.

The NIH letter says that “The consent form for ADAPT contains information on each of the adverse events described in these sections [Contraindications, Warnings, and Precautions sections] of the FDA-approved label with the exception of ‘Hematological Effects’ and ‘Preexisting Asthma.’” Just to state that a problem may occur (as done in ADAPT) is not the same as providing details of the signs and symptoms of that event or giving volunteers some estimate of its frequency. For example, the celebrex label says, ”Patients should be informed of the warning signs and symptoms of hepatotoxicity (e.g., nausea, fatigue, lethargy, pruritus, jaundice, right upper quadrant tenderness, and ‘flu-like’ symptoms). If these occur, patients should be instructed to stop therapy and seek immediate medical therapy.”

The NIH underscores the fact that only patients who haven’t had “serious ulcer disease or uncomplicated peptic ulcer with symptoms in the 28 days prior to beginning the study” will be included in the trial.  However, this does not guarantee a population at low risk for these complications, particularly when the population is elderly[8] and will be taking the drug for longer than in any previous clinical trial. Furthermore, one cannot rely on exclusion criteria to avoid health issues that can arise very quickly (even with minimal symptoms) in the course of a 7-year trial in a population of elderly patients, even with monitoring every 6 months.

It is not encouraging that the Steering Committee for ADAPT decided to change the rate of gastrointestinal (GI) ulceration in the informed consent from an already misleadingly low value to that of the rate seen in the relatively few patients treated so far in the ADAPT trial. It is the ethical responsibility of the investigators to provide the most complete information to their trusting patients (such as that provided in the FDA-approved labeling) and not try to find a rationale to use the lowest possible numbers. To the extent that the early ADAPT data are provided, they must be placed in the context of the lack of drug-specific ulceration rates and the lack of long-term follow-up.

Although we did not state that thrombotic tendency and delay in bone and ligament healing were in the FDA-approved label, these are all potential risks of which investigators and patients need to be aware. FDA routinely uses data from animal toxicology studies such as these to determine what should be monitored in clinical trials. Investigators will be unlikely to register these as adverse events unless they know that there is a theoretical basis for their occurrence.

The basic question is, how can one possibly know, a priori, which complications “pose a practical risk” when, for the first time ever in a clinical trial, elderly patients are being given an NSAID for seven years? Is it not better that physicians be alert for what could even remotely be considered a drug-related adverse event so that the drug could be stopped before the patient incurs serious harm? In regard to both duration of treatment and age of patients, this trial enters uncharted territory.

The ADAPT protocol states that COX-2 selective inhibitors “cause fewer adverse events.”[9] However, a recent editorial[10] notes that “many questions [concerning COX-2 safety] remain unanswered” and “at present it is still difficult to give patients an honest, accurate, and understandable account of the balance between relief of pain … and the likelihood of serious adverse effects …” The blithe assumption that the COX-2 drugs are safer is, at present, insupportable.

CONCLUSIONS

The question remains: should ADAPT continue in its current form?

We feel that the answer is no for the following reasons:

1. based on animal and epidemiological data, the two drugs being tested are less likely to work than ibuprofen
2. safety problems related to celecoxib use, because of the newness of the drug, are only partially known, and
3. the ADAPT trial is still in an early stage and patients could be re-randomized into a new trial employing ibuprofen and placebo.

To subject healthy people to long-term NSAID use with all the potential adverse effects, without arranging to test the one drug thought most likely to be of benefit, seems highly unethical.

With only two arms in an ibuprofen vs. placebo trial, the numbers of patients needed to be recruited are probably about the same as those remaining to be recruited for the rest of the ADAPT trial. With limited resources, why not use them in the manner most likely to benefit the patients themselves and society at large? If trials are to be done, they should be based on drugs that have the clearest safety profile and are most likely to be effective. Finally, both patients and investigators need to be fully informed of all possible risks.

Sincerely,

Elizabeth Barbehenn, Ph.D.
Research Analyst

Peter Lurie, M.D., M.P.H.
Deputy Director

Sidney M. Wolfe, M.D.
Director
Public Citizen’s Health Research Group

---