How did the regulators parry these new proposals and questions? Eight EMA officials and nine FDA officials took part in the discussion about preclinical trials in autosomal-dominant AD at an 8 November 2010 meeting held in London at Europe’s government drug agency. Read these excerpts and divine for yourself how they think about risk and benefit at this point. The excerpts are notes from this first of several conversations, not binding statements. EMA and FDA statements appear by agency, not by its individual representative.

Question: Can ADAD drugs receive orphan drug status?

FDA: AD itself by far exceeds the maximum prevalence of 200,000; ADAD stays well below it, so it is possible in principle. To obtain orphan drug designation, you’d have to make the case that ADAD is a subset of AD that needs different treatments. Or you’d have to make the case that ADAD is different from AD in general.

EMA: We take the same view. Since we consider the condition together with the drug, the drug would have to be developed specifically for the subset and could not be developed for the broad condition. That would exclude the use in the broad condition.

Question: what about existing and exploratory products that are not specific to ADAD?

FDA: To get an orphan indication, you’d have to make the case that the product is specific to that genetic form, i.e., to fewer than 200,000 patients.

Question: What impact would designating ADAD as orphan have on trial design?

FDA: The rules for qualifying a drug are the same for an orphan or non-orphan condition. You have to have safety and efficacy data. In theory, there is no difference between orphan or not orphan, though in practice, we do impose slightly different standards. If the condition is very rare, we may not require two independent studies. In this particular setting, we may require a single study.

EMA: We look more to the strength of the evidence, less to whether you have two trials.

FDA: So do we. We can apply the one-study standard plus totality of evidence, which can be supplied from other studies.

FDA: We can’t say in advance what outcomes would support approval. A very robust outcome, a clinical outcome, and a biomarker outcome would be appropriate in a setting like this.

The other clinical issue that might be raised when we talk about study design is this question of great interest to you: approval based on a surrogate only, without a concomitant clinical outcome. We have a rule and part of a law that allows us to approve on the basis of an unvalidated surrogate. ADAD might be a place where that could be considered. We do not think at the moment that would be appropriate, but it is a potential approach for study design.

EMA: On the possibility of getting orphan disease designation, either the drug is specified for ADAD or, if the drug is applicable for both ADAD and LOAD, it is unlikely that orphan designation will come forward. We are looking at a product that can be developed for both conditions—you spoke of doing ADAD trials as proof of concept because there has been so much difficulty in sporadic AD trials. ADAD trials would be an opportunity not only to give the patients from these families a chance to participate and have a treatment, but also to create new trials as proof of concept for LOAD. So the orphan pathway seems difficult.

If the drug is not specific to ADAD, in the first instance it might be studied in ADAD as proof of concept. But it also might come as full development of ADAD and then later also go for sporadic AD. So we have to discuss if you have the tools for development in familial AD. For that we can accept development with just one clinical trial. We can see a situation of doing just a single trial in ADAD in part because you showed us stable information for biomarkers in FAD. On the other hand, there are a few troubling data coming from sporadic AD that might make trial design extremely difficult. [Editor’s note: This refers to increased, then decreased, atrophy in responders to AN1792 immunotherapy.] My question here is how similar or different ADAD and sporadic AD are in this regard.

FDA: We are certainly open to approving a drug just for inherited AD. That is an option. Whether that disease is ultimately different from sporadic is something we can discuss. The genetics describe a discrete subset that would be appropriate for development and approval independent of whether the drug then works in sporadic AD, too. That is feasible from a regulatory standpoint.

EMA: We agree. ADAD and sporadic AD are the same clinically; we see no reason to separate them. We view ADAD as a cluster of AD. We want it to be no longer neglected in trials.

Question: How about placebo controls?

FDA: Placebo is appropriate in this setting and should be used. That is caught up in the question, How long can you keep people on placebo in this setting? The trial must be long enough.

EMA: Let’s discuss the alternative of high-dose versus low-dose drug. That can give the same result as placebo control, but is more acceptable to patients.

FDA: We would accept that. We would accept any trial that is designed to show a treatment difference clearly. High dose/low dose is more acceptable to patients, but it is a bit deceptive because the stated goal is to show a difference. This has ethical implications of its own. It becomes a semantic exercise where you need to show a clear difference and then use an effective and a likely non-effective dose, so it’s a slippery argument.

The bigger issue is that historical controls were suggested in the presentations, and we insist that concurrent control is appropriate.

EMA: We need to be pragmatic in these discussions. In this field, everything has failed in the last 10 years. We are having nasty surprises with trials that have shown deleterious effects. We cannot assume when putting a drug in a trial that it is going to be effective, or that there will be no problem. To do a trial in a small population, the only patients that exist in the world, will be a unique opportunity. We cannot play around with designs that have confounding factors, so we cannot get around placebo.

A number of good markers show when the familial patients are likely to develop dementia. These markers are very reliable from a variety of data. If the patients really want to partner in this adventure, they have to accept these rules. It is extremely difficult to succeed here, so we cannot introduce additional confounds. Low dose/high dose in essence is a partial randomization. We cannot afford that.

Rawlins: Huntington’s is a different situation. Currently, a significant cohort of historical controls is being collected. In HD, the mechanism of neurodegeneration in pre-manifest and manifest is the same. And drugs work on both. For pre-manifest HD, placebo-controlled trials are a non-starter.

FDA: That is in the context of where there is an effective drug for manifest disease. That is a pre-manifest study you do not need to do. If a drug exists and works in manifest, and pre-manifest is just a milder stage, then you do not need to study it. You want to avoid doing a study in a particular population because you can do it, but that study does not lend itself to interpretation. That would be a bad outcome.

Rawlins: For ADAD, how about a futility trial to check that your treatment is ineffective and move on to the next?

FDA: There is no regulatory objection to that.

Rawlins: ...and compare treatment with historical control.

FDA: That is problematic. It depends on the effect size. If you have a drug that stops disease in its tracks, and you can clearly see that in a reasonable period of time, then yes, but we do not expect that from AD drugs.

EMA: Agreed.

William Thies, U.S. Alzheimer’s Association: For ADAD, there would be no difficulty recruiting to a placebo-controlled trial. If you confuse the results in this group, it will confound to an extent we cannot accept.

Nick Fox, University College, London: We have an FAD support group. Their message is a very clear willingness to take part in a placebo-controlled study. There are no disease-modifying treatments for AD at the moment that they would forgo by taking placebo. Their willingness may change when an effective drug shows up. [For more on this support group, see Part 4 of this series.]

Low dose in lieu of placebo is a sleight of hand. It may only be useful to protect people from finding out their status through side effects. A real issue we have to deal with is how we let people participate who do not want to know their status and might find out through the side effects.

Question: Can we define endpoints in ADAD trials?

EMA: Despite small sample size, the DIAN biomarker data are robust in characterizing this population. The MRI data recap what Fox has produced, and the CSF data are in line with what we know from CSF biomarkers in sporadic AD. These data support the bulk of biomarker knowledge we have, and this knowledge is very good. We can predict relatively well for an individual in a family when he or she is going to develop symptoms. That is extremely useful to design a trial.

Where I think things become complex is in the definition of tools to define the endpoint, because the data for amyloid imaging and MRI atrophy rates for this population are as complex as they are for sporadic AD. They have the same type of problems as endpoints, as showing drug response as they have for the sporadic disease. For instance, we know atrophy can go up and down in sporadic AD and are unsure what is good and what is bad. We know amyloid imaging can change, but we do not know if this is good or bad. So it is probably better to rely on a cognition measurement. Which one? It would have to be sensitive enough. So far, data from ADNI suggest the CDR sum of boxes as the most sensitive outcome. Would we be able with small sample size to have a result with that? That is a question.

FDA: For sporadic AD, we require an effect not only on a cognitive measure, but also on a global measure. In ADAD pre-manifest disease, we would want a clinical outcome. But we are open in conjunction with an effect on a surrogate or even multiple surrogate markers, to use a cognitive outcome on the subtle early change instead of the clinical.

You can make the argument that in this setting, a change in subtle cognitive testing even before patients complain of impairment together with biomarker would be adequate.

Paul Aisen, University of California, San Diego, ADNI, DIAN: This is a very interesting discussion. To clarify, in the ADNI database, we find the CDR sum of boxes to be the most efficient measure of decline only in the prodromal pre-dementia, what many in the U.S. might call the MCI phase. As far as ADAD is concerned, we are hoping to move earlier than this prodromal phase. DIAN is interested in moving our clinical trials into the preclinical or asymptomatic stage. In that stage, the CDR sum of boxes shows no change with time. It is not useful as an outcome measure.

But we have seen in PAQUID and ADNI that it is possible to see subtle cognitive change in this asymptomatic population. As the FDA has suggested, it may be very feasible to link our efficacy biomarker to a subtle cognitive change as a way of demonstrating a sign of clinical manifestation.

We do not have a means to show a global clinical measure. By definition, you can’t do that in an asymptomatic population. But with the biomarker and the subtle cognitive change, we have a way to move forward.

Bruno Dubois, Hospital de la Salpêtrière: I completely agree. The CDR sum of boxes represents clear functional decline. We need to move earlier. My only point of discussion with Paul is that I view these patients with the early cognitive change as symptomatic. I take that as the first symptom.

Aisen: You have been very helpful in clarifying the lexicon. The field needs help there. I am not talking here about the complaints that would bring the patients in to see the clinician. The subtle cognitive change I mean is before that. But yes, it is a symptom.

EMA: We are discussing several issues at the same time. One is doing trials in asymptomatic genetic families. That is totally new. Another is doing trials in patients that are pre-MCI/eMCI/at-risk patients, which is also new.

These are two populations. Is it wise to mix them in the same trial? If you do, at least stratify. Can the endpoint be the same for them both? Even in terms of safety, the risk/benefit may be different for these two populations.

The number of patients is small and you want maximal sample size, but putting the two populations together carries risks. If you move forward, come to scientific advice with EMA’s CHMP and discuss the protocol. These issues go beyond the scope of the forum here today. We are eager to discuss details.

FDA: Our comments pertain to the ADAD population, not the sporadic population. Whether they would apply to the latter is a larger question that would require considerable thought. To clarify: Accepting biomarker and an early cognitive change would be acceptable at this point for ADAD.

EMA: But even within the families, you have two different populations. People in the pre-dementia stage and people years earlier than that are in different stages of evolution.

Question: What is your feeling about the current state of the biomarker data?

EMA: We have worked hard in the last year to try to validate the biomarkers. We are looking closely at specificity and sensitivity.

FDA: There is good evidence that they track disease closely; some are more sensitive than others. They are useful for choosing folks for trials. The only position we have taken is that they are not ready for use as sole outcome measures. We are not saying which one is best. We think as many as possible should be incorporated in trials so we can learn how they move in response to treatment. We have no favorite at this point.

Question: In closing, can the regulators say they would make a decision based on biomarker and cognition?

EMA: Yes.

FDA: Yes, this is an outcome package we would consider.

This is Part 3 of a four-part series. See also Part 1, Part 2, Part 4. View PDF of entire series.

Comments

  1. Placebo or Historical Controls: Mathematical Model Offers a Better Choice
    I read this informative series about the expert conference organized by EMA with interest. Among the issues discussed, this meeting touched on the question of whether studies with potential therapeutic agents in carriers of autosomal-dominant AD mutations could serve as a model for simplified clinical testing of new medications against pre-symptomatic stages of the more common sporadic forms of AD. My comment pertains to Part 3 of this series, which addresses the topic of placebo-controlled studies as part of clinical trials of new medications. I notice that this discussion, which comprised primarily questions by participants and answers by regulators, failed to distinguish between early and late clinical development phases of new medications. However, this differentiation is essential, because Phases 1, 2, and 3 of clinical development tackle different questions based on quite different knowledge bases about the therapy at hand.

    My colleagues and I take the view that long-term use of placebo in studies of new compounds that are in advanced development, typically Phase 3, is highly ethically problematic. The Declaration of Helsinki (Seoul, 2008, paragraph 32) allows use of placebo instead of the best current proven intervention in situations “where no current proven intervention exists; or where for compelling and scientifically sound methodological reasons the use of placebo is necessary to determine the efficacy or safety of an intervention and the patients who receive placebo or no treatment will not be subject to any risk of serious or irreversible harm. Extreme care must be taken to avoid abuse of this option.“

    With regard to long-term treatment of pre-symptomatic stages of all forms of AD, the first condition (i.e., lack of current proven treatments) is met; however, this is not true for the second condition. Without doubt, patients with autosomal-dominant AD face a high risk of serious or irreversible harm if treated with placebo for extended periods of time (18 months or longer, depending on the mechanism of action of the intervention studied). The same applies to patients with amnestic MCI who show specific neuropsychological, neurochemical, or brain imaging markers of pre-symptomatic AD. In the spirit of the Declaration, it is therefore unacceptable to knowingly expose such trial participants for extended periods of time to an ineffective intervention (placebo) when studying a treatment that in prior clinical trials has shown clear potential for clinically relevant efficacy and acceptable tolerability. These latter elements are customarily established in Phase 1 and early Phase 2 trials, which are shorter in duration and do require placebo control. For pre-registration trials, so-called pivotal studies, however, the scientific community and regulatory authorities would be well advised to consider novel study designs. Such designs should combine maximal protection of patients against ineffective treatments with the ability to foster scientifically valid testing of new candidate drugs.

    In an attempt to realize an alternative to placebo-based study designs that is both ethically and scientifically sound, our group has developed mathematical models to reliably forecast clinically relevant endpoints and disease trajectories of AD patients in pre-symptomatic stages (MCI subjects). Our models make use of medical history, biological, and neuropsychological measures that are routinely established at baseline of every therapeutic study. Model-based forecasted endpoints and trajectories of decline constitute the background—the “simulated placebo group”—against which potential drug effects can be contrasted. We call this alternative study design the Placebo Group Simulation Approach (PGSA). Our ideas about it were presented at the recent CTAD congress in Toulouse (Poster 25, Abstract in J Nutrition Health & Aging 14: S16, 2010; we will gladly send a PDF file of the poster to colleagues interested in our concept).

    First results using data from the ADNI database clearly demonstrate that empirically established and mathematically modelled endpoints and disease trajectories show high concordance in large samples of pre-symptomatic AD patients. We are currently attempting to validate our models using several independent datasets. Based on the first encouraging findings, we hope that the PGSA will replace placebo-controlled long-term studies in advanced stages of development of new anti-AD drugs. Needless to say, trials offering treatment with active drug to all participants are also easier to perform and less costly.

    (Previously at Novartis, the author was a project leader in the development of the cholinesterase inhibitor drug rivastigmine; see, e.g., Spiegel, 2002 and Almqvist et al., 2004.)

    References:

    . [Rivastigmine: a review of its clinical effectiveness]. Rev Neurol. 2002 Nov 1-15;35(9):859-69. PubMed.

    . Preserved cognitive function after 12 months of treatment with rivastigmine in mild Alzheimer's disease in comparison with untreated AD and MCI patients. Eur J Neurol. 2004 Apr;11(4):253-61. PubMed.

Make a Comment

To make a comment you must login or register.

References

News Citations

  1. London: Families Talk About Treatment Trials
  2. London: Europe-U.S. Regulators Mull Prevention Trial in Familial AD
  3. London: What, No Argument? Speakers Agree on Trials for Familial AD

Other Citations

  1. View PDF of entire series.

Further Reading