On April 1 in Washington, D.C., a meeting convened by the FTD Treatment Study Group offered an assembly of academic and industry scientists, funders, representatives of patient organizations, and caregivers the rare opportunity to collectively pick the brains of senior scientists from the Food and Drug Administration and the European Medicines Agency. How should we develop therapies for the rare and heterogeneous set of diseases that make up frontotemporal dementia, they wanted to know?
As no drugs are approved to treat FTD, none have undergone a formal regulatory process. No framework exists to guide the various drug development programs that are sprouting up across the United States and Europe. The EMA in January 2016 invited public comment on a draft guideline for drug development in Alzheimer’s disease and other dementias. However, it does not clarify concrete questions such as what role fluid or imaging markers can play as supportive or primary evidence for the registration of drugs developed in small, heterogeneous populations of patients. The FDA in 2015 issued a draft guidance on rare diseases, but it does not address FTD specifically at all. This absence of a status quo is both a challenge and an opportunity, said Philipp von Rosenstiel of Biogen, Cambridge, Massachusetts. The conversation in D.C. helped scientists gauge—and try to influence—how regulators currently think about these devastating forms of dementia.
In previous discussions, scientists had laid out how FTD differed from Alzheimer’s, for example by its bewildering lack of correspondence between an underlying pathophysiology and its phenotypic expression (see Part 1 of this series). They established that FTD is distinct from AD in its biomarker profile, and that much more work is needed there (Part 2). In this discussion, the scientists peppered regulators with questions to get a feel for how their own preclinical or clinical programs might be perceived. Many drug sponsors meet with regulators privately for advice based on data on a given drug in development, but in group conversations such as this, mentioning specific investigational drugs is largely taboo. Instead, the dynamic tends to be one where regulators answer questions with general principles, generating nuanced “it depends” answers that don’t pin down the regulator but nonetheless offer valuable insight into how regulatory thinking evolves along with advances in the field.
Maria Isaac of the EMA, with Billy Dunn and Nick Kozauer of the FDA, urged FTD researchers to absorb the lessons of failed trials in AD. Above all, develop biomarkers, they said. “You will find us very receptive to their use,” Kozauer said. Apply biomarkers to characterize the natural history of FTDs. Use them to ascertain that each participant has the underlying etiology the drug is targeting. Regulators recommended that scientists develop creative outcome measures, such as individualized ones, but urged the field to stay laser-focused on showing that a drug effect is meaningful to the patient.
If this sounds like a high bar, the regulators also emphasized that they can be flexible and that they want to see therapies succeed as much as everyone in the room. “Some of us have treated these patients as clinicians. We are very aware these are horrible diseases,” Kozauer said. “Together, we have a fantastic opportunity here. We have talked internally about the important need in FTD and are enthusiastic about greater attention to this area. By learning from AD, we should be able to bring greater efficiency to the process,” Dunn said. Most of all, regulators emphasized the need for individual players to avoid working in silos. “Share resources and data. It gets us there faster,” Dunn said.
Sometimes academic or industry scientists blame regulators for slowing innovation in drug development, but Dunn disputed this notion. “There is no disconnect between academics, industry, and regulators. We are all on the same team. We each play different roles, but we are here today to push for robust progress toward this goal,” Dunn said. Likewise, Isaac told the group,” I am happy to discuss your programs with you.”
Besides Biogen, biopharma companies that dispatched researchers to this meeting include A&G Pharmaceutical, Alector, Asceneuron, Bristol-Myers Squibb, Cogentis, Cydan, Forum, Genentech, Janssen, Lilly, Merck, Novartis, Qanterix, TauRx, as well as several contract research organizations. For paraphrased excerpts of the main discussion points, see Q&A below.
Q: What is the single thing that would propel drug evaluation in FTD the fastest?
A: Finding biomarkers that allow you to positively demonstrate the presence of the underlying hallmark lesions in vivo, and to discriminate between the etiological types of FTD. You need tau biomarkers, TDP-43 biomarkers, FUS biomarkers.
Q: How do you evaluate drug risk in a fatal neurodegenerative disease where there is no therapy?
A: We place those conditions into a special category. We understand the tolerance for risk is very high. That said, we need to be sure that we can contextualize and describe for the patients in a responsible way what the risks and benefits are. It’s tough to give a bright line because we consider this on a case-by-case basis. We need to see the evidence but assure you that we think hard about this.
Q: Do you think differently about risk tolerance in Phase 1 than registration?
A: Yes. In an approval decision, where we have efficacy data and can consider risk versus benefit, we can be flexible on how many people we need to see evidence on. According to the International Council of Harmonisation, it can be as few as 100 patients exposed for a year. Earlier in development, we are primarily concerned about risk because you know very little about the drug’s benefit. We can be somewhat flexible in how many exposures, and how long, you need in order to go forward. But we have had cases where a drug seemed safe early on and with more exposures bad things did happen.
Q: Long toxicology studies really slow things down. Meanwhile people are suffering.
A: We understand you want to find a streamlined approach. We have specific experience where, in an effort to accelerate, we allowed shorter-than-normal toxicology studies to support long-term human studies. That can work, but we have also had situations where toxicity emerged in longer-term animal studies after the drug was already in humans, and was so severe that the clinical program had to be stopped. We are eager to bring drugs forward but are aware of both sides of the issue.
Q: Does previous human exposure count?
A: Yes. It is extremely important. If adequate, that data can be used to support clinical studies in the absence of certain pre-clinical studies.
Q: How do you balance an educated, informed population against a paternalistic approach? How important is the patient’s knowledge and decision?
A: We favor an informative approach. We have heard from this patient community that they want information and want to able to make a decision. We do not say this is right and wrong for you.
Q: To illustrate the desperation we are facing: Those of us who have done trials in FTD hear both patients and families say it would be better if the person died more quickly rather than slowly. They cannot deal with what the disease does to them anymore. They say if a drug has serious side effects for the patient, that could not possibly be worse than what they are living already.
A: We understand and have great hope for the trials being started in these diseases.
Q: What endpoints could support approval? Is it necessary to demonstrate an effect on a functional or global outcome measure, or could a composite scale suffice?
A: It could. The bottom line is you have to do something meaningful for the patient—meaningful both in terms of the content, i.e., what the scale assesses, and in terms of how it assesses it. In FTD we know so little that we will not be prescriptive about recommended endpoints. Remember that even when we have considerable experience with an endpoint, the FDA or EMA remain open to proposals for other endpoints.
Q: Please elaborate.
A: Our openness about endpoints is true for many diseases, even the ones where we know much more. In those diseases we may ask you to include the assessment that we know to be the innovative one, but here we do not even have that. As you choose endpoints, our advice is to keep it simple. We really just want to see something that is meaningful for the patient. That is what we are after. We are very open to how you do that.
Q: What makes a scale meaningful?
A: Some confusion may stem from the limited experience with using composite scales in predementia trials as a single endpoint. We discuss this in our draft guidance for early Alzheimer’s. For example, the CDR is a composite in a sense. Its appeal is that as a clinician-rated scale it assesses cognitive domains through the lens of how a person functions. It is not just neurometrics out of context. If you measure neuropsychology without life context, then you are in trouble. In a trial you need to assess how someone is functioning. Do not cobble together a composite with some pure neuropsychology elements and some questions about daily activities. Build a composite where each individual element by itself is clinically meaningful.
Q: Is the Progressive Supranuclear Palsy Rating Scale (PSP-RS) sufficient as a global outcome measure for PSP?
A: A general caveat here is that some of these scales may be valid for following clinical progression but may not be ideal for use an as endpoint in clinical trials. We see promise in the PSP-RS but have very little experience with it. It may need to be adapted somewhat to serve as a trial outcome measure. It could be pared down to its more meaningful parts. We are happy to discuss the specifics of measures that sponsors propose.
Q: Do we need a co-primary? For example, would language be sufficient in aphasia?
A: If you assess an important cognitive domain like language in a way that shows the drug improves the person’s ability to function, then that is sufficient. But your language measure has to be meaningful.
Q: How about wearable devices to record outcomes?
A: We are very open to that. You will find us more flexible than you think. The clinical meaningfulness of what the device measures is where you will get scrutiny from us.
Q: We have been working on patient-reported outcomes for corticobasal degeneration. Many CBD patients have poor insight. They cannot communicate outcomes or improvement, so it is the caregivers who can tell.
A: We encounter that all the time. We understand that in the early stage, some people retain insight but the variability of this can cause problems in a trial. For example, assume the drug worked. Patients who take the drug may retain insight into their deficits while patients on placebo may lose insight and incorrectly perceive that they are doing fine. In that situation it might appear that patients taking the drug are doing worse than the patient taking placebo.
Q: And this means?
A: That despite our fervent belief in the patient’s voice, cognitive impairment can affect the reliability of that voice. We are engaged with the PRO consortium on this topic.
Q: How do you view volumetric MRI as an outcome marker versus cognition or a clinical measure?
A: As potential supportive data.
Q: Could accelerated approval based on a biomarker related to the disease mechanism be applied under subpart H? For example, one genetic cause of FTLD is haploinsufficiency of the progranulin gene, which causes low progranulin protein levels in brain. Could approval be pursued on the basis of restoration of CNS progranulin levels combined with a reduction for a fluid biomarker of disease activity, e.g., neurofilament light chain?
A: It is very tempting to jump in on this. At this point we do not understand the pathophysiology of progranulin and NFL well enough to base approval on them. Biomarkers will be incredibly valuable here, it’s just that we need to learn more first. Consider amyloid as a cautionary tale. There was interest in using amyloid as a surrogate for approval. It seemed so logical at the time. But we needed to learn more and based on what we know now, the relationship between amyloid and clinical benefit is not clear.
Q: What sort of knowledge would make you feel more comfortable? What can we do short of 10 years of study?
A: Understand the pathophysiology of the biomarker, and link a clinical benefit to change in that marker.
Q: Our community is excited about NFL because many labs have similar findings, and NFL makes sense in terms of neurodegeneration. What more do you need to see on it?
A: We need to have data to deduce that reducing NFL is reasonably likely to predict meaningful benefit. We can’t provide a specific answer without more data, but it seems based on the breadth of our experience that additional work should be done. In general, do not think that accelerated approval implies a lower standard. The evidence of an effect on a biomarker outcome that is reasonably likely to predict a clinical benefit must be robust. If you can obtain a clinical benefit in a well-designed study, that is what you should do. It does not help to invoke accelerated approval just because. True surrogacy is very hard to achieve.
Q: Let’s say we have data to show low progranulin causes disease, and that NFL levels rise with disease. If we then see with a drug an increase in progranulin and a reduction of NFL, I’d say that’s great. Let’s see what happens. If not subpart H, can we then invoke expanded access?
A: We are willing to consider proposals for expanded access. Or in Europe you can go for conditional approval.
Q: What if the primary endpoint missed statistical significance but goes in the right direction, and NFL goes in the right direction, and FDG PET goes in the right direction?
A: You talk about consistency of effect on different biomarkers that are all in the pathway of disease. This can help a study as part of the totality of data.
Q: Do genetically defined labels extend to all clinical phenotypes of a given mutation? In other words, if we define a label based on the presence of pathogenic C9ORF72 expansion, would the label extend from C9ORF72 ALS to C9ORF72 ALS-FTD and C9ORF72 FTD?
A: We try to generalize the results as far as is scientifically reasonable, but you need to show that the mutation is pathogenic for each phenotype in question.
Q: Is the answer the same for pathology? For example, if you have a drug that treats, say, a tauopathy?
A: It depends on what you know about the conditions. Potentially yes, but it is complicated and you need to consider whether the tauopathy in the condition you want to extend the label to accounts for as much of the clinical disease as it does in the original condition and that treating the tauopathy is beneficial.
Q: With the different mutations, pathologies, and phenotypes, the number of available patients for any kind of homogeneous patient group is small. How do you deal with that?
A: There is no minimum number. We can’t get too specific as the information and your outreach are evolving, but we understand you can only do what you can do. Just know that a limited pool of patients is not prohibitive for drug development. We have approved drugs for small populations.
Q: Can a PSP clinical trial performed in an adaptive platform study serve as a registration trial intended to provide primary efficacy evidence to support approval?
A: With the important caveat that the devil is in the details, the answer is yes. Primary efficacy data can be generated from such an approach.
Q: How about the pooled placebo used in such trial platforms?
A: There are details to consider, and we all know placebo groups can go bad in lots of ways. But yes, it can be acceptable. There are efforts underway that use pooled placebo. You do want high throughput and to streamline trials. We encourage that. If you are concerned that you can only get exploratory evidence with that, that is not the case.
Q: In PSP, can a single clinical trial be sufficient for approval?
A: Since the question is “can,” the short answer is “yes.” Typically, the evidence comes from two studies. But if a single trial is large, shows a big effect, survives multiple sensitivity analyses, hits multiple domains etc., then it can be sufficient. We have guidance on that.
Q: Can we prospectively use domains that are abnormal at baseline in individual subjects, as opposed to a broad spectrum scale for everyone?
A: Yes. We are open to patient-specific outcomes. If you individualize endpoints and specify that in the trial, then that could be a useful approach. It is intrinsically attractive as you want to treat the thing that is bothering that patient the most. We hear back from sponsors that patient-specific outcomes can be tricky to operationalize. You’ll have to work out a lot of details about the statistical analysis plan and other aspects. Ask us and we will advise.
Q: With this approach, do you still also want a unifying global outcome that is applicable across all these subtypes? The spectrum of symptoms is so wide.
A: That depends on the case. Not necessarily. One potential way to deal with the heterogeneity is to develop a disease-independent measure that captures the totality of how a patient is doing.
Q: What preclinical data will the FDA want to see for FTD caused by progranulin mutations? There are no good disease models yet.
A: We often get presentations of drug efficacy in animal models. Be aware that that may not be a large regulatory consideration in drug development. From our standpoint, animal data are most often used to inform safety considerations. Animal efficacy is often important to sponsors to make decisions about moving forward.
Q: In an ideal world, what evidence would you want?
A: You would show us this: “We caught these people bound to develop this disease early. Here is how they are already changing preclinically, here is what is going to happen to them in roughly this amount of time, here is how our therapy has affected these disease-related biomarkers, and here is why that is reasonably likely to delay people’s progression.” Bring us these data. We are in the same boat, and don’t want to be too late for these patients.
Q: A final word?
A: Do good science. We will advise along the way.