Late last month in Phoenix, a roomful of senior industry and academic scientists, and regulatory and statistical advisers, dug deep into the details of a proposed pre-symptomatic trials initiative put forward by Eric Reiman, Pierre Tariot, and their colleagues at the Banner Alzheimer’s Institute in Arizona’s state capital (see Parts 1, 2, 3 of this series). After near-unanimous praise for the initiative, the attendees delved into specifics and in the process exposed topics that need more preparation before any trial can actually start. Selecting a drug, setting the right dose, and designing the trial to greatest effect were among the hot button issues, as were questions about how to clear a regulatory path and persuade companies to hand over precious drugs to a public-private study they don’t control. None of these issues appeared grave enough to derail the initiative; rather, they define the current cutting edge of how academia and industry forge new collaborations in order to move toward prevention research. Parts 4 and 5 summarize the tough issues.
Drug Selection: Can We Have Privacy, Please!
If anyone was hoping that representatives from 19 companies would meet in one room, openly discuss the pros and cons of their investigational treatments, and vote on the best candidate for a public-private pre-symptomatic trial, those hopes fizzled quickly when the discussion neared the issue. Pharma scientists who had been discussing general issues about the Alzheimer’s Prevention Initiative (API) with remarkable candor, quickly clammed up.
“I am not allowed to discuss [my company’s] or any other drugs.”
“Can’t have this conversation.”
What’s more, some academic leaders joined in, saying that their various advisory roles and confidentiality agreements with pharma companies precluded them, as well, from discussing investigational drugs in the assembled group. In essence, different people in the room knew different parts of the full truth about any given drug. Pharma people tend to know what is public, confidential to their own company, and purchased from competitive intelligence companies; some academics know only what is published and their own research; other academics may know what is published, their own research, plus confidential information from one or several companies, but not all companies. Taken together, none of these groups is in a position to talk fairly about drugs, and the last has a hard time intellectually managing their conflicts in this regard, Lon Schneider wrote to ARF.
Things took an almost comical turn for a moment when a prospective trial leader insisted, “At some point, the rubber hits the road and we need to discuss the actual drugs. Do we have to exclude the world leaders on these drugs from the discussion?” Awkward silence. Next question: “Can we at least talk about what’s publicly known about these drugs, or do all our discussions have to stay in the abstract?” On that, both industry and academics acknowledged that, in practice, they could not tell apart clearly enough what was public and what wasn’t about a given treatment, therefore would not discuss them at all. “We cannot discuss individual drugs here without getting hopelessly mired in conflict of interest,” said one pharma representative to nods around the room.
And so it was. The entire day passed without a specific treatment name having been uttered by any of the attendees. But the stalemate did not last long. It gave way to consensus that it’s a question not of will but of process—and really not so difficult. In essence, an independent group of academic scientists can solicit advice from companies about what kind of information they need in order to choose the best candidate drug for a trial—deep safety, pharmacokinetic, pharmacodynamic data, performance on specific genetic backgrounds, etc. Then this group can invite companies confidentially to nominate a compound supported by this information. The academic group conducts its own due diligence on the nominations and makes its selection. This has been done before, and partial models, such as the ADCI or the TURNS process in schizophrenia research, can guide the API.
Another difficult question concerned how to preserve precompetitive space as the API proceeds. How to define shared gains once a drug has come into play will require more thought. “This is very different from ADNI,” several scientists cautioned. The day saw some discussion about whether a drug company, which views an investigational treatment that has consumed millions of dollars in development cost as a precious “asset,” would yield control of that asset to a public-private consortium. However, that this issue is shifting became clear when others pointed to prior industry-public agreements on how to share control, for example, those negotiated by ADCS, while some other pharma representatives assured the API scientists that “If you build it, we will come.” A growing number of pharma scientists appeared to take the view that testing their drugs in earlier stages would not put an asset at unacceptable risk, but on the contrary, might in fact save an asset whose performance in clinically ill patients did not live up to expectations. All agreed that pharma scientists have more work to do to communicate these issues to senior management, where decisions about which trials to move forward are made.
Similar public-private arrangements have been negotiated in the past for proprietary drugs in the areas of estrogen replacement therapy, stroke, sickle cell anemia, and HIV prevention, though those arrangements involved primarily marketed drugs. Finally, several industry researchers suggested that positioning participation in API as a benevolent act toward a societal crisis could free industry from some of the usual proprietary restraints in releasing investigational drugs.
Safety: How Much Is Enough?
The question of how much safety data are sufficient for pre-symptomatic trials generated intense discussion. In general, if trial participants are considered healthy yet expected to take a drug for a long time, the safety bar is high, too high for most investigational AD drugs to pass at present. Reiman noted that safety is paramount in drug selection, saying “The ‘go’ decision for us is not whether we have efficacy in clinically affected patients, but whether we have safety.” Typically, drugs tested in prevention trials are already approved by the FDA for other indications, but the group assembled in Phoenix appeared most interested in investigational treatments, primarily but not exclusively anti-amyloid agents. This more ambitious goal raised questions such as these:
- Can Phase 2 safety data possibly be sufficient?
- If the trial is to start in 2011, will more than a handful of candidates even have more than Phase 2 safety?
- Will the FDA require carcinogenicity tests on any chosen drug?
- Regulatory agencies typically require at least 1,500 long-term exposure patients—how many treatments even have that?
- Who is more able to withstand any drug side effects anyway—a cognitively healthy younger person or a person diagnosed with AD?
Some scientists felt that trial participants at highest risk of imminent AD required less stringent safety data than “regular” cognitively normal people. They argued that trial participants be involved in weighing their AD risk against possible drug risks. Some argued that people at risk for AD judge this issue differently than do their doctors and drug sponsors, who act to preclude legal consequences of drug side effects. Yet, there was no consensus on how to include people’s rationalization of their personal risks. For example, Tariot asked, “If you ask 68-year-old ApoE4 homozygote people if they’d accept the risk of vasogenic edema, some will say yes, others will say no. How do we handle that?” Others suggested that the less ambitious approach of enrolling participants whose cognition was already beginning to show signs of trouble would shift the risk equation toward treatment while also increasing the number of participants who will decline on placebo, i.e., make it easier to detect a clinical drug effect. This would not constitute pre-symptomatic treatment; rather, it would reflect the notion of conducting trials at what is sometimes called the early MCI (eMCI), or prodromal, stage of Alzheimer disease. Its downside, some noted, is that treatment may come too late to have its greatest possible effect.
At the end of the day, industry representatives agreed that they need to make haste in securing safety data. As one scientist put it, “We need to do now the appropriate safety and tolerability trials so we will be ready with that data by the time these prevention trials can be done.”
Efficacy: How Little Is Enough?
Must a drug have shown efficacy in clinical trials of mild to moderate AD in order to be chosen for a pre-symptomatic trial? Reflecting recent trends in AD research, most attendees said no, though they added that they’d feel more confident if the drug showed some signal, even a small one, in those trials. Even just for the purpose of setting the right dose, drug sponsors much prefer to have seen some sort of efficacy response to their drug. Everyone agreed on one bedrock requirement, however. To be selected, an API treatment must come with rigorous data showing that it has reached and engaged its intended molecular target, i.e., is biologically active in the brain. This can be done with biomarker readouts, but however it is accomplished, attendees insisted on solid target coverage data as an efficacy minimum if they are to take on the risk of pre-symptomatic research. This reflects, in part, lessons learned from the failure in Phase 3 trial of Flurizan. This drug was widely seen as a test of the amyloid hypothesis, but apparently went into Phase 3 without proof that enough of it reached and engaged its target, γ-secretase, in the human brain (see ARF related news story; ARF related news brief).
Beyond merely showing that the treatment engages its immediate molecular target, biomarker studies should show that it moves some of the most validated AD biomarkers in the expected direction, some scientists urged in Phoenix. One bold suggestion came up repeatedly: to inform the drug selection process, API scientists could request several candidate treatments and compare them side-by-side in short target engagement studies. These could use the biomarkers that most closely reflect the candidates’ underlying medical hypotheses. Even if drug companies balked at direct comparison of their “assets” and ran such brief studies individually on their compounds, they could still generate the necessary data on bioavailability, exposure, and indeed dose, for a larger, longer API trial. Moreover, genetic high-risk populations may have more pathology than people with sporadic AD; hence, doses needed to move biomarkers must be set directly in them.
Dose: Still a Black Box
Industry scientists suggested that a pre-symptomatic drug trial should test two doses versus placebo—provided, that is, that the sample sizes receiving a particular dose are large enough to evaluate the drug with adequate statistical power. How to set those doses in this pre-symptomatic setting is still quite an open question. The API scientists had drafted a study protocol for discussion purposes. Reviewing it, the invited group made clear they would like to see much more information about how to set the dose. Finding a safe and effective dose is especially important if the trials are to take as long as currently proposed, i.e., five years. Yet discussion about dose, indeed about many protocol details, can’t get overly specific until a drug is at hand and its mechanism of action can be taken into consideration, scientists cautioned. Hence, this discussion largely awaits another day.—Gabrielle Strobel.
- Phoenix: Vision of Shared Prevention Trials Lures Pharma to Table
- Phoenix: Can Alzheimer’s Prevention Initiative Break a Catch-22?
- Phoenix: Trials in Colombia and the U.S. for Those at Highest Risk?
- Phoenix: For Shared Prevention Trials, Devil Is in the Details
- Phoenix: Making Trials Work for Patient, Sponsor, Regulator
- Chicago: Flurizan Postmortem
- Paper Alert—Phase 3 Tarenflurbil Data Published