3 March 2010. Exactly how to correctly fashion a pre-symptomatic trial appears to be in the eye of the beholder. This was the upshot of trial design discussions at an industry advisory meeting hosted last month by Eric Reiman, Pierre Tariot, and their colleagues from the Banner Alzheimer’s Institute in Phoenix, Arizona, on their proposed Alzheimer’s Prevention Initiative (API). The API scientists fielded ample advice, including this: “You will get a lot of informed opinions but no consensus on any details. You’ll have to make your decision and go with it.” Here’s a short list of questions that came up repeatedly:

• Enroll people regardless of their brain amyloid load, or only people who have amyloid? Each approach had its proponents. The API leaders at present are leaning toward including all who meet a combined genotype/age requirement and then analyzing the trial outcomes relative to how much amyloid deposition they had.
• What to designate as a primary endpoint? One or more biomarkers, followed by cognitive/clinical secondary endpoints? Or vice versa?
• Should the trial analyze biomarkers and clinical outcomes together or separately? Each statistic has regulatory consequences.
• What would be the right clinical measure to determine if the drug has worked in people who were normal at the outset? This is the inherent conundrum of how one can prove prevention.
• What are these trials? Phase 2? Phase 3?
• Should API run pilot drug-biomarker studies and then hand the drug back to the sponsor for further development, or should the API reach for full-fledged, pivotal FDA registration trials?
• How do the API trials fit into a company’s overall development plan for its drug?

This last bundle of regulatory issues generated lively debate. Toward the end of the day, it appeared to settle around an Accelerated Approval pathway outlined by Jur Strobos, a regulatory consultant and former FDA official. This approach allows a drug onto the market for certain disorders such as AD based on non-clinical endpoints, on the condition that future data confirm that this endpoint in fact predicts clinical benefit. Not unlike a learner’s permit, in other words. The sponsor needs to submit a pre-specified statistical analysis plan. The actual longitudinal demonstration of the clinical outcome can come from the ongoing pivotal study, or from a parallel study that goes on while the drug is already on the market. If this last step fails in Phase 4, the sponsor must withdraw the drug.

For this to work, the mechanism of the drug and the direction in which it moves its attendant biomarkers must fit together in one consistent hypothesis, Strobos said. Moreover, the FDA tends to place value on quality of life measures that ask the study participants, rather than their physicians, whether the investigated drug helped them, Strobos added. This could be done with meta-memory tests, where people judge subjectively how well they remember things, or with informant-based interviews. One thing not to do: retrospective subset analysis. The FDA frowns on that, Strobos said. According to this pathway, the API researchers could chose a primary pre-specified imaging marker as an interim endpoint and use that for an approval application (see FDA page on accelerated approval).

Beyond that, there was a strong sense in the room that API should focus primarily on generating exciting science, not get overly concerned with obtaining regulatory support from the get-go. In other words, once data show a drug to have moved biomarkers in pre-symptomatic folks, regulators will be interested to find a path to approval. “If we can show that a given agent is effective in a prevention paradigm, then I am sure the regulatory agency can work with us,” one industry scientist said.

Some trial design issues stood out for the broad consensus they drew. First among them, industry scientists said that in order to justify API trials internally, they would prefer a shorter design that delivers at least some interim data after two years. “We need some early successes to keep enthusiasm and funding going,” one pharma representative said. The current trial design of five years, plus the time it takes to recruit and follow up, is too long, they said, if there is not a significant chance to learn something midway through. Drug sponsors do not want to treat for so long with a drug that might not work. Even the Alzheimer's Disease Neuroimaging Initiative (ADNI)—a simpler proposition to begin with—offered short-term data to look at, which kept industry support going. Some attendees worried about the prospect of a fuzzy outcome at the end, where there is a hint but not proof of clinical efficacy. What to do then? Keep treating with study drug for another five years in the hope that the separation between placebo and study drug will be clearer? In this situation, a strong primary biomarker that closely reflects the medical hypothesis of the treatment can generate some confidence, these advisers said.

Attendees around the room voiced dissatisfaction with the standard parallel arm trial designs and the standard primary/secondary outcomes. Calling these “archaic,” Lon Schneider from the University of Southern California in Los Angeles argued for more exploration of creative alternatives such as rolling starts and multivariate outcomes. These could even be customized to individual participants if they had undergone a period of intense observation prior to drug treatment, perhaps in a registry that uses social networking media, during which their biomarker levels were established or had even shown a trajectory of change. There’s more than one advantage to that approach: “If you get trajectory of change on individuals, then the recruitment for a treatment trial will be a nanosecond,” quipped Schneider (see expanded comment below). The Dominantly Inherited Alzheimer Network (DIAN) is taking this approach of natural history preceding drug treatment. It could be broadened to other populations if large registries of cognitively normal study volunteers were built from which companies could then recruit participants for observational, dose-finding, or target engagement studies prior to moving on to formal treatment trials. This goal drew unanimous support.

Furthermore, industry scientists urged the API researchers to bolster the scientific data on which they base their conversion rates and subsequent power calculations for the ApoE trial. Frequently, the power seen in observational studies does not map onto subsequent trials, and trial planning needs to account for that. In terms of cognition, the API leaders heard advice to fine-tune their tests of whether trial participants were developing memory problems during the course of the trial to the drug at hand: i.e., use tests that stress hippocampal function if it’s an anti-tau drug, or tests that probe attention if it’s an anti-amyloid drug.

Finally, this point resonated throughout the day: Whatever specific design the API leaders settle on for their initial trials, science will learn a huge amount, and these trials will galvanize interest in prevention studies throughout the field. Pharma representatives from Europe asked to be included in the ApoE4 initiative as well.

Reiman concluded the day by saying: “Both the stakes and opportunity at hand could not be greater. If it's a choice between an incremental approach to this problem or a bold one, we will get further if we are bold. Evaluating drugs in the highest-risk populations is not only good for companies, but also good for the world, and we will move this process forward while trying to address each of the concerns and issues raised here today.”—Gabrielle Strobel.

This concludes a five-part series. See also Part 1, Part 2, Part 3, Part 4. See also a PDF of the entire series.