This concludes a two-part story. See also Part 1.
6 August 2008. Companies have dealt with the uncertainty created by the Phase 2 problem in clinical trial design (see accompanying story) in several ways. They either looked for hints of clinical efficacy in Phase 2 data, or used biomarkers, or combined both. In his plenary lecture on the last morning of the ICAD conference last week in Chicago, Paul Aisen of UCSD cited two recent examples, Alzhemed and Flurizan, to illustrate how well this has worked. Both drugs sank in Phase 3. Neurochem ran a short, small Phase 2 and used CSF Aβ and tau as a biomarker to give them some confidence; Myriad ran a longer, larger Phase 2 without biomarkers. The Alzhemed trial showed a change in the CSF biomarker; the Flurizan trial was negative but later during the follow-on period showed a separation, though there was no placebo comparison at that point anymore.
Two current high-profile immunotherapies, i.e., the passive vaccines by Elan/Wyeth and Eli Lilly and Co., similarly tried to glean meaning from Phase 2 trials. Elan reported efficacy across cognitive measures in a subgroup of patients, as well as a signal on CSF biomarkers. This company took a look-see at this data during the still blinded Phase 2 trial to decide whether to go ahead with Phase 3. Lilly took a perhaps even riskier leap, deciding to move into Phase 3 based on an extensive biomarker analysis but without any hint of a clinical benefit in Phase 2 (see subsequent ICAD stories on these experimental therapies).
In the context of the recent failures of Alzhemed and Flurizan, Aisen cited a recent Newsweek column that came up frequently in conversations with scientists at ICAD. Aisen took exception to the charge that a narrow-minded bias in favor of the amyloid hypothesis blinkers the field. Presentations at ICAD in Chicago included results of clinical trials of two tau-based drugs, of lithium, of antioxidants, of curcumin, of a nicotinic agonist, of hormones, of gingko biloba, insulin, omega-3 fatty acids and various nutraceutical formulations. No home runs yet (though see details on trial results in subsequent story). Clinical trials of anti-inflammatory compounds to date have failed, and trials of statins and other compounds are ongoing. It’s not for lack of interest in alternatives, Aisen insisted. The two reasons behind recent high-profile disappointments were that the compounds were weak and that deep methodological problems need to be solved.
Toward this goal, Aisen reiterated a growing tide of consensus at the ICAD meeting, namely that future trials must enroll patients who are at an earlier stage in the neurobiological progression of AD. This requires changes in early detection and in the regulatory pathway to approval (see also ARF Live Discussion on early detection).
In practice, early detection means biomarkers. Several hundred presentations at ICAD dealt with various aspects of this growing area. The main candidate markers are plasma Aβ, CSF Aβ, tau, oxidative or inflammatory markers, and various modalities of brain imaging. Of these, CSF Aβ/tau, as well as structural, amyloid, and FDG-PET imaging appear to be the most robust at present for early detection, Aisen said. This impression is based not only on individual studies, but also on preliminary data from the Alzheimer’s Disease Neuroimaging Study (ADNI), which compares these markers in the same set of people. ADNI allows scientists to calculate how large hypothetical clinical trials must be to detect a treatment effect when using these markers to measure outcome. In a separate presentation at ICAD, Michael Donohue of University of California, San Diego, presented some of these power calculations. The upshot is that if these measures were incorporated into trials of people with MCI as baseline covariates, studies could be at least a third smaller. This would enable the community to conduct more trials on a greater variety of compounds and would place less pressure on the success of any given trial.
Using these markers to detect people prior to dementia and to power trials accordingly, will become doable within the next two years, Aisen said. Also possible will be efforts to exclude people with other forms of dementia from trials. Thinking bigger, however, all this is but a step toward the larger goal of developing a surrogate marker, Aisen said. Biology dictates the search for such a higher-level marker because the pathology begins years to decades before people become demented. This means primary prevention trials would take too long to detect a cognitive or clinical effect. Trials cannot follow people for 15 or 20 years.
“We need to develop biomarkers as surrogate outcome measures. This is the future of the development of disease-modifying drugs in AD,” said Aisen.
How can the field anoint a marker to this vaunted status? The steps are clear, Aisen said. Clinical research must establish a close relationship between the biomarker and the disease neurobiology and progression. This work is well underway. Most importantly, the field has to demonstrate that the effect of a drug on the biomarker correlates with cognitive and clinical benefit. This must be done in more than one trial, with more than one agent, i.e., with a secretase inhibitor and an immunotherapy. “We need to work harder to strengthen the link between biomarker and outcome to elevate them to the status of surrogate marker. Unless we do that, the case for a surrogate marker is unacceptable to the FDA,” Aisen said.
On this goal, too, the leading candidates at present include the CSF Aβ42 and tau/phospho-tau markers, brain atrophy and amyloid imaging, even certain neuropsychology tests, Aisen said. (It is important to remember that one biomarker may not serve all purposes—markers for asymptomatic prediction, progression, a drug hitting its target must all be individually defined and may each turn out to be different ones, scientists said.) As a next step, prediction and progression markers must be established in asymptomatic or mildly symptomatic patients.
Aisen left the field with a vision for how to test drugs in a brain that still has a fighting chance to stem disease, before it has suffered extensive damage. That is to screen people in mid- and late life for abnormalities in this future surrogate marker. This would detect preclinical AD neurobiology. Then treat to reduce the person’s risk for AD. In essence it’s to treat the surrogate marker in a person who could be said to have “surrogate AD,” or, to quote a term by Jeffrey Cummings from UCLA, treat “AD without dementia.”
This is in the future, Aisen said, but not a distant one. It is clearly within reach. ADNI, as well as some individual institutional research programs, such as that at Washington University, St. Louis, are encouraging in that their data are beginning to link both neuroimaging and CSF biomarkers to a risk of decline in a subset of people with normal cognition at baseline. This is the research front at this point, Aisen said. The field must continue to build the case linking a potential surrogate to AD diagnosis and progression in mild AD, show that a drug effect on the potential surrogate is linked with clinical effect, then establish a link between the potential surrogate in asymptomatic people and later disease—and all along the way, gradually persuade regulators that this data warrants updating the agency’s guidelines.
This goal requires sustained cooperation between otherwise competing groups in academia and industry. Several efforts are underway, or just gearing up, to bring people together around shared goals in so-called pre-competitive settings. Aside from ADNI, these include the Dominantly Inherited Alzheimer’s Disease Network (aka DIAN, see subsequent story), as well as the Alzheimer’s Study Group (ASG)), the Alzheimer’s Association Research Roundtables. Advocacy groups such as ACT-AD help, as well. Finally, Aisen said, a European initiative has helped to push AD diagnosis, backed with amyloid biomarkers, toward a very mild symptomatic stage (Dubois et al., 2007; Gauthier et al., 2008). This is an improvement over the standard NINDS-ADRDA criteria of possible and probable AD (McKhann et al., 1984), which are viewed as outdated yet remain in wide use. AD starts insidiously with a presymptomatic phase, and that is where detection ultimately must go to find it.—Gabrielle Strobel.
This concludes a two-part story. See also Part 1.