CTAD saw extensive discussion of cognitive outcome measures. Besides being brief, such tests should reflect the expected mechanism of action of the drug, Schneider said. He cautioned against using composites in which some tests have practice effects, as those may hide a drug response indicated by other tests in the mix. In fact, Keith Wesnes, of United BioSource Corporation, Goring-on-Thames, U.K., reported at CTAD that half of the cognitive tests used in ADNI showed practice effects. The list includes the Boston Naming and Trail Making tests, and others. ADNI-1 controls improved over six years’ time on those tests, even though prior research has established that cognitive function wanes with age; hence, the controls should have declined. In essence, Wesnes said that tests with practice effects show a longitudinal element, i.e., worse performance with age, when given once to people of different ages. When given repeatedly, as in ADNI, the practice effect can mask age-related decline.
The effect size of this "training" was in the range of that seen with acetylcholinesterase inhibitor drugs, hence, may swallow up small drug effects in clinical trials. The training effect calls the value of these tests in clinical trials into question, Wesnes claimed. “These tests are useful for cross-sectional use in longitudinal natural history studies; they are not suitable for drug studies in preclinical AD,” he said. Practice effects have been shown before (e.g., Wilson et al., 2002).
What about the most commonly used cognitive battery, the ADAS-cog? Schneider called it a "utility infielder," that is, a perfectly respectable, albeit overplayed, generalist. “It is used automatically whenever we test a drug. Whether Phase 2, Phase 3, no matter what the drug is,” Schneider said. Expectations have slid from a three-point ADAS-cog improvement in six months with the symptomatic drugs of the 1990s to “we’d be happy if we saw any improvement over 18 months,” Schneider said. “We should not continue to rely on it to advance drugs from Phase 2 to 3.”
On a separate note, Schneider advised against enriching Phase 3 trials for ApoE4 subsequent to analyzing Phase 2 trials by ApoE genotype. Rosiglitazone and bapineuzumab are two examples where analysis of Phase 2 data indicated a different drug response in ApoE4 carriers than non-carriers, but subsequent pharmacogenomic Phase 3 programs failed. Studies have shown that ApoE4 and CSF Aβ or brain amyloid are closely linked; glucose metabolism and connectivity change earlier in carriers than non-carriers as well. In ADNI, 89 percent of people with MCI due to AD who carried an ApoE4 allele had a positive CSF Aβ biomarker. Schneider’s argument against testing drugs separately in clinical trial samples enriched by ApoE4 status is that carriers essentially have more advanced disease than non-carriers, not necessarily a qualitatively different type of AD that would predict a different drug response.
To study enrichment by ApoE status, Schneider, with Richard Kennedy and Gary Cutter at the University of Alabama, Birmingham, simulated running trials in various samples ranging from all ApoE4 carriers to no ApoE4 carriers in a database of placebo data from 18 pooled ADCS trials. They found no differences in the trials’ efficiencies. This leads to the contrarian conclusion that selecting prodromal AD patients for a trial based on CSF Aβ42 or ApoE4 status may identify more advanced cases of MCI, but not necessarily enhance statistical power. As trials are adding genetics or evidence of amyloid to their inclusion criteria, it might be preferable at this point to use these markers as an explanatory variable, not to stratify by ApoE status or to exclude people based on their CSF Aβ42, Schneider suggested. These findings are consistent with van Dyck’s report that only 19 percent of potential MCI patients are severe enough, with both memory impairment and biomarker positivity, to be included in prodromal AD trials, Schneider wrote to Alzforum.
Many trials enroll a wide age range, typically from 55 to 90 years. This practice may be democratic but hazardous to the outcome, according to Dominic Holland of the University of California, San Diego. Holland studied how age affects disease progression in ADNI. He found that cognitive decline and atrophy tend to slow with advanced age in people who have MCI or AD, while, on the contrary, this decline speeds up with age in healthy controls. This implies that it is difficult to distinguish AD from normal aging in the oldest old, and that including very old people in a trial reduces statistical power. Younger cohorts with a tighter age range might better enable trials to measure drug effects, whereas studies in older people could help confirm safety and tolerability of a therapy, Holland said (Holland et al., 2012). Alas, a large proportion of dementia prevalence occurs in people older than 80 (Brayne and Davis, 2012).—Gabrielle Strobel.
This is Part 1 of a seven-part series. See also Part 2, Part 3, Part 4, Part 5, Part 6, and Part 7. Read a PDF of the entire series.