Cohort effects in mild to moderate AD trials: an empirical assessment utilizing data from the Alzheimer’s disease cooperative study

Abstract

AbstractBackgroundHeterogeneity in patient populations can impact clinical trial conclusions and generalizability. Long recruitment windows may exacerbate this in Alzheimer’s disease (AD) trials, which often recruit over extended periods.MethodWe used data from 8 Alzheimer’s Disease Cooperative Study (ADCS) clinical trials (DHA, Homocysteine, Huperzine, Intravenous Immunoglobulin, Resveratrol, Valproate, NSAIDs, and FYN Kinase inhibitor) in patients with mild to moderate AD to investigate cohort effects by time‐of‐trial‐enrollment.For each participant, we assigned a study‐ and site‐specific enrollment time defined as the number of days between a patient’s screening and the first screening visit at the site. We used study‐ and site‐specific time origins to account for calendar time differences between studies and, within a study, differences in site times to activation.To investigate differences in baseline characteristics, we split our sample into tertiles by enrollment time and compared sex, race and ethnicity, age, years of education, and baseline scores on the Alzheimer’s Disease Assessment Scale ‐ Cognitive Subscale (ADAS‐Cog).To assess cohort effects on trial outcomes, we used the 12‐month within‐subject change in ADAS‐Cog. We regressed the 12‐month change in ADAS‐Cog on a patient’s enrollment time while controlling for their baseline score and demographic information and allowing for clustering between patients in the same study.ResultGroups were demographically similar at baseline across enrollment tertiles. We found a significant difference in 12‐month change in ADAS‐Cog for patients with different enrollment times: later enrolled patients had an expected 12‐month change in ADAS‐Cog 0.146 points higher than similar patients who enrolled one quarter (91.2 days) earlier (95% CI: 0.082‐0.210). This magnitude of difference was roughly on par with an additional year of education. We also found a negative relationship with age, with older patients tending to experience a smaller 12‐month decline.ConclusionIn these 8 trials, we found minimal evidence of cohort effects that would meaningfully change study generalizability or treatment effect. Future analyses will account for the overall pace of accrual in the trials and assess for potential differential effects in fast vs. slow‐accruing studies.