The Impact Treatment Imbalance and Patient Enrichment has on Study Sample Size in Early/Mild Alzheimer’s Disease

Abstract

AbstractBackgroundDisease modification trials in Alzheimer’s Disease (AD) are typically conducted in Early AD patients, combination of patients with mild cognitive impairment (MCI) and mild AD (mAD), despite vastly different progression rates between these two AD groups. A large sample size (SS) is required to show efficacy due to low MCI APRs. Furthermore, treatment group imbalances in patients likely to progress (PLP) may lead to underpowered studies. We quantified the impact on SS of: (1) potential treatment group imbalances in PLPs and (2) PLP enrichment strategies.MethodRoughly 10% of patients diagnosed with MCI progress to dementia within 1 year (Mitchell, 2009). Patients with mAD progress at an APR of 25% (Davis, 2018). To assess the impact on SS, we considered a 1:1 randomized study, stratified by equally sized AD strata (MCI, mAD) with 90% power and 1‐sided 0.025 level Cochran Mantel Haenszel (CMH) test to detect lower APRs in the experimental arm compared to placebo for an odds ratio (OR) = 0.5. To assess the impact of treatment group imbalances in PLPs, we computed SS for various departures from a balanced study. SS calculations were performed after enriching the APR in the MCI or both AD strata.ResultSS increases as more PLPs are assigned to the experimental arm than the control arm. For example, SS is 852 when PLPs are balanced 50%:50% between arms and 932 with a 55%:45% imbalance. When the MCI population is enriched to a 20% APR, SS is 672 for a balanced study and 732 with a 55%:45% imbalance. The following figure displays the impact of a PLP enrichment strategy on SS by increasing APRs by 5% in each AD strata, assuming no treatment group imbalances in PLPs.ConclusionPrediction (Davis, 2018) and machine learning (James, 2021) models have been developed to classify PLPs. Using such tools to stratify patients by PLP status at randomization minimizes the risk of having an underpowered study. The same tools can be used to enrich both the MCI and mild AD populations with PLPs to reduce SS requirements.