Giulia Lestini (1), Amy Racine (1), Ines Paule (1), Chrystel Feller (1), Kostas Biliouris (2), Etienne Pigeolet (1), Cristina Lopez Lopez(1), Ana Graf (1), Angelika Caputo (1)
(1) Novartis Pharma AG, Basel, Switzerland; (2) Novartis Institute for Biomedical Research, Cambridge, MA, USA
Objectives: Optimal clinical trial designs to evaluate innovative treatments[1] are needed to test disease-modifying interventions in the Alzheimer’s disease (AD) prevention setting. The aim of our work was to develop a model to support the design of such clinical trials in participants at risk of developing mild cognitive impairment (MCI) or dementia due to AD. In particular, we wanted to assess the performance of two different endpoints in terms of statistical power. These are: i) the time to MCI or dementia diagnosis and ii) the cognitive decline, as measured by a recently developed composite cognitive score, the Alzheimer’s prevention initiative preclinical composite score (APCC)[2], intended to detect subtle cognitive changes in early stages prior to MCI or dementia diagnosis.
Methods: We first developed a time to event (TTE) model describing the time to first diagnosis of MCI or dementia using parametric survival functions. The TTE model was fitted to longitudinal data from the National Alzheimer’s Coordinating Center (NACC) and data collected from three cohort studies (ROS, MAP and MARS) of aging and dementia at the Rush Alzheimer’s Disease Center. As a second step, mixed-effects models describing the progression of APCC scores over time were developed and fitted to two subpopulations in the Rush cohorts. A nonlinear-mixed effects model was fitted to the so named “progressors” subpopulation, i.e. subjects with first diagnosis of MCI or dementia within eight years, and a linear-mixed effects model was fitted to the “non-progressors”, i.e. subjects who either were not diagnosed or had a diagnosis only after eight years.
A time scale anchored at the time to MCI or dementia diagnosis predicted by the TTE model was used for the progressors model.
Clinically relevant covariates were tested for statistical significance in both models using backward elimination based on Akaike’s information criterion. The estimation of APCC and TTE models parameters was performed using the R software. Simulations based on these models were performed to assess the power of a clinical trial using either the TTE or change in the APCC as endpoints. Different scenarios assuming different treatment effects expressed in terms of hazard ratio, i.e. assuming a reduction in the risk of MCI or dementia diagnosis, were simulated.
Results: In the TTE model, a Weibull survival function performed best among several other candidate functions, and it included age and APCC at entry of the study, APOE-ε4 status, and number of years of education as covariates. Predicted survival probabilities were adequately distributed around the Kaplan-Meier survival curves derived from the observed Rush and NACC data. Moreover, relevant diagnostic plots confirmed the quality and good predictive performance of the two APCC models. Both progressors and non-progressors models included APCC at entry of the study, APOE-ε4 status, number of years of education and gender as covariates. Furthermore, the progressors model included age at event whereas the non-progressors model included age at entry of the study as covariates. Trial simulations showed that an overall power greater than 80% could be reached with a realistic hazard ratio and reasonable sample size. The power to show a treatment effect was higher with a TTE endpoint than with a change in APCC for all tested scenarios.
Conclusions: This two-step modelling approach of time to MCI or AD diagnosis and APCC decline has been successfully used to design clinical trials in the AD prevention setting. The model shows good internal validity and allows comparing the performance of different endpoints in terms of statistical power. Further refinements of the model, e.g. including amyloid-beta and tau as covariates are objectives of future research.
References:
[1] Reiman EM, Langbaum JB, Tariot PN. Alzheimer’s Prevention Initiative: a proposal to evaluate presymptomatic treatments as quickly as possible. Biomark Med (2010) 4(1): 3–14
[2] Langbaum JB et al. An empirically derived composite cognitive test score with improved power to track and evaluate treatments for preclinical Alzheimer’s. Alzheimer’s & Dementia (2014) 10(6):666-74
Reference: PAGE 27 (2018) Abstr 8480 [www.page-meeting.org/?abstract=8480]
Poster: Drug/Disease Modelling - CNS