A MODEL FOR ALZHEIMER’S DISEASE IN THE PREVENTION SETTING
Author(s)
Caputo A1, Racine A1, Paule I1, Feller C1, Savelieva M1, Hummel N2, Karcher H2, Lopez Lopez C1, Graf A1
1Novartis Pharma AG, Basel, Switzerland, 2Analytica Laser, Loerrach, Germany
OBJECTIVES: Shifting the focus of clinical trials testing disease-modifying interventions against Alzheimer’s disease (AD) from the dementia stages of the disease to pre-symptomatic stages may increase the likelihood of success for these trials. The aim of this research was to develop a model for the pre-symptomatic time course in the AD prevention setting to inform clinical trial design. METHODS: We developed a statistical model describing time to first diagnosis of mild cognitive impairment (MCI) or AD diagnosis using a Weibull parametric survival function and the progression of the Alzheimer’s Prevention Initiative Preclinical Composite (APCC, see Langbaum et al. 2014), a measure for cognitive decline, using a non-linear empirical function. We chose model covariates based on clinical relevance, goodness of model fit and statistical tests. We trained the model on databases which included healthy as well as cognitively impaired and demented subjects. RESULTS: We identified age, apolipoprotein E ε4 status, APCC at baseline and education level as important model covariates. Patient simulations showed a good fit between model predictions and observed values, for both time to first diagnosis and progression of APCC. Simulations also showed that an enrichment strategy focusing on elderly participants yielded a higher power for a given hazard ratio of the investigated interventions. CONCLUSIONS: The 2-step model linking APCC decline and time to MCI or AD diagnosis is the first AD disease progression model for pre-symptomatic stages of the disease. It exhibits good internal validity and can be used in the context of optimizing design of clinical trials in the prevention setting. Further refinements of the model, e.g. including biomarkers such as amyloid-beta and tau as covariates and covering other relevant endpoints, external validation of the model, and incorporation into a health economic model to evaluate interventions in the prevention setting, are objectives of future research.
Conference/Value in Health Info
2017-11, ISPOR Europe 2017, Glasgow, Scotland
Value in Health, Vol. 20, No. 9 (October 2017)
Code
PRM144
Topic
Methodological & Statistical Research
Topic Subcategory
Confounding, Selection Bias Correction, Causal Inference, Modeling and simulation
Disease
Neurological Disorders