Lead Investigator: Sudhir Sivakumaran, Critical Path Institute
Title of Proposal Research: Development of model-informed drug development tools for Alzheimer’s Disease
Vivli Data Request: 7587
Funding Source: Critical Path Institute is supported by the Food and Drug Administration (FDA) of the U.S. Department of Health and Human Services (HHS) and is 54.2% funded by the FDA/HHS, totaling $13,239,950, and 45.8% funded by non-government source(s), totaling $11,196,634. Support for the CPAD consortium is also provided by consortium members, including AbbVie Inc., Bioclinica, Biogen, Eisai, Eli Lilly and Company, F. Hoffman La Roche, GE Healthcare, Imeka, IXICO plc., Janssen Research & Development LLC, Merck, Sharp & Dohme, Novartis Pharmaceuticals Corporation, Takeda Pharmaceuticals, Unlearn.AI, Inc, Alzheimer’s Association, Alzheimer’s Research UK, and the Cure Huntington’s Disease Initiative (CHDI) Foundation.
Potential Conflicts of Interest: None
Summary of the Proposed Research:
Recent estimates indicate that the number of people worldwide living with dementia is expected to rise to 139 million cases in 2050. Alzheimer’s disease (AD) accounts for an estimated 60% to 80% of cases of dementia. Despite the looming public health crisis due to AD, therapies available in the U.S. for the last two decades only treated the symptoms of the disease. There is a pressing need for effective therapeutic interventions and there is a growing interest in intervening earlier in the disease process. This requires a robust understanding of disease progression across all stages of disease. The Critical Path for Alzheimer’s Disease (CPAD) is a nonprofit, pre-competitive consortium of the Critical Path Institute (C-Path). CPAD focuses on integrating precompetitive patient-level data from high quality Alzheimer’s disease (AD) clinical trials and observational studies, and transforming those data into actionable knowledge to advance novel regulatory-grade quantitative drug development tools. We develop mathematical models of disease progression that will enable a user to simulate clinical trials in Alzheimer’s disease. The trial simulations will allow the user to identify the appropriate patients for a trial as well as understand how the trial should be designed in terms of number of patients. The simulations are meant to help companies design their trials more effectively, reducing trial cost and time, and reduced patient burden.
We plan to develop mathematical models of disease progression that will enable a user to simulate clinical trials in Alzheimer’s disease. Modelling approaches for disease progression models will utilize mixed-effects modeling as it readily has several advantages including incorporation of fixed observable effects to explain variability, incorporation of random unexplained variability between subjects with flexibility to include additional hierarchies (inter-study variability), flexibility in handling repeated measurements within individuals with different amount of and different spacing of time points, and ability to easily generate predicative simulations through Monte Carlo sampling. We will incorporate Clinical Outcomes Assessments and biomarkers (dependent variables). The selection of these dependent variables will be informed by the availability of relevant information, as well as the clinical meaningfulness, interpretability, as well as regulatory and drug development relevance of each proposed dependent variable. We will include apolipoprotein E4 allele (APOE4), sex, baseline age, baseline severity, other genetic and demographic data, as well as multiple biomarker candidates as covariates. The trial simulations will allow the user to identify the appropriate patients for a trial as well as understand how the trial should be designed in terms of number of patients. The simulations are meant to help companies design their trials more effectively. The trial simulation tool will benefit patients by reassuring them that they will be able to enroll in more efficiently designed clinical trials in the future.
Statistical Analysis Plan:
We plan to develop a comprehensive set of disease progression models across the continuum of the disease utilizing available data. Disease progression models serve as the backbone for drug-disease-trial models, which are used for clinical trial simulation. Drug-disease-trial models include disease-relevant covariates, (e.g. biomarkers and patient demographics), as baseline prognostic indicators and trial-specific information (e.g. dropout information) that can inform each of the various components in the model. For example, natural history data to inform underlying disease progression, placebo arm data to inform about magnitude, onset and offset of placebo response in controlled clinical trials, estimates of various drug effects (magnitude, time to onset, and durability), rate and magnitude of drop-outs in the trials. The combination of these components gives sponsors the ability to optimize trial design by informing inclusion criteria, enrichment strategies, stratification approaches, and sample size calculations. No single data source can provide all these elements and therefore requires multiple integrated datasets.
We performed exploration, including at the meta-data level, of Alzheimer’s disease datasets that contain extensive fluid and imaging biomarker information, which are key covariates to inform disease progression models being developed by CPAD. This will enable identification of sources of variability and account for heterogeneity in disease progression through a harmonized and standardized quantitative analysis process. Observations with missing sex or race will be regarded as a separate category, and we will fill missing baseline ages with population’s average age of corresponding study. Additional solutions will be applied based on missing structure, e.g., the proportion of missing data.
Data will be pooled across biomarker assays and imaging modalities to define modeling and analysis specifications that incorporate key aspects of fluid and imaging biomarkers. We will maintain the structure and the independence of the studies by keeping study identifiers associated with subject level data. Modelling approaches for disease progression models will utilize mixed-effects modeling as it readily has several advantages including incorporation of fixed observable effects to explain variability, incorporation of random unexplained variability between subjects with flexibility to include additional hierarchies (inter-study variability), flexibility in handling repeated measurements within individuals with different amount of and different spacing of time points, and ability to easily generate predicative simulations through Monte Carlo sampling. Modeling will be conducted in NONMEM (Pirana) and R.
General model development will follow established best practices according to five main steps:
1. Selection of the base structural model to describe the natural disease progression, including two levels of random effects (inter-individual and residual variability).
2. Inclusion of candidate predictors (i.e., covariates) of natural disease progression using a “full model” approach based on clinical plausibility.
3. Incorporation of inter-study variability (data permitting).
4. Evaluation of model performance during key model building steps with simulation-based diagnostics.
5. External validation using contemporary clinical trial data sources.
Knowledge from published longitudinal models will be considered for the selection of candidate base structural models. For the natural disease component, various linear and nonlinear models will be tested depending on the nature of the longitudinal trajectories in the observed data. Random effects will be included on relevant structural parameters, e.g. additive on intercept and rate for linear models. Error models will be explored that best account for residual unexplained variability, e.g. (additive, proportional, or combined). Additional error models will be explored that account for potential heteroskedasticity due to floor or ceiling effects, such as beta regression.
Effect of LY3202626 on Alzheimer’s Disease Progression as Measured by Cerebral ¹⁸F-AV-1451 Tau-PET in Mild Alzheimer’s Disease Dementia
Data Contributor: Lilly
Study ID: NCT02791191
Sponsor ID: 16223
Effect of γ-Secretase Inhibition on the Progression of Alzheimer’s Disease: LY450139 Versus Placebo
Data Contributor: Lilly
Study ID: NCT00594568
Sponsor ID: 7666
Effect of LY450139 a y-Secretase Inhibitor, on the Progression of Alzheimer’s Disease as Compared With Placebo
Data Contributor: Lilly
Study ID: NCT00762411
Sponsor ID: 11271
Open-Label Extension for Alzheimer’s Disease Patients Who Complete One of Two Semagacestat Phase 3 Double-Blind Studies (H6L-MC-LFAN or H6L-MC-LFBC)
Data Contributor: Lilly
Study ID: NCT01035138
Sponsor ID: 5930
Sudhir Sivakumaran, Yashmin Karten, Nicholas Cullen, Corissa Lau, Eileen Priest, Hazel White, Klaus Romero, Michael Irizarry. Accelerating drug development through precompetitive data sharing and collaboration in the Critical Path for Alzheimer’s Disease (CPAD) Consortium. AD/PD 2023, Abstract P0325/#231, p325.