BACKGROUND: The credibility of economic evaluations of emerging Alzheimer’s disease (AD) therapies will depend on addressing persistent structural and data challenges with transparent and evidence-based approaches. This study describes methodological challenges recurring throughout the AD modeling literature and recommends practical solutions for future researchers.
METHOD: Systematic literature reviews of economic models in AD were supplemented by a targeted review of recently published models. The findings from the reviews and the approaches used in recent models were synthesized to identify remaining structural and data challenges. Data availability and anticipated health technology assessment requirements were used to inform potential solutions to these challenges.
RESULT: We identified four fundamental challenges for models used to evaluate emerging AD therapies: (1) quantifying heterogeneity in population characteristics and the natural history of disease progression and dementia onset; (2) assembling evidence supporting the extrapolation of short-term trial endpoints to long-term benefits; (3) addressing data gaps in costs, utilities, and mortality by disease severity accounting for age; and (4) bridging differences in populations, endpoints, and care settings across datasets and markets. We recommend analyses of existing longitudinal datasets to understand variability and prognostic factors within and between datasets, matching patient characteristics and progression patterns across datasets and clinical trials (challenges 1, 4). Existing longitudinal datasets also can be used to generate evidence of correlation between short-term clinical trial endpoints and long-term progression in support of model extrapolations (challenge 2). Long-term follow-up of clinical trial participants should be planned to validate findings from these analyses. Previously collected data on costs, utilities, and mortality should be reanalyzed to isolate AD-specific estimates across disease severity levels accounting for the impact of aging (challenge 3). Crosswalks between study populations, symptom measures, and biomarker criteria should be created to support adaption of economic models to local markets (challenges 1, 3, 4). Finally, models should be developed with flexibility in their structural assumptions to address uncertainty inherent in these recommended solutions.
CONCLUSION: Pragmatic solutions tailored to the available data and the anticipated profiles of therapies in late-stage development should be pursued to support near-term decision making and inform the design of long-term evidence-generation efforts.
