| Autor: |
Atri, Alireza, Ivkovic, Milana, Jönsson, Linus, Gustavsson, Anders, Léon, Teresa, Hahn‐Pedersen, Julie Hviid, Handels, Ron, Raket, Lars Lau |
| Zdroj: |
Alzheimer's & Dementia: The Journal of the Alzheimer's Association; Dec2023 Supplement 21, Vol. 19 Issue 21, p1-4, 4p |
| Abstrakt: |
Background: Progression models for repeated measures (PMRM1) are a class of statistical models that enable estimation of slowing of progression of Alzheimer's disease, expressed as percentage or time delay. Similarly, accelerated failure time (AFT) models can estimate relative expected time until progression when evaluating disease‐modifying therapies. We compared these methods to mixed‐effects models for repeated measures (MMRM) and Cox time‐to‐event analysis on simulated trial data, evaluating their power and the difference in interpretation when expressing outcomes as a measure of time. Method: Primary outcomes were Clinical Dementia Rating (CDR) Sum of Boxes (MMRM and PMRM) and time to progression to dementia as assessed by CDR global (Cox regression and AFT models). Weibull and log‐logistic distributions were assumed for random error distribution in AFT models. Subject‐level placebo arm trajectories were jointly simulated based on estimated CDR mean trajectories and joint temporal correlation structure of 538 amyloid‐positive MCI patients meeting typical disease‐modifying trial inclusion criteria from the ADNI study. Active arm trajectories were simulated to show an average 20% slowing of disease progression vs placebo at each visit. A thousand simulations were conducted across scenarios, varying number of subjects per arm (200‐700) and clinical trial duration (18‐36 months). Result: Power of PMRM was shown to be consistently greater than that of MMRM (Table 1). The AFT and Cox models' power rarely exceeded 40%. PMRM accurately estimated the underlying 20% slowed progression. With greater sample sizes and at least 2 years of follow‐up, the AFT models indicated a slowing of progression, but the median estimated slowing was 5%. The MMRM and Cox model found median estimated reduction in decline of 25% and median estimated hazard ratios of 0.88‐0.91, respectively. Conclusions: PMRM and AFT models estimate slowing of progression. Compared to the MMRM and Cox model, these time‐based models better capture the effect of disease‐modifying therapies and facilitate translation of results to patients and their care partners. Among all methods studied, the PMRM offered the best combination of interpretability and power. References: 1. Raket, LL. Progression models for repeated measures: Estimating novel treatment effects in progressive diseases. Statistics in Medicine. 2022; 1‐ 21. https://doi.org/10.1002/sim.9581 [ABSTRACT FROM AUTHOR] |
| Databáze: |
Supplemental Index |
| Externí odkaz: |
|
