|
Abstract:
|
Rapidly progressing patients, although rare, have influenced results in Alzheimer’s disease clinical trials and contributed to the lack of reproducibility in study results that was cited by FDA as a major factor in interpreting results from a recent BLA. The present research explored the impact of rapidly progressing patients and the highly skewed data resulting from them in a simulation study patterned after data from the BLA mentioned above. Results indicate that as expected the central limit theorem provides for unbiased estimation from large samples. However, averaging results across repeated samples masks differences between individual study results. Rapid progressors inflate variance and an imbalance in rapid progressors had a marked influence on results from MMRM. In contrast, robust regression, like MMRM was unbiased, but provided more stable estimates than MMRM, especially when the frequency of rapidly progressing patients was not balanced across treatment arms. These results suggest more work is needed from a diverse pool of sponsors and researchers to understand if something other than MMRM should be the standard analytic approach in Alzheimer's clinical trials.
|
