Optimal design and analysis of AD treatment

点击翻译按钮获取中文摘要

DESCRIPTION (provided by principal investigator): The Food and Drug Administration has recently relaxed its rules governing statistical analysis plans for clinical trials of investigational new drugs to allow rate of change analysis. Specifically, analyses that estimate slope using all longitudinal data points, not just first and last, have now been reported for investigational new drugs to treat Alzheimers disease. A representative example is the recently reported Alzheimers disease treatment trial of bapineuzumab, which tested the effect of treatment on rate of decline on a global cognitive assessment scale. The analysis used a modied intent to treat (MITT) sample, restricting the analysis to subjects with at least one follow-up observation. Sample size considerations are poorly developed for this analysis plan, and the many assumptions implicit in this analysis have never been formally tested with representative Alzheimers disease data. Two unique data resources are available to explore these issues. One data resource is the accumulated clinical trial data from the Alzheimers Disease Cooperative Study (ADCS). The ADCS, which performs clinical trials of non-licensible treatments not under the purview of the FDA, has over ten years of experience with the rate of change analysis. The second data resource is the Alzheimers Disease Neuroimaging Initiative (ADNI) cohort, which was created expressly for the purpose of informing the design of future Alzheimers disease treatment trials and secondary prevention trials of mild cognitively impaired (MCI) subjects. With this in mind, we propose the following Specific Aims: Specific Aim 1. Using data from the Alzheimers Disease Cooperative Study (ADCS) and the Alzheimers Disease Neuroimaging Initiative (ADNI), to accurately determine statistical sample size requirements for Alzheimer treatment trials and secondary prevention (MCI) trials using standard clinical and neuropsychometric outcomes. Specific Aim 2. Using data from ADCS and ADNI, to describe the potential relative utility of various biomarkers proposed as surrogate outcome measures for Alzheimer treatment trials and secondary prevention (MCI) trials. Specific Aim 3. Using data from ADCS and ADNI, to test the validity of the statistical assumptions implicit in the MITT rate of change analysis, specically, the assumption of random dropout and the assumption of linear progression over time. Specific Aim 4. Using data from ADCS and ADNI, to explore the performance of standard MITT analyses using mixed effects models or generalized estimating equations relative to alternative methods that are robust to failures of the random dropout assumptions. PUBLIC HEALTH RELEVANCE: Clinical trials that are too small are noninformative and prone to 'false negatives', that is, erroneous conclusions that an effective treatment is ineffective. Trials that are too large incur unnecessary study subject burden and cost, a not inconsequential concern, as Alzheimer treatment trials require hundreds of subjects and millions of dollars to perform. We will use accumulating data from NIH sponsored cohort studies and treatment trials to determine optimal samples size for future clinical trials.