Abstract:
|
Various methodologies have been proposed for statistical validation of surrogate endpoint; the evaluation of performance of those methods is needed for practitioners to apply these methods effectively. Acute myeloid leukemia (AML) is a heterogeneous disease associated with poor clinical outcomes. Now sensitive molecular biology tests become available to measure minimal levels of cancer cells in bone marrow samples (i.e., minimal residual disease), which is a potential surrogate endpoint for OS. For establishing surrogacy, it is important to quantify 2 types of correlations: individual-level and trial-level correlation. Establishing trial-level correlation is usually more challenging. A simple and initiative approach is weighted least square regression using aggregated data; alternatively, a sophisticated statistical modeling approach proposed by Burzykowski(2004) can be applied using subject level data. In this work, we evaluate these two approaches using simulated data under scenarios which are practical in AML settings. . In addition, we also recommend statistical criteria for surrogacy based on simulation results.
|