No Thumbnail Available
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- Clinical trials for rare diseases face significant challenges, including limited patien populations, variability in symptoms, and the complexity of trial designs such as those measuring outcomes at baseline versus the end of the treatment period without a control group (Kandi and Vadakedath 2023). The Generalized Pairwise Comparison (GPC) methodology offers a potential solution by allowing the inclusion of multiple endpoints in the analysis, thereby eliminating the need to select a single primary endpoint (Buyse 2010). In this thesis, we employ linear mixed models to simulate the data and evaluate the performance of the GPC approach, focusing on a comparison between the matched and unmatched methodologies. The two methodologies measure different aspects of treatment effects, the matched method measure Γ, the proportion of individuals who show improvement under treatment compared to their baseline. While the unmatched method measure ∆, the expected improvement of a patient across the entire population when is under treatment rather than the standard care. Unless the two groups are nearly independent, both methods exhibit similar statistical power. However, due to the interpretability of Γ, we advise to use the matched method in a baseline versus end-of-treatment design, because correlation between the two measures is expected. Applying these methods to real data from the Sirolimus trial, we find a strong concordance between the two approaches, further validating their use. While GPC shows great promise as a tool for analyzing rare disease trials, one major limitation is its current inability to account for the longitudinal nature of the data. Future research should aim to extend the GPC methodology to incorporate longitudinal data, thereby enhancing its utility in clinical trials.