Prognosis studies examine outcomes and identify factors associated with or predictive of observed outcomes. In rare diseases, prognosis studies play an especially important role in informing patients and guiding treatment. Yet the current prognosis literature is plagued by design issues limiting clarity of information.
Objectives: This thesis aims to improve on the design and analysis of prognosis studies in rare diseases. The objectives are: 1) to evaluate the quality of current Systemic Lupus Erythematosus (SLE) prognosis literature and identify areas of deficits; 2) to apply the longitudinal study design to a childhood-onset SLE (cSLE) cohort to study longitudinal evolution of organ damage and disease activity; and 3) to explore the application of modern longitudinal analytic methods in analyzing an observational cohort.
Methods: A systematic review of the literature was performed. A cSLE cohort was studied longitudinally. Longitudinal modelling of an irregular-visit-schedule observational cohort was performed using marginal and random effect models, along with Bayesian latent class growth mixture modelling.
Results: I have identified study design elements at high risk of bias within the SLE literature. Using the longitudinal design, I demonstrated that cSLE patients accrued damage throughout their disease courses and determined prognostic factors with clear temporal predictive relationships with damage evolution. I applied latent class technique to group patients’ disease activity trajectories into more homogeneous classes and identified early factors that predicted patterns of disease activity evolution.
Conclusions: I have demonstrated that the design and analysis of prognosis studies can be improved in rare diseases. I identified areas of deficits in prognosis study design. I have successfully applied the longitudinal study design and modern longitudinal analytic techniques to provide more precise prognostication in cSLE. Given this body of work, we now have the tools to design robust prognosis studies to achieve maximal gain in clarity of information