Abstract:
|
The Glasgow Outcome Scale-Extended (GOS-E), is often selected as the primary outcome for clinical trials of traumatic brain injury (TBI). The common practice is to use a sliding dichotomy of the GOS-E that incorporates subject’s baseline prognosis, which results in loss of statistical information due to the collapse of adjacent categories. To improve on power, a recent paper by Yeatts et al. proposed a sliding scoring of the GOS-E. However, drawback is treating ordinal as continuous endpoint based on normal approximation without continuity correction. We proposed analyzing the sliding scoring of the GOS-E as either ordinal or continuous with continuity correction. We used retrospective data from a TBI trial for prospective Bayesian design using dichotomous, continuous, or ordinal model, and evaluated the design operating characteristics and estimation efficiency (bias, variance, MSE) via simulations. We also illustrated a connection between the ordinal data and time-to-event (TTE) data censored to integers allowing use of Bayesian software that utilize TTE-based models.
|