Opportunities for more powerful statistical analyses of ordinal scales: A scoping review of neurological trials

Abstract

AbstractBackground and ObjectivesOrdinal scales are widely adopted as outcome measures in neurological randomized controlled trials (RCTs). There have been extensive discussions about appropriate statistical analysis strategies of ordinal neurological outcomes. We aimed to evaluate which statistical methods have been used to test and estimate treatment effects from ordinal outcomes in recent RCTs across a range of acute neurological diseases.MethodsWe searched for RCTs in five acute monophasic neurological diseases (stroke, traumatic brain injury (TBI), subarachnoid hemorrhage (SAH), meningitis, and Guillain-Barré Syndrome (GBS)) published in high-impact journals between January 1, 2015 and November 1, 2023. Trials had to report on an ordinal scale as the primary or secondary outcome. Two independent reviewers assessed whether/how investigators (1) delt with the ordinal nature of outcomes, (2) assessed and reported key assumptions,(3)utilized longitudinal measurements, (4) adjusted for prognostic variables.ResultsWe included 70 RCTs for treatment evaluations in stroke (n=36), TBI (n=13), SAH (n=10), meningitis (n=7), and GBS (n=4). In 46/70 (66%) trials, investigators retained the full ordering information, commonly analyzed by a proportional odds model (33/46 trials, 72%). The proportional odds assumption was not addressed in 23/33 (62%) of these trials. In 22/70 (31%) trials, the ordinal outcome was dichotomized, with notable disagreement on the cut-point within neurological diseases. In 41/70 (59%) trials, the ordinal outcome was assessed at multiple time points, while some form of longitudinal data analysis was performed in only three (7%) of these 41 studies. The time point chosen for analysis was inconsistent within neurological diseases.DiscussionThe current practice of analyzing ordinal outcomes is often suboptimal in neurological trials according to modern statistical standards. Dichotomization and focus on a single arbitrary time point are still common, while more efficient analysis strategies exist. Further research needs to clarify the balance between maximizing the statistical power and assumptions made in approaches that better leverage ordinal information.