Effect of Two Glasgow Outcome Scale-Extended Scoring Methods on Traumatic Brain Injury Clinical Trial Design: A...

The Glasgow Outcome Scale-Extended (GOSE) is the most frequently used outcome measure for traumatic brain injury (TBI) clinical trials. The GOSE may be administered several ways, the choice depending on the purpose of the research. For example, the GOSE can be administered to reflect functional limitations attributed to the overall injury, including extracranial injuries (GOSE-All), or to discount limitations attributed to extracranial injuries (GOSE-TBI). In this investigation, we assessed the effect of using GOSE-All versus GOSE-TBI in clinical trial design. We estimated the impact of the differences in assessment strategy on sample size and power for a clinical trial of an intervention that affects only TBI-related limitations. Inclusion criteria based on TBI severity and extracranial injury severity were examined, as were primary assessments at 2 weeks or 3, 6, or 12 months after injury. Data from 2,288 participants in the prospective observational Transforming Research and Clinical Knowledge in Traumatic Brain Injury (TRACK-TBI) study were used to simulate the effects. If the trial were analyzed by a Mann-Whitney test comparing GOSE-All scores between treatment groups, sample size would need to increase 8-158% to account for the apparent decreased effect of a treatment that affects only the brain injury. If the sample size were not adjusted, power to detect a treatment effect would decrease from 80% to as low as 41%. If the outcome were dichotomized (favorable=GOSE 8 if including only patients with Glasgow Coma Scale [GCS]=13-15, GOSE 5-8 if GCS = 3-12), the sample size would need to increase 6-165%. The ratios of sample size are largest when the trial population consists of people with milder brain injuries and decrease with time since injury in those with GCS 13-15. It is crucial for researchers, given the aims of their studies, to decide in advance whether the classification of the GOSE should be based on effects attributed to the brain injury, despite the fact that extracranial injuries may not have allowed one to experience the extent of limitation due to the TBI, or all injuries, including extracranial injuries, and to power their studies accordingly. Instructions to the respondent and outcomes examiner need to be clear about what causes of disability are to be included. The assessment method should be accounted for in the power and sample size calculations, clearly indicated in the protocol and publications and documentation accompanying shared data, and emphasized in the training of the outcome examiners so all are collecting the desired information.