Traumatic Brain Injury: What Is a Favorable Outcome?

The Glasgow Outcome Scale-Extended (GOSE) ¹ is the most widely used outcome measure in traumatic brain injury (TBI) research, ² frequently serving as the primary end-point in studies investigating the natural history of recovery from TBI, accuracy of predictive biomarkers, and treatment efficacy. The GOSE also facilitates clinical prognostic counseling by specifying functionally based recovery milestones. The scale is ordinal in nature, comprised of eight discrete categories that are hierarchically organized to capture the full spectrum of TBI outcome, from death (GOSE = 1) to return to pre-injury baseline (GOSE = 8). Although the GOSE produces a single categorical score, outcomes are often dichotomized by either assigning a cut-point at a fixed value or using a sliding scale informed by initial injury severity. ³ Scores falling above and below the cut-point are grouped into “favorable” and “unfavorable” outcome categories, respectively. The cut-point used to define a “favorable” outcome after TBI is not standardized and is determined by clinicians and investigators rather than by patients and caregivers.

While dichotomization simplifies data analysis and interpretation, this approach also decreases the precision of the GOSE by reducing eight distinct categories into two broad categories. For example, for a cut-point of 4, “unfavorable” recovery encompasses persons with outcomes ranging from death (GOSE = 1) to part-time supervision in the home (GOSE = 3). Conversely, “favorable” recovery includes persons with outcomes ranging from inability to function independently outside the home (GOSE = 4) to resolution of all symptoms (GOSE = 8). There is no standardized cut-point for dichotomizing the GOSE, which complicates comparison of findings across studies. Further, cut-points invite subjective interpretations of the potential for recovery after TBI, and may influence vital decisions about goals of care and access to inpatient rehabilitation. We offer an immediate, practical solution to improve the use of the GOSE in research and clinical settings, and highlight the need for patient- and caregiver-centered TBI outcome measures.

We identified several important limitations associated with dichotomizing the GOSE and ascribing “favorable” and “unfavorable” outcome labels to scores falling above and below the cut-point. Statistically, dichotomization may reduce the power to detect a significant relationship between a variable and an outcome, underestimate the variation present within each group, and mask non-linear relationships within the data. ^{4 –7} Further, there is no consensus as to what constitutes a “favorable” outcome after TBI. The selection of a cut-point and the designation of an outcome as “favorable” or “unfavorable” is inherently subjective and reflects biases, which may be influenced by ephemeral personal, cultural, and societal value systems. ⁸ Even if consensus around a GOSE cut-point could be achieved, the terms “favorable” and “unfavorable” are loaded with emotional valence and reflect neither the recovery potential of some individuals with “unfavorable” outcome, nor the substantial long-term disability experienced by some individuals with “favorable” recovery. Finally, cut-points are not guided by patient or caregiver preferences. This increases the risk that the border between “favorable” and “unfavorable” outcome will be influenced by the personal values of individual providers.

As testament to the variability in how “favorable” outcome has been defined in the TBI literature, in studies published between 2014–2019, we found six different GOSE cut-points for “favorable” outcome, ranging from GOSE = 3 to GOSE = 7, (Table 1 ^9–14), a testament of the variability in GOSE dichotomization. None of the identified studies provided an empirically based rationale for selection of the cut-point, making it impossible to compare results across investigations. Further, none of the studies considered caregiver or patient perspectives. We believe this is paternalistic and disregards the foundational principle of respect for persons, which stipulates that those with diminished autonomy are entitled to special protections. Prior studies have shown that some individuals with mild TBI report a negative quality of life, ¹⁵ while those with substantial permanent disability often report acceptable quality of life. ¹⁶ This phenomenon, known as the “disability paradox,” further emphasizes the need to better understand patient and caregiver preferences and perspectives on outcome after TBI.

A comprehensive review of TBI randomized control trials published in 2016 found that 88% of methodologically robust studies used the GOSE, or its predecessor, the GOS, as the outcome measure. ³ Because of its historical legacy, ease of use, and acceptance by the U.S Food and Drug Administration, the GOSE will likely continue to serve as a primary outcome measure in TBI trials. However, the absence of a standardized, evidence-based, and consumer-informed GOSE dichotomy is consequential for both research and clinical practice. In the research setting, TBI trial results may be biased, difficult to interpret, and incomparable across studies. In clinical practice, an early “unfavorable” neurologic prognosis communicated by the provider is a common factor underlying the decision to withdraw life-sustaining therapy. ¹⁷ Dichotomizing the GOSE may ultimately cause harm to individual patients and complicate efforts to improve outcome after TBI.

A practical alternative to arbitrary outcome dichotomization that can be implemented immediately in TBI studies is to report results as the proportion of subjects achieving each of the eight GOSE milestones. Specifying the percentage of subjects who fall into each GOSE category facilitates comparison of findings across studies, improves statistical robustness, ^4,7 and provides clinically translatable information that can be incorporated into prognostic counseling. Although reporting all GOSE categories is preferred over dichotomization, if dichotomization is unavoidable (e.g., when there is missing data across categories) the cut-point should be based on patient and caregiver perspectives, rather than on investigator beliefs. Limitations of the GOSE as a TBI outcome measure are well-documented, ¹⁸ and recent gap analyses published by investigators from the Neurocritical Care Society Curing Coma Campaign suggest that novel, precise TBI outcome measures are required to improve prognostic modeling ¹⁹ and assessment of treatment efficacy. ²⁰ Future studies should, therefore, use mixed methods to establish a patient- and caregiver-centered TBI outcome metric and a definition of “favorable” outcome. For example, a qualitative study design that includes interviews or focus groups may engage key stakeholders to identify the breadth of potential meaningful outcomes, as well as the critical elements of a “favorable” outcome. Subsequently, the proposed outcomes may be validated using quantitative surveys in a larger sample. ^21,22 In the absence of such a patient- and caregiver-centered measure, an analytic approach that provides the likelihood of achieving each of the eight GOSE categories will improve the rigor of TBI trials, minimize bias in outcome reporting, facilitate transparent clinical prognostic counseling, and accommodate diverse value systems.