Early diagnosis and prediction of traumatic brain injury (TBI) is essential for determining treatment strategies and allocating resources. This study evaluated the predictive accuracy of Glasgow Coma Scale (GCS) verbal, motor and eye components alone, or in addition to pupil size and reactivity, for TBI.


A retrospective cohort analysis of data from 51 425 severely injured patients registered in the Trauma Registry of the German Society for Trauma Surgery from 1993 to 2009 was undertaken. Only directly admitted patients alive on admission and with complete data on GCS, pupil size and pupil reactivity were included. The unadjusted predictive roles of GCS components and pupil parameters, alone or in combination, were modelled using area under the receiver operating characteristic (AUROC) curve analyses and multivariable logistic regression regarding presence of TBI and death.


Some 24 115 patients fulfilled the study inclusion criteria. Best accuracy for outcome prediction was found for pupil reactivity (AUROC 0·770, 95 per cent confidence interval 0·761 to 0·779) and GCS motor component (AUROC 0·797, 0·788 to 0·805), with less accuracy for GCS eye and verbal components. The combination of pupil reactivity and GCS motor component (AUROC 0·822, 0·814 to 0·830) outmatched the predictive accuracy of GCS alone (AUROC 0·808, 0·800 to 0·815). Pupil reactivity and size were significantly correlated (rs = 0·56, P < 0·001). Patients displaying both unequal pupils and fixed pupils were most likely to have TBI (95·1 per cent of 283 patients). Good outcome (Glasgow Outcome Scale score 4 or more) was documented for only 1929 patients (8·0 per cent) showing fixed and bilateral dilated pupils.


The best predictive accuracy for presence of TBI was obtained using the GCS components. Pupil reactivity together with the GCS motor component performed best in predicting death. Copyright © 2011 British Journal of Surgery Society Ltd. Published by John Wiley & Sons, Ltd.