Objective. To determine the prognostic value of multimodal evoked potentials (EPs) and event-relatcd (ERPs) potentials in coma (Glasgow Coma Score < 8), after severe traumatic brain injury (TBI). Design: Prospective, longitudinal stud:a of neurophysiological responses recorded during traumatic coma. Setting: Intensive Care Unit, Frenchay Hospital, Bristol, UK. Participants: Fifty-four comatose TBI patients (age range 1-80 years, mean 36.4). Methods. Neurophysiological responses were recorded from 11 scalp, electrodes with earlobe reference Conduction times were measured for brainstem auditory, Hash visual and somatosensory, short-latency EPs. Peak latencies and amplitudes were determined for long-latency components of visual and auditory ERPs, generated by passive ''oddball'' paradigms. These neurophysiological and various clinical parameters were correlated with patient outcome using Pearson's coefficient. Main outcome measure. Three month Glasgow Outcome Scale (GOS). Results and conclusion. Highly significant (P < 0.001) correlations exist between long-latency ERP components and 3-month outcome. Short-latency EPs, brainstem (wave I-V) and somatosensory conduction times, also correlate significantly with the GOS (P < 0.01). Of the clinical measurements, pupillary response patterns, APACHE II and Glasgow Coma Scores (GCS) correlate significantly with outcome, as do the retrospective measures of duration of coma and post-traumatic amnesia (PTA) in survivors. Unfortunately, due to variance of long-latency responses, even in controls, absolute values cannot be relied upon as prognosticators. The presence of ''mismatch negativity'' predicted the return of consciousness (89.7% sensitivity and 100% specificity) and preceded changes in GCS. Its latency was the single best indicator of 90-day outcome from coma (r = -0.641).
