Review of the use of somatosensory evoked potentials in the prediction of outcome after severe brain injury

OBJECTIVEReview the predictive powers of somatosensory evoked potentials (SEPs) in severe brain injury. DATA SOURCESPublications in the scientific literature, manual review of article bibliographies, and questioning workers in the field. STUDY SELECTIONStudies addressing the prediction of outcome after severe brain injury using SEPs. DATA EXTRACTIONTo determine the outcome of patients with either normal or bilaterally absent SEPs as categorized using the Glasgow Outcome Scale into favorable outcomes (good or moderate disability) or unfavorable outcomes (severe disability, vegetative, or dead). Studies were included if they were in English and allowed the determination of outcomes for all patients with normal or bilaterally absent SEPs. Papers were not considered if subjects were neonates, consisted of abstracts where all necessary details were unavailable, were case reports or duplications of other published studies, or dealt only with brain dead subjects. DATA SYNTHESISFor all studies (n = 44), positive likelihood ratio, positive predictive value, and sensitivity were 4.04, 71.2%, and 59.0%, respectively, for normal SEPs (predicting favorable outcome) and 11.41, 98.5%, and 46.2%, respectively, for bilaterally absent SEPs (predicting unfavorable outcome). Summary receiver operating characteristic curve analysis detected a cut-off criterion effect for only blinded studies of bilaterally absent SEPs. Twelve patients (12/777) were identified with bilaterally absent SEPs who had favorable outcomes. These false positives are typically pediatric patients or have suffered traumatic brain injuries. We suggest criteria for the use of bilaterally absent SEPs in the prediction of poor outcome, which include absence of focal lesions, subdural or extradural fluid collections, and no decompressive craniotomy in previous 48 hrs. Using these criteria the data suggest that the false-positive rate is