The objective of the study was to test the hypothesis that dynamic cerebral pressure-autoregulation is associated with the outcome of patients with severe head injury and to derive optimal criteria for future studies on the predictive value of autoregulation indices. Repeated measurements were performed on 32 patients with severe head injury. Arterial blood pressure (ABP) was measured continuously with an intravascular catheter, intracranial pressure (ICP) was recorded with a subdural semiconductor transducer and cerebral blood flow velocity (CBFV) was measured with Doppler ultrasound in the middle cerebral artery. Transfer function analysis was performed on mean beat-to-beat values, using ABP or CBFV as input variables and CBFV or ICP as the output variables. A dynamic index of autoregulation (ARI) ranging between 0 and 9 was extracted from the CBFV step response for a change in ABP. No significant differences between survivors and non-survivors were found due to mean values of ICP, ABP, CPP, CBFV, pCO2, GCS, age or heart rate. The transfer functions between ABP-ICP and CBFV-ICP did not show any significant differences either. The median [lower, upper quartiles] ARI was significantly lower for non-survivors compared with survivors [4.8 (0.0, 5.9) v. 6.9 (5.9, 7.4), p= 0.004]. The correlation between ARI and GOS was also significant (r=0.464, p=0.011). Cohen's coefficient was optimal for a threshold of ARI= 5.86 (kappa 0.51, p=0.0036), leading to a sensitivity for death of 75%, specificity=76.5%, odds ratio =9.75 and overall precision = 75.8%. The difference in ARI values between survivors and non-survivors persisted when results were adjusted for GCS (p = 0.028). A similar analysis for the Marshall CT scale did not reach significance (p = 0.072). A logistic regression analysis confirmed that apart from the ARI, no other variables had a significant contribution to predict outcome. In this group of patients, death following severe head injury could not be explained by traditional indices of risk, but was strongly correlated to indices of dynamic cerebral pressure-autoregulation extracted by means of transfer function analysis. Future studies using a prospective design are needed to validate the predictive value of the ARI index, as estimated by transfer function analysis, in relation to death and other unfavourable outcomes.