BACKGROUND: Hyperventilation may cause brain ischaemia after traumatic brain injury. However, moderate reductions in PaCO(2) are still an option in the management of raised intracranial pressure (ICP) under some circumstances. Being able to predict the ICP-response to such an intervention would be advantageous. We investigated the ability of pre-hyperventilation ICP and cerebrospinal compensatory reserve to predict the reduction in ICP achievable with moderate hyperventilation in head injured patients. METHODS: Thirty head injured patients requiring sedation and mechanical ventilation were investigated. ICP was monitored via an intraparenchymal probe and intracranial cerebrospinal compensatory reserve was assessed using an index (R(ap)) based on the relationship between mean ICP and its pulse amplitude. Measurements were made at a constant level of PaCO(2) during a 20-minute baseline period. The patients were then subjected to an acute decrease in PaCO(2) of approximately 1 kPa and, after an equilibration period of 10 minutes, measurements were again made at a constant level of PaCO(2) for a further 20 minutes. A multiple linear regression model, incorporating baseline PaCO(2), ICP, and R(ap) was used to identify the relevant predictors of ICP reduction. FINDINGS: Baseline ICP and R(ap) were both significant predictors of ICP-reduction (p=0.02 and 0.001 respectively) with R(ap) being the more powerful parameter. CONCLUSIONS: A model based on cerebrospinal compensatory reserve and ICP can predict the achievable ICP-reduction and may potentially be used to optimise patient selection and intensity of hyperventilation.