Nitrous oxide anesthesia has been implicated as contributing to the development of delayed tension pneumocephalus following surgery performed in the sitting position. The authors tested the hypothesis that withdrawal of nitrous oxide anesthesia administered during formation of an intracranial gas cavity would lead to a decrease in intracranial pressure (ICP) as N2O diffuses from the cavity back into the blood. Ten halothane-anesthetized rabbits were prepared for measurement of supracortical ICP and arterial blood pressure (BP) and for intracranial volume alterations via a cisterna magna infusion catheter. Hyperventilation (Paco2 = 28-30 mmHg) and mannitol were used to shrink the brain to accommodate intracranial infusion of either air or lactated Ringer's (LR) solution, which was used to elevate ICP to between 10-15 mmHg from a baseline ICP of 2.1 +/- 2.5 mmHg over a period of 8 to 10 min. Following stabilization at an elevated ICP, inhalation of nitrous oxide (75%) was either initiated or withdrawn (if already present during the induced ICP increase) and the subsequent changes in mean ICP and BP were recorded. Following ICP elevation with LR to 10 +/- 1 mmHg, initiation of 75% N2O administration resulted in no change in ICP and modest increases (P less than 0.05) in BP and cerebral perfusion pressure (CPP = BP - ICP) after 4 min. However, when ICP was raised (to 12 +/- 3.5 mmHg) with intracranial air infusion, subsequent initiation of 75% N2O inhalation caused an abrupt ICP increase to 22.3 +/- 9 mmHg (from control P less than 0.001).(ABSTRACT TRUNCATED AT 250 WORDS)