BACKGROUND: Transcranial Doppler sonography (TCD) assessment of intracranial blood flow velocity has been suggested to accurately determine intracranial pressure (ICP). OBJECTIVE: We attempted to validate this method in patients with communicating cerebrospinal fluid systems using predetermined pressure levels. METHODS: Ten patients underwent a lumbar infusion test, applying 4 to 5 preset ICP levels. On each level, the pulsatility index (PI) in the middle cerebral artery was determined by measuring the blood flow velocity using TCD. ICP was simultaneously measured with an intraparenchymal sensor. ICP and PI were compared using correlation analysis. For further understanding of the ICP-PI relationship, a mathematical model of the intracranial dynamics was simulated using a computer. RESULTS: The ICP-PI regression equation was based on data from 8 patients. For 2 patients, no audible Doppler signal was obtained. The equation was ICP = 23*PI + 14 (R = 0.22, P < .01, N = 35). The 95% confidence interval for a mean ICP of 20 mm Hg was -3.8 to 43.8 mm Hg. Individually, the regression coefficients varied from 42 to 90 and the offsets from -32 to +3. The mathematical simulations suggest that variations in vessel compliance, autoregulation, and arterial pressure have a serious effect on the ICP-PI relationship. CONCLUSIONS: The in vivo results show that PI is not a reliable predictor of ICP. Mathematical simulations indicate that this is caused by variations in physiological parameters.