Reliability of the Blood Flow Velocity Pulsatility Index for Assessment of Intracranial and Cerebral Perfusion Pressures in Head-Injured Patients

BACKGROUND: It has been postulated that the Gosling pulsatility index (PI) assessed with transcranial Doppler (TCD) has a diagnostic value for noninvasive estimation of intracranial pressure (ICP) and cerebral perfusion pressure (CPP). OBJECTIVE: To revisit this hypothesis with the use of a database of digitally stored signals from a cohort of head-injured patients. METHODS: We analyzed prospectively collected data of patients admitted to the Cambridge Neuroscience critical care unit who had continuous recordings of arterial blood pressure, ICP, and cerebral blood flow velocities (FVs) using TCD. PI was calculated (FVsys-FVdia)/FVmean over each recording session. Statistical analysis was performed using Spearman rank correlation, receiver-operator-characteristics methods, and modeling of a nonlinear PI-ICP/CPP graph. RESULTS: Seven hundred sixty-two recorded daily sessions from 290 patients were analyzed with a total recording time of 499.9 hours. The correlation between PI and ICP was 0.31 (P<.001) and for PI and CPP -0.41 (P<.001). The 95% prediction interval of ICP values for a given PI was more than ±15 mm Hg and for CPP more than ±25 mm Hg. The diagnostic value of PI to assess ICP area under the curve ranged from 0.62 (ICP>15 mm Hg) to 0.74 (ICP>35 mm Hg). For CPP, the area under the curve ranged from 0.68 (CPP<70 mm Hg) to 0.81 (CPP<50 mm Hg). Probability charts for elevated ICP/lowered CPP depending on PI were created. CONCLUSION: Overall, the value of TCD-PI to assess ICP and CPP noninvasively is very limited. However, extreme values of PI can still potentially be used in support of a decision for invasive ICP monitoring.