Model could replace invasive intracranial pressure monitoring
MedWire News: Researchers have developed a model that uses minimally invasive measurements to give near-realtime estimation of intracranial pressure (ICP) in patients with traumatic brain injury.
The model has "clinically acceptable accuracy" relative to invasive parenchymal ICP measurement, and "performs better than the invasive epidural and subdural measurements that are still used in current clinical practice," say Thomas Heldt (Massachusetts Institute of Technology, Cambridge, USA) and team.
The model calculates ICP from two routine measurements: peripheral arterial blood pressure (ABP), via minimally invasive radial artery catheterization, and from middle cerebral artery (MCA) blood flow velocity, measured noninvasively using transcranial Doppler.
The researchers validated the model by examining 45 sets of data from 37 comatose patients who had measurements of ABP, MCA flow velocity, and ICP. Overall, the model estimated ICP with a mean error of 1.6 mmHg and a standard deviation of error (SDE) of 7.6 mmHg.
When the analysis was limited to patients who had bilateral MCA flow velocity recordings available, the mean error fell to 1.5 mmHg and the SDE to 5.9 mmHg.
These measurements were obtained without the need for calibrating the model on specific patients or against a reference population, the team notes in Science Translational Medicine.
Using individual patient data, Heldt et al show that the model performs particularly well in three specific clinical situations: in a "plateau wave" when ICP rises sharply, remains at a plateau, and then falls; in severe progressive intracranial hypertension; and when ICP is close to the normal range.
The model was less accurate in the case of two successive plateau waves. In this instance, it tracked the invasive recordings closely during the first wave, but less so during the second wave, although it did capture its amplitude and duration.
In an accompanying commentary, Brahm Goldstein (Ikaria, Hampton, New Jersey, USA) and colleagues say: "It would be helpful to know the conditions or factors that could be used to determine a priori whether a given case will yield low error or high error." But they note that this "will not be an easy task."
However, they comment that absolute accuracy is less important than the ability to detect "large, rapid changes (such as ICP threshold values), which could then be used to alert physicians to the patient's need for immediate treatment or surgery."
They say: "The model developed by Kashif and colleagues may be able to estimate these rapid fluctuations and appears to be capable of quasi-real-time output with a lag time of less than 1 min."
By Eleanor McDermid