Continuous time-domain analysis of cerebrovascular autoregulation using near-infrared spectroscopy

Abstract

Background and purpose: Assessment of autoregulation in the time domain is a promising monitoring method for actively optimizating cerebral perfusion pressure (CPP) in critically ill patients. The ability to detect loss of autoregulatory vasoreactivity to spontaneous fluctuations in CPP was tested with a new time-domain method that used near-infrared spectroscopic measurements of tissue oxyhemoglobin saturation in an infant animal model.

Methods: Piglets were made progressively hypotensive over 4 to 5 hours by inflation of a balloon catheter in the inferior vena cava, and the breakpoint of autoregulation was determined using laser-Doppler flowmetry. The cerebral oximetry index (COx) was determined as a moving linear correlation coefficient between CPP and INVOS cerebral oximeter waveforms during 300-second periods. A laser-Doppler derived time-domain analysis of spontaneous autoregulation with the same parameters (LDx) was also determined.

Results: An increase in the correlation coefficient between cerebral oximetry values and dynamic CPP fluctuations, indicative of a pressure-passive relationship, occurred when CPP was below the steady state autoregulatory breakpoint. This COx had 92% sensitivity (73% to 99%) and 63% specificity (48% to 76%) for detecting loss of autoregulation attributable to hypotension when COx was above a threshold of 0.36. The area under the receiver-operator characteristics curve for the COx was 0.89. COx correlated with LDx when values were sorted and averaged according to the CPP at which they were obtained (r=0.67).

Conclusions: The COx is sensitive for loss of autoregulation attributable to hypotension and is a promising monitoring tool for determining optimal CPP for patients with acute brain injury.

Figures

Figure 1

Time trends of recordings from a single piglet. ICP, ABP, and CPP are shown in mm Hg; laser-Doppler red blood cell flux is in arbitrary units; and cerebral oximetry (NIRS) is expressed as a percent saturation of hemoglobin. Time on the x axis covers a spread of 4 hours and 10 minutes. Slow “B” waves of ICP are seen in the top tracing at low ABP before failure of autoregulation (solid arrow). The oximeter readout showed a more gradual decline relative to the laser-Doppler flux, which had a pattern more indicative of autoregulation (dashed arrows). A similar trend was observed in all 6 piglets.

Figure 2

A, Steady-state autoregulatory graph of laser-Doppler flux vs CPP in a single piglet. The breakpoint was defined as the division that resulted in regression lines with the lowest combined residual squared error (34 mm Hg in this piglet). B, Near-infrared spectroscopy (NIRS)-derived cerebral oximetry vs CPP. This relationship did not have the obvious plateau seen with laser-Doppler flux. However, the laser-Doppler index (LDx, ±SE, C) and the cerebral oximetry index (COx, D) were concordant, showing low values above a CPP of 35 mm Hg and high values below a CPP of 35 mm Hg (arrows).

Figure 3

Static autoregulation curves derived from 6 piglets (±SE). A, Laser-Doppler flux as a percent of baseline flux at 60 mm Hg. B, Cerebrovascular resistance (CVR), calculated as CPP/CBF from the same data set and expressed as a percentage of CVR at CPP of 60 mm Hg. C, Cerebral oximetry, measured by NIRS, shown as a percentage of baseline tissue oxyhemoglobin saturation. P<0.0001 by ANOVA for both laser-Doppler flux and oximetry curves. The average breakpoint of autoregulation, determined for individual piglets, was 29.7±5.5 mm Hg (vertical dashed line).

Figure 4

Average LDx (A) and COx (B) for the 6 piglets (±SE) stratified by the CPP at which they were measured. The horizontal dashed line shows the 90% sensitivity cutoff for detecting autoregulatory failure. The receiver-operator characteristics are compared between the LDx (C) and COx (D) calculations of 6 piglets, averaged for each 5 mm Hg increment of CPP. AUC is area under the curve. Confidence intervals for sensitivity and specificity and likelihood ratios are tabulated for different sensitivity levels for each index (E, F).

Figure 5

Linear regression (A, B) and Bland Altman plots (C, D) comparing LDx and COx for all data points (A, C) and averaged data points taken at the same CPP for each piglet (B, D). Agreement improves substantially by averaging, which implies a low signal-to-noise ratio for individual index measurements. Dashed lines are 95% CI (regression) and 95% limits of agreement (Bland-Altman).