The limitations of near-infrared spectroscopy to assess cerebrovascular reactivity: The role of slow frequency oscillations

Background: A total hemoglobin reactivity index (THx) derived from near-infrared spectroscopy (NIRS) has recently been introduced to assess cerebrovascular reactivity noninvasively. Analogously to the pressure reactivity index (PRx), THx is calculated as correlation coefficient with arterial blood pressure (ABP). However, the reliability of THx in the injured brain is uncertain. Although slow oscillations have been described in NIRS signals, their significance for assessment of autoregulation remains unclear. In the current study, we investigated the role of slow oscillations of total hemoglobin for NIRS-based cerebrovascular reactivity monitoring. Methods: This study was based on a retrospective analysis of data that were consecutively recorded for a different project published previously. Thirty-seven patients with traumatic brain injury and admitted to Addenbrookes Neurosciences Critical Care Unit between June 2008 and June 2009 were included. After artifact removal, we performed spectral analysis of the tissue hemoglobin index (THI, a measure of oxy-and deoxygenated hemoglobin) and intracranial pressure (ICP) signal. PRx and THx were calculated as moving correlations between ICP and ABP, and THI and ABP, respectively. The agreement between PRx and THx as a function of normalized power of slow oscillations (0.015-0.055 Hz) contained in the input signals was assessed performing between-subject and within-subject correlation analyses. Furthermore, the correlation between the THx values derived from the right and left sides was analyzed. Results: The agreement between PRx and THx depended on the power of slow oscillations in the input signals. Between-subject comparisons revealed a significant correlation between THx and PRx (r = 0.80, 95% confidence interval 0.53-0.92, P < 0.01) for patients with normalized slow wave activity >0.4 in the THI signal, compared with r = 0.07 (95% confidence interval -0.40 to 0.51, P = 0.79) in the remaining files. Furthermore, within-subject comparisons suggested that THx may be used as a substitute for PRx only when there is an at least moderate agreement (r = 0.36) between the THx values derived from the right and left sides. Conclusions: Our results suggest that the NIRS-based cerebrovascular reactivity index THx can be used as a noninvasive substitute for PRx, but only during phases with sufficient slow wave power in the input signal. Furthermore, a good agreement between the THx measures on both sides seems to be a prerequisite for comparison of a global (PRx) versus the more local (THx) index. Nevertheless, noninvasive assessment of cerebrovascular reactivity may be desirable in patients without ICP monitoring and help to guide ABP management in these patients.