A noninvasive estimation of mixed venous oxygen saturation using near-infrared spectroscopy by cerebral oximetry in pediatric cardiac surgery patients

Background: Near-infrared spectroscopy (NIRS) is a noninvasive optical monitor of regional cerebral oxygen saturation (rSO(2)). The aim of this study was to validate the use of NIRS by cerebral oximetry in estimating invasively measured mixed venous oxygen saturation (SvO(2)) in pediatric postoperative cardiac surgery patients. Methods: Twenty patients were enrolled following cardiac surgery with intraoperative placement of a pulmonary artery (PA) or superior vena cava (SVC) catheter. Five patients underwent complete biventricular repair - complete atrioventricular canal (n = 3) and other (n = 2). Fifteen patients with functional single ventricle underwent palliative procedures - bidirectional Glenn (n = 11) and Fontan (n = 4). Cerebral rSO(2) was monitored via NIRS (INVOS 5100) during cardiac surgery and 6 h postoperatively. SVO2 was measured from blood samples obtained via an indwelling PA or SVC catheter and simultaneously correlated with rSO(2) by NIRS at five time periods: in the operating room after weaning from cardiopulmonary bypass, after sternal closure, and in the CICU at 2, 4, and 6 h after admission. Results: Each patient had five measurements (total = 100 comparisons). SVO2 obtained via an indwelling PA or SVC catheter for all patients correlated with rSO(2) obtained via NIRS: Pearson's correlation coefficient of 0.67 (P < 0.0001) and linear regression of r(2) = 0.45 (P < 0.0001). Separate linear regression of the complete biventricular repairs demonstrated an r = 0.71, r(2) = 0.50 (P < 0.0001). Bland-Altman analysis showed a bias of +3.3% with a precision of 16.6% for rSO(2) as a predictor Of SVO2 for all patients. Cerebral rSO(2) was a more accurate predictor Of SVO2 in the biventricular repair patients (bias -0.3, precision 11.8%), compared with the bidirectional Glenn and Fontan patients. Conclusions: Regional cerebral oximetry via NIRS correlates with SvO(2) obtained via invasive monitoring. However, the wide limits of agreement suggest that it may not be possible to predict absolute values Of SvO(2) for any given patient based solely on the noninvasive measurement of rSO(2). Near-infrared spectroscopy, using the INVOS 5100 cerebral oximeter, could potentially be used to indicate trends in SVO2, but more studies needs to be performed under varying clinical conditions.