Abstract Background Measurement of cerebral venous oxyhemoglobin saturation (ScvO2 ) is considered a gold standard in assessing the adequacy of tissue oxygen delivery (DO2 ) after the stage 1 palliation (S1P), with SvO2  <30% often representing severely compromised DO2 . Regional oxygenation index (rSO2 ) based on near-infrared resonance spectroscopy (NIRS) frequently is used to screen for compromised DO2 , although its sensitivity to detect severe abnormalities in SvO2 is uncertain. Methods ScvO2 was measured by co-oximetry from the internal jugular vein as clinically indicated in 73 neonates after S1P. These values were compared with cerebral rSO2 (FORE-SIGHT; CASMED) via mixed effects model linear regression, Bland-Altman analysis, and sensitivity analysis. Because NIRS devices measure a composite of arterial and venous blood, we calculated an rSO2 -based ScvO2 designed to remove arterial contamination from the rSO2 signal: rSO2 -based ScvO2  = (rSO2 − arterial oxygen saturation × 0.3)/0.7. Results Among 520 time-matched pairs of ScvO2 and cerebral rSO2 , the slope of the relationship between rSO2 and ScvO2 (after we adjusted for effects of hemoglobin) was 0.37 ± 0.04 with only modest correlation (r2  = 0.39), and mean bias of +8.26. When ScvO2 was <30%, cerebral rSO2 was <30 in less than 1%, <40 less than 1%, and <50 in 45.7% of data points; specificity of rSO2 in the same range is >99%. Correction of rSO2 for arterial contamination significantly decreased mean bias (+3.03) and improved the sensitivity of rSO2 to detect ScvO2  <30 to 6.5% for rSO2  <30, 29% for rSO2  <40, and 77.4% for rSO2  <50. Conclusions Cerebral rSO2 in isolation should not be used to detect low ScvO2 , because its sensitivity is low, although correction of rSO2 for arterial contamination may improve sensitivity. Cerebral rSO2 of 50 or greater should not be considered reassuring, though values below 30 are specific for low ScvO2.