We investigated the in vitro performance of the CeVOX system for continuous monitoring of central venous oxygen saturation by spectrophotometry (Pulsion Medical Systems, Munich, Germany). Oxygen inflow into the system was varied, and oxygen saturation values measured by CeVOX were documented. Blood samples were simultaneously taken to assess oxygen saturation by co-oximetry, and values were compared by Bland-Altman analysis. Sixty-six data pairs were obtained at CeVOX and co-oximetry values of 16-99% and 5.5-100%, respectively. Overall, CeVOX values only slightly overestimated co-oximetry values (mean bias +2.4%), but limits of agreement (2 SD of bias) were wide (-11.8 to +16.6%). Saturation measured by CeVOX underestimated that measured by co-oximetry at higher oxygen concentrations and overestimated it at lower oxygen concentrations. There was a nearly linear correlation of the mean bias, suggesting a systematic error. We conclude that the current version of the CeVOX system does not reliably reflect oxygen saturation.