Minimally invasive prediction of ScvO2 in high-risk surgery:The introduction of a model Index of Oxygenation

INTRODUCTION: The purpose of this study was to examine the trilateral relationship between cardiac index (CI), tissue oxygen saturation (StO2) and central venous oxygen saturation (ScvO2) and subsequently develop a model to predict ScvO2 on minimal invasive manner in patients undergoing major elective surgery.METHODS: The continuous data of thirty-three high-risk noncardiac surgical patients from a randomized controlled trial (NCT01342900) were analyzed on a between- and within-patient basis using regression on group means, random-effects Generalized Least Squares (GLS) regression. Trend concordance was assessed using a four-quandrant plot. We developed a model Index of Oxygenation (IO) to predict ScvO2 based on CI, StO2, heart rate (HR), fraction of inspired oxygen (FiO2) and mean arterial pressure (MAP). The dataset was split by randomizing each patient into an estimation or validation subsample. Randomization was stratified by type of surgery. IO was validated using random-effects GLS regression, a Bland-Altman analysis and four-quadrant concordance.RESULTS: The patients were monitored for an average duration on 227.5 minutes, giving 7509 observations in total. We could not establish a positive significant association between StO2 and CI using regression analyses (slope of -1.08 (p=0.15; 95%CI -2.54 to 0.377; within-R2 0.01; figure1.) Trending concordance is non-existent (on a 5-minute basis: 56%) or very weak (on a 60-minute basis: 73%; figure 2). CI was found to be strongly associated with ScvO2 on a within-patient basis, but lacks predictive power in explaining between-patient differences. StO2, in contrast, was significantly associated with between-patient ScvO2 differences, but does not follow (short term) within-patient ScvO2 variability (figure 3). The mean difference or bias between ScvO2 and IO is 1.07% (95% limits of agreement -14.7% to 16.9%). Concordance for 5-minute and 60-minute trends was 71% and 90%, respectively. The IO model was translated into a linear prediction, which was then scaled back to the mean (78.5) and SD (6.49) of ScvO2 in the estimation sample, according to the following formula (the brackets denote standardized variables): Figure 4 displays the complete intraoperative ScvO2 data and the IO prediction for two patients in the validation sample.CONCLUSIONS: StO2 cannot be considered a flow dependent variable during high-risk surgery. We hypothesize that StO2 is a gauge of microcirculatory functioning more than a measure of systemic oxygen balance. IO is a better estimator is ScvO2 than either CI or StO2 alone and could potentially be used for minimally invasive monitoring of systemic oxygenation.