Background
Feasible estimations of perioperative changes in oxygen consumption (VO2) could enable larger studies of its role in postoperative outcomes. Current methods, either by reverse Fick calculations using pulmonary artery catheterisation or metabolic by breathing gas analysis, are often deemed too invasive or technically requiring. In addition, reverse Fick calculations report generally lower values of oxygen consumption.
Methods
We investigated the relationship between perioperative estimations of VO2 (EVO2), from LiDCO™plus-derived (LiDCO Ltd, Cambridge, UK) cardiac output and arterial-central venous oxygen content difference (Ca-cvO2), with indirect calorimetry (GVO2) by QuarkRMR (COSMED srl. Italy), using data collected 2017–2018 during a prospective observational study on perioperative oxygen transport in 20 patients >65 years during epidural and general anaesthesia for open pancreatic or liver resection surgery. Eighty-five simultaneous intra- and postoperative measurements at different perioperative stages were analysed for prediction, parallelity and by traditional agreement assessment.
Results
Unadjusted bias between GVO2 and EVO2 indexed for body surface area was 26 (95% CI 20 to 32) with limits of agreement (1.96SD) of -32 to 85 ml min−1m−2. Correlation adjusted for the bias was moderate, intraclass coefficient(A,1) 0.51(95% CI 0.34 to 0.65) [F (84,84) = 3.07, P<0.001]. There was an overall association between GVO2 and EVO2, in a random coefficient model [GVO2 = 73(95% CI 62 to 83) + 0.45(95% CI 0.29 to 0.61) EVO2 ml min−1m−2, P<0.0001]. GVO2 and EVO2 changed in parallel intra- and postoperatively when normalised to their respective overall means.
Conclusion
Based on this data, estimations from LiDCO™plus-derived cardiac output and Ca-cvO2 are not reliable as a surrogate for perioperative VO2. Results were in line with previous studies comparing Fick-based and metabolic measurements but limited by variability of data and possible underpowering. The parallelity at different perioperative stages and the prediction model can provide useful guidance and methodological tools for future studies on similar methods in larger samples.