Abstract
Background
The randomised goal-directed perfusion trial confirmed retrospective findings that a goal directed perfusion strategy to maintain oxygen delivery index (DO2i) during cardiopulmonary bypass >280 ml/min/m2 reduces the incidence of acute kidney injury (AKI). We developed a predictive model for AKI using data from the Australian New Zealand Collaborative Perfusion Registry to determine whether these findings could be validated in a real-world clinical setting and used to identify an optimal DO2i threshold for predictive diagnostic accuracy.
Methods
Data in 19,410 cardiopulmonary bypass procedures was randomly divided into training and validation datasets (n=9,705). Multivariate logistic regression was used to determine the best predictive models for AKI (RIFLE classification), incremental predictive value of minimum cardiopulmonary bypass DO2i and optimal threshold.
Results
Minimum DO2i was significantly associated with any AKI, Risk and Injury or greater class in both datasets (validation dataset; Any AKI OR 0.993, 95% CI=0.991-0.995, p<0.001; AKI-Risk OR 0.994, 95% CI=0.992-0.996, p<0.001, AKI-Injury or greater 0.993, 95% CI=0.991-0.996, p<0.001), representing on average a 7% increase in the likelihood of AKI for every 10 ml/min/m2 decrease in DO2i. Diagnostic accuracy was similar for both datasets with an optimal DO2i threshold of 270 ml/min/m2. The odds of any AKI were increased by 52% in those below the threshold (OR=1.52, 95% CI=1.29-1.77, p<0.001).
Conclusions
This study confirms previous findings that minimum DO2i during cardiopulmonary bypass is independently associated with AKI, supporting previous findings in a broader risk, multicentre cohort.