To estimate whether the association of transfusion and acute kidney injury (AKI) has a threshold of oxygen delivery below which transfusion is beneficial but above which it is harmful.

Retrospective study

Cardiovascular operating room and intensive care unit

Patients undergoing cardiac surgery with continuous oxygen delivery monitoring during cardiopulmonary bypass

None

Logistic regression was used to estimate the associations between oxygen delivery (mean, cumulative deficit, and bands of oxygen delivery), transfusion, and their interaction and AKI. A subgroup analysis of transfused and nontransfused patients with exact matching on cumulative oxygen deficit and time on bypass with adjustment for propensity to receive a transfusion using logistic regression. Nine hundred ninety-one of 4,203 patients developed AKI within 7 days. After adjustment for confounders, lower mean oxygen delivery (odds ratio [OR], 0.968; 95% confidence interval [CI], 0.949-0.988; p = 0.002) and transfusions (OR, 1.442; 95% CI, 1.077, 1.932; p = 0.014) were associated with increased odds of AKI by 7 days. As oxygen delivery decreased, the risk of AKI increased, with the slope of the OR steeper at <160 mL/m²/min. In the subgroup analysis, matched transfused patients were more likely than matched nontransfused patients to develop AKI (45% [n = 145] v 31% [n = 101]; p < 0.001). However, after propensity score adjustment, the difference was nonsignificant (OR, 1.181; 95% CI, 0.796-1.752; p = 0.406).

We found a nonlinear relationship between oxygen delivery and AKI. We found no level of oxygen delivery at which transfusion was associated with a decreased risk of AKI.