Temperature-adjusted perfusion-metabolic phenotyping during cardiopulmonary bypass: An exploratory study

Introduction Goal-directed perfusion during cardiopulmonary bypass (CPB) has traditionally focused on indexed oxygen delivery (DO 2 i), yet metabolic behavior during aortic cross-clamping likely reflects a more complex interaction among oxygen delivery, oxygen extraction (O 2 ER), and temperature-dependent oxygen demand. We aimed to identify temperature-adjusted perfusion-metabolic phenotypes during cross-clamping and to explore their association with early postoperative outcomes. Methods In this retrospective single-center study, perfusion data (restricted to the aortic cross-clamp period) were linked with clinical data at the case level. Variables’ trajectories were normalized in time, resampled at 15 equidistant points, and represented using DO 2 i, O 2 ER, and temperature-adjusted indexed oxygen consumption and carbon dioxide production corrected to 37°C with a Q10-based approach (VO 2 i_37eq and VCO 2 i_37eq). Principal component analysis and k-means clustering were applied. Logistic regression was used to assess the association between cluster membership and postoperative acute kidney injury (AKI) according to KDIGO criteria. Results Seventy patients were included (mean age 64.3 years, 35.7% females). A 3-cluster solution identified a low-extraction/low-metabolic-demand phenotype, a high-delivery/compensated-metabolic-demand phenotype, and a high-extraction/high-metabolic-stress phenotype. The latter showed the highest incidence of AKI of any stage (53.1%, p = 0.034), the highest incidence of stage 2–3 AKI (18.8%), greater postoperative creatinine increase, and longer intensive care unit stay. In multivariable logistic regression, this phenotype remained associated with postoperative AKI after exploratory multivariable adjustment. Conclusions Temperature-adjusted unsupervised analysis of perfusion trajectories identified three physiologically coherent phenotypes. The high-extraction/high-metabolic-stress phenotype was associated with the least favorable early postoperative renal profile, supporting a GDP framework beyond DO 2 i alone.