Lactate may carry information about kidney risk beyond its familiar role as a marker of metabolic stress. In a new study, Xue and colleagues combined retrospective clinical prediction models with rat cardiopulmonary bypass (CPB) and renal-cell experiments to investigate the relationship between lactate, protein lactylation, and acute kidney injury (AKI).
The clinical analysis identified lactate as an important predictor within a multivariable model. Experimental inhibition of lactate production reduced several measures of kidney injury. These are complementary findings, but they do not establish that pharmacologically lowering lactate prevents AKI in patients.
Study Design
The investigators analyzed 6,558 records across two MIMIC cohorts: 3,613 patients in MIMIC-IV and 2,945 in MIMIC-III. The MIMIC-IV cohort was divided into training and internal testing sets of 2,529 and 1,084 patients. MIMIC-III provided a separate validation cohort, with overlapping patients excluded according to the methods.
Although the summary refers to multicenter databases, these MIMIC cohorts originate from the same hospital. The validation therefore provides evidence across distinct historical cohorts rather than independent validation across different institutions.
Nine variables were selected for modeling: age, body mass index, lactate, SAPS II, potassium, serum creatinine, surgery type, blood urea nitrogen, and heart failure. Laboratory and clinical predictors were described as the latest available measurements within 24 hours before CPB. Eight machine-learning approaches were compared.
Key Clinical Findings
| Measure | Reported finding |
|---|---|
| MIMIC-IV cohort | 3,613 patients; AKI in 74.20% |
| MIMIC-III validation cohort | 2,945 patients; AKI in 91.82% |
| Best-performing model | Gradient boosting machine |
| Internal testing AUC | 0.898 (95% CI 0.876–0.919) |
| Validation AUC | 0.798 (95% CI 0.757–0.838) |
| Internal sensitivity / specificity | 81.7% / 88.4% |
These performance estimates apply to the complete multivariable model, not lactate alone. AKI rates describe the selected study cohorts and should not be treated as expected rates for all cardiac surgical patients.
SHAP analysis placed lactate among the leading contributors to model predictions, alongside body mass index, potassium, blood urea nitrogen, and surgery type. Feature importance explains how the model uses its inputs; it does not by itself prove that a feature causes the outcome.
The lower AUC in the validation cohort is clinically relevant. A model that discriminates well within its development setting may perform less well in a different population or period. Prospective testing and local calibration would be needed before clinical implementation.
What the Experimental Work Adds
Lactate accumulation and protein lactylation
In rats, CPB was followed by increased circulating and renal-tissue lactate, biochemical and histological evidence of kidney injury, and higher global lysine lactylation. Lactylation is a chemical modification of proteins associated with lactate metabolism; it is distinct from the blood lactate concentration routinely measured during perfusion.
The investigators also studied human proximal tubular epithelial cells exposed to oxygen-glucose deprivation and reoxygenation. Together with transcriptomic analyses, these experiments implicated altered glycolysis and lactate dehydrogenase A (LDHA) in the injury response.
Experimental inhibition of lactate production
Rat experiments used oxamate and 2-deoxy-D-glucose to inhibit aspects of lactate production and glycolytic metabolism. Treatment attenuated several injury-related findings, including renal dysfunction, oxidative stress, inflammatory signaling, and global protein lactylation. Cellular LDHA silencing provided an additional mechanistic approach.
These interventions remain experimental in this context. Systemic metabolic effects and off-target actions mean that improvement cannot be attributed solely to a kidney-specific lactylation pathway. The study did not test these drugs as AKI-prevention treatments in cardiac surgical patients.
Relevance to Perfusion Practice
The study offers a reason to view perioperative lactate as part of a broader assessment of renal vulnerability and metabolic stress. It also raises a research question: whether lactate-related cellular changes contribute to injury, rather than simply accompanying it.
For clinical interpretation, three distinctions matter:
- Prediction is not causation: an important model feature is not automatically a treatment target.
- A multivariable model is not a lactate cutoff: the reported AUC values do not validate a standalone lactate threshold.
- Experimental protection is not demonstrated patient benefit: animal and cell results cannot establish clinical efficacy or safety.
This paper does not establish a new pump-flow target, oxygen-delivery threshold, lactate treatment threshold, or renal-protection drug protocol.
Limitations and Evidence Assessment
The retrospective clinical analysis remains vulnerable to confounding and selection effects. The unusually high AKI prevalence, particularly in the validation cohort, limits direct generalization to routine cardiac surgical populations. Important intraoperative variables—including detailed CPB duration, cross-clamp time, and blood-pressure fluctuations—were unavailable.
The experimental findings strengthen biological plausibility but do not identify every relevant lactylated protein or establish a single causal pathway. Systemic inhibition cannot fully separate reduced circulating lactate exposure from reduced local kidney production. Prospective multicenter validation and further mechanistic work are needed before translating these results into treatment.
Source and Attribution
This original IPA educational summary is based on the open-access article by Xue and colleagues, distributed under the Creative Commons Attribution 4.0 license. Findings have been summarized and their presentation adapted. The cover is a conceptual illustration, not an anatomical or treatment guide.
Original Article
Xue D, Zhang M, Chang X, Wang T, Ji B, Xi C, Wang G. Lactate as a predictive biomarker and therapeutic target in cardiopulmonary bypass-associated acute kidney injury. iScience. 2026;29(11):117643. Issue date: November 20, 2026. doi:10.1016/j.isci.2026.117643.
