Millions worldwide require Renal Replacement Therapy (RRT) as a treatment essential for survival. However, optimizing RRT strategies via AI is challenging due to heterogeneous patient dynamics, missing data, and the absence of an AI-oriented health assessment criterion. We propose an AI-Oriented Comprehensive Normalized Assessment (CNA) for healthy status and apply it to optimize RRT strategies by using offline reinforcement learning (RL). The key idea of CNA is transforming vital-sign distributions into a standard normal space, enabling a unified, data-driven health-status score defined by deviations from referent intervals, which also provides an AI-oriented criterion to assess strategy quality and supports RL termination. We further design a structured 23-dimensional state representation that integrates 19 indicators with 4 RRT descriptors, and employ matrix decomposition to reconstruct missing vital signs, improving data completeness for learning. These components are incorporated into multiple offline RL algorithms and validated via systematic ablation studies on RRT feature subsets. Compared with physicians’ observed treatments, the best learned strategy reduces mortality from 13.2% to 5.0% (reducing 62.24%) and shortens average in-hospital stay from 308.5 to 250.1 h (reducing 18.93%), demonstrating both methodological innovation and the potential of CNA-guided RL to improve RRT outcomes in nephrology.

Sepsis remains a major global health challenge due to its high mortality rate and the difficulty of predicting its onset from nonspecific early symptoms. To address this, we explored the feasibility of using fine-tuned small-scale large language models (LLMs) to perform accurate sepsis prediction from structured clinical time-series data. We developed a Semantic Abstraction Rule Engine (SARE) that transforms numerical clinical data into clinician-readable textual descriptions through temporal feature extraction and semantic transformation. Using these representations, four lightweight LLMs were fine-tuned with Low-Rank Adaptation and evaluated on publicly available datasets. The fine-tuned models consistently outperformed traditional machine learning and temporal prediction methods, with notable improvements in sensitivity and area under the curve. The Gemma-2-9B model achieved an AUC of 0.9307 and a sensitivity of 0.8532 on the PhysioNet/Computing in Cardiology Challenge 2019 dataset, surpassing existing state-of-the-art approaches. These findings demonstrate that lightweight LLMs can provide accurate and computationally efficient predictions, offering practical support for early sepsis detection in resource-limited healthcare environments.

Background The JAK/STAT pathway plays a pivotal role in hepatic ischaemia/reperfusion (I/R) injury, a serious perioperative complication. Although signal transducers and activators of transcription 1 (STAT1) activation is known to drive I/R-induced injury, the specific post-translational modifications (PTMs) governing its activity in hepatic I/R remain poorly understood.Objective This study identifies SMYD2, SET and MYND domain Containing 2 (SMYD2) as a critical regulator of STAT1 and investigates the mechanistic basis of SMYD2-mediated PTMs in modulating STAT1 function during hepatic I/R.Design Using an integrated transcriptomic-proteomic approach and functional screening, we identified SMYD2 as a critical regulator of STAT1 activation in hepatic I/R injury. Clinical correlations linked SMYD2 expression to postoperative liver function, while loss-of-function and gain-of-function studies in vitro and in vivo validated its mechanistic role.Results Our findings demonstrate that SMYD2 modulates hepatic I/R injury through the JAK-STAT1 pathway. Clinically, elevated SMYD2 expression correlated with improved liver function and better surgical outcomes following hepatectomy. Mechanistic studies revealed that SMYD2 physically interacts with STAT1 and mediates its methylation at lysine 175 (K175), thereby inhibiting STAT1 phosphorylation and nuclear translocation. Both in vitro and in vivo studies demonstrated that SMYD2 overexpression alleviated hepatic I/R injury, whereas its genetic depletion or pharmacological inhibition exacerbated the damage.Conclusion This study establishes SMYD2 as a novel negative regulator of STAT1 activity through K175 methylation, providing new insights into the epigenetic control of STAT1 during hepatic I/R injury. Our findings reveal a previously unrecognised mechanism for fine-tuning STAT1 signalling in hepatic I/R injury, and targeting the SMYD2-STAT1 axis may present a promising therapeutic strategy for mitigating I/R-associated liver damage.

Privacy-sensitive and distributed characteristics of multi-center medical data bring severe obstacles to centralized modeling for accurate early prediction of sepsis. Federated learning (FL) has attracted growing attention as a promising framework for collaborative model development, as it allows multiple institutions to jointly train predictive models without directly sharing or centralizing raw data. Nevertheless, its practical performance, robustness, and privacy-preserving benefits remain insufficiently evaluated using real-world clinical datasets. To bridge this gap, this study systematically examines the application of federated learning to multi-center sepsis prediction. The experimental dataset consists of 648 clinically screened samples collected from three tertiary hospitals in China, with rigorous inclusion and exclusion criteria. We establish a centralized training paradigm as the performance baseline, and then implement a horizontal federated learning framework for distributed collaborative modeling. Extensive experimental results demonstrate that the federated learning-based model achieves highly comparable prediction accuracy to the centralized counterpart, while fundamentally avoiding privacy leakage. Further privacy security analysis verifies that malicious attackers cannot reconstruct the original patient data from the transmitted model parameters, indicating strong resistance against data reconstruction attacks. This work not only validates the practicality and security of federated learning in clinical sepsis prediction, but also provides a reliable and feasible solution for privacy-preserving multi-center medical collaboration.

Background Postoperative agitation (EA) is a common complication in pediatric patients, and its early identification is crucial for improving perioperative safety. This study aims to identify the risk factors for EA and develop an interpretable machine learning model.Methods This multicenter retrospective study included 445 pediatric patients. Data from 321 patients from one center were used for model development, and 124 patients from another center were selected as an independent validation set. The development dataset was randomly divided into training and validation sets in a 8:2 ratio. Feature selection was performed using the Least Absolute Shrinkage and Selection Operator (LASSO) regression. Six machine learning algorithms were used to build the prediction model: Logistic Regression (LR), Support Vector Machine (SVM), Multilayer Perceptron (MLP), Random Forest (RF), Extreme Gradient Boosting (XGB), and Light Gradient Boosting Machine (LGBM). Model performance was evaluated based on the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, Brier, and F1 score. The interpretability of the model was analyzed using the SHapley Additive Explanations (SHAP) method.Result The incidence of EA in the development cohort and external validation cohort was 29.5% and 25.8%, respectively. Following feature selection, five clinically relevant predictive variables were identified: parental education level, ALT level, postoperative analgesic pump use, antagonist administration, and the number of suctioning maneuvers during extubation.In internal validation, the support vector machine (SVM) model achieved the best performance, with an AUC of 0.918 (95% confidence interval [CI] 0.844-0.973). In external validation, the MLP performed optimally, with an AUC of 0.705 (95% CI 0.590-0.804), accuracy of 0.718, sensitivity of 0.645, specificity of 0.780, F1 score of 0.571, and Brier score of 0.190.Given that external validation represents the gold standard for assessing model generalizability, MLP was chosen as the candidate model for clinical application. Using the optimal SVM model derived from internal validation, SHAP analysis demonstrated that shorter recovery time, analgesic pump use, higher parental education level, elevated ALT, no antagonist use, and absence of suctioning during extubation were significant risk factors for EA.Notably, recovery time-an intraoperative indicator-was excluded from the final clinical model to ensure that all predictive variables could be obtained prior to emergence from anesthesia.Conclusion This study developed and externally validated an interpretable machine learning model for predicting the risk of emergence agitation (EA) in children undergoing elective surgery, incorporating five preoperative or postoperative readily available clinical variables: parental education level, ALT level, postoperative analgesic pump use, antagonist administration, and the number of suctioning maneuvers during extubation. The model exhibited moderate discriminatory performance, and SHAP analysis further clarified the contribution and underlying mechanisms of key risk factors. This model may serve as a preliminary decision-support tool for individualized risk stratification of pediatric EA. Nevertheless, future multicenter prospective studies are warranted to validate its generalizability and clinical utility prior to routine implementation.

The conventional common bile duct ligation (CBDL) model suffers from high postoperative mortality (frequently exceeding 50%), which severely limits longitudinal monitoring of disease progression and complications such as hepatopulmonary syndrome (HPS). We developed an improved CBDL (MCBDL) to preserve the continuity of bile duct, and compared MCBDL (n=40) with traditional CBDL (n=40) and sham operation (n=10) within 42 days. MCBDL captures cirrhosis and extrahepatic organ injury while significantly improving survival (72.5% vs. 45%, P=0.003) and attenuating systemic inflammation (IL-6: 117 ± 11 vs. 153 ± 26 pg/mL, P=0.047). Scanning electron microscopy revealed persistent secretory activity in ligated epithelia of both groups, but only CBDL permitted leakage from isolated stumps, creating a pathogenic cycle driving mortality. By preserving ductal continuity, MCBDL is associated with significantly reduced mortality while maintaining the hallmark features of cholestatic cirrhosis (METAVIR F4) and HPS-related extrahepatic complications, providing a stable platform for mechanistic studies.

In the intensive care setting, sepsis continues to be a major contributor to patient illness and death; however, its timely detection is hindered by the complex, sparse, and heterogeneous nature of electronic health record (EHR) data. We propose Triplet-GCN, a single-branch graph convolutional model that represents each encounter as patient–feature–value triplets, constructs a bipartite EHR graph, and learns patient embeddings via a Graph Convolutional Network (GCN) followed by a lightweight multilayer perceptron (MLP). The pipeline applies type-specific preprocessing – median imputation and standardization for numeric variables, effect coding for binary features, and mode imputation with low-dimensional embeddings for rare categorical attributes – and initializes patient nodes with summary statistics, while retaining measurement values on edges to preserve "who measured what and by how much". In a retrospective, multi-center Chinese cohort (N = 648; 70/30 train–test split) drawn from three tertiary hospitals, Triplet-GCN consistently outperforms strong tabular baselines (KNN, SVM, XGBoost, Random Forest) across discrimination and balanced error metrics, yielding a more favorable sensitivity–specificity trade-off and improved overall utility for early warning. These findings indicate that encoding EHR as triplets and propagating information over a patient–feature graph produce more informative patient representations than feature-independent models, offering a simple, end-to-end blueprint for deployable sepsis risk stratification.

Hepatopulmonary syndrome (HPS) is a condition characterized by pulmonary angiogenesis and refractory hypoxemia, often seen in patients with chronic liver disease. Its unclear mechanism means that liver transplantation is the only effective therapy. Agrimoniin, a compound from Pilosa ledeb, shows potential in protecting against liver cirrhosis via anti-angiogenic and anti-glycolytic effects. This study investigates agrimoniin as a potential integrated therapy for HPS-related liver and lung dysfunction. Using transcriptome data and an ICU cohort, we analyzed the role of glycolysis in chronic liver disease progression. HPS rats were established via common bile duct ligation, and serum metabolites were measured. The oxygen consumption rate and extracellular acidification rate were also detected. Rats were treated with agrimoniin (3 mg/kg/day or 8 mg/kg/day) at the early stage of HPS. Our results showed that imbalanced oxidative phosphorylation and glycolysis correlated with chronic liver disease progression and poorer outcomes. Decreased oxygen consumption rate and increased extracellular acidification rate, as well as increased glycolysis, were observed in the HPS group. Agrimoniin treatment improved liver and lung function by inhibiting pathological angiogenesis and glycolysis. Through TCM suite analysis, molecular docking, and dynamics simulations, PGC-1α was identified as a potential target of agrimoniin. Inhibiting PGC-1α blocked agrimoniin's benefits on angiogenesis and glycolysis flux. Thus, agrimoniin may be a potential integrated therapy for HPS by activating PGC-1α to inhibit glycolysis and angiogenesis.

Background This study aimed to investigate the therapeutic potential of a combined low-dose propofol (PPF) and salvianolic acid A (SAA) regimen in mitigating lipopolysaccharide (LPS)-induced cardiac dysfunction and ferroptosis in diabetic contexts, and to explore the role of the sirtuin 1 (SIRT1)/forkhead box O1 (FoxO1) signaling pathway.Methods Type 2 diabetes (DM) was induced in mice, followed by LPS administration to induce cardiac injury. The mice were randomly assigned to six groups: control, DM, control + LPS, DM + LPS, DM + LPS + high-dose PPF, and DM + LPS + low-dose PPF + SAA. Cardiac function was assessed via echocardiography, while ferroptosis was evaluated with BODIPY staining and transmission electron microscopy. In vitro, H9c2 cardiomyocytes were treated with high glucose and LPS. Ferroptosis was assessed with FerroOrange and JC-1 staining. Oxidative stress and inflammatory cytokines were measured using ELISA or flow cytometry, and protein expression of key markers was analyzed.Results Diabetes aggravated LPS-induced cardiac injury in mice evidenced as impaired myocardial function, which was concomitant with decreased cardiac expression of SIRT1, FoxO1, and GPX4 proteins; increased production of oxidative stress, pro-inflammatory cytokines, oxidized lipids, and damaged mitochondrial cristae as compared to NC + LPS group. The combined use of PPF and SAA enhanced cardiac SIRT1 and FoxO1 and ameliorated LPS-mediated cardiac dysfunction in diabetic mice, but these beneficial effects were abolished by inhibition of SIRT1 or FoxO1. In vitro, hyperglycemia and LPS-induced cellular injuries were ameliorated by PPF and SAA, respectively. Low dose of PPF in combination with SAA reduced the production of ROS and ferroptosis that were concomitant with increased SIRT1 and FoxO1 expression, effects that were either comparable or superior to those achieved with high-dose PPF alone, but these protective effects were reversed by silencing SIRT1 or FoxO1.Conclusion The combination of low-dose PPF and SAA effectively protects against LPS-induced cardiac dysfunction and ferroptosis in diabetic conditions by activating the SIRT1/FoxO1 pathway.

Hepatopulmonary syndrome (HPS) is a condition characterized by pulmonary angiogenesis and refractory hypoxemia, often seen in patients with chronic liver disease. Its unclear mechanism means that liver transplantation is the only effective therapy. Agrimoniin, a compound from Pilosa ledeb, shows potential in protecting against liver cirrhosis via anti-angiogenic and anti-glycolytic effects. This study investigates agrimoniin as a potential integrated therapy for HPS-related liver and lung dysfunction. Using transcriptome data and an ICU cohort, we analyzed the role of glycolysis in chronic liver disease progression. HPS rats were established via common bile duct ligation, and serum metabolites were measured. The oxygen consumption rate and extracellular acidification rate were also detected. Rats were treated with agrimoniin (3 mg/kg/day or 8 mg/kg/day) at the early stage of HPS. Our results showed that imbalanced oxidative phosphorylation and glycolysis correlated with chronic liver disease progression and poorer outcomes. Decreased oxygen consumption rate and increased extracellular acidification rate, as well as increased glycolysis, were observed in the HPS group. Agrimoniin treatment improved liver and lung function by inhibiting pathological angiogenesis and glycolysis. Through TCM suite analysis, molecular docking, and dynamics simulations, PGC-1u03B1 was identified as a potential target of agrimoniin. Inhibiting PGC-1u03B1 blocked agrimoniinu2019s benefits on angiogenesis and glycolysis flux. Thus, agrimoniin may be a potential integrated therapy for HPS by activating PGC-1u03B1 to inhibit glycolysis and angiogenesis.

With the continuous advancement of medical technologies, perioperative anesthesia management decisions are confronted with challenges of complexity and dynamicity, rendering traditional machine learning models insufficient for clinical needs. As one of the breakthrough applications in artificial intelligence, large language models (LLMs) may offer better help for intelligent anesthesia management. LLMs are capable of processing multidimensional and multi-source data, enabling more comprehensive prediction and intervention suggestions, thereby optimizing anesthesia management processes. This review summarizes the current applications of LLMs in anesthesiology and proposes methods for building specialized LLMs to tackle challenges in application. This work albeit mainly in the proposed stage aims to provide references for future development in this field and to promote in-depth research of LLMs in anesthesiology. Graphical Abstract

Objective: Sepsis is a life-threatening condition caused by severe infection leading to acute organ dysfunction. This study proposes a data-driven metric and a continuous reward function to optimize personalized heparin therapy in surgical sepsis patients. Methods: Data from the MIMIC-IV v1.0 and eICU v2.0 databases were used for model development and evaluation. The training cohort consisted of abdominal surgery patients receiving unfractionated heparin (UFH) after postoperative sepsis onset. We introduce a new RL-based framework: converting the discrete SOFA score to a continuous cxSOFA for more nuanced state and reward functions; Second, defining "good" or "bad" strategies based on cxSOFA by a stepwise manner; Third, proposing a Treatment Effect Comparison Matrix (TECM), analogous to a confusion matrix for classification tasks, to evaluate the treatment strategies. We applied different RL algorithms, Q-Learning, DQN, DDQN, BCQ and CQL to optimize the treatment and comprehensively evaluated the framework. Results: Among the AI-derived strategies, the cxSOFA-CQL model achieved the best performance, reducing mortality from 1.83

Sepsis is a common critical illness in intensive care medicine, affecting millions of patients globally each year. It has a high mortality rate and is one of the leading causes of death in intensive care units (ICUs). Despite significant advancements in sepsis research through artificial intelligence technologies in recent years that have reduced mortality rates, the current mortality rate in patients with sepsis remains as high as 25

Background There was no standardized clinical guideline of the application of preoperative anti- Hepatitis B Virus therapy in hepatitis B virus (HBV)-related hepatocellular carcinoma (HCC) patients receiving hepatectomy. This study explored HBV infectious impact on postoperative outcomes of HCC patients. Methods This retrospective study included 962 HCC patients who received hepatectomy from the Chronic Liver Diseases Perioperative Database (n = 360,767). Primary outcome was liver failure post-hepatectomy. Unsupervised consensus clustering classified HBV-related HCC patients into two clusters. A LightGBM model identified 15 factors linked to worse outcomes. Then, restricted cubic splines and logistic regression revealed that HBV DNA load was highly associated with liver failure. Results HBV-related HCC patients showed significantly higher liver failure (24.1% vs. 14.8%, p = 0.035) and severe complications (29.6% vs.18.5%, p = 0.019) compared to no HBV infection HCC patients after propensity score matching. The high HBV DNA load (≥ 3.679 log10 IU/mL) had 1.63-fold increased live failure risk (95% CI: 1.14–2.34). Preoperative antiviral therapy decreased live failure (OR: 0.30, 95% CI: 0.18–0.49) and severe complications (OR: 0.33, 95% CI: 0.22–0.51). The high HBV DNA load was associated with greater live failure risk among patients with either significant fibrosis [FIB-4 > 1.45 (OR: 2.01), APRI > 0.4 (OR: 1.87)] or impaired liver function [ALP > 147 U/L (OR: 3.89)]. Conclusion HBV positive with high DNA load is a risk factor of liver failure and severe postoperative complications in HCC patients undergoing hepatectomy. Antiviral therapy suppressing HBV DNA load below 3.679 log10 IU/ml should be recommended to hepatitis B virus infected patients before receiving liver surgery.

Objective To investigate the influencing factors for post-hepatectomy liver failure(PHLF)in hepatocellular carcinoma(HCC)patients with HBV infection at low viral load,and then construct a risk prediction model.Methods A total of 403 HCC patients who underwent initial hepatectomy in the First Affiliated Hospital of Army Medical University between January 1,2015 and March 1,2023 were recruited,and randomly assigned into a training set and a verification set in a ratio of 7:3.Lasso regression and multivariate logistic regression analyses were applied to screen the risk factors for occurrence of PHLF,and based on these identified factors,a nomogram prediction model was constructed.Receiver operating characteristic(ROC)curve analysis(area under the curve,AUC),calibration curve analysis,decision curve analysis,and clinical impact curve analysis were preformed to assess the predictive efficacy of the model.Results History of anti-viral therapy,history of drinking,logHBsAg,and international normalized ratio(INR)were independent influencing factors for the occurrence of PHLF in HCC patients with HBV infection at low viral load.The model established based on these indicators demonstrated excellent discriminative capabilities in both the training and validation sets,with an AUC value of 0.744 and 0.737,respectively.Calibration curve analysis indicated our model of high accuracy(training:P=0.995;validation:P=0.701),and decision curve analysis and clinical impact curve analysis displayed that our model provided greater clinical benefit.Conclusion Our model can effectively evaluate the risk of PHLF in HCC patients with HBV infection at low viral load,and shows good predictive performance,which has certain guiding significance for timely identification of high-risk populations.

IntroductionThough the importance raised attention, the clinical applications of methods for screening high-risk patients of sepsis after abdominal surgery were restricted. Therefore, we aimed to develop and validate models to screening high-risk patients of sepsis after abdominal surgery based on machine learning with routine variables.Material and methodsThe whole dataset was composed of three representative academic hospitals in China and Medical Information Mart for Intensive Care IV (MIMIC-IV) database. Routine clinical variables were implemented for model development. Boruta was applied for feature selection. Afterwards, ensemble learning and other eight conventional algorithms were used for model fitting and validation based on all features and selected features. The area under curves of the receiver operating characteristic curves (ROCAUCs), sensitivity, specificity, F1 score, accuracy, Net reclassification index (NRI), integrated discrimination improvement (IDI), Decision Curve Analysis (DCA), and calibration curves were used for model evaluation.ResultsA total of 955 patients undergoing abdominal surgery were finally analyzed (sepsis:285, non-sepsis:670). After feature selection, the ensemble learning model constructed by integrating k-Nearest Neighbor (KNN) and Support Vector Machine (SVM), yielded the ROCAUC of 0.892(0.841-0.944), the accuracy of 85.0% on the test data, and the ROCAUC of 0.782(0.727-0.838), the accuracy of 68.1% on the validation data, which performed best. Albumin, ASA score, Neutrophil-lymphocyte ratio, age, and glucose were the top features associated with postoperative sepsis by KNN and SVM.ConclusionsWe developed a new and potential generalizable model to preoperatively screening the high-risk patients of sepsis after abdominal surgery with the advantages of a representative training cohort and routine variables.

Modern anesthesiology has expanded beyond intraoperative care. It now integrates pain management, critical care, and emergency resuscitation. However, it still faces challenges like biological variability in drug responses, unpredictable intraoperative crises, and complex perioperative complications. Artificial intelligence (AI) emerges as a transformative force, can effectively enhance clinical quality and operational efficiency by extracting critical insights from vast amounts of healthcare data including electronic health records, vital sign waveforms, and imaging databases. AI applications in clinical anesthesia span the entire perioperative period, encompassing preoperative risk assessment, intraoperative physiological monitoring with adverse event prediction and visualized procedural guidance, as well as postoperative outcome forecasting and dynamic adaptive individualized treatment to enhance recovery after surgery. Beyond direct patient care, AI enhances operating room efficiency and revolutionizes anesthesia education. Despite progress, challenges persist in algorithm generalizability, data interoperability, and clinical validation. This review synthesizes the transformative role of AI across anesthesiology subspecialties, analyzes the barriers to implementation, and proposes strategic directions to bridge technological innovation with clinical optimization.

Background: With the growing demand for Total Hip Arthroplasty (THA), there is an urgent need to devise individualized perioperative management strategies to expedite patient recovery, particularly in scenarios involving multiple disciplines and concurrent decision-making. Objective:This study aims to construct a decision network and provide decision recommendations based on Bayesian Networks (BNs) for stratified risk groups. Methods: The MIMIC-IV database was employed to establish a developing dataset by extracting patients who underwent THA, followed by K-means clustering for patient phenotyping. A BN model for THA decisions were developed, and support degree was calculated according to conditional probability for decision recommendation. An external validation dataset from the multiple center big data platform was then utilized to validate the BN model's recommendations by calculating the conformity of risk-recommended protocols after phenotype classification. An online tool was developed for convenient application in clinical practice. Results: The developing dataset comprised 1701 admissions, while the validation dataset included 418 admissions. Patients in the developing dataset were divided into three phenotypes: a low-risk group (n=722), a moderate-risk group (n=673), and a high-risk group (n=306). Tailored recommendation protocols were formulated for each phenotype, covering preoperative and postoperative periods, based on support degree. Consistency in the distribution of differences among clustered indicators was observed in both datasets. Adherence to the recommended decisions was associated with reduced postoperative length of stay (PLOS) across all three decision protocols. Conclusions: Decisions recommended by Bayesian Networks for patients of different risk levels can decrease the PLOS and hasten recovery. This systematic approach, integrating risk stratification through clustering with an indicator network constructed by the BN model for decision recommendations, is adept at handling multi-decision tasks and supports clinical physicians in making informed choices regarding perioperative risk assessment and treatment plan selection. The registration has been completed on the website of the China Clinical Trial Registry Center (ChiCTR1900023927).

Background: Dexmedetomidine (Dex) may have anti-inflammatory properties and potentially reduce the incidence of postoperative organ injury. Objective: To investigate whether Dex protects pulmonary and renal function via its anti-inflammatory effects in elderly patients undergoing prolonged major hepatobiliary and pancreatic surgery. Design and Setting: Between October 2019 and December 2020, this randomized controlled trial was carried out at a tertiary hospital in Chongqing, China. Patients: 86 patients aged 60- 75 who underwent long-duration (> 4 hrs) hepatobiliary and pancreatic surgery without significant comorbidities were enrolled and randomly assigned into two groups at a 1:1 ratio. Interventions: Patients were given either Dex or an equivalent volume of 0.9% saline (Placebo) with a loading dose of 1 mu g kg(- 1) for 10 min, followed by 0.5 mu g kg(- 1) hr(- 1) for maintenance until the end of surgery. Main Outcome Measures: The changes in serum concentrations of interleukin-6 (IL-6) and tumour necrosis factor-alpha (TNF-alpha) were primary outcomes. Results: At one hour postoperatively, serum IL-6 displayed a nine-fold increase (P< 0.05) in the Placebo group. Administration of Dex decreased IL-6 to 278.09 +/- 45.43 pg/mL (95% CI: 187.75 to 368.43) compared to the Placebo group (P=0.019; 432.16 +/- 45.43 pg/mL, 95% CI: 341.82 to 522.50). However, no significant differences in TNF-alpha were observed between the two groups. The incidence of postoperative acute kidney injury was twice as high in the Placebo group (9.30%) compared to the Dex group (4.65%), and the incidence of postoperative acute lung injury was 23.26% in the Dex group, lower than that in the Placebo group (30.23%), although there was no statistical significance between the two groups. Conclusion: Dex administration in elderly patients undergoing major hepatobiliary and pancreatic surgery reduces inflammation and potentially protects kidneys and lungs.