Mechanistic basis of perioperative gut-liver axis disruption
Under ordinary conditions, the liver receives a steady flow of microbial and metabolic signals from the gut. Low-grade exposure helps maintain hepatic immune tolerance without preventing a response to pathogens. Surgery can disturb that balance. Anesthesia, tissue trauma, hypoperfusion, antibiotics, and reduced availability of enteral substrates can reduce microbial diversity, weaken tight junctions, and increase intestinal permeability3,4,7. When the barrier fails, lipopolysaccharide (LPS), bacterial DNA, extracellular vesicles (EVs), and other microbial products can enter the portal circulation. Kupffer cells and hepatic sinusoidal endothelial cells may then receive stronger inflammatory signals8,9,10.
Two metabolic circuits deserve particular attention after hepatectomy. Short-chain fatty acids (SCFAs) and related microbial products support hepatocyte proliferation, membrane phospholipid biosynthesis, and innate immune tolerance. Antibiotic exposure and germ-free states impair regeneration in experimental models, whereas restoration of commensal signals improves regenerative competence10,11,12,13. Bile acids form a second circuit. They act as signaling molecules as well as digestive detergents. Microbiota-dependent bile acid conversion regulates the farnesoid X receptor (FXR), Takeda G protein-coupled receptor 5 (TGR5), and intestinal fibroblast growth factor 15/19 (FGF15/19) signaling. In rodents, FGF15 and, in humans, FGF19 are produced in the ileum and reach the liver via the portal blood, where they act on hepatic pathways that affect proliferation, inflammation, and metabolic adaptation14,15,16,17,18,19,20. Dysbiosis may therefore harm recovery by permitting translocation and weakening regenerative signals.
Distinct clinical contexts in liver resection and transplantation
Hepatectomy and liver transplantation should be treated as related but distinct clinical settings. Compared with hepatectomy, LDLT and DDLT involve different biological challenges, including graft ischemia-reperfusion injury, immunosuppression, donor-derived microbial exposure, and antimicrobial pressure. After resection, outcome risk is strongly shaped by future liver remnant volume, underlying liver disease, biliary pathology, and the extent of operative injury. After transplantation, the recipient is exposed to allograft ischemia-reperfusion, immunosuppression, donor-recipient matching, graft size, and antimicrobial pressure. Living-donor and deceased-donor transplantation also differ in graft anatomy, donor-related procedures, organ allocation, and preservation. These differences affect microbiome perturbation, immune responses, complication profiles, and the safety of microbiota-directed interventions. Evidence should therefore be synthesized within, rather than across, these settings whenever possible.
Clinical consequences of liver surgery
Gut-liver axis injury helps explain why a local operation can lead to systemic postoperative illness. In hepatectomy, bacterial translocation and loss of microbial metabolic support may contribute to cholangitis, intra-abdominal infection, pneumonia, sepsis, delayed regeneration, and prolonged hospitalization, especially in patients with biliary obstruction, cirrhosis, malnutrition, sarcopenia, or extensive resection1,21,22,23,24. Liver transplantation adds a different set of stresses because dysbiosis, immunosuppression, donor-recipient factors, graft size, and allograft ischemia-reperfusion injury often coexist. Living-donor liver transplantation involves a planned partial graft and donor-related anatomical and perioperative considerations, whereas deceased-donor transplantation involves a deceased-donor graft and distinct allocation, preservation, and ischemia profiles. These settings should therefore be analyzed separately, and transplant findings should not be generalized directly to hepatectomy or across transplant types23,25,26,27.
The same biology may be relevant to PHLF, but the supporting evidence is less robust than that for postoperative infection. PHLF is defined by impaired synthetic, excretory, and detoxification function after liver resection6. Recent studies of microbial DNA and EV signatures suggest that host-microbial signals may complement established variables, including future liver remnant volume, bilirubin, coagulation function, and the severity of underlying liver disease8,9. However, current reports are few, observational, and not externally validated; they have not established incremental predictive value beyond these clinical parameters. They should therefore be regarded as hypothesis-generating tools for risk stratification, not as proof that perioperative microbiome modulation prevents PHLF.
Gut-liver-brain signaling also warrants attention. Hepatic encephalopathy (HE) is the best-established clinical manifestation of gut-liver-brain pathology in hepatology, and gut microbiota, ammonia handling, inflammation, bile acid signaling, and blood-brain barrier effects contribute to current models of HE3,4,28. Postoperative delirium after liver surgery may involve inflammatory, hepatic, microbial, and vagal routes, but it is not the same as HE. Direct evidence in the perioperative liver surgery setting is lacking, and delirium prevention should therefore be regarded as a new translational problem rather than an established microbiome-guided intervention29,30.
Evidence strength and clinical interpretation
To avoid treating all gut-liver axis evidence equally, clinical claims should be interpreted according to study type, surgical context, and endpoint. Table 1 summarizes representative trials, meta-analyses, perioperative-care measures, and emerging translational approaches. The most consistent clinical signal concerns postoperative infection, whereas PHLF biomarkers, bile acid strategies, FMT, and multiomics remain earlier-stage evidence.
Clinical stratification changes the implications of this review. Pro-/synbiotic strategies show the clearest infection-prevention signal in selected high-risk hepatobiliary surgery and transplantation cohorts, but the evidence does not establish one standard regimen. PHLF prediction, bile acid modification, FMT, delirium prevention, and multiomics-guided therapy remain under investigation and have not been integrated into routine perioperative practice.
Perioperative interventions targeting the gut-liver axis
Clinical data are strongest for pro-/synbiotic strategies, but these interventions are heterogeneous and should be interpreted in light of the clinical context and endpoint (Table 1). Randomized trials in biliary cancer hepatectomy reported lower infection rates with postoperative synbiotics plus enteral nutrition than with enteral nutrition alone (19%, 4/21 versus 52%, 12/23), and with preoperative plus postoperative synbiotics than with postoperative-only synbiotics (12.1%, 5/41 versus 30.0%, 12/40)21,22. Early addition of lactic acid bacteria and fiber reduced bacterial infection after liver transplantation compared with fiber alone (3% versus 48%) in a randomized trial23. The PREPRO trial in living-donor liver transplantation reported fewer overall infectious complications with synbiotics than with placebo (22% versus 44%; odds ratio (OR) 0.359, 95% confidence interval (CI) 0.150–0.858), but hospitalization, non-infectious complications, antibiotic use, and 30-day mortality did not differ (27). A 2020 systematic review of liver surgery and transplantation reported a pooled relative risk (RR) of 0.46 (95% CI 0.31–0.67), whereas a 2025 meta-analysis of randomized trials in liver surgery reported an OR of 0.34 (95% CI 0.25–0.45)1,24. These estimates support an overall infection signal across mixed surgical populations, not a universal hepatectomy effect. Strain, dose, timing, nutritional background, antibiotic exposure, surgical procedure, and patient risk differed across studies, preventing a single standard regimen.
Patient selection affects causal inference due to confounding. A low-risk patient undergoing a minor hepatectomy may need standard ERAS nutrition and routine antimicrobial prophylaxis. Patients with obstructive jaundice, biliary drainage, cirrhosis, recent broad-spectrum antibiotics, transfusion, diabetes, steatosis, or a small future liver remnant require closer management of infection, nutrition, cholestasis, and postoperative liver function. These are clinical risk strata, not mechanisms. Studies attributing outcomes to modulation of the gut-liver axis should specify how antibiotic exposure, transfusion, operative approach, inflow occlusion, blood loss, biliary drainage, immunosuppression, and adherence to ERAS were measured or controlled for.
Nutrition and ERAS measures are now available and have the strongest practical rationale for baseline care. Short-term fasting can support carbohydrate loading, early oral or enteral nutrition, and prevention of prolonged ileus, which may help maintain mucosal integrity and provide substrates for commensal metabolism5,25,31,32. Antibiotics remain necessary for prophylaxis and treatment, but overly broad-spectrum exposure may remove commensal signals required for regeneration7,12. Fecal microbiota transplantation (FMT) and bile acid-directed therapies may be considered for selected indications such as severe dysbiosis, recurrent infection, cholestasis, or metabolic dysfunction-associated steatotic liver disease. For general perioperative use, donor selection and timing, immunosuppression, drug interactions, pathogen transmission, and safety monitoring still require liver-surgery-specific trials17,33,34,35,36,37.
Bile acid-directed therapy has a mechanistic rationale but limited perioperative clinical evidence. FXR agonism, modulation of enterohepatic circulation, and careful management of cholestasis may affect inflammation and regenerative signaling14,15,16,18,19,20. Bile acid pools differ across normal parenchyma, steatosis, cholestasis, cirrhosis, and transplanted grafts. Before perioperative bile acid interventions are adapted from chronic metabolic liver disease protocols, pharmacokinetic, microbial, liver-function, and safety studies are needed in the relevant surgical populations.
Proposed precision recovery pathway
The pathway proposed here is a conceptual system for trial design and risk-adapted care, not an approved clinical standard. Preoperative risk stratification should combine conventional surgical variables with gut-liver axis risks, including cirrhosis, biliary drainage, cholestasis, recent antibiotic exposure, malnutrition, diabetes, steatosis, planned future liver resection, expected transfusion risk, and transplant-specific immunosuppression. Intraoperative and early postoperative plans should protect barrier function through stable hemodynamics, judicious inflow occlusion, normothermia, opioid-sparing analgesia when feasible, targeted antimicrobial therapy, and early oral or enteral nutrition5,7,25,31,32. These perioperative interactions and potential precision-recovery strategies are summarized in Figure 1.
High-risk patients may be suitable for targeted modulation, such as strain-defined probiotics or synbiotics, prebiotic fiber or medical nutrition, de-escalation of unnecessary antibiotics, and monitoring of bile acid and inflammatory profiles1,17,18,21,22,23,24. Multiomics and artificial intelligence should support these decisions rather than become the endpoint. Longitudinal metagenomics, metabolomics, bile acid profiling, immune markers, and electronic health record data may identify recovery phenotypes and signal deviations from the expected course. Prediction models should be transparent, reproducible, externally validated, separately calibrated for hepatectomy and transplantation, standardized across analytical platforms, and linked to interventions that can be tested prospectively.
The pathway should remain clinically feasible. Many centers do not perform stool metagenomics or bile acid profiling for every patient, and unstable patients cannot wait for complex assays. A tiered model is more practical: clinical risk stratification for all patients, focused microbial or metabolomic tests for selected high-risk cases, and research-grade multiomics in prospective trials. This approach can improve routine care while generating the evidence needed to build more precise protocols. Clinical implementation should also address turnaround time, laboratory reproducibility, data governance, cost, interpretability, and safety monitoring.
Future directions
Several gaps remain before translation. Studies use different sampling time points, sequencing methods, metabolite platforms, probiotic products, antibiotic regimens, surgical procedures, and outcome definitions. Multicenter perioperative cohorts should therefore collect stool, blood, bile, transfusion, antibiotic, operative, immunosuppression, and clinical-outcome data around shared surgical milestones. Trials should treat microbiota modulation as a set of defined interventions and recruit patients by clinical context and risk stratum, such as cholestasis, cirrhosis, major hepatectomy, marginal grafts, living-donor transplantation, or documented dysbiosis, rather than describing these categories as mechanisms. Studies should report null outcomes alongside infection, hospitalization, PHLF, delirium, death, adverse events, and microbiome-mediated safety signals. External validation, platform standardization, and independent replication should precede clinical implementation.