INTRODUCTION
Gastrointestinal surgery is undergoing a major transformation. It is no longer confined to pursuing a radical cure in oncology but is shifting towards a model that emphasizes precision, functionalization, and rehabilitation[1,2]. At present, stomach cancer remains the fifth most common cancer worldwide and the fourth leading cause of cancer-related deaths. Surgery is the main radical treatment method, but the occurrence of complications after surgery is still as high as 11.0% to 40.1%, which seriously affects the recovery and long-term prognosis of patients[3]. Against this backdrop, with the continuous advancement of minimally invasive techniques, the widespread adoption of the enhanced recovery after surgery (ERAS) concept, and the gradual clarification of the metabolic regulatory mechanisms of the body through bariatric and metabolic surgeries have jointly driven gastrointestinal surgery towards a higher level of integrated development[4].
In the surgical therapy of gastric and colorectal cancer, laparoscopic surgery has become a widely used standard surgical procedure. The robotic surgical system, with its advantages of three-dimensional high-definition vision, tremor filtering and flexible instrument operation, has gradually entered clinical practice[5,6]. Several meta-analyses and systematic reviews have shown that robotic surgery can significantly reduce the rate of conversion to open surgery and the rate of positive surgical margins, and the 30-day reoperation rate in middle and low rectal cancer. However, compared with laparoscopic surgery, there were no significant differences in terms of length of hospital stay and complications[7-9]. Total gastrectomy for gastric cancer can reduce severe complications, the rate of intra-abdominal infection, intraoperative blood loss and postoperative hospital stay. However, the 3-year overall survival rate and disease-free survival rate are similar to those achieved with laparoscopic total gastrectomy[10-13]. Therefore, robotic surgery has advantages in improving perioperative safety and pathological quality, but the long-term oncological outcomes have not surpassed those of laparoscopic surgery, and the cost-effectiveness issue still needs to be evaluated.
In terms of surgical strategies for special populations, the influence of viscera obesity on prognosis after colorectal cancer surgery is receiving increasing attention[14-16]. Nowadays, patients with visceral obesity have a higher the conversion rate to open surgery, fewer detected lymph nodes, longer recovery time of intestinal function, and a significantly increased overall incidence of complications[17]. Therefore, for patients with visceral obesity, precise preoperative risk assessment and individualized surgical strategy optimization are required.
The ERAS concept, through multimodal perioperative management measures, inhibits surgical stress responses and accelerates postoperative recovery, and has become the core paradigm of perioperative management in gastrointestinal surgery[18-20]. In the field of gastrointestinal surgery, the ERAS protocol can significantly reduce the length of hospital stay after surgery and lower the incidence of postoperative complications and readmission rates[21]. In radical gastrectomy for gastric cancer, ERAS can significantly shorten the hospital stay, accelerate the recovery of gastrointestinal function, and reduce the risk of complications. The refined implementation of the ERAS pathway is becoming a key link in improving the quality of gastrointestinal surgery[22].
Meanwhile, the risk prediction of postoperative complications is evolving from traditional statistical models to new tools driven by artificial intelligence[23,24]. The random forest model developed by Lin et al[25] can identify high-risk patients at an early stage and enable personalized interventions by predicting complications following radical gastrectomy for gastric cancer. The clinical application of this type of tool is expected to further enhance the precision level of perioperative management.
Bariatric and metabolic surgery in gastrointestinal surgery not only achieves significant weight loss and remission of type 2 diabetes, but also its systematic regulatory mechanism on the body’s metabolism has increasingly become a research hotspot[26-28]. Previous studies have confirmed that bariatric metabolic surgery is closely related to significant changes in the intestinal microbiota. Alterations in microbiota diversity, remodeling of bile acid metabolism, and enrichment of beneficial microbiota are all associated with postoperative metabolic improvement[29]. It is also proposed that the “microbiota-bile acid-metabolism” axis plays a key role in postoperative metabolic benefits[30-32].
In the field of oncology surgery, the rise of conversion therapy represents an important trend of the cross-integration of local treatment and systemic treatment throughout the body[33-35]. For initially unresectable upper gastrointestinal tract and hepatobiliary and pancreatic malignant tumors, achieving tumor downstaging through systemic chemotherapy, targeted therapy and other means before surgical resection can significantly prolong the survival period of patients[36].
The precision and functionalization transformation of gastrointestinal surgery, as illustrated in Figure 1, is reflected in the continuous optimization of minimally invasive techniques, systematic innovation in perioperative management, and the in-depth intersection of surgical and metabolic regulation. Throughout this review, we focus on a central question: How can gastrointestinal surgery balance precision (achieving radical oncological resection) with functional preservation (maintaining postoperative quality of life)? By integrating the latest research in the above three dimensions, this article aims to provide theoretical references for the evolution of modern gastrointestinal surgery from “radical resection” to the “radical-functional-rehabilitation” trinity goal, and looks forward to the application prospects of intraoperative navigation technology and artificial intelligence-assisted decision-making systems in future individualized surgical treatment.
Figure 1 Conceptual framework of precision and functional preservation in gastrointestinal.
ERAS: Enhanced recovery after surgery; AI: Artificial intelligence; RCTs: Randomized controlled trial.
THE IN-DEPTH APPLICATION OF MINIMALLY INVASIVE SURGICAL TECHNIQUES
The continuous evolution of minimally invasive surgical techniques is one of the core driving forces promoting the precision and functionalization transformation of gastrointestinal surgery. In recent years, the new generation of minimally invasive platforms represented by laparoscopic surgery and robotic surgical systems, with their advantages of three-dimensional high-definition vision, tremor filtering and flexible instrument operation, are profoundly changing the treatment landscape of gastrointestinal tumor surgery[37-39].
Evidence-based progress in robotic gastrectomy
Stomach cancer is the fifth most prevalent cancer worldwide and the third leading cause of cancer-related deaths. Surgery remains the primary radical treatment for it[40-42]. Although laparoscopic gastrectomy has established its standard status in early gastric cancer, technical challenges of laparoscopic surgery in terms of lymph node dissection thoroughness and complex digestive tract reconstruction remain in locally progressive gastric cancer[43]. The introduction of robotic surgical systems aims to overcome the inherent limitations of conventional laparoscopy in terms of field of view, instrument flexibility, and operating precision.
In recent years, several high-quality systematic reviews have compared robotic gastrectomy with laparoscopic gastrectomy in terms of perioperative outcomes and long-term survival rates. A study incorporating 72 trials (n = 30081) showed that robot-assisted gastrectomy had higher overall survival [hazard ratio (HR) = 0.89] and lower complications [odds ratio (OR) 0.77] than laparoscopic gastrectomy[44]. Robotic surgery also reduced intraoperative bleeding [mean difference (MD) -37.45 mL], increased lymph node yield (MD 1.88), lowered conversion to open surgery (OR 0.44), and accelerated recovery. However, it lengthened operative time (MD 35.53 minutes). Notably, subgroup analyses further revealed populations in which robotic surgery was advantageous: Robotic gastrectomy had better overall and disease-free survival in patients with stage I/II gastric cancer, and robotic surgery also showed significant benefits in total gastrectomy, patients with a body mass index ≥ 25 kg/m2, and patients undergoing in vivo gastrointestinal tract reconstruction.
The results of another meta-analysis similarly support this conclusion. This analysis included 1 randomized controlled trial (RCT) and 14 propensity score-matched studies (n = 5079)[45]. Although robotic gastrectomy takes longer (MD 19.82 minutes), it outperforms laparoscopy in reducing blood loss (MD -28.91 mL), shortening hospital stay (MD -0.69 days), and lowering overall complications [relative risk (RR) 0.82] and severe complications (RR 0.71). In addition, the study found that robotic surgery could shorten the time of first flatus after surgery, the time to first liquid food, the time of first soft food, suggesting that robotic surgery could help accelerate the recovery of postoperative gastrointestinal function.
A meta-analysis of 31 studies (n = 12401) further validated these findings. Robotic gastrectomy was associated with lower complications (OR 0.81), particularly pancreatic complications (OR 0.376)[46]. It also increased lymph node yield [weighted MD (WMD) 2.03], shortened time to first flatus (WMD -0.105 days), and reduced bleeding (WMD -20.09 mL). However, operative time was longer (WMD 40.192 minutes) and costs were significantly higher (WMD RMB 19141.68). Notably, there is little difference in the long-term oncological outcomes between these two surgical approaches. This is because the two groups of patients were very similar in terms of 3-year overall survival (OR 1.030), 5-year overall survival (OR 0.862), and mortality rates.
A meta-analysis of distal gastrectomy (20 studies, n = 5447) showed that robot-assisted surgery performed better than laparoscopy in complications, blood loss, and distal margin distance. However, it was associated with longer operative times and higher costs[47]. The two groups of patients showed virtually identical results in several aspects, including the distance of the proximal resection margin, the incidence of severe complications, the incidence of anastomotic fistula, the time to first flatus, and the length of hospital stay. This result suggests that the advantages of robotic surgery in distal gastrectomy are mainly in terms of refined operation and complication control rather than macroscopic recovery speed.
However, a more conservative evaluation was provided by a study published by Salem et al[48]. The analysis, incorporating 90 studies involving a total of 65296 patients, revealed a potential association between robot-assisted gastrectomy and a lower incidence of Clavien-Dindo grade ≥ II complications (OR 0.74), with low heterogeneity (I² = 21.4%). More importantly, only three studies were RCTs, and the GRADE assessment of the evidence quality was rated as “very low”, indicating that the existing evidence is subject to a high risk of bias. In particular, the study highlights that despite the potential benefits of robotic surgery, it should be promoted with caution in the absence of high-quality RCT evidence, and that safe promotion requires standardized training, competency benchmarking, and limiting the impact of industry funding.
While the results derived from these two meta-analyses show the superiority of robotic surgeries in terms of reduced intraoperative bleeding and complications in the short term, substantial heterogeneity of the existing evidence should be considered while analyzing these studies. In the first place, the nature of the studies involved is clearly different, with RCTs and predominantly retrospective propensity score-matched studies being compared to each other in terms of bias management, with the second type probably exaggerating the clinical benefits associated with surgical interventions. Secondly, the patients’ stage at the time of the study is diverse, and it appears that the survival benefits provided by robotic procedures manifest themselves only for patients with early-stage gastric cancer, while this issue remains to be determined for patients with locally advanced tumors.
The question that arises here is why there does not exist any survival advantage of robotics over traditional approaches in the long run. Most likely, this is because, on the one hand, the prognosis in the long term depends mostly on the biological properties of the tumors and their clinical staging, while on the other hand, robotic technology will have no influence on the surgical procedure itself, as long as a good-quality lymphadenectomy will be performed. Thus, the point about the superiority of robots in terms of curing the patients cannot hold any water; however, the main advantage of the technology consists of the fact that it allows the patient to recover from the surgery safely. In other words, the benefit of robotic surgery lies in its ability to reduce surgical stress when the disease can be cured using traditional methods.
Systematic evaluation of robotic colorectal surgery
Minimally invasive surgery has become a standard treatment for colorectal cancer[49,50]. However, rectal cancer surgery is one of the most promising fields for the application of robotic surgical systems due to the narrow anatomy of the pelvis, the stringent requirements for precision in total mesorectal excision, and the technical challenges of anus-preserving surgery for low rectal cancer[51,52].
In the field of rectal cancer: Multiple randomized trials have compared robotic and laparoscopic surgery for colorectal cancer. An article involving 7 RCTs and a total of 1731 patients conducted a systematic comparison between robotic colorectal surgery and laparoscopic surgery[53]. The findings indicate that robotic surgery leads to a significant reduction in the time required for bowel function (MD -0.62 days) to return to normal, reduced the incidence of severe complications (OR 0.70), decreased the conversion rate to open surgery (OR 0.27), and lowered the positive rate of circumambulation margin (OR 0.62). However, robotic surgery takes longer (MD 22.7 minutes), while laparoscopic surgery has a slight advantage in terms of the number of lymph nodes detected (MD +0.53). There were no significant differences between the two groups in terms of total mesorectal excision integrity, R0 resection rate and anastomotic leakage. The findings of this research indicate that robotic surgery offers significant advantages in terms of bowel function recovery, surgical precision, and intraoperative effects, supporting its broader application in the treatment of colorectal cancer.
A sequential meta-analysis of 14 RCTs involving 2867 patients with colorectal cancer further strengthened the above conclusion[54]. This study found that compared with laparoscopic surgery, robotic surgery significantly reduced the conversion rate to open surgery (RR 0.54) and the positive rate of circumferential margin (RR 0.65). Sensitivity analysis showed that robotic surgery also had a slight advantage in the time to first flatus (MD -0.13 days). However, the two groups showed comparable intraoperative blood loss, postoperative complications, and cancer recurrence rate. It is worth noting that this meta-analysis clearly indicates that although robotic surgery has advantages in reducing the conversion rate to open surgery and the positive rate of circumambulation margins, its postoperative complications and oncological outcomes are comparable to those of laparoscopic surgery, and the operation time of robotic surgery is significantly prolonged (MD 49.4 minutes).
In the field of right-sided colon cancer: With the promotion of the concept of complete mesocolic excision (CME), the clinical application of robotic CME is increasing day by day. A study comprising 7 trials and 733 patients compared the perioperative and pathological outcomes of robotic CME with those of laparoscopic CME[55]. The results showed that the two groups of patients were comparable in terms of postoperative complications (OR 0.95), blood loss (MD -4.6 mL), length of hospital stay (MD -0.10 days), postoperative intestinal obstruction (OR 0.88), and the number of lymph nodes detected (MD 2.05). Although robotic CME was linked to a lower conversion rate to open surgery (OR 0.07), this finding is based on a small number of cases and low-certainty evidence. Moreover, operative time was significantly longer in the robotic group (MD 45.6 minutes). The two study groups were comparable for disease-free survival (OR 1.25) and overall survival (OR 1.00).
The impact of the learning curve: A systematic review and meta-analysis that included 6 RCTs and 17 observational studies, focusing on the early learning stage of robotic colorectal surgery, provided an important supplement[56]. This study found that even in the early stage of the adoption of robotic technology, robotic surgery could significantly shorten the postoperative hospital stay (MD -0.132 days) and significantly reduce the conversion rate to open surgery (OR 1.480). The two experimental groups are similar in terms of 30-day mortality, early postoperative complications, readmission rate and reoperation rate, suggesting that robotic surgery can achieve clinical safety comparable to laparoscopic surgery at the early learning curve stage.
Optimization of surgical strategies for special populations
Obese patients: Obese patients are a high-risk group in gastrointestinal surgery. Obesity not only increases the technical difficulty of laparoscopic surgery (limited field of vision, blurred anatomical landmarks, and difficult vascular separation), but is also associated with a higher incidence of postoperative complications[57,58]. A study on obese rectal cancer patients (including 5 studies with a total of 499 patients) indicated that robotic rectal cancer surgery could significantly shorten the length of hospital stay (WMD -1.67 days), reduce the overall postoperative complication rate (OR 0.41), and lower the readmission rate (OR 0.37) compared with laparoscopic surgery, but the operation time was longer (WMD 41.38 minutes)[59]. There were no significant differences between the two groups in terms of intraoperative blood loss, laparotomy conversion rate, number of lymph nodes detected, positive rate of circumferential resection margin and anastomotic leakage. This study suggests that robotic surgery may be a safer treatment option with potentially greater benefits for patients with obesity-related colorectal cancer.
Visceral obesity: As a special phenotype of obesity, visceral obesity has received increasing attention from research. Unlike subcutaneous fat, visceral fat has higher metabolic activity and is closely related to the difficulty in exposing the surgical field and the limited operating space[60]. A large-scale study involving 11129 patients with colorectal cancer assessed the effect of visceral obesity on postoperative results[61]. The results showed that patients with visceral obesity had a higher laparotomy conversion rate, fewer detected lymph nodes, and a longer recovery time of intestinal function. In terms of complications, the overall incidence of complications in patients with visceral obesity was significantly increased, specifically including anastomotic leakage, intestinal obstruction, intra-abdominal abscess, incision infection, and pulmonary complications. It is worth noting that overall survival and disease-free survival were virtually identical between the two groups, suggesting that visceral obesity primarily affects perioperative safety rather than long-term oncological results.
Therefore, in clinical practice, for patients with visceral obesity, more accurate risk prediction is needed before the operation, more refined surgical strategies should be adopted during the operation, and complication monitoring should be strengthened after the operation. Robotic surgery, with its three-dimensional high-definition field of view and flexible instrument operation, may have unique application value in this special group of people, but more high-quality RCTs are still needed for verification. In addition, preoperative body composition analysis (such as visceral fat area measurement) can be used as a routine item for preoperative assessment of colorectal cancer to guide the formulation of individualized surgical plans.
INNOVATIONS IN PERIOPERATIVE MANAGEMENT
The innovation of perioperative management is another key pillar promoting the precise and functional transformation of gastrointestinal surgery. In recent years, the extensive promotion and in-depth application of the ERAS concept have greatly optimized the traditional perioperative management model, systematically suppressing surgical stress responses and accelerating postoperative recovery from the perspective of multimodal intervention[62,63]. Meanwhile, with the rapid development of artificial intelligence technology, postoperative complication risk prediction models based on machine learning are gradually moving from research tools to clinical applications, providing brand-new technical support for individualized perioperative management[64,65].
Evidence-based basis and clinical effects of the ERAS pathway in gastrointestinal surgery
The ERAS concept was first proposed by Professor Henrik Kehlet from the University of Copenhagen in Denmark in 1997. The core of this approach lies in implementing a series of perioperative optimization measures through multidisciplinary collaboration and based on evidence-based medicine. It covers core links such as preoperative consultation and nutritional support, multimodal analgesia with opioid frugality, goal-directed fluid therapy, temperature protection, early ambulation and early oral feeding, aiming to reduce surgical stress response, maintain organ function, shorten hospital stay and improve patient prognosis[66]. After nearly three decades of development, the ERAS pathway has accumulated rich evidence-based evidence in the field of gastrointestinal surgery, and its clinical effects have been confirmed by multiple high-quality meta-analyses and systematic reviews.
A meta-analysis conducted by Dong et al[67] included 13 RCTs and cohort studies, involving a total of 5603 patients. The study systematically assessed the impact of ERAS on postoperative recovery in patients undergoing gastrointestinal surgery. The study findings indicate that the ERAS protocol significantly reduces the length of postoperative hospital stay (MD -3.16 days) and significantly lowers the incidence of postoperative complications (RR 0.70) and the readmission rate (RR 0.75). However, the study also noted that ERAS had no significant effect in reducing reoperation rates and mortality (RR 0.59), suggesting that further high-quality research is needed to clarify the role of this protocol in these rare clinical outcomes.
McKechnie et al’s research[68] focused on the application of the ERAS pathway in emergency intra-abdominal surgery, filling the evidence gap regarding the application of ERAS in emergency situations. This study included a total of 20 studies, involving 1615 patients who received the ERAS protocol and 1933 patients who received conventional treatment. In the upper gastrointestinal surgery subgroup, the length of postoperative hospital stay was significantly shorter for patients in the ERAS group (MD = 3.35 days), and the risk of complications showed a downward trend (RR = 0.56). In the lower gastrointestinal surgery subgroup, the ERAS group also demonstrated clinical benefits such as shortened hospital stay (MD 2.80 days) and significantly reduced risk of complications (RR 0.66). It is worth noting that although the ERAS group showed a favorable trend in postoperative intestinal obstruction, which was not statistically significant. Similarly, the difference in the risk of readmission between the two groups was not significant. But the GRADE evidence rating for this study is “low to very low”, suggesting that the existing data is still limited by imprecision and that larger sample size studies are needed for verification in the future.
In the field of acute perforation of the gastrointestinal tract, the meta-analysis published by Wang et al[69] included 27 trials, involving a total of 1864 patients. The results showed that patients in the ERAS/fast-track surgery group had significantly lower stress markers (C-reactive protein levels) and a significantly lower incidence of complications compared to the conventional treatment group. In this group, the time to first ambulation, first postoperative flatus, and initiation of oral intake were all significantly earlier, and both the length of hospital stay and medical costs were significantly reduced. This discovery indicates that even in emergency surgery scenarios, the ERAS pathway remains safe and effective, but its promotion needs to be appropriately adjusted according to the characteristics of emergency surgeries.
The systematic review written by Buhl et al[70] provides an important reference for implementing the ERAS pathway in patients with intestinal obstruction. This review included a total of 16 studies and found that 56% (9/16) of the studies adopted more than 10 ERAS intervention measures, with multimodal systemic analgesia being the most common. This discovery reveals significant heterogeneity in the composition of ERAS programs among different studies, emphasizing the importance of standardized reporting and the implementation of ERAS intervention measures. At the same time, it is pointed out that further research on the ERAS guidelines for emergency laparotomy is urgently needed.
The regulatory role of ERAS compliance on clinical outcomes
Although the clinical value of the ERAS pathway has been widely recognized, the compliance of different medical institutions with the ERAS protocol varies significantly, resulting in uneven clinical outcomes. A recent article that included 35 studies involving a total of 34921 patients, covering 11 types of surgeries such as colorectal (74%), liver, and orthopedics, systematically evaluated the impact of ERAS protocol compliance on postoperative outcomes[71]. The study found that the average hospital stay of patients in the high compliance group was shortened by 2.46 days [95% confidence interval (CI): -3.18 to -1.74], and the risk of complications was reduced by 56% (OR 0.44), but no significant differences were achieved in subgroups such as gastric surgery and orthopedic surgery. In terms of secondary outcomes, although the recovery time of intestinal function (WMD -1.12 days) and readmission rate (OR 0.81) in the high compliance group showed an improving trend, they did not reach statistical significance. Although the mortality rate decreased by 56% (OR 0.44), the CI was relatively wide (0.21-1.17), and larger-sample studies are needed for confirmation.
This study further reveals the gap problems existing in the implementation of ERAS. Although the benefits of the plan are clear, the global average compliance rate is only 40% to 92.2%. The main obstacles are insufficient multi-disciplinary collaboration and so on. Researchers suggest enhancing compliance through a theory-driven implementation framework and establishing a centralized auditing system to monitor long-term effects.
At the guideline level, in recent years, multiple academic organizations have released ERAS guidelines or consensus based on the latest evidence. In 2025, the ERAS Society released the latest guidelines for perioperative care in elective colorectal surgery. This guideline is not a simple revision of previous versions but an independent update written from scratch based on a brand-new methodological framework, comprehensively covering all aspects of perioperative care in colorectal surgery[72]. In 2024, the ERAS Committee of Korea released the ERAS guidelines for colorectal cancer. These guidelines are based on a systematic review, with all key issues focusing on RCTs. If there are less than two RCTs, propensity score matching studies are included, and the recommendation strength and evidence level are marked for each recommendation[73]. In the same year, the European Society for Endoscopic Surgery and the American College of Gastrointestinal Endoscopic Surgeons jointly released evidence-based recommendations and expert consensus on perioperative optimization for elderly patients[74]. The consensus reached through this systematic review and meta-analysis, which addressed 24 predefined key questions, clearly states that minimally invasive surgery and the ERAS protocol are recommended perioperative management strategies for elderly patients, particularly for colorectal surgery. It is worth noting that, except for the ERAS protocol in colorectal surgery, all other recommendations are conditional, with evidence certainty ratings of very low to very low.
According to the current state of knowledge from the field of implementation science, there are some specific strategies that can help to improve compliance with the ERAS program, namely: (1) Forming a multidisciplinary ERAS team; (2) Developing standard checklists for each perioperative period (before surgery, during surgery, after surgery); (3) Regularly conducting audit and providing feedback about the clinical results and the level of compliance with the program to the medical staff; (4) Making the ERAS orders part of the electronic records of patients by default; and (5) Conducting training programs for rotating residents and nurses.
A postoperative complication risk prediction model based on machine learning
Although the ERAS pathway significantly improved the average prognosis of patients in gastrointestinal surgery, individualized risk prediction of postoperative complications remains one of the core challenges in perioperative management. The mechanisms underlying postoperative complications are complex and involve the interaction of multiple factors, including the patient’s baseline condition, anesthetic stress, surgical trauma, and perioperative management[75,76]. Most traditional risk assessment tools are based on linear models such as logistic regression, making it difficult to fully capture the nonlinear relationships and interaction effects among these factors[77,78]. The rapid advancement of machine learning in recent years has opened up new avenues for predicting the risk of postoperative complications. Algorithms such as random forest, XGBoost, and support vector machine have demonstrated superior predictive performance compared to traditional models in multiple studies.
In terms of predicting overall postoperative complications of gastric cancer and colorectal cancer: Lu et al[79] published a study based on a national multi-center prospective database that analyzed 3926 patients with gastrointestinal cancers from the PACAGE database (covering 20 medical centers), including 2271 cases of gastric cancer and 1655 cases of colorectal cancer. The overall postoperative complication rates were 18.1% and 14.8% respectively, and intra-abdominal infection was the most common type of complication. The research team developed and validated multiple machine learning models. The results showed that the random forest model achieved the highest average area under the curve values in the prediction of overall complications and different types of complications in both the gastric cancer and colorectal cancer groups, and consistent results were verified in the external cohort. This study also developed application-based clinical tools to facilitate their application in clinical practice. The researchers pointed out that this model has good predictive performance for overall complications and infectious complications, and can support clinical decision-making and the formulation of individualized treatment strategies.
In the prediction of anastomotic leakage: In recent years, a number of high-quality machine learning studies have emerged. Shiwakoti et al[80] published a prospective multicenter validation study, developing an anastomotic leakage prediction model for patients with gastric adenocarcinoma who underwent total or proximal gastrectomy. The uniqueness of this study lies in its adoption of a multi-center prospective validation design, providing a higher level of evidence support for the clinical translation of the model. Kang et al[81] focused on predicting anastomotic fistulas following rectal cancer resection and developed an interpretable machine learning model based on serum calcium levels. This study found that postoperative serum calcium levels are a key feature for predicting anastomotic leakage, and the SHapely Additive exPlanations value analysis of the model provides good interpretability.
It is worth noting that the application of deep learning technology in the field of anastomotic leakage prediction has also made breakthrough progress. A prospective study based on intraoperative laparoscopic video developed a novel deep learning model for predicting colorectal anastomotic leakage, representing a technological leap in artificial intelligence from structured data analysis to image analysis. Another study analyzed the images of the anastomotic site during laparoscopic gastrectomy through a deep learning model to predict anastomotic leakage or residual leakage. The results suggest that deep learning can predict the risk of leakage in real time during the operation. These studies collectively indicate that the prediction of anastomotic leakage is evolving from single-factor analysis to multi-modal dynamic integration.
In the prediction of specific complications such as gastroparesis after CME surgery: A study published in 2024 developed four machine learning models (XGBoost, random forest, support vector machine, and K-nearest neighbor) using data from 1146 colon cancer patients (95 of whom developed gastroparesis) to predict the risk of gastroparesis after CME surgery[82]. The study collected 34 variables, including demographic characteristics, chronic diseases, preoperative examination results, surgical methods and intraoperative details. The results show that the XGBoost model has the most outstanding predictive performance, with an area under the curve of 0.976 in the training set, 0.906 in the validation set, and 0.77 in the external validation, confirming the model’s strong generalization ability. This study also identified key risk factors for gastric atony, including older age, prolonged surgery, significant surgical technique, intraoperative blood loss, low serum protein levels, diabetes, anemia, and hypothyroidism. These findings provide important decision-making basis for the prevention and early intervention of gastroparesis after CME surgery.
Despite these machine learning models having shown good performance for prediction, there are a number of challenges that need to be taken into account when applying them in clinical practice. The first challenge concerns the availability of input variables; some models depend on variables that are not commonly and systematically captured in daily clinical settings, such as some intraoperative variables or special lab results. The second challenge is related to model interpretability. While black-box algorithms like XGBoost and random forest have good precision, they do not make intuitive decisions and therefore may not be trusted by surgeons. Although there are some ways of determining the importance of different features, using these results in making clinical decisions is difficult. Finally, implementation can be a challenge due to the necessity to integrate the tool with electronic health records. Future research should prioritize prospective external validation and the development of implementation strategies to overcome these barriers.
Considering these constraints, how do we differentiate between models that have a better clinical outlook? To facilitate this process, we suggest that readers consider a straightforward three-point framework. The first criterion is that of external validation. is there evidence that the model has been validated through independent datasets? The second criterion involves prospective validation, has the model been prospectively validated in the clinic setting? The third criterion concerns interpretability, do clinicians know why the model predicted what it did?
Applying this framework to the models cited in this review, none satisfies all three criteria. The random forest model by Lu et al[79] has external validation (tested in an external cohort) but lacks prospective testing and has limited interpretability. The XGBoost model by Shiwakoti et al[80] has prospective validation (multi-center prospective design) and moderate interpretability via SHapely Additive exPlanations analysis, but requires external validation in different populations. The interpretable model by Kang et al[81] offers good interpretability (serum calcium as a key feature) but needs both external and prospective validation. Overall, the model by Shiwakoti et al[80] appears closest to clinical application among those reviewed, as it has undergone prospective multi-center validation. However, all models remain at the proof-of-concept stage, and no model is yet ready for routine clinical deployment without further validation.
The synergistic drive of artificial intelligence risk prediction and ERAS
The collaborative integration of artificial intelligence-driven risk prediction models and ERAS pathways is becoming a new paradigm for individualized perioperative management. The standardized protocol adopted by the traditional ERAS pathway is that all patients receive the same combination of intervention measures[83]. However, there are significant differences in baseline risks among different patients, and a uniform ERAS protocol is difficult to meet individualized clinical needs. The value of artificial intelligence risk prediction models lies in their ability to conduct precise risk assessments for patients before or in the early postoperative period, thereby guiding individualized adjustments to ERAS programs[84,85]. For patients predicted to be at low risk, ERAS intervention measures can be appropriately streamlined, the length of hospital observation can be shortened, and the rehabilitation process can be accelerated. For patients predicted to be at high risk, it is necessary to enhance perioperative monitoring, optimize nutritional support and analgesia regimens, and extend hospital stays to prevent the occurrence of complications.
This “risk stratified driven individualized ERAS” model has demonstrated feasibility in preliminary studies. For instance, the prediction of overall postoperative complications by the aforementioned random forest model and the prediction of anastomotic leakage by the XGBoost model can both provide quantitative risk assessment for clinicians before surgery or in the early postoperative period, thereby guiding the optimization of ERAS protocols and the rational allocation of perioperative resources[86]. In the future, with the continuous iteration of predictive models and the development of real-time monitoring technologies, artificial intelligence is expected to achieve a transformation from static risk prediction to dynamic risk monitoring. For instance, by analyzing the dynamic change trend of C-reactive protein after surgery, it can issue real-time warnings for the occurrence of anastomotic leakage, thereby promoting the development of perioperative management towards greater precision and individualization.
THE CROSS-INTEGRATION OF SURGERY ANG METABOLIC REGULATION
The boundaries of surgical operations are expanding from simple anatomical resection to a broader field of functional regulation. The remodeling effect of bariatric and metabolic surgery on the body's metabolic network, the fundamental improvement of prognosis for patients with advanced tumors through conversion therapy, and the continuous enhancement of surgical accuracy and safety by intraoperative navigation technology jointly outline a future picture of the deep intersection of gastrointestinal surgery and metabolic regulation.
The metabolic regulatory mechanism of bariatric and metabolic surgery
Bariatric and metabolic surgery is the most representative example in the intersection of surgery and metabolic regulation. Metabolic weight-loss surgery not only achieves significant weight loss but also demonstrates exceptional efficacy in remission of type 2 diabetes, outperforming drug therapy[87,88]. However, the exact molecular mechanism of its metabolic benefits remains one of the most core scientific issues in the field of metabolic surgery.
In 2025, the Clinical Issues Committee of the American Society for Metabolic and Bariatric Surgery published a comprehensive review examining the regulatory impact of bariatric metabolic surgery on the gut microbiota and its contribution to metabolic enhancement[89]. Reviews indicate that the gut microbiota is a key regulator of metabolic health in obesity and type 2 diabetes. Factors such as diet, genetics, the environment, and the host’s physiological state induce changes in the microbial composition, which in turn affect insulin sensitivity, energy absorption, fat storage, and systemic inflammation. Metabolic bariatric surgery is closely related to the specific transformation of the intestinal microbiota. Restructuring of bile acid metabolism, changes in microbial diversity, and the enrichment of beneficial microbial communities are all associated with good metabolic outcomes. The gut microbiota may also interact with host metabolism through molecular pathways such as the glucagon-like peptide-1 (GLP-1) signaling pathway. This review holds that the bidirectional interaction between the microbiota and bile acids, namely the “microbiota-bile acid-metabolism” axis, may constitute the core mechanism for metabolic improvement after bariatric surgery, providing a brand-new perspective for future therapeutic targets.
The research published by Ma et al[90] provided key evidence in the differential regulation of bile acid profiles by specific surgical procedures. In a rat model of type 2 diabetes, we compared the impacts and metabolic correlations of a single-anastomotic duodenal-ileal bypass combined with sleeve gastrectomy (SADI-S) vs simple sleeve gastrectomy (SG) on the bile acid profile. The results showed that compared with the sham operation group, both SADI-S and SG greatly improved dyslipidemia, hyperglycemia and β-cell integrity, among which the therapeutic effect of SADI-S was more prominent. In terms of bile acid profile remodeling, SADI-S induced a broader increase in serum bile acids, contain important regulators including chenodeoxycholic acid and choline. More importantly, elevated levels of various serum bile acids were significantly negatively correlated with fasting blood glucose, triglycerides, and glycated hemoglobin, while being positively correlated with GLP-1 levels. The changes in the gut microbiota are associated with specific alterations in bile acids, jointly supporting the key role of the “microbiota-bile acid-metabolism” axis in postoperative metabolic benefits. SADI-S and SG surgeries result in distinct, surgery-specific changes in bile acid composition, a phenomenon that is significantly associated with metabolic improvements in patients with type 2 diabetes. The strong correlation between particular bile acid types and metabolic parameters highlights their potential as mediators and treatment targets.
A meta-analysis published in 2024 focused on the integration effect of hormonal changes after bariatric surgery[91]. This meta-analysis found that the total postprandial GLP-1 level increased after the operation and was associated with the improvement of blood glucose control indicators. However, fasting GLP-1, fasting glucose-dependent insulinotropic polypeptide, postprandial glucose-dependent insulinotropic polypeptide, fasting ghrelin and fasting glucagon all showed a downward trend, but none of them reached statistical significance. The research also found that significant changes occurred in the fat around the liver and pancreas, inflammatory markers, microRNA and intestinal microbiota after bariatric surgery. The findings of this meta-analysis highlight the complexity of the underlying mechanisms: Although the studies have identified several potential pathways, further research is required to ascertain the primary mechanisms underlying the remission of type 2 diabetes following bariatric surgery.
Implications of the “microbiota-bile acid-metabolism” axis for clinical surgery exist, but it is crucial not to conflate associations with causations. The current literature on the topic relies mainly on observational studies which report the association between the alterations in microbiota composition and the positive changes in metabolic parameters following bariatric surgery. Evidence for the causation is based predominantly on animal studies and few human mechanistic experiments. Consequently, while the associations are strong, the mechanism itself remains uncertain and requires further investigation. More precisely, it is unknown whether the changes in microbiota cause the positive outcomes or just occur concurrently with them.
Tumor conversion therapy in surgical operations
Transformative therapy represents a fundamental change in the concept of surgical treatment. From passively accepting the judgment of “unresectable” to proactively achieving tumor downstaging through systemic treatment and then performing radical resection[92]. This strategy has demonstrated particularly significant clinical value in advanced gastric cancer and liver cancer, and it is a model of the cross-integration of local treatment and systemic treatment.
In the field of gastric cancer: Multiple meta-analyses have provided a solid evidence-based basis for conversion therapy. The 2025 article by Chan et al[93] included 36 research involving 3177 patients with stage IV gastric cancer: 1273 in the conversion surgery group and 1904 in the non-surgical group. The study systematically compared the outcomes of long-term survival between conversion surgery following systemic therapy and systemic therapy alone. As can be seen from the study results, the median overall survival for patients receiving conversion surgery is between 14.4 and 60.0 months, compared to 4.7 and 19.9 months in the systemic treatment alone group. Overall survival in favor of conversion surgery was observed among the predominantly retrospective data pooled (HR = 0.36, 95%CI: 0.32-0.40), and progression-free survival was also significantly higher in the conversion surgery group (HR = 0.38, 95%CI: 0.31-0.46). At this stage, the results obtained cannot be used as an independent criterion for the effectiveness of conversion therapy since this approach is characterized by selection bias. Specifically, the patients in the study received conversion surgery due to their positive response to previous chemotherapy and better tumor biology.
In terms of postoperative complications, the incidences of anastomotic leakage, abdominal abscess and postoperative bleeding were 5.4%, 3.6% and 2.0% respectively. However, the study also clearly pointed out that the quality of the evidence was relatively low, mainly due to the fact that the included studies were mainly retrospective studies, and the selection criteria for conversion surgery were heterogeneous, which might be biased towards patients with good tumor biological behavior.
Another meta-analysis published by Wu et al[94] in 2025 further supports the above conclusion. This study included 12 observational cohort studies to evaluate the efficacy differences between surgical and non-surgical treatments after conversion therapy for advanced gastric cancer. The results showed that conversion surgery significantly improved the overall survival rates at 1 year, 3 years and 5 years (RR 0.38, 0.64 and 0.77) as well as the progression-free survival rates at 1 year and 3 years (RR 0.57 and 0.67). There was virtually no difference in the incidence of adverse events between the two groups. However, consistent with the research of Chan et al[93], this meta-analysis also pointed out that the survival benefits in observational studies might reflect inherent prognostic differences, the surgical group was essentially chemotherapy responders, and their good biological behavior might partially explain the improvement in survival.
In the conversion therapy of liver cancer: Wang et al’s research[95] indicates that the 5-year overall survival rate of patients with unresectable liver cancer who successfully underwent resection after conversion therapy was 24.9% to 57.0%, which was comparable to the 30% to 50% survival rate of patients with initially resectable liver cancer. More notably, R0 resection after conversion therapy is an independent protective factor for the long-term survival of patients with unresectable liver cancer. However, the heterogeneous nature of advanced liver cancer has led to the lack of a unified consensus or guideline in this field. The selection of indications for conversion therapy, protocol optimization, and surgical timing remain the current research focus.
Another review published in 2023 also confirmed that patients who underwent radical surgery after conversion therapy had significant survival benefits compared with the chemotherapy alone group, the group with successful downstaging after conversion therapy but no surgery, and the group that directly received surgery[96]. However, research also indicates that the success rates of conversion therapy vary greatly, ranging from 0.8% to 60%. The conversion plan combining local treatment and systemic treatment has demonstrated significant clinical advantages, with a conversion rate of up to 60%, an objective response rate of 96%, and a disease control rate as high as 100%. It is worth noting that patients who have undergone conversion therapy experience more complications than those who have undergone direct surgery, including hepatotoxicity, bone marrow suppression, local adhesions, increased fragility of blood vessels and liver tissues, and liver edema, all of which increase the difficulty of subsequent surgeries.
Intraoperative navigation techniques in surgical operations
The continuous innovation of intraoperative navigation technology is an important technical driving force for promoting the precise transformation of gastrointestinal surgery[97]. Among them, indocyanine green (ICG) fluorescence imaging technology has become one of the most clinically valuable techniques in the field of gastrointestinal surgery due to its unique advantages in lymphatic tracing, anastomotic blood supply assessment and tumor boundary demarcation[98].
In 2025, the American College of Gastrointestinal Endoscopic Surgeons published a systematic review and meta-analysis on ICG fluorescence image-guided surgery, providing an important basis for the high-level evidence-based evaluation of this technique[99]. This meta-analysis included literature up to October 2022 (with the application of colorectal anastomosis extended to September 2024), covering eight prespecified key clinical questions, including thoracic duct identification, detection of distant cancer metastasis and primary cancer, lymph node acquisition, and anastomotic perfusion assessment, etc. The results of the meta-analysis showed that ICG fluorescence imaging reduced the leakage rate of colorectal anastomosis by 42% (OR = 0.58, 95%CI: 0.44-0.75). In colorectal anastomosis, ICG guided the change of the intraoperative cross-sectional point (OR = 35.15, 95%CI: 8.72-141.77), meaning that fluorescence imaging could change the intraoperative decisions of surgeons in real time. In terms of lymph node acquisition, ICG fluorescence imaging increased the average number of lymph nodes detected in gastrointestinal cancer surgeries by 6.32 (95%CI: 4.43-8.22). The conclusion of this review is clear: ICG fluorescence image-guided surgery has improved clinical outcomes in specific surgical applications, especially showing significant benefits in the acquisition of lymph nodes in malignant tumors and the reduction of colorectal anastomotic leakage. However, its effectiveness varies depending on the surgical context and clinical issues, and more high-quality studies are needed to fill the evidence gap.
In the field of colorectal cancer: A meta-analysis involving 15 studies and 3453 patients further verified the clinical value of ICG[100]. The results showed that the risk of anastomotic leakage in the ICG group was significantly reduced (RR 0.50, 95%CI: 0.37-0.67), the length of hospital stay was significantly shortened (standardized MD -0.31, 95%CI: -0.54 to -0.08), and the number of detected lymph nodes was significantly increased (standardized MD 0.19, 95%CI: 0.02-0.36). However, there were no statistically significant differences between the two groups in terms of operation time, intraoperative blood loss and the overall incidence of postoperative complications.
In the field of gastric cancer: The value of ICG fluorescence lymphangiography in lymph node dissection during minimally invasive gastric cancer surgery is increasingly prominent[101]. Studies have shown that ICG fluorescence-guided navigation technology can significantly increase the number of lymph nodes removed in gastric cancer patients, with particularly notable results for patients with early-stage gastric cancer who require precise lymph node staging[102,103]. This technique is particularly applicable to robotic or laparoscopic radical gastrectomy for gastric cancer, which can improve the accuracy of lymph node staging without prolonging the operation time or increasing complications[104]. It is worth noting that the application of ICG fluorescence imaging in the conversion therapy of advanced gastric cancer (assisting in the precise resection of primary tumors and liver metastases) is also a cutting-edge direction worthy of attention. Theoretically, it can bring a higher R0 resection rate to patients.
However, the ICG fluorescence imaging technique also has certain disadvantages. First, it is associated with a relatively complicated learning process as misinterpretation of the signals is possible due to the lack of standard procedures for the application of this method in practice. For example, the best protocol for the detection of gastric cancer lymph nodes has yet to be developed. Second, it is rather expensive and limited in terms of availability since only a few medical institutions possess the required equipment. Third, its application may vary depending on a particular case. Thus, for instance, the use of ICG reduces the likelihood of anastomotic leakage after colorectal surgery by 42%, but there are no long-term results showing a positive impact on life expectancy in gastric cancer.
CONCLUSION
This review systematically sorts out the three core dimensions of the precision and functionalization transformation of gastrointestinal surgery. In the field of minimally invasive surgery, robotic surgical systems have demonstrated multiple advantages in the minimally invasive treatment of gastric cancer and colorectal cancer, including reducing the overall incidence of postoperative complications, decreasing intraoperative blood loss, lowering the conversion rate to open surgery, and improving pathological quality indicators. However, high-quality RCTs in the existing evidence are still quite limited. The evidence certainty ratings of most meta-analyses are “low” or “very low”, and the problems of prolonged operation time and increased cost of robotic surgery have not been effectively addressed. In the field of ERAS, the ERAS pathway has accumulated sufficient evidence-based evidence in shortening hospital stays, reducing complications and readmission rates through multimodal perioperative optimization measures, and high compliance is the key guarantee for achieving its maximum clinical benefits. Meanwhile, the postoperative complication risk prediction model based on machine learning has demonstrated superior performance compared to traditional models in predicting specific outcomes such as overall complications and anastomotic leakage. Its collaborative integration with the ERAS pathway, by precisely predicting risks and guiding individualized perioperative management, represents the future direction of perioperative management. In the intersection of surgery and metabolic regulation, bariatric metabolic surgery achieves metabolic benefits beyond weight loss through a complex regulatory network of the “microbiota-bile acid-metabolism” axis. However, the differences in the mechanisms of action among different surgical methods and the establishment of the dominant mechanisms still require more high-quality research to clarify. Conversion therapy offers a bridge from “unresectable” to “radical resection” for patients with advanced gastric cancer and liver cancer. Existing meta-analysis evidence consistently indicates that conversion surgery can significantly improve long-term survival, but the issue of selection bias in observational studies urgently needs to be verified by RCT evidence. The clinical value of intraoperative ICG fluorescence navigation technology in lymph node acquisition and anastomotic leakage prevention has been supported by high-level evidence-based evidence, providing strong technical support for precision surgery.
Overall, the precision and functionalization transformation of gastrointestinal surgery is continuously advancing in three directions. Firstly, technological upgrades, such as robot platforms and intraoperative navigation technologies, are redefining the precision boundaries of surgery. However, the cost-effectiveness ratio and the lack of high-level evidence remain practical obstacles to their widespread application. Secondly, process optimization and the deep integration of the ERAS path with artificial intelligence risk prediction are expected to achieve a paradigm leap in perioperative management from “standardized plans” to “individualized strategies driven by risk stratification”. However, the lack of standardized assessment tools for ERAS compliance and insufficient prospective verification of artificial intelligence models are the main challenges currently faced. Thirdly, the boundaries are expanding. The metabolic regulatory mechanisms of bariatric and metabolic surgery and the survival benefits of conversion therapy are broadening the cognitive boundaries of traditional surgery, and interdisciplinary collaboration will become the core driving force for the continuous development of this field. In the future, the implementation of high-quality multi-center RCTs, the establishment of interdisciplinary collaboration mechanisms, and the systematic accumulation of real-world data will be the key paths to drive the above-mentioned innovations from evidence-based to clinical practice.
