Evaluation of outcomes of robotic liver surgery using the IWATE criteria.

The use of robotic technology in liver surgery is increasingly common and gaining acceptance among surgeons. Despite over 20 years of experience and a growing body of literature suggesting its advantages, there is still a need for more scientific data to fully validate its effectiveness, define its indications more clearly, and prevent potential pitfalls. Currently, the classification of surgeries in the literature is quite heterogeneous, often grouping together resections of different complexities. As a result, proposed classifications lack a clear clinical correlation, making comparison of results often difficult. In 2014, the Second International Consensus Conference on Laparoscopic Liver Resections (ICLLR) introduced the IWATE criteria to categorize the various difficulties encountered in laparoscopic surgeries ( 1 ). Our study seeks to evaluate whether the IWATE criteria are applicable to robotic liver surgery by comparing the outcomes associated with these criteria and establishing a clearer clinical correlation. We present this article in accordance with the STROBE reporting checklist (available at https://hbsn.amegroups.com/article/view/10.21037/hbsn-24-379/rc ).

Based on a dataset of 232 cases, we conducted a retrospective analysis concentrating on 188 liver resections executed by a single surgeon (P.C.G.) within the Division of Minimally Invasive and Robotic Surgery at the University of Illinois at Chicago. Forty-four cases were excluded from the original dataset, as they were performed at a different institution and are considered part of the surgeon’s learning curve. The study period extended from June 2007 to May 2023, and the analysis was limited to patients who underwent robotic liver surgery. All patients admitted to the medical center for elective liver resections were routinely assessed for robotic surgery and evaluated for general surgical risk and operability. The exclusion criteria for robotic surgery were similar to those for open hepatectomies and included general contraindications to anesthesia, cardiac or respiratory insufficiency, the need for vascular resections, and an American Society of Anesthesiologists (ASA) physical status higher than III. Liver resection for benign tumors was considered only in the presence of highly symptomatic lesions or when the malignant nature was unclear based on the preoperative work-up. Resection of colorectal liver metastases was deemed feasible if peritoneal carcinomatosis or unresectable extrahepatic disease was absent. Patients with hepatocellular carcinoma (HCC) were eligible for liver resections only if they had well-compensated cirrhosis (Child-Pugh class A/B, low grade), no signs of severe portal hypertension (esophageal varices less than F2), and a platelet count of at least 80×10 9 /L. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study received approval from the Institutional Review Board of the University of Illinois at Chicago (No. 2022-1008) (Title—Retrospective/Prospective Study of Surgical Outcomes). Inform consent was obtained from the patients. Comprehensive data pertaining to patients, surgical procedures, pathological findings, and postoperative outcomes were extracted from the electronic medical record. The information collected encompassed patient demographics, surgical specifics, pathological details, as well as 30-day mortality and morbidity. Surgical data included the duration of the operation, blood loss, conversion rates, reason for conversion, intraoperative and perioperative complications, and length of hospital stay (LOS). Operative time (OT) was calculated from the pneumoperitoneum induction to the closure of the last port incision, including the docking time for the robotic system. Intraoperative blood loss was determined as the balance between fluids instilled in the peritoneal cavity and aspirated fluids during the operation. Resection margins were evaluated only in cases of malignant pathology. Postoperative morbidity was categorized according to the Clavien-Dindo classification ( 2 ). Post hepatectomy liver failure and bile leak were defined according to the ISGLS classification ( 3 , 4 ). The liver resections were classified in accordance with the Couinaud classification established by the International Hepato-Pancreato-Biliary Association. Patients were divided into four groups according to the IWATE criteria ( 1 ), and outcomes were compared. The detailed surgical technique was previously described in another study ( 5 - 8 ). In general, an intraoperative laparoscopic ultrasound was always performed to localize the lesion, confirm/adjust the resection strategy and value the margins. The parenchymal transection was performed with the use of the robotic Harmonic scalpel and laparoscopic vascular staplers for the liver core. Procedures that reduced vascular inflow during resection, such as the Pringle maneuver, were not used. However, in cases of large hemangiomas, selective arterial clamping was performed. Intrahepatic parenchymal hemostasis was achieved through selective and precise control of bleeding points using techniques such as selective suturing, harmonic scalpel, clipping, or stapling, combined with superficialization/exposure of the section line (marsupialization technique). Categorical variables were described as proportions of the denominator population and reported as percentages. Continuous variables were tested for normality using the Kolmogorov Smirnov/Shapiro-Wilk test and presented as median [interquartile range (IQR)]. Univariate analyses were performed using Pearson Chi-Square for categorical variables and the Kruskal-Wallis non-parametric test for continuous variables. The association of morbidity (Clavien-Dindo) and the IWATE criteria was evaluated in three ways using three separate Pearson Chi-Square tests: (I) using the raw score; (II) grouping the score into minor and major complications; and (III) performing a subgroup analysis in the converted patients. For variables that were significant in the omnibus test, post-hoc pairwise comparisons were conducted using Dunn’s test for continuous variables, and Chi-Square test for categorical variables. To control for the inflated type I error rate due to multiple comparisons, P values were adjusted by false discovery rate (FDR) via the Benjamini-Hochberg procedure, with an FDR threshold of 5%. Cramer’s V for categorical variables and Spearman’s rank for continuous variables were used to assess the correlation between intraoperative/postoperative variables and IWATE criteria (low =1, intermediate =2, advanced =3, expert =4). The analysis was performed with SPSS Statistics software (IBM Corp. Released 2023. IBM SPSS Statistics for Macintosh, Version 29.0.2.0 Armonk, NY: IBM Corp.) and R (version 4.4.0). A P value <0.05 was considered significant.

A total of 188 cases performed between June 2007 and May 2023 were included in our analysis. Following the IWATE criteria, the four groups consisted of low, 13 patients; intermediate, 75 patients; advanced, 45 patients; expert, 55 patients. Table 1 describes the cohort characteristics. Age did vary significantly (P=0.04), with the Low group averaging 46 years and the Intermediate, Advanced, and Expert groups averaging 59, 56, and 57 years, respectively. Previous chemotherapy showed significant differences (P=0.046) across the expertise levels, with the Expert group having the highest percentage of individuals who had received prior chemotherapy (45.5%). IQR, interquartile range; BMI, body mass index; ASA, American Society of Anesthesiologists; LAR, low anterior resection. Table 2 provides an overview of lesion characteristics across different expertise levels. Malignant pathology is categorized into different subtypes, including colorectal metastasis, HCC, gallbladder cancer, cholangiocarcinoma, and other metastatic lesions (3 from pancreatic cancer, 2 from breast cancer, 2 from melanoma, 2 from lung cancer, 1 from neuroendocrine tumor, 1 from pheochromocytoma). The distribution of these malignant pathologies varied significantly among the expertise groups; more in detail, colorectal metastasis and gallbladder cancer showed significant differences (P=0.005, and P=0.001, respectively), with the Expert and Intermediate groups having the highest prevalence. Benign pathology includes adenomas, hemangiomas, focal nodular hyperplasia (FNH), cystic lesions, and other indications such as 2 liver endometriosis, 2 biliary hamartomas, 2 leiomyomas, 1 echinococcal cyst, 1 complicated liver cyst, 1 living donor, 1 perivascular epithelioid cell tumor. These indications exhibited variations across the expertise levels. Among these conditions, hemangioma showed significant differences (P=0.04), with the Advanced group demonstrating the highest prevalence. Data are presented as n (%). *, heavily symptomatic patients or unclear diagnosis. HCC, hepatocellular carcinoma; FNH, focal nodular hyperplasia. Table 3 outlines the surgical procedures undertaken, highlighting statistically significant distinctions among each class except for bisegmentectomy VI–VII. Data are presented as n (%). Regarding the intraoperative outcomes reported in Table 4 , there was a significant progression of the OT among the categories, from 148 minutes for the LOW class to 350 minutes for the EXPERT class (P<0.001). The same meaningful trend across the different categories was also observed in the estimated blood loss (EBL) (from 75 to 400 mL) (P<0.001) and the rate of intraoperative transfusions (from 0 to 14) (P=0.001). IQR, interquartile range; EBL, estimated blood loss. There were 18 conversions (9.6%), occurring exclusively in the EXPERT and Advanced groups, with a significant distribution (P<0.001). Among those cases, a distinction can be made between emergency conversions for acute conditions and elective conversions due to the inability to complete the operation minimally invasively in the best interest of the patient. Within the first group, four cases were connected to difficult bleeding control. In the second group, six cases were attributed to inability to achieve adequate exposure, three cases resulted from difficulties in assessing the oncological margins, four were prompted by diaphragmatic tumor infiltration near the vena cava, and one was caused by a stapler misfire on a hepatic vein, requiring conversion to ensure the safe removal of the stapler. Table 5 depicts the postoperative outcomes. The total median LOS was 5 days, varying from 3 to 7 days according to the group (P2). There were 5 minor bile leaks, two of them were managed in a conservative way leaving in place the surgical drain, two were treated with endoscopic retrograde cholangiopancreatography (ERCP) and biliary stent placement and in one patient a biloma was drained with interventional radiology. Four cases of post-hepatectomy liver failure were identified according to the ISGLS criteria. Among these cases, three were classified as grade A and were effectively managed through conservative treatment, while one case was classified as grade C and necessitated intensive care unit recovery. Only one case needed reoperation and it was due to a small bowel obstruction after a lateral sectionectomy for colorectal metastasis associated with a synchronous low anterior rectal resection. Seven patients required readmission in a 30-day follow-up, accounting for 3.7% of all cases. Among them, two were readmitted due to abscesses that were managed with interventional radiology drainage, one for hematoma, one for ascites which necessitated drainage, one for self-limited rectal bleeding, and two patients with a small bowel obstruction. Notably, we registered a zero 30-day mortality rate among all treated patients. IQR, interquartile range. Table 6 presents the pathology results, indicating a significant difference in the median major diameter across the categories. The overall R0 rate remained consistent at 96.3%, showing no statistically significant variation. IQR, interquartile range. In Table 7 is reported a post-hoc pairwise analysis performed on our results that confirms in most cases the significance of difference between groups, even though often it is not observed between Advanced and Expert groups. All P values are FDR-adjusted: *, P<0.05; **, P<0.01; ***, P<0.001. Age and previous chemotherapy did not show significance in post-hoc analysis after FDR adjustment. EBL, estimated blood loss; FDR, false discovery rate. A correlation analysis between IWATE criteria and Intraoperative/postoperative variables was performed ( Table 8 ); IWATE criteria was significantly correlated with intraoperative transfusion (r=0.29), conversion (r=0.37), overall OT (r=0.58), EBL (r=0.55), and LOS (r=0.48). We also investigated the association between the IWATE score and complications ( Tables 9,10 ), including within the subgroup of converted patients, but did not find any significant results. † , Cramer’s V correlation; ‡ , Spearman’s rank correlation. EBL, estimated blood loss. Data are presented as n (%). † , Clavien-Dindo I, II; ‡ , Clavien-Dindo IIIa, IIIb, IV, V. Data are presented as n (%). † , Clavien-Dindo I, II; ‡ , Clavien-Dindo IIIa, IIIb, IV, V. NA, not applicable.

Since the first worldwide series of robotic liver resections (RLRs) performed and documented by Giulianotti et al. in 2003 ( 9 ) there has been an increased interest in robotic approach to liver surgery even though the transition from open to minimally invasive surgery was not as fast as expected ( 10 ). This could be attributed to several factors as a steep learning curve, the lack of hepato-specific robotic instruments, the need for specialized training, and the risk of intraoperative vascular complications. Although there has been a delayed adoption of minimally invasive hepatic surgery due to organ-specific difficulties, robotic surgery has demonstrated a faster adoption rate compared to laparoscopy, likely due to the technical benefits associated with the robotic system ( 11 , 12 ). One of the challenges encountered worldwide so far is comparing results from different centers and defining the indications for robotic minimally invasive procedures more precisely. The lack of classification of different surgeries makes it difficult to compare results and evaluate the challenges associated with training and the progress of robotic liver surgery compared to open and laparoscopic approaches. Our goal is to establish a better correlation between surgical outcomes and the degree of difficulty associated with these surgeries. Ban et al. ( 13 ) initially introduced a difficulty score for laparoscopic liver resections. Subsequently, in 2014, during the 2nd International Consensus Conference on Laparoscopic Liver Resection (ICCLLR) held in Morioka, Japan, the IWATE criteria were developed ( 1 ). These criteria utilize six preoperative factors to assess the difficulty level of a hepatectomy. Specifically, they consider tumor size and location, proximity to major vessels, extent of liver resection, the use of Hand Assisted Laparoscopic Surgery or hybrid techniques, and liver function. These criteria have been largely applied to laparoscopic liver resections and have been recently used for robotic surgery with different conclusions ( 14 - 18 ). One advantage of our study, which involves a single-center experience with a single surgeon, is the elimination of variabilities connected with different individual surgical skills. Another advantage is represented by the exclusion of cases done while overcoming the learning curve. In this study, the learning curve was acquired during the first forty-four cases performed at another Institution, excluded from this analysis. The benefit of eliminating the learning curve is that the perioperative outcomes in the analyzed groups reflect the specific complexity of the procedure rather than differences in the level of training among surgeons. Hence, validating a classification system like the IWATE, which correlates with the complexity level, could streamline the adoption of a gradual approach to conducting RLRs. Such a classification could serve as a valuable tool for guiding surgeons in training, providing them with clear indications of procedural complexity levels ranging from low to expert. By delineating these levels, a stepwise approach to RLRs can be established, offering a structured pathway for skill development, and minimizing the likelihood of procedural complications. Moreover, the establishment of this classification system holds significant importance in facilitating comparative analyses across multicenter international studies. With a common framework in place, researchers could more accurately assess and compare outcomes, morbidity rates, and mortality rates across different centers, regions, and techniques. This standardization not only enhances the reliability of study findings but also fosters collaboration and knowledge sharing within the global surgical community. Our study yielded significant differences in operative outcomes, underscoring the validity of the classification system employed. Specifically, we observed significant variations in OT and EBL among the analyzed groups. Comparison of these findings with existing literature revealed that OT was similar in low and intermediate-complexity cases but slightly prolonged in more difficult cases ( 14 , 15 ). EBL was slightly different among the compared studies. Xie et al. ( 19 ) reported similar OT for robotic procedures in low and intermediate-complexity cases, although in their analysis, laparoscopic and open resections exhibited shorter OTs. Moreover, EBL was higher in the open resection group but akin to our results in the robotic technique. Furthermore, our analysis indicated that in complex cases, OT in robotic hepatectomies closely mirrored our results, although slightly longer than in the open resection group. EBL was comparable between the two studies within the robotic procedures and smaller when compared to open liver resection (OLR). When analyzing the conversion rate, we found that this data was lower in the Intermediate group compared to the Advanced and Expert group. Our conversions were predominantly elective and only a small number due to acute conditions. These conversions happened solely in the advanced and expert groups (mostly in the latter). The initial robotic approach also in these converted cases offers some advantages, allowing some of the steps to be performed minimally invasively (mobilization of the liver, intraoperative ultrasound, localization of the lesion, hilar dissection, and control). An increased adoption of planned “hybrid” liver surgeries may potentially lead to an expanded number of less invasive liver resections. This trend for hybrid approaches is also emerging in other surgical fields ( 20 ). In our analysis of post-operative outcomes, we found a significant association between LOS and the advanced and expert groups. However, the post-hoc pairwise analysis did not reveal any significant differences between these two groups. This outcome is linked to the higher intensity of care needed by these patients, which naturally leads to longer hospital stays. The observed low rate of complications could be attributed to the technical advantages afforded by the robotic system in liver surgery. Notably, magnified 3D vision and stability of the system allow the definition of anatomical details, continuous dynamic retraction facilitated by the fourth arm throughout the procedure, tremor filtration, and the intricate endo-wristed movements enabling a superior level of precision, particularly in managing vascular and hilar structures ( 5 ). The remarkable precision of this system allows for the minimization or highly selective application of ischemic inflow control like the Pringle maneuver. Regardless of the ongoing debate in the literature concerning the risk-to-benefit ratio of the Pringle maneuver, it remains a fact that avoiding intraoperative generalized parenchymal ischemia may better preserve the functionality of the residual liver tissue ( 21 ). This can mainly be achieved with controlled blood loss and a low rate of transfusions, as demonstrated in the present study. The lack of statistically significant differences in complications between the groups could be attributed to the current sample size. With a larger sample, the trend of increasing complications may become statistically significant. Therefore, further studies with larger sample sizes are needed to confirm these findings. Furthermore, in our series, we observed a very small number of R1 resections. Only 5 out of 135 specimens presented positive margins and were classified as Advanced/Expert, while no positive margins were observed in resections with Low/Intermediate complexity. The small number of procedures doesn’t allow for a precise analysis. Although not statistically significant, this data suggests that resections with a lower level of difficulty are associated with a very low risk of positive margins. This further suggests that the surgical indications for Low/Intermediate surgeries classified with the IWATE criteria would result in low rates of R1 resections and could guide the surgeon according to their level of expertise. In the post-hoc pairwise analysis, we found that the difference between the Advanced and Expert groups is often not statistically significant. This may be due to the sample size or the ability of the robotic system to smooth out the challenges typically encountered in traditional and laparoscopic surgery, as mobilization of the liver and clear vision of the dissection line, thus narrowing the gap in difficulty between these two groups. However, additional data are required to validate this hypothesis. In literature, there are significant series of laparoscopic liver resections where the IWATE classification has shown a valid correlation ( 22 , 23 ). If it is confirmed that the IWATE criteria demonstrate the same level of correlation in robotic surgery, adopting this classification system will provide a better opportunity for a more homogenous comparison of robotic and laparoscopic techniques. This study has some limitations, specifically the moderate number of treated patients, the bias connected with a single operator, and its retrospective nature.

In summary, the increasing transition to robotic liver surgery underscores the need to adopt a standardized classification system like the IWATE criteria. Our study supports the validity of this classification, showing consistent results in perioperative outcomes. While the technical advantages of robotics can contribute to low complication rates, barriers to wider adoption persist due to specialized training, a difficult learning curve, and a lack of specific robotic instrumentation. Nevertheless, implementing a classification system based on complexity could facilitate a stepwise approach to robotic liver surgery, potentially overcoming adoption barriers and enhancing patient care.

The article’s supplementary files as