Robotic Surgery for Biliary Tract Cancer

Simple Summary Complete surgical resection of biliary tract cancer provides the best chance at long-term survival. These surgeries are complex and can be associated with high morbidity and prolonged recovery. Minimally invasive approaches have been shown to improve some of these outcomes in other cancers. However, there is a paucity of data in biliary cancer, and even less is known about the outcomes of surgery utilizing a robotic platform. The present review reports the pooled outcomes of robotic surgery for 259 patients with biliary tract cancer. These outcomes are often equivalent to or improved compared to contemporary data on open surgery. The published data lack prospective or randomized studies. Thus, to evaluate any of the potential benefits of robotic surgery for biliary tract cancer, higher-quality studies are needed. Abstract Biliary tract cancer consists of cholangiocarcinoma (CC) and gallbladder cancer (GBC). When resectable, surgery provides the best chance at long-term survival. Unfortunately, surgery for these tumors is associated with long operative times, high morbidities, and prolonged hospital stays. Minimally invasive surgery has been shown to impact selected outcomes, including length of stay, in other diseases, and robotic surgery may offer additional advantages compared to laparoscopic surgery in treating bile duct cancers. This is a systematic review of robotic surgery for biliary tract cancer. Predetermined selection criteria were used to appraise the literature. The PRISMA guidelines were followed. In total, 20 unique articles with a total of 259 patients with biliary tract cancer undergoing robotic surgery met the inclusion criteria. For CC and GBC, respectively, the weighted average operative time was 401 and 277 min, the estimated blood loss was 348 and 260 mL, the conversion rate to open was 7 and 3.5%, the all-cause morbidity was 52 and 9.7%, the major morbidity was 12 and 4.4%, the perioperative mortality was 1.4 and 0%, the length of stay was 15 and 4.8 days, the positive margin rate was 27 and 9%, and the number of lymph nodes retrieved was 4.2 and 8. Robotic surgery for biliary tract cancer appears non-inferior to open surgery when compared to the published contemporary data. However, the current literature on the topic is of low quality, and future prospective/randomized studies are needed.


Introduction
Biliary tract cancer consists of intra-hepatic, peri-hilar, and distal cholangiocarcinoma (CC), along with gallbladder cancer (GBC). The prognosis remains poor (10% overall 5-year survival), and the best chance at long-term survival is surgical resection. Unfortunately, a large proportion of patients are not surgical candidates (80%) at diagnosis. However, in a select group of patients, median overall survival can reach 51 months after resection [1].
CC at the hilum may be classified by the Bismuth-Corlett classification which describes the extent of biliary duct involvement. Type 1 represents the biliary tumor located proximal to the confluence, Type 2 involves the confluence, Type 3a involves both the confluence and the right hepatic duct, Type 3b involves both the confluence and the left hepatic duct, and Type 4 extends into both the right and left hepatic bile ducts [2]. This classification may dictate the type of surgical resection that will be performed. An extra-hepatic biliary resection with biliary-enteric reconstruction is possible for type 1 and 2. However, type 3 will also require a hepatic resection of the caudate lobe or associated hemi-liver. Resections for type 4 tumors are limited by the future liver remnant and reconstructable ducts. Some tumors may also require vascular resection and reconstruction. Localized intrahepatic cholangiocarcinoma may be resected, and, finally, distal CC may require a pancreaticoduodenectomy.
GBC can be classified as incidental, referring to an intraoperative (diagnosis made on frozen biopsy after the gallbladder has been removed but before the surgery is finished) or on permanent section (after the gallbladder has been removed and the surgery finished), or as de novo, referring to the pre-operative diagnosis of GBC. TNM staging is critically important for prognosis but also for management. Incidental GBC with Tis or T1a tumors (limited to lamina propria), and with negative margins does not require an additional liver resection [3]. Surgical management of higher T-stages and de novo GBC, however, may require radical cholecystectomy, which consists of a cholecystectomy, hepatectomy (Segments 4B/5), portal lymphadenectomy, and negative cystic duct margin.
Recently, Sheetz et al. reported that among 169,404 patients, between 2012 and 2018, the use of robotic surgery surged from 1.8% to 15% [4]. Given these rapid changes, there are concerns regarding the broad and indiscriminate implementation of a new surgical platform with limited data. Whereas numerous comparative studies are ongoing for common surgical procedures, the use of a robotic platform for rare and more technically complex surgeries, including biliary tract cancer, is limited.
Indeed, in order to achieve an R0 resection for biliary tract cancer, a surgeon must have a complete understanding of the biliary anatomy and be ready to perform a biliary enteric anastomosis that carries a potential risk of stricture or leakage, a partial or major hepatectomy with risk of hemorrhage or post-hepatectomy liver failure, a vascular resection and reconstruction with risk of hemorrhage or ischemia, a thorough lymphadenectomy of the porta hepatis with risk of achieving inadequate lymph node sampling, and other associated procedures, such as hepatic ultrasonography and bile duct exploration. These technical aspects of the surgery can be challenging, and require proper training, exposure, retraction, and instrumentation, much of which is difficult to do minimally invasively with robotic surgery. Despite these challenges, the first robotic resection of CC was reported by Giulianotti et al. in 2010 [5], and the first robotic GBC resection was reported by Shen et al. in 2012 [6]. There may be some potential advantages to a minimally invasive approach in selected patients, however, and particularly a robotic platform as it pertains to bilio-enteric anastomoses given 3-D magnification and wristed instrumentation. In the present study, we aim to create the most up-to-date systematic review of the literature on robotic surgery for biliary tract cancer.

Search Strategy
A literature search was conducted by two independent researchers (LK and PB) and included all peer-reviewed manuscripts reporting on robotic interventions on cholangiocarcinoma and gallbladder cancer between the first reported case in March 2010 and December 2021. Exclusion criteria included: non-biliary pathology, no original data, no data on robotic technique, non-English manuscript, duplicate data, and staged procedures, e.g., ALLPS. If more than one study was published on the same cohort, only the study with the largest cohort size was included to avoid overlapping populations. The search and selection process, data extraction, and critical appraisal of the selected studies were conducted independently by LK and PB. In the event of discordant findings, a third reviewer was consulted to achieve consensus (AVM).

Data Extraction
The data extraction was pre-determined and included the author, Pubmed ID, study type, year of publication, number of patients, type of tumor (intra-hepatic vs. extra-hepatic for cholangiocarcinoma, Bismuth-Corlett for hilar cholangiocarcinoma, and T-stage for GB cancer), age, gender, body mass index, biliary stent placement, biliary drainage procedure, procedure type, use of ICG, operative time, estimated blood loss, conversion to open, all morbidity, major morbidity (Clavien-Dindo ≥ 3), mortality, length of stay, margin status, and number of lymph nodes retrieved.

Data Management
Data are reported as a mean with standard deviations for continuous variables and frequency with percentage for categorical variables. When a study did not provide these estimates but provided the raw data of individual patients enabling the calculation of mean and standard deviation, the results were annotated with a superscript letter "C". When a study provided median with interquartile range, these estimates were converted to mean with standard deviation using the method by Hozo et al. [7].
The weighted mean was calculated by: Wi with X representing the data values to be averaged, W the weights applied to each X value, and n the number of terms to be averaged. A meta-analysis was performed with the use of the Cochrane Review Manager 5.4.1. A random effect model was selected. The mean differences with the associated 95% confidence interval were computed. The results were reported in a forest plot.

Critical Appraisal
Independent critical appraisal was performed by two reviewers (LK and PB) using the Joanna Briggs Institute critical appraisal checklist for case reports and case series. After the resolution of any discrepancies between the two reviewers, a traffic light plot was created using the Robvis tool [8]. For the four comparative studies included in this study, we performed a Newcastle-Ottawa Scale critical appraisal. This systematic review was registered with the unique identifying number: "researchregistry7559".This study was performed in accordance with the Preferred Reporting Items for Systematic Reviews (PRISMA) guidelines.
There were two comparative studies for robotic CC. Yang et al. compared 70 robotic liver resections (only 6 of which included CC) to 252 open liver resections. They reported longer operative time with the robotic approach, but shorter length of stay and lower blood loss [12]. Xu et al. is the only study comparing the robotic (n = 10) and open (n = 32) approach strictly in patients with CC. They reported longer operative time (703 vs. 475 min) and higher morbidity (90 vs. 50%). There were three major morbidities and one mortality from post-hepatectomy liver failure in the robotic group (Table 2). Given these unfavorable outcomes, the group recommended against the robotic approach for hilar CC. However, it is worth noting that all patients in this cohort underwent extensive resection, including major hepatectomy with biliary resection and reconstruction [17]. BC "Bismuth-Corlette"; BMI "Body mass index"; Proc "Procedure"; ICG "Indocyanine green or reported using Firefly TM system"; Conv. "Conversion"; Mort. "same admission/30-day mortality"; LOS "Length of stay"; LND "lymph node"; CS "case series"; CR "case report"; RC "retrospective cohort"; M "male"; BS "biliary stent"; BD "biliary drainage procedure"; Cs "Intra-operative cholangioscopy"; iCC "intra-hepatic cholangiocarcinoma"; hCC "hilar cholangiocarcinoma". Superscripts: c Calculated, e Estimated, mp Limited data due to mixed pathology in study. Procedure types:  Among the 113 patients with gallbladder cancer, 13 were Tis or T1, 64 were stage T2 or greater, and 36 were unspecified. The weighted average age was 63 years old. Whereas 70 patients (62%) underwent en-bloc cholecystectomy with central hepatectomy and portal lymphadenectomy, 32 patients (28%) underwent completion central hepatectomy with lymphadenectomy post a prior cholecystectomy. In total, 11 patients had unspecified procedures. Three studies reported the intraoperative use of indocyanine green in 45 patients. The weighted average operative time was 277 min. The weighted average estimated blood loss was 260 mL. The weighted average conversion to open was 3.5%. The weighted average morbidity rate was 9.7%, while the weighted average major morbidity rate was 4.4%. There was no perioperative mortality reported. The weighted average length of stay was 4.8 days. The weighted average rate of margin positivity was 9%. In total, 9 out of 10 studies reported on the number of lymph nodes retrieved, with an average of 8 lymph nodes retrieved (Table 3) Table 2). A meta-analysis of those three studies revealed a decrease in estimated blood loss with a mean difference of 360 mL (95% CI: −762,38), and a significant decrease in length of stay with a mean difference of 3 days (95% CI: −4.9, −1.0). There was no clear difference in operative time (Figure 2).
The critical appraisal of all included studies was performed with the Joanna Briggs Institute checklist. Most case series did not have clear inclusion criteria, and few reported consecutive inclusions, resulting in an increased risk of selection bias ( Figure 3A). With regards to case reports, three out of four studies did not have clear description of potential adverse events ( Figure 3B). Furthermore, the quality of the four comparative studies included in Table 3 were also evaluated with the Newcastle-Ottawa Scale, and each scored 7 out of 9. The major limitation of these studies was the lack of adjustment for potential confounders. BMI "Body mass index"; Proc "Procedure"; ICG "Indocyanine green or reported using Firefly TM system"; Conv. "Conversion"; Mort. "same admission/30-day mortality"; LOS "Length of stay"; LND "lymph node"; CS "case series"; CR "case report"; RC "retrospective cohort"; M "male". Superscripts The critical appraisal of all included studies was performed with the Joanna Briggs Institute checklist. Most case series did not have clear inclusion criteria, and few reported consecutive inclusions, resulting in an increased risk of selection bias ( Figure 3A). With regards to case reports, three out of four studies did not have clear description of potential adverse events ( Figure 3B). Furthermore, the quality of the four comparative studies included in Table 3 were also evaluated with the Newcastle-Ottawa Scale, and each scored 7 out of 9. The major limitation of these studies was the lack of adjustment for potential confounders.

Discussion
The use of robotic surgery is expanding at an impressive rate. Adoption is, in part, attributed to several technical aspects in comparison to traditional open or laparoscopic surgery, such as improved "ergonomics, decreased surgeon fatigue, tremor filtration, seven degrees of motion, 3-dimensional vision," and overlay of near-infrared fluorescence (NIRF) imaging [27]. Indeed, Wee et al. recently published a systematic review supporting improved ergonomics and decreased physical strain on the surgeon and trainee with robotic surgery compared to both open or laparoscopic approaches [28]. However, any clinical benefits to the patients have not been well established. Indeed, as of January 2019, in all surgical fields, only 18 randomized control trials on robotic surgery had been published [27]. Furthermore, the financial costs and learning curve remain potential limitations for widespread adoption, particularly around highly technical procedures such as complex hepatobiliary surgery.
The present study is a systematic review of robotic surgery for cholangiocarcinoma and gallbladder cancer. Although there have been previous review articles on minimally invasive surgery for these pathologies, this is the first review strictly focused on a robotic approach. Given the known technical complexity, high morbidity, and mortality of biliary tract surgery for cholangiocarcinoma and gallbladder cancer, it is not surprising to find a limited number of studies on the topic reporting a robotic approach.
When evaluating other reported cohorts of open surgery for CC, the present pooled analysis for robotic surgery does not appear inferior. Indeed, reported in a cohort of 440 patients with hilar CC, Nuzzo et al. a morbidity rate of 47%, mortality rate of 10%, and an R0 resection rate of 77% [29]. Furthermore, the present pooled results for robotic surgery may be compared to the established international benchmark results published in 2021 [30]. Among the 708 cases of hilar CC qualifying as benchmark cases, the median operative time was 432 min, estimated blood loss was 852 mL, length of hospitalization was 16 days, rate all complications was 76%, rate of major complications 57%, in-hospital mortality was 3%, R0 rate was 57%, and lymph node retrieval was 4 nodes [30]. Tang et al. performed a meta-analysis of open vs. MIS surgery for hilar CC which included 9 studies (7 laparoscopic, 2 robotic) and revealed increased operative time and cost with MIS, but a decrease in blood loss, pain, and length of stay [31].
Similarly, when evaluating other reported cohorts of open surgery for GBC, the present pooled analysis for robotic surgery also does not appear inferior. Indeed, Cao et al. reported outcomes of 61 radical cholecystectomies for T1b/T2 GBC and reported an operative time of 216 min, estimated blood loss of 256 mL, length of stay of 11 days, and lymph node retrieval of 8. Furthermore, the same group compared their outcomes to another 61 radical cholecystectomies performed laparoscopically and did not identify any significant differences in perioperative outcomes [32]. Navaro et al. reported 43 open radical cholecystectomies for T2b GBC with an average operative time of 211 min, estimated blood loss of 208 mL, length of stay of 12 days, all morbidity of 11%, and 12 retrieved lymph nodes. Interestingly, this same group also compared these outcomes to laparoscopic radical cholecystectomy and reported a significant decrease in operative time, estimated blood loss, and length of stay. However, these benefits were at the cost of a significant decrease in lymph node retrieval [33].
In summary, the pooled perioperative outcomes of robotic surgery for biliary tract cancer appear non-inferior, if not improved, compared to contemporary studies reporting outcomes of open surgery for either CC or GBC. Certainly, the present study has numerous limitations. Given the limited amount of data on this topic, this review is based primarily on low-quality retrospective case reports or case series with high risk of publication bias and patient selection bias. Indeed, given the unconventional role of robotic surgery in biliary tract cancer, it is likely that the few patients selected to proceed with this novel surgical approach had good physiological reserve and anatomical features, which may account for the observed outcomes. Only four studies comparing robotic to open surgery for biliary tract cancer were identified, therefore limiting our meta-analysis to gallbladder cancer. Furthermore, there remains significant heterogeneity in the surgical management of biliary tract cancer with respect to the extent of resection, approach to biliary drainage, and biliary-enteric reconstruction, rendering published data difficult to compare. Finally, a major factor confounding operative outcomes of robotic surgery pertains to the surgeons' learning curve and prior experience with the technology. This is a factor that is often not reported and not able to be controlled for in this study.
There is no evidence to suggest that robotic surgery for complex hepatobiliary surgery is slowing down. It is therefore imperative to have higher-quality, prospective, comparative, and randomized studies on this topic. Furthermore, research needs to be dedicated to optimizing the learning curve of robotic surgery. Adjunct technologies that may further facilitate the robotic operator also need to be investigated and further developed. Indeed, NIRF imaging with indocyanine green (ICG) enables visualization and characterization of the biliary anatomy may assist with lymphatic clearance around the porta-hepatis, and with resection of the cystic duct margin. Moreover, NIRF imaging with ICG is an example of a tool that may be better applied on a minimally invasive platform than in a traditional open surgery [22,34].