Background

Perihilar cholangiocarcinoma (pCCA) is a devastating disease with an annual incidence of one to two cases per 100,000 individuals [1]. Radical surgery is the most important potentially curative treatment for pCCA. Meanwhile, radical surgery is also undoubtedly one of the most difficult and sophisticated skills for surgeons [2]. Recent advances in surgical techniques and perioperative management have offered increased resectability and improvement in surgical outcome; however, post-operative morbidities and mortality remain a problem [3].

Laparoscopic surgery (LS) has been increasingly used in all types of hepato-pancreato-biliary resections including pancreatectomy and hepatectomy [4,5,6]. In liver surgery, LS presented improved postoperative outcomes compared with open approach [7]. Recently, there has been increasing enthusiasm for performing laparoscopic pancreaticoduodenectomy (PD), which has been demonstrated to be feasible and safe, with outcomes that are comparable to open PD [8]. The surgical procedures for pCCA entail perihilar biliary tract resection with extended hepatectomy, which requires concomitant vascular resection and reconstruction or even combined hepatopancreatoduodenectomy. A major challenge is the complex anatomical location of the lesion, which is in close proximity to the portal vein (PV), hepatic arteries, and liver parenchyma [9, 10]. Thus, there are few reports on laparoscopic resection for pCCA due to the complexity of the surgical procedure. Till now, LS resection for pCCA was considered to be still in its infancy [11]. However, the available evidence on LS for pCCA is limited; LS cannot be considered as a dogmatic contraindication to pCCA. It thus is imperative to undertake large-scale multicenter analyses to investigate in the technical feasibility and safety of LS for pCCA [12].

Herein, we compared perioperative outcomes of LS and open operation (OP) in 654 pCCA patients in China before and after a propensity score-based analysis. To our knowledge, this is the largest series of LS compared with OP for pCCA to be investigated to date.

Methods

Patients and data collection

A retrospective review of institutional databases from 11 hospitals in China identified 645 pCCA patients who underwent radical surgery between January 2013 and January 2019, including 256 patients underwent LS and 389 underwent OP. An intention-to-treat design was used, such that cases converted to OP were included in the LS cohort. All cases were histologically confirmed pCCA. The exclusion criteria were as follows: (1) patients with peritoneal seeding or metastasis to the liver, para-aortic lymph nodes, or distant sites; (2) patients with non-adenocarcinoma histology; and (3) patients with incomplete clinical data. The detailed study flow was shown in Fig. 1 and Supplementary Table 1. Ethical approval was given for the study by the ethics committee/Institutional Review Board of Ethics Committee of Tongji Hospital (approval number: TJ-IRB20220531). The need for informed consent was waived by the ethics committee/Institutional Review Board of Ethics Committee of Tongji Hospital (approval number: TJ-IRB20220531) because of the retrospective nature of the study. This research was also registered at ClinicalTrials.gov (identifier: NCT05402618; date of first registration: 02/06/2022). All the work followed the ‘Sex and Gender Equity in Research-SAGER-guidelines’ and has been reported in line with the STROCSS criteria [13].

Fig. 1
figure 1

Study flow

Full size image

Operative technique

LS was defined as total laparoscopic surgery (TLS), where both resection and digestive reconstruction were completed laparoscopically. The hepatectomy with en bloc resection of the caudate lobe and extrahepatic bile duct and regional lymph node dissection was applied in both laparoscopic and open surgery. Extended resections (left/right trisectionectomy), arterial vascular encasement, vascular resection/reconstruction, biliary stent in place were needed for patients based on Bismuth stage.

Variables and definitions

Data collection included demographic, clinical and oncologic data. Patients were staged according to the National Comprehensive Cancer Network (NCCN) Guideline Version 2.2019 Edition tumor node metastasis (TNM) clinical staging system for pCCA [14]. The local extension of the disease was expressed mainly according to the Bismuth–Corlette classification [15]. Operation time was defined as the time from skin incision or trocar placement to complete skin closure. Intraoperative blood loss (IBL) was carefully recorded by the anesthetist using a vacuum system. Vessel reconstruction was defined as any repair or replacement of major vessels during surgery. Main short-term outcome in this study was postoperative length of stay (LOS), which was defined as the time from surgery to normal discharge. Patients who stayed longer than 30 days or less than three days will be considered as censored. Other short-term evaluations including morbidity and mortality, defined as any complication or death, respectively, which occurred during hospitalization or within 90 days of discharge. Hospital reoperation within 30 days was recorded and postoperative morbidity was evaluated according to the Clavien-Dindo (CD) classification system [16]. Major postoperative complications such as bile leakage, postoperative hemorrhage, and liver failure were classified as previously reported [17,18,19]. Surgical failure was defined as severe complication (CD ≥ III), or received reoperation during hospitalization, or dead within 30 days after surgery.

Statistical analysis

Patients that underwent LS were matched in a 1:1 ratio to patients that underwent OP based on the propensity score model [20], with age, BMI, sex, ASA physical status, CA 19–9, preoperative Tbil, AST, ALT, albumin, Bismuth-Corlett type, hepatectomy, operative time, digestive reconstruction, IBL, vascular resection and reconstruction, positive margin, number of lymph nodes, and TMN stage being selected as the covariates. The nearest neighbor matching without replacement was performed with caliper width setting at 0.2.

Continuous data are presented as means with standard deviations (SDs) or medians with interquartile ranges (IQRs) and compared with independent-samples t-test or Mann–Whitney U test. Categorical data are presented as numbers and percentages and were compared using the Chi-square or Fisher’s exact test, as appropriate. Logistic regression analysis was used to access the risk factors of adverse postoperative complications (CD stage ≥ III). Cox regression model were used to assess the risk factors associated with postoperative LOS, with the proportional hazard assumption being tested by weighted residual score method. Furthermore, considering the postoperative complication would prolong the LOS during hospitalization, the Fine-Gray model were used to explore the prognostic factors of length of stay, with the adverse postoperative complications and death within 30 days being the competing risk events. Continuous variables will be converted into two or multiple category dummy variables according to the normal range during the regression analysis. All the statistical procedures were conducted using SAS software version 9.40 (SAS Institute, Inc., Cary, NC). Two-sided hypothesis testing with a predetermined level of P < 0.05 was considered statistically significant.

Results

Participants and baseline characteristics

Baseline characteristics of demographics and preoperative comorbidity of the included 645 resections of histologically confirmed pCCA are listed in Table 1. In raw cohort, most baseline characteristics were comparable in LS and OP group, except for more female patients (49.61% vs. 40.36%; P = 0.021), better hepatic function index (lower TBIL, AST, and ALT levels), and lower TNM stage being observed in LS group than in OP group. To overcome the bias from retrospective study, the propensity score matched analyses was conducted, in which 141 pCCA patients undergoing LS were matched with 141 patients undergoing OP. All baseline characteristics were balanced after PSM.

Table 1 Baseline characteristics before and after propensity score matching
Full size table

Perioperative outcomes

Before PSM, 19 patients (7.42%) converted to laparotomy. The mean operative time of the LS group was similar with OP group (353.4 vs. 342.1 min, P = 0.176). More patients received hepatectomy in LS compared with OP (63.67% vs 55.53%, P = 0.013), especially more left hemihepatectomy in LS (47.66% vs 36.76%). Fewer hepaticojejunostomy (39.06% vs 61.44%, P < 0.0001), biliary plasty (28.91% vs 48.84%, P < 0.001), vascular resection (8.59% vs 23.39%, P < 0.0001) and fewer hepaticojejunostomy (43.48% vs 65.48%, P < 0.0001) were performed in the LS group compared with OP group. R0 rate, IBL, transfusion rate and volume, as well as caudate lobectomy were similar between groups (Table 2).

Table 2 Intraoperative outcomes before and after propensity score matching
Full size table

There were no significant differences with regards to the common postoperative complications between the two groups, except lower heart failure (0.78% vs. 3.60%, P = 0.0244) and severe complications in LS group (12.11% vs. 22.88%, P = 0.0006). Patients in LS group required shorter postoperative drainage tube keep time (PDTK) (median [IQR], 8.00[5.00 ~ 11.00] vs. 9.00[6.00 ~ 14.00], P < 0.0001), and postoperative LOS (median [IQR], 13.00[10.00 ~ 18.00] vs. 15.00[12.00 ~ 23.00], P < 0.0001) than OP patients. There was no significant difference on reoperation rate or death within 30 or 90 days. The similar incidence of some most common postoperative complications between two surgical groups were still comparable in matched cohort (Table 3).

Table 3 Postoperative outcomes according to different BMI level before and after propensity score matching
Full size table

Postoperative outcomes according to Bismuth type

In the present study, 328 patients of pCCA had a low Bismuth type (Bismuth I/II) and 317 had a high Bismuth type (Bismuth III/V), with significantly different surgical characteristics and postoperative short-term outcomes. The median LOS was much shorter in the low Bismuth type (15 day; 95%CI, 14 ~ 16) than in the high Bismuth type (18 day; 95%CI, 17 ~ 20, P < 0.0001). Furthermore, shorter LOS and PDTK, as well as less transfusion during surgery were observed in patients who underwent LS compared with those who underwent OP among patients with Bismuth types I-II, with other postoperative complications being comparable between the two groups. Among patients with Bismuth types III-V, those who underwent LS showed comparable or better short-term outcomes than the OP group, such as less total postoperative complications (39(32.23)86(43.88), P = 0.0393), shorter length of stay (median days, 14 vs. 17 days), and lower rates of severe complications (10.74% vs. 26.02%), demonstrating guaranteed safety of laparoscopic resection for high Bismuth types of PHC. Besides, a significantly less vascular resection and biliary plasty were observed in the LS group (Supplementary table 2–7).

Subgroup analysis for surgical failure and LOS

Subgroup analysis about the surgical failure and LOS was performed according to age, gender, BMI level, ASA score, tumor size, hepatectomy, vascular resection, and Bismuth stage. After PSM, the rate of surgical failure was similar in LS compared to OP group in total (4.30% vs 7.20%, P = 0.131) and in all subgroups (Fig. 2). However, the subgroup analysis about LOS showed that patients would benefit from LS for rapid recovery, regardless of age, tumor size and severe postoperative complications. Other patients, such as those who were underweight, had ASA score of I or III, or underwent vascular resection, had comparable LOS in LS and OP group (Fig. 3).

Fig. 2
figure 2

Forest plot of subgroup analysis regarding to surgical failure. Surgical failure was defined as severe complication (CD ≥ III), or received reoperation during hospitalization, or dead within 30 days after surgery.

Full size image
Fig. 3
figure 3

Forest plot of subgroup analysis regarding to length of stay

Full size image

Risk analysis of adverse surgical outcome and length of stay

The risk analysis showed that higher ASA score, more intraoperative blood loss, and more vascular reconstruction were identified independent risk factors of postoperative complications among both matched and unmatched samples (adjusted odds ratio (AOR) > 1, P < 0.05 for all) (Table 4). Besides, we found the OP, higher level of CA19-9, as well as some common postoperative complications, and reoperation were associated with longer LOS after adjusting other covariates in matched cohort (Table 5).

Table 4 The risk factors analysis of postoperative adverse outcomes according to univariate and multivariate analysis using raw cohort and matched cohort
Full size table
Table 5 The Cox proportional hazard model analysis of length of stay using raw cohort and matched cohort
Full size table

Furthermore, the Fine Gray model were conducted to further investigate the association between surgical methods and the LOS, with postoperative complications, received reoperation during hospitalization, or dead within 30 days after surgery being identified as competing events when analyzing the LOS. After adjusting competing risks, the open surgery was found to be an independent risk factor of prolonged LOS in both unmatched and matched analysis (Supplemental Table 8).

Discussion

The short-term outcomes of this retrospective study showed that the LS was favored in LOS and function recovery details such as the duration of postoperative drainage tube keep than OP for pCCA patients. Moreover, the mortality rates, postoperative complication rates, and oncological outcomes were not significantly different between two surgical methods. This is the largest real-world pCCA data to date comparing LS and OP for pCCA. It showed that LS, although challenging, is feasible and safe for pCCA patients. We also performed the PSM to eliminate potential biases and found that the LS group was beneficial for postoperative functional recovery, with no differences in complications and death rate compared to the OP group. The multivariate analysis further confirmed that LS was safe for perioperative complications and could significantly shorten the postoperative LOS.

Although several cumulative evidence with meta-analysis have shown the applicability of LS in pCCA [21,22,23,24], the technique is still restricted to a minority of highly experienced surgeons and specialized institutions, and lack of large sample evidence from multicenter. LS has been used for all Bismuth-Corlette types, although it is predominantly used in patients with low stage tumors [25,26,27]. The Bismuth III-IV accounted for 47.26% in this study, and equal safety and better postoperative recover, reflected by shorter LOS and PDTK, were also found in the LS group compared to OP group in this sample. Furthermore, the feasibility of the laparoscopic approach is attested by a conversion rate (19/256 cases, 7.4% before PSM; and 11/141 cases, 7.8% after PSM), which was much lower than the generally reported surgeries for pCCA [28] and also lower for major hepatectomies [29, 30]. In fact, although the surgical management of pCCA involves most complex technical procedures, such as portal infiltration or arterial encasement, or disease infiltration along the bile ducts, the requirements of an urgent conversion are detected with relative low frequency. Since the surgical field can be easily accessible to operator’s views and thus, some operative procedures with a high degree of technical complexity were more manageable in laparoscopy.

To date, only retrospective studies with small sample size reported the perioperative outcomes of LS and OP regarding the pCCA. The efficacy and safety of LS for pCCA is largely controversial [11, 12, 31, 32]. The only comparative study conducted by Xu et al. reported the robotic resection compared unfavorably to traditional open resection in operative time. No significant difference was found in blood loss, mortality, or length of postoperative hospital stay. While the hospital expenditure was much higher in the robotic group, and the tumor recurrence-free survival was inferior in the robotic group [33]. A study summarized a series of laparoscopic procedures showed there was significantly lower blood loss, fewer intraoperative and postoperative blood transfusions, and shorter LOS in LS group. Both overall and lymphadenectomy-related morbidity were lower in the LS group [34]. In this study, there were no significant differences in operative time, IBL, postoperative complications, or death between the two groups, but shorter LOS in LS group. Besides, the surgical methods, no matter open or laparoscopic, were not associated with adverse postoperative complications in both matched and unmatched samples. Furthermore, the LS showed to be a protective factor of adverse postoperative complication than OP in the univariate analysis. These evidence demonstrated that LS is safe and feasible for pCCA patients.

The most important potential advantage of LS is the increased adoption of postoperative functional recovery pathways and shorten hospital stay [35]. The LS technique can not only allow functional recovery for simple operations, such as LC and herniorrhaphy, but also for complex surgeries, such as LPD, laparoscopic hepatectomy, or laparoscopic gastrectomy [36,37,38,39]. Our study observed a protective effect of LS in shortening the LOS with or without considering the postoperative complications being the competing risk events. Perioperative complications would potentially influence the short-term outcome evaluation with life-threatening events, and thus affect the assessment of net cumulative effect of LS in shortening the LOS. Neglecting the existence of these competing risk events will lead inaccurate calculations of the cumulative discharge rate when using the classic survival analysis methods [40, 41]. Until now, few studies considered using the Fine-Gray model to explore the postoperative benefits from LS. PSM was another statistical method to control bias in this study. It has been proposed as a method to overcome selection bias and increase the evidence level in observational and randomized studies [42, 43]. In this study, we demonstrated that fewer hepaticojejunostomy, less vascular resection, shorter LOS, and a trend towards fewer severe complication (CD stage ≥ III) in LS group compared to OP before and after PSM. The results indicated that LS could reliably shorten hospital stays and provide benefits regarding postoperative functional recovery in pCCA patients.

Furthermore, we found that high ASA score, more intraoperative blood loss and hepaticojejunostomy were consistent risk factors of adverse surgical outcome. Besides, more intraoperative blood loss was identified as a risk factor for prolonged LOS in both the common risk model and competing risk model. These results indicated that better preoperative condition and less intraoperative blood loss were associated with better short-term prognosis among pCCA patients. During surgical management of pCCA, most technical complexities are related to the management of portal infiltration, or arterial encasement, or disease infiltration along the bile ducts. Along with these conditions and related risks of vascular injury are generally more manageable in laparoscopy since the surgical field is easily accessible being close to operator’s view and bleedings can be stopped with the use of a temporary clamp if emergency situation occurred. Therefore, the LS would be better to manage intraoperative bleeding, potentially bringing benefits for short-term outcome as recently reported [28, 44]. The preoperative condition of the patient is the most concerned issue in the determination of surgery. Patient selection is a common phenomenon in clinical practice, and it is also a key challenge that still puzzled most surgical experts. In our study, we can find that the rate of vascular resection, postoperative biliary fistula, and liver hepatic insufficiency are very low, suggesting an obvious patient selection prior to surgical management. In daily practice, patients with pCCA will be asked to receive ICG test, liver function testing, and 3D imaging examination to help surgeons calculate the liver volume, assess the liver function, and present exact anatomy of the lesion before surgery. Some surgeons would be relatively conservative to perform radical surgery for pCCA when encountering complex resection and reconstruction, especially for laparoscopic surgery. Despite the difficulty, enthusiasm is growing among surgeons regarding laparoscopic radical perihilar cholangiocarcinoma surgery. The selection of suitable patients for laparoscopic radical perihilar cholangiocarcinoma to help surgeons navigate the learning curve will be the focus of our future research.

To our knowledge, this is the largest case series of pCCA comparing LS and OS techniques. However, some limitations must be noted. First, the retrospective nature of the study is at risk of selection biases and unexpected recall bias cannot be completely ruled out. Second, the multicentric nature of the study is at risk of different selection criteria applied in the participating Institutions. While, due to the complexity of the procedure, there is a paucity of laparoscopic resection available now and this large-sample analysis can provide several important information about the laparoscopic procedures of pCCA. Third, all laparoscopic resections performed by experienced hands were still in the early and exploratory stages and is performed only in select cases. In addition, the centers included in this study were all high-volume referral centers in China. The experience may be difficult to generalize to the surgeons with less intensive training. Nonetheless, given the ongoing debate regarding the complicated procedure, and the current shift toward value-based healthcare reimbursement models, our findings are relevant and timely.

Conclusions

In summary, this extensive, multi-center study demonstrated that LS does not seem to increase the intraoperative and postoperative risks of pCCA. This report thus serves as a foundation for national protocols aimed at safely implementation of LS in pCCA patients. Larger prospective cohort series and prospective randomized studies in multiple countries are warranted to further investigate this topic.