Preoperative Liver Function as a Predictor of Postoperative Liver Failure in Perihilar Cholangiocarcinoma

Abstract

Background/Aim

Operative mortality following surgical intervention for perihilar cholangiocarcinoma (PHCC) is attributable to clinically relevant liver failure (CRLF). Nonetheless, the risk factors for CRLF in patients undergoing preoperative portal vein embolization (PVE) remain undetermined. This study aimed to investigate the risk factors for CRLF following major hepatectomy with PVE in patients with PHCC.

Patients and Methods

We retrospectively assessed patients with PHCC. Post PVE, a value for rate of plasma disappearance of indocyanine green×residual liver volume (%) (ICGKF) ≥0.05 warranted curative resection.

Results

In total, 346 patients were reviewed. CRLF developed in 27.5% of the patients. Operative time >660 min, blood loss >1,900 ml, and procedures other than left hepatectomy were independent risk factors for CRLF. Subsequently, 167 patients undergoing PVE were included in the analysis. Pre-PVE ICGKF <0.05, long operative time, and high intraoperative bleeding were independent risk factors for CRLF. The mortality rate for patients with all three risk factors was 25%. The incidence of CRLF after hepatectomy with extended bile duct resection for perihilar cholangiocarcinoma was 27.5%.

Conclusion

Prolonged surgery, and high blood loss are associated with an elevated mortality rate following PHCC surgery, particularly when accompanied by low pre-PVE ICGKF.

Introduction

Post-hepatectomy liver failure (PHLF) is one of the most critical complications after surgery for perihilar cholangiocarcinoma (PHCC) (1,2). Previous studies have reported the incidence of PHLF after PHCC surgery, that is, major hepatectomy with extrahepatic bile duct resection (EBDR), as being 17-24% (2-5). Reportedly, the mortality rate after PHCC surgery was 2-17%, with most cases attributed to PHLF (2-4).

Marked differences in postoperative clinical courses have been noted between cases undergoing simple hepatectomy and those who underwent hepatectomy with EBDR. Furthermore, the frequency of PHLF is higher after hepatectomy with EBDR than that after simple hepatectomy (6). Takagi et al. (6) reported that EBDR might adversely affect the postoperative outcomes of hepatectomy with EBDR as the liver regeneration rate was considerably lower than that after simple hepatectomy. Furthermore, their study demonstrated in a rat model that the mechanism for regeneration was slower after intrahepatic cholangiojejunostomy in a rat model (7). Therefore, simple hepatectomy and hepatectomy with EBDR must be considered separately in the discussion of the incidence or clinical course of PHLF.

A few postoperative complications, including post-operative bleeding, and sepsis, can cause severe liver dysfunction and even mortality in the late postoperative period (8). In 2021, van Keulen et al. (9) described clinically relevant liver failure (CRLF) after hepatectomy with EBDR and classified CRLF into two types: primary and secondary. They defined primary CRLF as PHLF without any underlying postoperative cause and secondary CRLF as PHLF that was caused by an underlying postoperative complication. They evaluated 253 patients who underwent hepatectomy with EBDR for presumed PHCC and reported the incidence of CRLF to be 20.9% and the 90-day mortality rate to be 17.0%. Since then, as only one study has reported CRLF after hepatectomy with EBDR, which was divided into two types of CRLF, the frequency of and risk factors for primary and secondary CRLF remain unclear.

Most studies have analyzed the risk of liver failure for all types of hepatectomy (2,10,11). However, left lobectomy (H1/2/3/4/-B) has reportedly been associated with a significantly higher residual liver volume and a lower risk of developing liver failure compared with other techniques. Therefore, preoperative portal vein embolization (PVE) is often used to increase residual liver volume, especially in Asia, when procedures other than left lobectomy are performed (2,12,13). However, remarkably few studies have evaluated the risk of liver failure limited to PVE cases (14).

This retrospective cohort study aimed to determine the incidence of and risk factors for CRLF after hepatectomy with EBDR for PHCC after PVE.

Patients and Methods

Ethics. This retrospective study was approved by the Institutional Review Board of Hokkaido University (020-0076), conducted according to the Declaration of Helsinki, and registered at the University Hospital Medical Information Network Center Clinical Trial Registry (UMIN000040384).

Data collection. The data for this study were collected and analyzed from June 2023 to December 2023.

Hepatectomy type. The types of hepatectomy were defined based on the New World terminology (15).

Patients. This study included 346 consecutive patients with PHCC who underwent hepatectomy with EBDR with or without pancreaticoduodenectomy at Hokkaido University (Department of Gastroenterological Surgery II) between January 2000 and December 2022.

Preoperative diagnosis. Previously described methods for preoperative PHCC diagnosis were used (16). Briefly, PHCC was diagnosed based on clinical presentation, imaging studies using enhanced computed tomography (CT), magnetic resonance imaging, direct cholangiography, and biopsy samples.

Biliary drainage. Previously described biliary drainage procedures were used in this study (16). The aim of preoperative biliary drainage was complete drainage of the future remnant liver. Preoperative biliary decompression was performed to reduce the serum bilirubin level to <2 mg/dl in all patients with jaundice and to control segmental cholangitis.

Assessment of remnant liver function, PVE, and surgical indications. The liver function of all patients was preoperatively evaluated using methods similar to those used by the Nagoya University Group (11). To evaluate whole liver function, the indocyanine green (ICG) retention rate at 15 min and plasma clearance rate of ICG (ICGK) were calculated using linear regression analysis of plasma ICG concentration, as previously described (11). The future remnant liver volume (%RLV) was calculated using preoperative CT scan volumetry. ICGKF was calculated using the following formula: %RLV×ICGK.

Preoperative PVE. Preoperative PVE of the liver lobe to be resected was indicated when subsequent segment, S1/5/6/7/8-B (right hepatectomy), S1/4/5/6/7/8-B (right trisectionectomy), or S1/2/3/4/5/6-B (left trisectionectomy), was decided. We did not perform PVE in some patients with postal vein stenosis or occlusion by tumor. The PVE methods used at our Institution have been described elsewhere (16). Two weeks after PVE, ICGKF was re-evaluated, and surgery was performed. The scheduled surgery was postponed when the ICGKF value did not meet the Nagoya criterion i.e. ICGKF ≥0.05. CT volumetry and ICG tests were performed every 2 weeks until the ICGKF value satisfied the criterion (11). For a few patients with ICGKF <0.05, surgery was performed only if the patient provided informed consent.

Surgery. Surgery for PHCC was performed using previously described procedures (16). Briefly, radical resection included en bloc excision of the liver hilum with (extended) hemi-hepatectomy, including the caudate lobe in most cases; complete lymphadenectomy of the hepatoduodenal ligament; and excision and reconstruction of each or both, portal vein and hepatic artery, if needed. In addition, selected patients with invasive cancer in the intrapancreatic bile duct underwent hepatopancreatoduodenectomy.

Postoperative management. We used fresh frozen plasma in cases where patients had hemorrhagic diathesis or a prothrombin time and international normalized ratio of >1.8. Patients with edema and pleural effusion or ascites were treated with diuretics until their weight returned to their respective preoperative levels.

Definitions of postoperative complications. Postoperative complications with a Clavien-Dindo classification of grade IIIa or more were defined as morbidity (17). All in-hospital mortalities after surgery were defined as mortality.

Definition of CRLF. The International Study Group for Liver Surgery criteria were used in this study, and only grades B and C were considered clinically relevant (18). Primary and secondary CRLFs were defined based on the definitions of a previous study (9). Definition of these CRLF were as follows: primary liver failure was defined as liver failure without any underlying postoperative cause; and secondary liver failure was defined as liver failure caused by an underlying postoperative complication.

Statistical analyses. Statistical analyses were performed using JMP version 15.2 (SAS, Cary, NC, USA). Chi-squared test, Fisher’s exact test, and Mann-Whitney U-test were used, as appropriate. A value of p<0.05 was considered statistically significant. Clinical characteristics of CRLF were assessed using the receiver operating characteristic (ROC) curve analysis, and the area under the curve (AUC) for these factors were compared. The association between two parameters was analyzed using Pearson product-moment correlation coefficient. All variables with a value of p<0.1 in univariate analysis were included in a multivariate analysis using a logistic step-by-step regression model. Results are presented as odds ratios (ORs) with 95% confidence intervals (CIs).

Results

Patients. We evaluated 349 consecutive patients with PHCC, with or without pancreaticoduodenectomy. Three patients with early mortality (until postoperative day 3) due to postoperative bleeding were excluded from the analysis because their clinical data were not available to evaluate CRLF. Therefore, we finally assessed 346 patients. Table I summarizes the patient characteristics, operative procedures, and surgical outcomes.

Table I. Patient characteristics and univariate analysis of clinically relevant liver failure (CRLF).

HPD: Hepatopancreaticoduodenectomy; ICGK: rate of plasma disappearance of indocyanine green; ICGKF: ICGK×%future remnant liver volume (RLV) calculated by computed tomography; PVE: portal vein embolization; S: liver segment. Hepatectomy type: C: central; L: left; L3: left trisectionectomy; R: right hepatectomy; R3: right trisectionectomy; Morbidity: Clavien-Dindo classification III-V; Statistically significant p-values are shown in bold.

Incidence of and risk factors for primary/secondary CRLF. Of the 346 patients, 87 (25.1%) met our criteria for CRLF, and 15 patients died due to CRLF.

Table II illustrates the timing and number of patients with CRLF. The causes of secondary CRLF were vascular thrombosis in 3, postoperative bleeding in 9, and infectious complications, such as cholangitis, abscess formation, and pneumonia in 20. The mortality rates due to primary and secondary CRLF were 12.7% and 25%, respectively. There was no significant difference regarding mortality rate between primary and secondary CRLF (p=0.144).

Table II. Timing and causes of clinically relevant liver failure (CRLF) (N=87).

*Mortality is shown in parentheses.

Risk factors for CRLF (entire cohort). The ROC curve analysis showed the cut-off values for the risk factors for CRLF to be operative blood loss >1,900 ml (AUC=0.743), operative time >680 min (AUC=706), and RLV <56% (AUC=677). ICGK and ICGR15 had low AUCs (0.543 and 0.531, respectively). Although preoperative ICGKF had a moderately high AUC (0.692), the cut-off value (0.0876) was inadequate for clinical usage, and %RLV was a confounding factor. Figure 1 shows the mortalities and preoperative ICGKF.

Figure 1.

Distribution of mortality and the preoperative ICGKF (rate of plasma disappearance of indocyanine green×% remnant liver volume) for the entire cohort undergoing major hepatectomy with extrahepatic bile duct resection for perihilar cholangiocarcinoma. ICGKF data were missing for seven patients. N indicates the total number of patients in each group. The plot illustrates that lower preoperative ICGKF values were associated with elevated postoperative mortality, however, a high proportion of patients with ICGKF ≥0.07 died. The findings imply that ICGKF should be regarded not as a definitive indicator of achieving zero mortality.

Table I summarizes the risk factors for 87 patients with CRLF. Univariate analysis revealed that preoperative biliary drainage, preoperative RLV<56%, preoperative PVE, operative time of >680 min, operative blood loss of >1,900 ml, hepatectomy other than left hepatectomy (H1/2/3/4/5/8-B), concomitant portal vein resection, and hepatopancreaticoduodenectomy were risk factors for CRLF. However, the type of hepatectomy (H1/2/3/4/5/8-B vs. all others), preoperative biliary drainage, PVE, preoperative ICGKF, hepatopancreatico-duodenectomy, and concomitant portal vein resection were confounding factors. Multivariate analysis revealed that an operative time of >680 min (OR=2.94, 95% CI=1.63-5.42, p=0.007), operative blood loss of >1,900 ml (OR=2.99, 95% CI=1.71-5.31, p<0.001), and hepatectomy other than left hepatectomy (H1/2/3/4/5/8-B) (OR=5.05, 95% CI=2.77-9.70, p<0.001) were independent risk factors for CRLF.

Next, we evaluated the risk factors for primary and secondary CRLF. Univariate and multivariate analyses showed that operative factors (operative time and bleeding) and preoperative cholangitis were independent risk factors for primary CRLF. Furthermore, preoperative liver volume (RLV <56%) and operative time were independent risk factors for primary CRLF.

Risk factors for CRLF in PVE cases. We evaluated the risk factors for CRLF for all patients who underwent PVE. Results are presented in Table III. One hundred sixty-seven patients who underwent PVE were assessed. ROC analysis revealed that the cut-off values of pre-PVE ICGKF and preoperative ICGKF were 0.05 (AUC=60) and 0.087 (AUC=57), respectively. Figure 2 shows the mortality and preoperative ICGKF. The interval between PVE and surgery was not associated with CRLF (AUC=430).

Table III. Risk factors for clinical liver failure after portal vein embolization.

ICGK: Rate of plasma disappearance of indocyanine green; ICGKF: ICGK×%future remnant liver volume (RLV) calculated by computed tomography; PVE: portal vein embolization. Morbidity: Clavien Dindo classification III-V. Statistically significant p-values are shown in bold.

Figure 2.

Distribution of postoperative mortality according to preoperative ICGKF (rate of plasma disappearance of indocyanine green×% remnant liver volume) among patients undergoing portal vein embolization (PVE). N indicates the total number of patients in each group. Pre-PVE ICGKF data were missing for 16 patients. The figure demonstrates that mortality was not limited to cases with low ICGKF; several deaths also occurred in patients with preoperative ICGKF ≥0.08. Thus, while lower ICGKF values are associated with an increased risk of postoperative death, defining surgical indication based solely on ICGKF remains difficult.

Univariate analysis revealed that pre PVE ICGKF <0.05, operative time of >680 min and operative blood loss of >1,900 ml were risk factors for CRLF (Table III). Multivariate analysis showed that these factors were independent risk factors for CRLF: pre PVE ICGKF <0.05 (OR=2.66, 95% CI=1.19-5.97; p=0.018), operative time of >680 min (OR=2.81, 95% CI=1.27-6.25; p=0.011), operative blood loss of >1,900 ml (OR=3.44, 95% CI=1.61-7.37; p=0.001), respectively.

As risk factors accumulated, the likelihood of developing CRLF increased, and the risk of death increased when three of the extracted risk factors overlapped (Figure 3). The mortality rate with the concurrent presence of the three risk factors was 25%.

Figure 3.

Mortality rate according to the accumulation of independent risk factors for clinically relevant liver failure (CRLF) after hepatectomy with portal vein embolization (PVE) in patients with perihilar cholangiocarcinoma. The presence of all three risk factors, namely pre-PVE ICGKF <0.05, operative time >680 min and intraoperative blood loss >1,900 ml, was associated with a 25% postoperative mortality rate.

Discussion

This study demonstrated that the incidence rate of CRLF after hepatectomy with EBDR was 25%. One out of three cases of CRLF was caused by postoperative complications (secondary CRLF). CRLF is associated with surgical factors and liver function (ICGKF). Our results indicated that among the cases in which surgery was performed based on the criterion of preoperative ICGKF ≥0.05, those with ICGKF <0.05 before PVE were at particularly high risk, and when this risk was compounded with surgical factors, the risk of death notably increased.

Patients who undergo hepatectomy with EBDR have a significantly higher morbidity rate than those who undergo simple hepatectomy; this can be attributed to multiple reasons (6). Hepatectomy with EBDR for perihilar malignancies requires more complicated procedures (lymphadenectomy around the hepatoduodenal ligament, caudate lobe resection, intrahepatic cholangiojejunostomy, and vascular resection) and longer operative time. Intrahepatic jejunostomy has adverse effects on liver regeneration after surgery (6). Furthermore, postoperative complications, such as surgical site infection followed by conditions such as anastomotic insufficiency and pancreatic fistula, can occur a few weeks after surgery (19,20). Therefore, it is important to determine the incidence of and risk factors for CRLF after hepatectomy with EBDR, which differ from those of simple hepatectomy.

Our results in this study did not contradict the usefulness of PVE. Several previous studies have shown that PVE reduced mortality in PHCC surgery (5,13,21). Our previous study comparing the difference between eastern and western high-volume centers for PHCC demonstrated the differences in several clinical procedures and results (5). In particular, mortality rates differed between the Eastern and Western institute. The lower PVE rate in the studies based in Western centers, which was far below that of our study, would explain the differences in mortality. As a follow up to our study, Olthof et al. (21) evaluated 1667 patients with PHCC in 20 high-volume Western centers and found that only 20% of the patients underwent PVE. Moreover, they demonstrated that PVE before hepatectomy with EBDR was associated with a lower incidence of PHLF, biliary leakage, abscess formation, and mortality. Franken et al. (13) also reported that PVE reduced postoperative mortality (from 16.3% to 2%) in Western high-volume centers. These data support the clinical benefits of PVE for patients with small remnant volume (5,21).

However, a previous experimental study raised the following question: Does the increase in liver volume achieved after PVE or in associating liver partition and portal vein embolization for staged hepatectomy contribute to improved liver function (22). These procedures can increase RLV over a short period of time. However, experimental studies have shown that the liver undergoes temporary ‘edema’ when stressed (22). This is due to fluid movement into a third space, such as the Disse space, hepatocyte interstitium, and Gleason’s sheath, accompanied by sinusoidal endothelial cell swelling and fenestration widening. This can lead to increased liver volume and may not reflect liver regeneration and improved liver function. In contrast, other groups demonstrated that PVE can lead to hepatic microvascular density, and oxygenation and vascular endothelial growth factor increased by PVE (23). However, clinically, PVE plays an important role by reducing the risk of liver failure.

Our data suggest that the interval between PVE and surgery had no significant association with incidence of CRLF. We were of the opinion that there may be a bias pertaining to the interval after PVE because we tended to extend the time from PVE to surgery in high-risk patients, such as those who developed cholangitis or had concerns about insufficient liver volume. Therefore, the results of this study cannot be evaluated independently of this bias.

The assessment of liver function before surgery is also an issue. Preoperative assessment prevents CRLF. The difference in the results between Eastern and Western centers seems to be related to whether PVE was performed and whether or not liver volume was assessed. van Kuren et al. (9) reported that preoperative liver volume assessment was crucial for reducing CRLF incidence, as approximately 30% of the patients in their study had not undergone preoperative liver volume assessment. In their study, only 53.4% of the patients underwent both RLV and functional evaluation.

ICG has been used widely to evaluate hepatic functional reserve before hepatectomy in Japan. It provides valuable information for clinical decision-making concerning indications or the extent of hepatic resection required in patients with damaged livers (11). Preoperatively, which means post PVE, ICGKF is widely used in Japanese high-volume centers for PHCC. Nagino et al. (24) advocated patients with ICGKF ≥0.05 as suitable candidates for hepatectomy with EBDR. Yokoyama et al. (11) reported a validation study of this criterion. They showed that although the mortality rate was significantly lower in patients with ICGKF ≥0.05, postoperative deaths still occurred when ICGKF was 0.05-0.08. The present study’s ROC curve analysis revealed the accurate cut-off value of ICGKF as 0.0876 for postoperative mortality in the whole cohort. Although this AUC value is remarkably low and can only be used as a reference, the cut-off value was also 0.0876 in the study of cases in which PVE was performed. If an ICGKF of >0.08 or 0.09 were used for patient selection, more than half of the patients would not undergo surgery owing to expected mortality. The number of patients with ICGKF ≥0.08 who died after surgery was not negligible (n=2) (Figure 2). Therefore, defining the surgical indication completely using only the ICGKF is difficult. Another possible problem is that ICGKF assumes that all liver functions are the same. Uesaka et al. (25) suggested differences in function of the left and right sides of the liver, especially after PVE. Therefore, whether ICGKF is suitable for evaluating remnant liver function is unclear. Consequently, an alternative modality for patient selection is required. We previously reported an alternative preoperative method for liver function evaluation widely used in Japan: ^99mTc-galactosyl human serum albumin liver scintigraphy helps predict postoperative mortality after hepatectomy with EBDR (4). We showed that a liver receptor index (LHL15) of <0.903 was an independent risk factor for predicting PHLF-related mortality. Furthermore, Olthof et al. (26) stated that mebrofenin hepatobiliary scintigraphy, widely used in Europe, was useful for patient selection for PHCC surgery. A cut-off value for the liver function was set at 8.5%/min, which resulted in a negative predictive value of 94% and a positive predictive value of 41%. A recent study demonstrated that the hepatocellular uptake index (HUI) obtained by gadoxetate disodium-enhanced magnetic resonance imaging facilitated quantitative estimation of liver function and correlated well with the ICG clearance test (27). The HUI is calculated directly from the future RLV, which implies that assessment using this index is more accurate for the quantification of future liver remnant function, especially in patients with regional liver heterogeneity. Notake et al.’s HUI of remnant liver would be useful in predicting CRLF (AUC=885) (27). Future studies are warranted to evaluate which modalities are useful.

According to a recent study on right tri-sectionectomy, the addition of surgical factors in combination with poor liver function was associated with higher surgical mortality. Hayashi et al. (14) showed that RLV/body weight <0.65% (OR=11.7; p=0.005), age ≥65 years (OR=31.7; p<0.001), and blood loss ≥25 mL/kg (OR=22.1; p=0.004) were independent risk factors for CRLF. Patients with all three risk factors died after surgery. Our findings are similar to their results. Our data indicate that the presence of a single risk factor does not significantly increase mortality; however, when multiple risk factors are present together, the overall risk of mortality rises substantially (Figure 3). These studies suggest that surgeons should consider shortening the operative time and reducing operative bleeding. Longer operations result from difficult surgical procedures, such as concomitant vascular resection or hepatopancreatoduodenectomy. However, a shortened operative time would reduce both operative bleeding and surgical invasiveness. A recent report from Nagoya University revealed that adjusted blood loss >20 ml/kg increased severe postoperative complications (28).

Another key to reducing CRLF and associated mortality is to reduce severe postoperative complications, especially vascular-related complications (bleeding and thrombosis), which are associated with high mortality (8,19). Our data suggest that although these complications were rare, the mortality rate associated with them was high. Moreover, Kawakatsu et al. (20) suggested that cumulative minor postoperative complications, such as edema and arrhythmia, played a key role in the Comprehensive Complication Index (CCI). Even minor complications can lead to major complications, which collectively result in a higher CCI (20). They suggested that a cumulative CCI >40 could be considered as severe morbidity. Moreover, the 360-day cumulative mortality rate was significantly higher in the severe morbidity course than in the mild and moderate morbidity courses.

Study limitations. Firstly, the retrospective study design is inherently subjected to selection bias. However, we enrolled consecutive patients from 2000, resulting in a more homogeneous cohort than that was included in most previous studies. We also followed-up patients over a longer period of time, during which the treatment strategy has continued to evolve. Secondly, because patients with remarkably poor ICGKF were initially excluded from being surgical candidates, all study patients met a minimal level of ICG excretion. This selection bias should be considered when interpreting the present results as the value of ICGKF may be underestimated.

In conclusion, the incidence rate of CRLF after hepatectomy with EBDR for PHCC was 27.5%. Approximately 30% of the CRLF cases were caused by postoperative complications (secondary CRLF), Surgical factors and ICGKF <0.05 before PVE were independent risk factors for the development of CRLF. The co-existence of all three risk factors leads to a remarkably high risk of mortality.

Conflicts of Interest

The Authors declare no conflicts of interest for this article.

Authors’ Contributions

Conceptualization: TN, YY and SH; methodology: TN and SH; data curation: TN, SY, MW, KT, AM, YN, TA; writing -original draft preparation: SY, YY and TN; writing - review and editing: MW, KT, AM, YN, TA, T, N; visualization: SY and TN; supervision: SH; project administration: TN. All Authors read and agreed to the published version of the manuscript.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Artificial Intelligence (AI) Disclosure

During the preparation of this manuscript, a large language model (Grammarly, Grammarly Inc, CA, USA) was used solely for language editing and stylistic improvements in select paragraphs. No sections involving the generation, analysis, or interpretation of research data were produced by generative AI. All scientific content was created and verified by the authors. Furthermore, no figures or visual data were generated or modified using generative AI or machine-learning-based image enhancement tools.