Impact of portal hypertension on short‐ and long‐term outcomes after liver resection for intrahepatic cholangiocarcinoma: A propensity score matching analysis

Abstract Objective We explored the impact of clinically significant portal hypertension (CSPH) on short‐ and long‐term outcomes of intrahepatic cholangiocarcinoma (ICC) after liver resection (LR). Methods Data of 352 ICC patients with cirrhosis who underwent LR were extracted from the Primary Liver Cancer Big Data (PLCBD) between 2005 and 2015 and reviewed. A nomogram based on logistic analyses was developed to illustrate the influencing factors of post‐hepatectomy liver failure (PHLF). The impact of CSPH on long‐term survival was explored through propensity score matching (PSM) analysis, log‐rank test, Cox proportional hazards model, and Kaplan–Meier curves. Results A total of 106 patients had CSPH, and 246 patients did not. A nomogram established based on GGT level, CSPH, intraoperative blood loss, and multiple tumors had an area under the receiver operating characteristic curve of 0.721 (95% confidence interval [CI] = 0.630–0.812), which displayed a better PHLF predictive value than the MELD score (0.639, 95% CI = 0.532–0.747) and Child–Pugh score (0.612, 95% CI = 0.506–0.719). Moreover, the patients with CSPH had worse overall survival (OS) rates than the patients without CSPH in the whole cohort (p = 0.011) and PSM cohort (p = 0.017). After PSM, multivariable Cox analyses identified that CSPH was an independent risk factor for OS (hazard ratio = 1.585, 95% CI = 1.107–2.269; p = 0.012). Conclusion CSPH is a significant risk factor for PHLF and OS in ICC patients with cirrhosis after surgery. Selecting the proper patients before operation can effectively avoid PHLF and improve the prognosis of ICC.


| INTRODUCTION
Intrahepatic cholangiocarcinoma (ICC), the second leading primary liver cancer, has been increasing in incidence worldwide during the past decades, accounting for 5%-20% of the primary liver cancer population. 1,2 Although the exact pathogenesis of ICC remains unclear, current studies have demonstrated that multiple causative risk factors, including hepatitis virus infection, hepatolithiasis, and cirrhosis, are involved in the development of this malignancy. 3,4 Cirrhosis can be observed in 27.8%-50.5% of patients with ICC, [5][6][7] which is largely a result of viral hepatitis, alcohol abuse, and hepatolithiasis. 8 Liver resection (LR) remains the potential curative therapy for patients with ICC. 9 However, many challenges remain in the management of LR. 10 On the one hand, the resectability of ICC is still low, and patients with unresectable tumors have an extremely short survival time. 11 On the other hand, even after LR, the risk of tumor recurrence and metastases is still high, resulting in a poor survival outcome. 12,13 The currently reported 5-year overall survival (OS) rates after hepatectomy range from 20% to 35%. 14,15 The effectiveness of liver transplantation, transarterial chemoembolization, and ablation remains to be determined in patients with ICC. 16 A recent study has summarized the clinical efficacy of futibatinib, a FGFR antagonist, and pointed out that ICC patients with FGFR2 gene fusions or other rearrangements could acquire a survival benefit from futibatinib. 17 However, this conclusion needs further studies to be confirmed in the future.
In general, some cirrhotic patients present with clinically significant portal hypertension (CSPH) as time goes by, which is considered a contraindication for surgery. However, LR is also performed in selected oriental patients with CSPH because of a lack of other effective therapies for ICC. [5][6][7] Although the effectiveness and safety of LR in these patients have been reported, 7 the impact of CSPH on short-and long-term survival after LR in those patients remains unclear.
This study sought to verify the influence of CSPH on short-term survival, especially the incidence of posthepatectomy liver failure (PHLF), in ICC after surgery. The impact of CSPH on long-term survival in patients with ICC after hepatectomy was also explored through propensity score matching (PSM).

| Patient selection
Data on 394 cirrhotic patients who underwent R0 resection with histopathologically confirmed ICC were extracted from the Primary Liver Cancer Big Data (PLCBD) in Fujian province from September 2005 to December 2015 and reviewed. Of these patients, 42 were excluded as follows: 10 for extrahepatic distant metastasis, 4 for receiving preoperative anticancer treatments, 23 for lost to follow-up, and 5 for incomplete clinical data. The remaining 352 patients were enrolled in the analysis.

| Preoperative work-up and liver resection
Before surgery, the patients were routinely asked for demographic information. Routine laboratory examinations and imaging studies were also tested. Upper gastrointestinal endoscopy or barium enema examination of the upper digestive tract and cardiopulmonary function were also routinely performed. WHO classification criteria are the basis of the clinical diagnosis of ICC. 15 Surgical indications were as follows: (1) good performance (ECOG score 0-2) without significant heart, kidney, lung, and other important organ diseases; (2) liver function within Child-Pugh grade A or B7, without refractory ascites; (3) tumor localized to the liver segment or lobe with no evidence of distant tumor metastasis and removal of the liver parenchyma with the possibility of maintaining a liver remnant of ≥50% using measurements from preoperative CT/MRI; (4) no history of variceal bleeding, cirrhosis mild to moderate, and esophageal and gastric varices moderate to severe without bleeding tendency, namely, red color signs; and (5) platelet count ≥75 × 10 9 /L. Detailed surgical procedures were similar to those previously reported. 18,19 Regional lymphadenectomy was routinely performed for pre-operatively or intra-operatively diagnosed ICC. Histopathological examination of the tumor specimens was routinely performed.

| Definitions
The presence of preoperative CSPH was defined as the presence of esophagogastric varices or splenomegaly (diameter >12 cm) with a platelet count <100 × 10 9 /L, which is based on the definition of standard surrogate criteria proposed by the Barcelona Clinic Liver Cancer (BCLC) classification. 20

| Follow-up protocol and endpoints
After hepatectomy, patients were followed up once every 2 months in the first 2 years and once every 3-6 months thereafter. The patients with CSPH were asked if they had any tendency or signs of bleeding, including upper gastrointestinal hemorrhage. The diagnosis of ICC recurrence was the same as that of the initial ICC diagnosis. Treatment options after relapse were discussed by a multidisciplinary team as previously reported.
OS and time to recurrence (TTR) were the primary endpoints of this study. The secondary endpoint was PHLF. OS was calculated as the interval between the date of LR to the date of last follow-up or patient death. TTR was computed from the date of LR to the date of identification of disease recurrence. Surgical morbidity, including PHLF and mortality, was also observed. This study was censored on 30 October 2018.

| Statistical analysis
Continuous variables were described as the median and interquartile range (IQR) or mean ± standard deviation (SD), and Student's t-test or Mann-Whitney U test was used to compare differences. Categorical variables were described as the number of cases and percentage, and chisquared test or Fisher's exact test was used to compare the differences between the two groups if necessary. A nomogram was established using the "rms" package of R, version 3.1.1 (http://www.r-proje ct.org/). The performance of the predictive model was assessed using the receiver operating characteristic (ROC) curves and the corresponding area under the curve (AUC). The Kaplan-Meier method was used to plot survival curves, and the log-rank test was used for comparison. The Cox proportional hazards model was used to analyze independent prognostic factors of OS and tumor recurrence.
A one-to-one PSM was used to balance the differences in baseline clinicopathological features between the CSPH and non-CSPH groups. 24 Covariates included in the PSM analysis were TBIL, AST, PT, Child-Pugh score, intraoperative blood loss, and tumor number to calculate the propensity score by logistic regression. Individuals in the CSPH group were matched to those in the non-CSPH group by a 1:1 nearest neighbor matching with a caliper of 0.2 and without replacement to realize the closest estimated propensity score values between the two groups in PSM. 25 After PSM, a new cohort was created with minimal differences in patient clinicopathological characteristics between the two groups.
Statistical analyses were performed using SPSS (IBM SPSS for Windows, version 22.0; IBM Corp.) and PSM for SPSS version 22.0 (R statistical package version 2.15; Cornell University, USA). All reported p values were bilateral, and statistical significance was considered at p < 0.05.

| Baseline characteristics
Among 352 patients with cirrhosis who underwent LR, 106 (30.1%) patients had CSPH, whereas the remaining 246 (69.9%) patients did not ( Figure 1). Table 1 shows the baseline demographic and clinicopathological characteristics of patients with or without CSPH. After PSM analysis, the baseline characteristics between the two groups were balanced well.
Among these 352 patients, 27 (7.6%) patients had PHLF, 20 patients presenting with PHLF 90 days after surgery. Another seven patients with PHLF were identified more than 90 days after surgery. Detailed information on complications at 90 days after surgery is listed in Table 2.

| Establishing the predictive nomogram for PHLF
The nomogram of the PHLF probabilistic prediction model was constructed (Figure 2A) using the four factors mentioned in Table 3. The points of each factor were calculated by their weight of ORs. Then, the final score was calculated to acquire the probability of PHLF in a patient with ICC after surgery. In addition, the calibration curve showed good agreement between the likelihood of our PHLF nomogram and the actual observed incidence of PHLF in patients with ICC ( Figure 2B).
Using the nomogram, we calculated the nomogram score for every patients and then compared it with the Child-Pugh and MELD scores to assess the performance of PHLF prediction through the ROC curves. Figure 2C shows that the AUC value of the nomogram score was 0.721 (95% CI: 0.630-0.812), which was greater than that of the MELD score (0.639, 95% CI: 0.532-0.747) and the Child-Pugh score (0.612, 95% CI: 0.506-0.719).

| Impact of CSPH on long-term prognosis in the PSM cohort
Of the 198 patients of the matching cohort, 134 (67.6%) patients exhibited tumor recurrence and 126 (63.6%) patients died.

| Subgroup analysis in the PSM cohort
We divided the PSM cohort into two subgroups: group tumor diameter ≥5 cm and group tumor diameter <5 cm. Kaplan-Meier curve showed that the OS significantly improved in the non-CSPH group compared with the CSPH group with a tumor diameter ≥5 cm (p = 0.048). However, in patients with a diameter <5 cm (p = 0.141), no significant difference in OS was found between the two groups ( Figure S1).

| DISCUSSION
In this study, we identified CSPH and other factors significantly associated with PHLF in resectable ICC with cirrhosis and then developed a nomogram to predict postsurgical PHLF. We also analyzed the influence of CSPH on long-term survival outcomes after LR. Analysis following PSM showed that CSPH was a significant risk factor for OS after hepatectomy, although it had no impact on recurrence.
Currently, the availability of LR for the treatment of hepatocellular carcinoma (HCC) patients with CSPH T A B L E 2 Surgical complication in the whole cohort remains controversial. 26 In accordance with the guidelines of the American Association for the Study of Liver Diseases/European Association for the Study of the Liver, 27,28 portal hypertension is considered a contraindication to LR in patients with HCC. This opinion has been supported by several regional studies over the past few years, showing that patients with HCC who are indirectly diagnosed 29,30 with portal hypertension or directly diagnosed 31,32 by hepatic venous pressure gradient (HVPG) frequently suffer from severe PHLF and poor long-term prognosis. However, several studies from different countries reported no difference in short-and long-term outcomes between portal hypertension and normal HCC patients. 33,34 In 2009, Cucchetti et al. retrospectively analyzed and compared the postoperative results and long-term survival between 89 HCC patients with CSPH and 152 HCC patients with non-CSPH. After PSM, the patients with CSPH had the same intraoperative course, the incidence of postoperative liver failure, morbidity, and survival rates as the patients with non-CSPH. 33 Santambrogio et al. divided 223 HCC patients with BCLC stage A into two groups according to the presence (n = 63) or absence (n = 160) of portal hypertension and compared the prognosis of the two groups. They reported that the short-and long-term outcomes in the patients with portal hypertension were similar to those in patients with normal portal venous pressure. 29 T A B L E 3 Uni-and multivariable logistic regression analyses of independent risk factors of PHLF in the whole cohort At present, the number of studies on short-and longterm prognosis of ICC patients combined with CSPH after surgery is still limited as compared with those in HCC. Studies of ICC have mostly focused on the role of cirrhosis in the prognosis of patients after surgery. 6,7,35,36 Clinically, the safety and receptivity of ICC patients undergoing surgical treatment should be seriously considered in the context of CSPH. However, the impact of CSPH on postoperative outcomes, especially PHLF, has never been reported in patients with ICC. Therefore, our study is necessary and meaningful for revealing the impact of CSPH on short-and long-term prognosis in ICC patients with cirrhosis after hepatectomy.
Cucchetti et al. and He et al. reported that CSPH is not an independent risk factor for OS after PSM matching. 33,34 In contrast, our study showed that the CSPH group had significantly worse OS than the non-CSPH group, which was an independent prognostic hazard of OS after multivariable Cox regression analyses. Moreover, in the tumor diameter ≥5 cm group, subgroup analysis showed that the CSPH group had worse OS than that of the non-CSPH group. Similarly, Zheng et al. analyzed the survival outcomes of 355 patients with HCC and found that the longterm survival rate of patients without portal hypertension was significantly better than that of patients with portal hypertension in the cirrhosis subgroup. 30  compensatory cirrhosis who undergo surgery for HCC. 20 Several reasons might explain this discrepancy. First, the definitions of CSPH or portal hypertension before surgery used in the aforementioned studies were different. Second, the number of patients and study periods in different articles were heterogeneous. Third, the assessment criteria of patients' liver function and prognosis were different and not comparable.  PHLF is a serious complication of LR. The estimation and PHLF prediction could be important before treatment in cirrhotic patients. In this study, the risk of PHLF depended on several preoperative and intraoperative events, for example, GGT level, CSPH, intraoperative blood loss, and tumor numbers. GGT level was a preoperative indicator of liver function.
Recently, some studies have reported that impaired preoperative indicators, such as GGT level, platelet count, and bilirubin, are important predictors of impaired postoperative liver function. 37,38 In 2017, Hu et al. 39 reported that high serum GGT level is an important risk factor in predicting liver failure risk after hepatectomy for patients with HCC, which is consistent with our results. We found that GGT >65 IU/ml increases the risk of PHLF after surgery in cirrhotic ICC in our PHLF prediction model.
The effect of portal hypertension on postoperative complications and PHLF in HCC is controversial. 40 Previous studies on HCC have also verified that portal hypertension is an important determinant of PHLF. 30,41 However, no studies have revealed the impact of CSPH on PHLF in postoperative patients with ICC. Therefore, the present study is the first to reveal that CSPH is an important risk factor of PHLF. In addition, patients with ICC in the CSPH group had a significantly higher chance of overall complications than those in the non-CSPH group.
More and more findings are suggesting that intraoperative blood loss is highly correlated with the occurrence of PHLF. 42,43 Researchers have found that intraoperative blood loss can increase the probability of ischemic reperfusion injury, leading to morbidity and mortality after liver surgery. 44,45 In addition, transfusion may induce immunosuppressive effects, which are harmful and may lead to poor perioperative outcomes in cancer patients undergoing surgery. 46,47 These studies may provide clues to explain the relationship between blood loss and PHLF. However, further research is needed to elucidate this mechanism.
Multiple tumors indicate that patients will undergo a longer surgery time and that more normal liver tissues will be removed during surgery. A sufficient future liver remnant (FLR) volume guarantees that severe postoperative complications do not occur in liver surgery. 48 In the present study, the cirrhotic ICC patients with multiple tumors had a higher risk of PHLF after surgery compared with their counterparts.
Our study may serve as a practical guide for the surgical treatment of patients with CSPH in the future. This study can also be used to select suitable patients for surgery. On the one hand, CSPH patients who meet the proper surgical requirements will not only benefit from a good prognosis, but will also benefit from a reduced occurrence of PHLF. Many factors determine whether or not patients with CSPH can receive surgical treatment, and these need further exploration by researchers. This is only a retrospective study from a single-center, and more global multicenter data or prospective studies are needed to verify our research conclusions or generate new ideas. On the other hand, surgeons must pay attention to CSPH because of its significant impact on the perioperative period and prognosis. More and more new discoveries are expected in the future as the understanding of CSPH is deepened. In the next 5 or even 10 years, from the aspect of surgical operation, more unresectable CSPH patients can benefit from continuous technological progress in surgical techniques, advanced medical equipment (e.g., laparoscopic and robotic surgery), and virtual reality technology applied in precision LR. From the aspect of comprehensive treatment, with the rapid development of the system therapy in liver cancer, immunotherapy and targeted therapy system has become the main treatment of advanced liver cancer. These treatment methods enrich the options for advanced liver cancer, especially for patients with CSPH. Patients with multiple tumors and tumors larger than 5 cm may benefit from neoadjuvant therapy and sequential surgical treatment. This finding needs to be confirmed by further clinical studies.
Our study has several limitations. First, this is a singleinstitution retrospective study. More patients from more institutions need to be studied to confirm our conclusions. Second, ICG-R15 and FLR are important indicators used to assess liver function and volume in clinical practice. However, in this study, these data were missing for various reasons, resulting in an imperfect study. These two variables must be added in our future studies. Third, CSPH was diagnosed indirectly by clinical criteria; thus, we could not determine the real influence of HVPG measurement as a diagnostic criterion for ICC patients with CSPH. Last, given the limitations of the PSM approach, our study still had the potential for selection bias, although we used PSM in an attempt to decrease these biases.

| CONCLUSION
The presence of CSPH is significantly associated with PHLF and a worse OS in patients with ICC after hepatectomy. CSPH patients with high levels of CEA, GGT, multiple tumors, and large tumor diameter (≥5 cm) preoperatively are not recommended to undergo surgical treatment to avoid postoperative liver failure and poor prognosis.

ETHICS STATEMENT
The study protocol was performed in accordance with the guidelines outlined in the Declaration of Helsinki. The Institutional Review Board of Mengchao Hepatobiliary Hospital of Fujian Medical University approved the study (ID: 2018-048-01). The patients/participants provided their written informed consent.