From March 2008 to December 2019, 341 HCCA patients undergoing radical resection at the Southwest Hospital and PLA Middle Military Command General Hospital were enrolled. All procedures were performed according to R0 resection criteria. This study was approved by the Ethical Committee (Approval No. KY20211215), and all patients or their guardians provided written informed consent when informed of potential risks or adverse effects during the clinical process.

In this study, preoperative epidemiologic features like age, gender, height, weight, BMI, heart disease and diabetes, laboratory tests like white blood cells, platelet, liver transaminase, bilirubin, albumin, alkaline phosphatase, coagulation and tumor markers, clinic-pathological features like liver cirrhosis, hepatic artery invasion, portal vein invasion, lymph node metastasis, tumor differentiation, and Tumor Node Metastasis (TNM) stage were collected.

The inclusion criteria of this study included the following: patients with pathological confirmation of HCCA; patients received R0 resection; expected survival time longer than 3 months; patients had a clear record of accepting preoperative biliary drainage or not; and PTBD was the unique approach of biliary drainage. This R0 resection implied the histologically negative margins in the operation. Moreover, the proximal end of the bile duct greater than 5 mm from the edge of the tumor and the distal end at the upper edge of the pancreas were resected in the operation. The entire caudate lobe and surrounding liver parenchyma greater than 15 mm surrounding the bile duct axis were removed. For patients with Bismuth–Corlette III or IV HCCA, hemihepatectomy or extended hemihepatectomy and caudate lobectomy were conducted. The lymph nodes of groups 8, 12, and 13 were conventionally dissected.

The criteria of accepting PTBD in this study included the following: the level of total bilirubin was higher than 100 µmol/L or the symptom of jaundice has sustained for more than 4 weeks, or those patients needed to accept large scale hepatectomy (over thee Couinaud segments).

These patients who accepted R0 resection did not routinely accept adjuvant therapy after operation. All individuals were closely followed up in outpatient clinic after discharge. Basic measurements including liver function and blood routine examination, examination of serum tumor markers, and imaging were performed.

Continuous variables were presented as the mean and standard deviation or median and interquartile range. Categorical variables were presented as amounts and percentages. Comparisons between clinicopathological factors were made using Student’s t-test or analysis of variance or Kruskal–Wallis test for continuous variables and χ2-test for categorical variables. The log-rank test and t-test were used to compare the differences between those patients accepting PTBD or not.

Univariate and multivariate logistic regression were used to analyze the risk factors associated with portal vein invasion and lymph node metastasis. Odds ratios (OR) derived from logistic regression models represent the likelihood of pathological outcomes per unit increase in exposure, with p-values <0.05, indicating statistically significant associations. The Cox proportional-hazards model was used to identify relationship between clinical characteristics and survival. Kaplan–Meier curves were used to estimate overall survival (OS) and relapse-free survival (RFS), and log-rank test was used to evaluate statistically significant differences. Causal mediation analysis was performed to investigate the mediation effect of PTBD on OS and RFS.

All statistical analyses were performed in the STATA (version 14), and p < 0.05 was considered statistically significant. The figure was re-arranged in the Adobe Illustrator (version 2019).

In this study, a total of 341 HCCA patients who accepted radical resection were included. As shown in Table 1 , the differences of basic epidemiologic features like age, gender, weight, and chronic diseases between the PTBD group and the non-PTBD group were not statistically significant (p > 0.05). In the PTBD group, the number of white blood cells (7.3 vs. 6.3 × 10⁹/L, p = 0.001) and platelets (261.3 vs. 232.5 × 10⁹/L, p = 0.005) was significantly higher than that in the non-PTBD group. Also, levels of bilirubin, including total bilirubin (317.5 vs. 116.6 µmol/L, p = 0.001) and direct bilirubin (167.8 vs. 60.1 µmol/L, p = 0.001), were significantly elevated in the PTBD group, reflecting the presence of preoperative jaundice as a crucial indication for PTBD acceptance. Furthermore, tumor marker CA242 levels were significantly higher in the PTBD group compared to the non-PTBD group (82.3 vs. 42.9 U/mL, p = 0.002), underscoring its diagnostic relevance in PTBD patients.

Particularly, the proportion of patients with portal vein invasion (39.9% vs. 25.8%, p = 0.006) and lymph node metastasis (45.4% vs. 33.7%, p = 0.027) was significantly higher in the PTBD group compared to that in the non-PTBD group. Additionally, there were more patients with poorly differentiated tumors (25.5% vs. 14.1%, p = 0.010) in the PTBD group than that in the non-PTBD group.

To evaluate potential preoperative risk factors associated with portal vein invasion, univariate analysis was conducted initially ( Table 2 ). Significant factors identified included liver cirrhosis (OR = 4.4, P = 0.001), total bilirubin > 42 µmol/L (OR = 2.1, P = 0.036), Alanine aminotransferase (ALT) > 42 IU/L (OR = 2.2, P = 0.032), Alanine aminotransferase (AST) > 42 IU/L (OR = 2.8, P = 0.018), Alanine aminotransferase (ALP) > 188 U/L (OR = 4.9, P = 0.009), PTBD (OR = 1.9, P = 0.006), and Bismuth–Corlette type III–IV classification (OR = 2.4, P = 0.002). These factors were found to be statistically significant in association with portal vein invasion. Afterward, these factors with p-values less than 0.1 in univariate analysis and those considered clinically important were included in the multivariable logistic regression. As shown in Table 2 , previous PTBD (OR = 2.3, P = 0.002) was an independent risk factor for portal vein invasion.

In the univariate analysis, both total bilirubin and PTBD showed statistically significant differences. However, the level of total bilirubin was excluded from the multivariate analysis because it was a key determinant for deciding PTBD, which was central and critical to this study. Including both variables could introduce confounding effects (14). This approach helped avoid multicollinearity and ensured the clarity of PTBD’s impact on outcomes (15).

Similarly, univariate and multivariate logistic regression analyses were performed to explore the potential relationships between these peri-operative factors and lymph node metastasis, with the results presented in Table 3 . In the univariate analysis, PTBD (OR = 1.6, P = 0.028) and tumor diameter > 2.5 cm (OR = 1.7, P = 0.016) were the only two factors significantly associated with lymph node metastasis. Afterward, these factors with p-values less than 0.1 in univariate analysis and those considered clinically important were included in the multivariable logistic regression. Following multivariate analysis, CA242 > 20 IU/mL (OR = 1.9, P = 0.016), PTBD (OR = 1.9, P = 0.008), and tumor diameter > 2.5 cm (OR = 1.8, P = 0.022) emerged as three independent risk factors associated with lymph node metastasis.

This study aimed to investigate the relationship between preoperative percutaneous transhepatic biliary drainage (PTBD), portal vein invasion, lymph node metastasis, and clinical survival in patients with hilar cholangiocarcinoma. The Kaplan–Meier curves were drawn, and the log-rank test was used to evaluate differences between two groups. As presented in Figures 1A, B , these patients who did not accept preoperative PTBD showed a significantly longer OS (p = 0.039) than the PTBD group, whereas the RFS did not show a statistical difference between two groups (p = 0.082). Consistent with other studies, those patients without portal vein invasion and lymph node metastasis had a significantly better prognosis both in OS and RFS (all p < 0.001), which was shown in Figures 1C–F .

Furthermore, univariate and multivariate Cox regression models were employed to investigate potential risk factors associated with OS and RFS. Although preoperative PTBD emerged as a significant risk factor for OS (HR = 1.3, P = 0.039) in the univariate analysis, this association did not retain statistical significance (HR = 0.99, P = 0.919) in the multivariate analysis ( Table 4 ). Moreover, the preoperative PTBD was not directly related to RFS in the univariate and multivariate analysis ( Table 5 ). However, the portal vein invasion and lymph node metastasis were two significant risk factors associated with OS and RFS (all p < 0.01). Subsequently, causal mediation analysis was performed to investigate the mediation effect of PTBD on OS and RFS. The analysis revealed that portal vein invasion accounted for 30.1% (p = 0.030) of PTBD’s effect on OS and 36.4% (p = 0.022) on RFS. Similarly, lymph node metastasis mediated 35.9% (p = 0.038) of PTBD’s effect on OS and 18.1% (p = 0.034) on RFS, respectively. These findings indicated that the effect of PTBD on survival was partly mediated by portal vein invasion and lymph node metastasis.

In HCCA patients, the presence of postoperative complications was an important factor affecting hospital mortality and long-term survival (16, 17). In this study, the comparison of postoperative complications is between the PTBD group and the non-PTBD group, and the results were presented in Table 6 . The proportion of pleural effusion (28.0% vs. 14.6%, p = 0.004) in the PTBD group was significantly higher than that in the non-PTBD group. Additionally, the mean length of hospitalization was 26.7 days, which was obviously longer than that in the non-PTBD group (21.8 days, p = 0.002). The presence of other complications including fever, infection, bile leakage, cholangitis, liver failure, and bleeding was similar in the PTBD group and the non-PTBD group.