This study included 76 patients who underwent their first LT for PSC at a tertiary referral center between January 1996 and July 2024. For the analysis of rPSC, 12 patients who died within 1 year post-transplant and two recurrent cases for whom sufficient clinical information could not be obtained were excluded. Ultimately, 62 patients were enrolled in the final analysis. The study protocol was approved by the institutional ethics committee (No. R1473-8) and performed in compliance with the Helsinki Declaration.

Demographic characteristics, perioperative findings, and long-term outcomes were retrospectively collected from medical records. In addition to risk factors for rPSC, we collected microbiological and treatment-related variables, including the number of submitted culture tests, the types of organisms detected, the number of positive detections, and antimicrobial use. Data on first-time liver transplantation cases were included in the analysis.

rPSC was defined according to the criteria proposed by Graziadei et al [10]. Specifically, patients with a preoperative diagnosis of PSC were considered to have recurrence if, at 90 days or more after LT, they exhibited either multiple intrahepatic bile duct strictures on cholangiography or magnetic resonance cholangiopancreatography (MRCP), or histological findings consistent with fibrous cholangitis or fibro-obliterative lesions. Other potential causes, such as ischemic cholangitis, chronic ductopenic rejection, and anastomotic stricture within 90 days post-transplantation, were excluded.

For Kp detection in rPSC cases, Kp positivity was defined as those with Kp identified in blood, bile, or ascitic fluid cultures obtained between LT and rPSC diagnosis, whereas in non-rPSC cases, as those with Kp identified in same clinical cultures between LT and the end of follow-up. The same criteria were applied to the evaluation of other bacterial species and the number of clinical culture tests performed. Only clinical cultures with species-level identification were included; those identified only to the genus level were excluded.

All procedures and postoperative management followed the standard protocol of our institution in all cases. Roux-en-Y hepaticojejunostomy was the standard biliary reconstruction method for PSC cases. Biliary stents were placed intraoperatively and externally drained through the blind end of the duodenum. Patients were discharged with the stent in place, which was removed under fluoroscopic guidance after approximately 3 months if no complications occurred. Preoperative prophylactic antibiotics typically included ampicillin and second-generation cephalosporins. If preoperative clinical cultures were positive, antibiotic selection was adjusted based on sensitivity testing. Antibiotics were continued for 2 days after LT and discontinued if no signs of infection were observed. When infection was suspected, based on fever, elevated liver enzymes, or inflammatory markers, blood, bile, ascites, and urine cultures were obtained, and empirical antibiotics administered. In addition, clinical cultures were submitted at the surgeons’ discretion. The same protocol was applied in the case of readmission due to fever or similar symptoms. Bile samples were collected aseptically from a stent placed intraoperatively until its removal. Thereafter, further bile samples were collected via endoscopic retrograde cholangiopancreatography or percutaneous transhepatic cholangial drainage. Ascitic fluid was collected via intraperitoneal drains or paracentesis. Immunosuppressive regimens consisted of dual or triple therapy combining a calcineurin inhibitor with steroids, mycophenolate mofetil, and everolimus, which were administered from 1 month after transplantation. Tapering of immunosuppressive agents during the maintenance phase was performed at the surgeon’s discretion. Rituximab was administered as induction therapy for ABO-incompatible living donor liver transplantation (LDLT) or in patients with preoperatively strong positive donor-specific antibodies. Histological diagnoses of acute and chronic rejection were made according to the Banff criteria.

Continuous variables were expressed as medians with interquartile ranges (IQR) and compared using the Mann–Whitney U test. Categorical variables were compared using the chi-square test or Fisher’s exact test. Kaplan–Meier method and log-rank test were used for survival analyses. Factors associated with rPSC were evaluated using a Cox proportional hazards model, with death or re-transplantation treated as censoring events. Cut off values for continuous variables were determined by receiver operating characteristic curve analysis and the Youden index or based on previously published studies and clinical relevance. Variables considered clinically and statistically relevant were included in the multivariable analysis, with model selection via the backward stepwise method. In addition, bootstrap resampling with 1,000 iterations was conducted to assess the robustness and stability of the multivariable model estimates. All statistical analyses were performed using R software (version 4.3.0).

Among 76 patients who underwent initial LT for PSC at our institution during the study period, the median age was 36.0 years (IQR, 25.8–49.0 years), 66 were adults (86.8%), and 36 were male (47.4%). LDLT was performed in 64 cases (84.2%), deceased donor liver transplantation in eight cases (10.5%), and domino transplantation in four cases (5.3%). ABO-incompatible LT was performed in eight cases (10.5%). The median follow-up period was 75.1 months (IQR, 37.9–168.0 months; mean, 105.1 months). rPSC occurred in 24 patients (31.6%), with a median time to recurrence of 55.0 months (IQR, 26.4–76.9 months). Kaplan–Meier curves for overall survival (OS) and recurrence rates among the 76 patients are shown in Figure 1. The 5-, 10-, and 15-year OS rates were 74.3%, 68.4%, and 56.1%, respectively (Figure 1A). The 5-, 10-, and 15-year recurrence rates were 20.9%, 50.1%, and 56.3%, respectively (Figure 1B). Primary graft survival (PGS) rates were 69.6% at 5 years, 50.1% at 10 years, and 25.0% at 15 years.

The background characteristics of the 62 patients included in the recurrence analysis are summarized in Table 1. rPSC occurred in 22 patients (35.5%), and 40 patients (64.5%) were in the non-recurrent group. The median age in the rPSC group was 37.0 years (IQR, 26.5–47.5), and 41.0 years (IQR, 27.5–50.0) in the non-rPSC group, with no significant difference. There were no significant differences between both groups in the proportion of adult cases, sex, preoperative MELD score, Mayo PSC risk score, Child–Pugh classification, history of ulcerative colitis (UC), or human leukocyte antigen (HLA)-DR15 positivity. Regarding donor factors, the median donor age was 48.0 years (IQR, 37.0–55.0) in the rPSC group and 41.0 years (IQR, 33.0–51.0) in the non-rPSC group (p = 0.12), with no significant difference. There were no significant differences between both groups in sex mismatch, ABO-incompatible transplantation, first-degree-related donors, or number of HLA mismatches. The presence of donor HLA-DR15 also did not significantly differ between the groups (rPSC vs. non-rPSC = 12 cases [57.1%] vs. 13 cases [35.1%], p = 0.18). Surgical factors, including LDLT, operative time, blood loss, cold ischemia time, warm ischemia time, and biliary reconstruction method, were not significantly different. Postoperative factors, including major complications, biliary complications, CMV infection, acute rejection, and immunosuppressive regimen at 1 year post-LT (alone/dual/triple), also showed no significant differences between the groups. On the other hand, active postoperative UC was significantly more frequent in the rPSC group (rPSC vs. non-rPSC = 10 cases [45.5%] vs. six cases [15.0%], p = 0.02). As for bacterial cultures, the median number of clinical cultures submitted postoperatively (including blood, bile, and ascitic fluid) was 26.5 (IQR: 11.3–46.3) in the rPSC group and 27.5 (IQR: 19.0–35.5) in the non-rPSC group, with no significant difference (p = 0.66). However, the detection rate of Kp was significantly higher in the rPSC group (seven cases [31.8%] vs. three cases [7.5%], p = 0.03). No significant differences were observed between the groups in the identification of three or more bacterial species or the use of antibiotics.

Kaplan–Meier curves for PGS and OS stratified by recurrence status among the 62 patients are shown in Figure 2. The 5- and 10-year PGS rates in the rPSC group were 80.4% and 49.0%, respectively, with a median graft survival of 97.2 months. In the non-rPSC group, the 5- and 10-year PGS rates were 85.9% and 73.3%, respectively, with a median graft survival of 229.2 months. PGS was significantly lower in the rPSC group (p = 0.003, Figure 2A). OS stratified by recurrence status showed no significant difference between both groups: in the rPSC group, 5-, 10-, and 15-year OS rates were 90.0%, 80.0%, and 57.4%, respectively, with a median survival of 236.4 months; in the non-rPSC group, the corresponding rates were 85.8%, 80.4%, and 73.1%, with a median survival of 229.2 months (p = 0.78, Figure 2B).

Table 2 summarizes the results of univariate and multivariate Cox proportional hazards models assessing variables associated with rPSC. In the univariate analysis, postoperative active UC (p = 0.03), Kp detection (p = 0.03) and the identification of three or more bacterial species in clinical (p = 0.02) were significantly associated with rPSC. In the multivariate analysis, Kp detection (HR: 3.83, 95% CI: 1.44–10.21, p = 0.007) and three or more bacterial species in clinical cultures (HR: 0.31, 95% CI: 0.12–0.79, p = 0.01) were independently associated with rPSC. Even when the multivariate analysis was restricted to culture data obtained within 1 year after LT, similar results were observed. Kp detection (HR: 3.47, 95% CI: 1.26–9.54, p = 0.02), the identification of three or more bacterial species in clinical cultures (HR: 0.35, 95% CI: 0.14–0.86, p = 0.02), and postoperative active UC (HR 2.52, 95% CI 1.05–6.07, p = 0.04) remained independently associated with rPSC. Bootstrap resampling analysis (1,000 iterations) showed similar results, with Kp detection (median HR: 4.10, 95% CI: 1.57–14.71), identification of three or more cultured bacterial species (median HR: 0.29, 95% CI: 0.09–0.73), and postoperative active UC (median HR: 2.43, 95% CI: 0.89–7.53).

Table 3 presents an exploratory analysis of bacterial species detected in at least three patients. Only Kp showed a significant association with rPSC (p = 0.03). Supplementary Table 1 lists all other bacterial species detected in the study cohort, as well as their respective detection counts.

Figure 3 shows the types and timing of Kp detection specimens. Among the 10 patients in whom Kp was detected, Kp was found in bile in nine cases, blood in three cases, and ascitic fluid in four cases. In nine of the 10 cases, Kp was detected within 1 year post-LT, and in six cases within 3 months post-LT. Among the 10 patients with Kp detection, Kp was identified from clinical cultures obtained during suspected cholangitis episodes accompanied by symptoms such as fever or abdominal pain and/or abnormal blood test findings in four of seven rPSC patients and two of three non-rPSC patients. Figure 4 shows recurrence-free survival (RFS) stratified by Kp detection. The 5- and 10-year RFS rates were 70.0% and 28.0% in the Kp (+) group, and 84.9% and 56.7% in the Kp (−) group, respectively. RFS was significantly worse in the Kp (+) group (p = 0.03).