1. Introduction
Hepatocellular carcinoma (HCC), the most prevalent form of primary liver cancer, remains the third leading cause of cancer-related mortality
[1]. In China, HBV infection has historically been the predominant etiological factor for HCC. However, a notable epidemiological transition has emerged, characterized by a decline in HBV prevalence—largely attributed to widespread HBV vaccination programs—and a rising predominance of nonviral risk factors, particularly alcohol consumption, which poses an escalating public health concern
[2],
[3].
Liver transplantation represents the most effective curative approach for HCC; however, its success is substantially compromised by high recurrence and metastasis rates
[4],
[5]. Notably, HBV reactivation is a frequent consequence of immunosuppressive therapy administered post-transplantation for malignancies, exacerbating HCC recurrence
[6]. This association was first demonstrated in our previous study
[7]. The systemic effects of HBV and alcohol extend beyond the liver, inducing immunological perturbations and multi-organ pathology, which may be further exacerbated by post-transplant immunosuppressive regimens
[8],
[9]. Mechanistically, alcohol accelerates HBV-related HCC progression by activating the RAD21-mediated DNA repair pathway, which subsequently triggers the casein kinase 2 alpha 1 (CSNK2A1)–insulin-like growth factor II receptor (IGF2R) serine 2484 axis to enhance endogenous cholesterol biosynthesis. Cholesterol, in turn, directly binds to CSNK2A1, establishing a self-reinforcing cycle that prioritizes fatty acids as the primary carbon source for oxidative phosphorylation. This metabolic shift leads to excessive reactive oxygen species generation, further driving tumorigenesis. Given this interplay, alcohol-related disease (ALD) is hypothesized to be a critical factor in promoting HBV reactivation. Our previous research established a nomogram to predict HBV reactivation in a Chinese patient cohort
[7]. Building on this foundation, the present study integrates metabolic indicators and ALD-related clinical parameters—both readily available preoperatively—considering growing evidence that metabolic factors influence HBV reactivation
[6].
The supplementary analysis in this study aims to evaluate the impact of ALD on HBV reactivation in liver recipients with HBV-related HCC, utilizing a metabolic perspective based on a real-world database from China.
2. Materials and methods
2.1. Population
This study is a post hoc analysis of a previously described HBV reactivation cohort comprising liver transplant recipients with liver cancer between 2015 and 2020 (Fig. S1 in Appendix A, clinical trial identifier: NCT06114251)
[7]. A total of 920 patients who underwent orthotopic liver transplantation for liver cancer were initially registered. Inclusion criteria encompassed preoperative HBV diagnosis and histopathologically confirmed HCC. Exclusion criteria were applied to 467 patients due to the absence of an HBV background, lack of documented alcohol intake, synchronous malignancies, re-transplantation, concurrent portal vein tumor thrombus, or an overall survival (OS) of less than 90 days. Consequently, the final analysis included 453 eligible patients.
Ethical approval for the HBV reactivation cohort was obtained from the Ethics Committee of two centers (Shulan (Hangzhou) Hospital and the First Affiliated Hospital, Zhejiang University School of Medicine), consistent with prior research (approval IDs: No. 2020-510 and No. KY2021014). Data were anonymized at individual facilities before being aggregated for centralized analysis. This study adhered to the transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) guidelines (Table S1 in Appendix A)
[10].
2.2. Data acquisition and clinical management
Pre-transplant clinical data were systematically collected from the most recent assessments. Predictors of HBV reactivation are summarized in
Table 1. Post-transplant monitoring commenced three months after surgery, during the stable phase, with routine surveillance of key HBV markers, including hepatitis B surface antigen (HBsAg), hepatitis B e antigen (HBeAg), and HBV DNA. Recipients were followed until December 31, 2022, with a maximum follow-up duration of eight years.
All recipients received a standardized immunosuppression regimen, comprising interleukin-2 receptor antagonists, mycophenolate mofetil, Tacrolimus or Sirolimus, and a glucocorticoid with an early tapering strategy. Standard prophylaxis against HBV recurrence included lifelong administration of low-dose nucleoside analogs (e.g., entecavir or tenofovir) and hepatitis B immunoglobulin, in accordance with clinical guidelines for liver transplantation in HCC management in China
[11]. Post-transplant alcohol consumption and relapse were rigorously documented due to their potential impact on transplant outcomes. Throughout the follow-up period, strict adherence to antiviral therapy was meticulously monitored to ensure optimal long-term graft survival.
2.3. Data pre-processing
Initial data preprocessing involved rigorous cleansing to mitigate distribution skew. Records with missing patient data below 5% at the row level and missing clinical features below 5% across all patients at the column level (Table S2 in Appendix A) were excluded. This stringent threshold was met by all patient datasets and features, ensuring data robustness, with 453 recipients and 25 distinct clinical attributes retained. Remaining missing values were imputed using the k-nearest neighbor algorithm with default settings, implemented via the “impute.knn” function from the R package “impute” (version 1.70.0).
2.4. Model development, feature selection, and model explanation
The preprocessed dataset was randomly partitioned into an 80% training/validation set and a 20% external test set to minimize overfitting. Benchmark analysis, incorporating grid search, evaluated 21 machine learning models using both five-fold and ten-fold cross-validation (CV) to ensure model stability across different sample sizes. Model performance was assessed using the concordance index (C-index).
Feature importance and model interpretability were analyzed using the Shapley additive explanation (SHAP) method via the R package “survex” (version 1.2.0)
[12]. Feature selection followed a sequential strategy, progressively incorporating variables from an initial subset of 3 to all 25 available clinical indicators, aligning with their ranked importance. The final model, optimized for clarity while preserving predictive accuracy, was selected based on a comprehensive set of performance metrics, including area under the receiver operating characteristic (AUC), sensitivity, specificity, F1-score, accuracy, positive predictive value (PPV), and negative predictive value (NPV).
Global interpretability of the machine learning model was derived from SHAP values, offering insights into HBV reactivation risk. Lower survSHAP values indicated heightened susceptibility to HBV reactivation, whereas elevated values corresponded to a reduced risk. The aggregated influence of each feature, quantified through cumulative survSHAP values, provided a dynamic and nuanced assessment of their contributions to HBV reactivation propensity.
2.5. Recipient stratification and clinical implications
Each recipient was assigned a crank score using the “predict_newdata” function from the R package “mlr3” (version 0.19.0). The scores were then normalized to a 0–120 scale, designated as the alcohol-modified hepatitis B virus reactivation index (AMBRI). This scale was divided into eight equal intervals of 15 points each to determine the optimal threshold for stratifying recipients based on post-transplant HBV reactivation risk.
The clinical significance of AMBRI stratification was primarily assessed in liver transplant recipients experiencing HBV reactivation, with additional evaluations of OS and tumor recurrence-free survival (RFS). The impact of alcohol consumption on AMBRI-based risk stratification was statistically compared using the χ2 test.
2.6. Statistical analysis
Categorical variables were presented as counts (percentages) and compared using the χ2 test or Fisher’s exact test, as appropriate. Continuous variables were expressed as medians [interquartile ranges (IQRs)] and analyzed using the Kruskal–Wallis test. Survival curves were compared via the Kaplan–Meier method with log-rank testing. In cases where survival curves intersected, landmark analysis was applied using the R package “jskm” (version 0.5.2). Receiver operating characteristic (ROC) curves were compared via DeLong’s test using the “compare” function from the R package “pROC” (version 1.18.0). Hazard ratios (HRs) for each subgroup were visualized in a forest plot using the “ggforest” function from the R package “survminer” (version 0.4.9).
Propensity score matching (PSM) was conducted using a 1:2 matching ratio, applying the default nearest-neighbor matching algorithm from the R package “MatchIt” (version 4.5.5). Propensity scores (PSs) were estimated via multivariable logistic regression, incorporating recipient age, sex, serum alpha-fetoprotein (AFP) level, post-transplant alcohol consumption/relapse, and cold ischemia time (CIT) as covariates. Adjustment for a single sex factor was avoided to prevent imbalance in other key prognostic variables for patients with HCC. The PSM cohorts demonstrated well-balanced baseline characteristics.
Sensitivity analyses were performed under the following conditions: ① excluding recipients who received HBsAg-positive grafts; ② excluding recipients with post-transplant alcohol consumption or relapse; ③ excluding recipients with a follow-up duration of less than five years.
A P-value < 0.05 was considered statistically significant across all analyses. AUCs were used to assess the model’s predictive performance, with the optimal threshold determined by maximizing the Youden index (balancing sensitivity and specificity). Further validation included decision curve analysis (DCA) and the Precision–Recall (P–R) curve. All statistical analyses were conducted using R software (version 4.2.1).
3. Results
3.1. Patient baseline characteristics
A total of 453 HBV-related liver transplant recipients met the study’s inclusion criteria (Fig. S1), comprising 113 patients (24.9%) with a concurrent pre-transplant diagnosis of ALD and HBV infection (AB-HCC) and 340 patients (75.1%) with HBV infection alone (B-HCC). Baseline characteristics and surgical outcomes for both the entire cohort and the 1:2 PSM cohort are detailed in
Table 1.
Before matching, the AB-HCC group exhibited a significantly higher proportion of males (100.0% vs 87.9%, P < 0.001). Following stepwise 1:2 PSM, 113 AB-HCC cases were successfully matched to 226 B-HCC cases. Among pre-transplant factors, AB-HCC patients demonstrated significantly lower serum total cholesterol (TC) (before matching: 2.80 [1.99, 3.68] vs 3.20 [2.43, 4.04], P = 0.001; after matching: 2.80 [1.99, 3.68] vs 3.20 [2.45, 4.00], P = 0.002), lower high-density lipoprotein (HDL) (before matching: 0.74 [0.50, 1.00] vs 0.92 [IQR: 0.63, 1.21], P < 0.001; after matching: 0.74 [0.50, 1.00] vs 0.91 [0.61, 1.16], P = 0.001), and lower low-density lipoprotein (LDL) (before matching: 1.43 [0.92, 2.12] vs 1.63 [1.20, 2.30], P = 0.006; after matching: 1.43 [0.92, 2.12] vs 1.65 [1.23, 2.22], P = 0.008) compared to B-HCC patients. Analysis of post-transplant outcomes revealed a higher incidence of HBV reactivation in the AB-HCC group (before matching: 23.9% vs 14.4%, P = 0.028; after matching: 23.9% vs 13.7%, P = 0.028) and a greater proportion of recipients with liver metastasis (before matching: 27.4% vs 15.9%, P = 0.010; after matching: 27.4% vs 13.7%, P = 0.003). Notably, post-transplant alcohol consumption did not differ significantly between the AB-HCC and B-HCC groups (before matching: 7.1% vs 3.8%, P = 0.243; after matching: 7.1% vs 4.4%, P = 0.441).
3.2. Surgical outcomes between AB-HCC and B-HCC groups
Both unmatched and PSM datasets were utilized to assess the impact of ALD on HBV-related HCC outcomes in liver transplant recipients. Notably, patients with a pre-transplant diagnosis of ALD and HBV infection (AB-HCC) exhibited inferior reactivation-free survival, OS, and tumor RFS compared to those with HBV infection alone (B-HCC). This trend remained consistent across both unmatched and 1:2 PSM cohorts (
Fig. 1; log-rank
P = 0.017 vs 0.018; 0.011 vs 0.015; 0.0032 vs 0.0043, respectively). In the 1:2 PSM cohort, the one-year and five-year HBV reactivation-free survival rates were 82.1% vs 90.1% and 74.9% vs 85.4%, respectively (
Fig. 1(b)). The one-year and five-year OS rates were 90.2% vs 94.2% and 56.2% vs 70.5%, respectively. Similarly, the one-year and five-year tumor RFS rates were 69.9% vs 79.2% and 47.8% vs 63.3%, respectively.
To further validate these findings, sensitivity analyses were conducted by excluding recipients who received HBsAg-positive grafts or reported post-transplant alcohol consumption. In the first sensitivity analysis, statistically significant disparities were observed in HBV reactivation (P = 0.0032), OS (P = 0.012), and tumor RFS (P = 0.0022, Fig. S2(a) in Appendix A). The second sensitivity analysis reinforced these findings, demonstrating a robust statistical difference in HBV reactivation (P = 0.005), OS (P = 0.016), and tumor RFS (P = 0.0088, Fig. S2(b) in Appendix A).
Subset analyses stratified by HBV reactivation status revealed no significant difference in tumor RFS (
Fig. 2(a); log-rank
P = 0.67) or OS (
Fig. 2(b); log-rank
P = 0.31) among recipients who experienced HBV reactivation. However, in recipients without HBV reactivation, ALD further worsened tumor RFS (
Fig. 2(c); log-rank
P = 0.026). The negative impact of ALD on OS persisted beyond 12 months (
Fig. 2(d); landmark
P = 0.032), whereas no significant difference was observed within the first post-transplant year (
Fig. 2(d); landmark
P = 0.366). Comparable trends across the full cohort were illustrated in Fig. S3 in Appendix A.
3.3. Explainable machine learning method to predict HBV reactivation status
Leveraging clinical data to identify risk factors for predicting post-transplant HBV reactivation represents a compelling strategy. The machine learning framework is outlined in Fig. S4 in Appendix A. Pre-processed data were partitioned into a training/validation cohort (
n = 362, 80%) with five-fold and ten-fold internal CV and an external test cohort (
n = 91, 20%), with no significant baseline differences, ensuring clinical comparability (Table S2 in Appendix A). Benchmarking across 21 machine learning algorithms identified surv.cforest as the top-performing method in the training/internal validation set (C-index: 0.619 for five-fold CV, Fig. S5(a) in Appendix A; 0.632 for ten-fold CV, Fig. S5(b) in Appendix A). Feature importance rankings are presented in Fig. S6 in Appendix A. To construct a parsimonious yet high-performing model, the clinical feature set was systematically reduced from 25 to 3, while maintaining comparable predictive performance with the top 12 features across multiple models in the training/validation cohort (Fig. S7(a) and Table S3 in Appendix A). The optimized 12-feature model achieved an AUC of 0.914, with a sensitivity of 0.921, specificity of 0.839, Youden index of 0.760, PPV of 0.547, NPV of 0.980, accuracy of 0.854, and an F1 score of 0.686, effectively predicting HBV reactivation risk within five years post-liver transplantation (Fig. S7(a) and Table S3). Comparative analysis revealed superior performance of the 12-feature model over the 3-feature model (DeLong
P < 0.001), with no significant difference in predictive power compared to the 25-feature model (DeLong
P = 0.521, Fig. S7(b) in Appendix A). Furthermore, the 12-feature model exhibited substantial net benefit and favorable threshold probability, with an area under the P–R curve comparable to the 25-feature model, reinforcing its high clinical utility (Figs. S7(c) and S7(d) in Appendix A). External validation in the test cohort confirmed robust predictive performance, yielding an AUC of 0.884, underscoring the model’s stability and generalizability (Fig. S7(e) in Appendix A). The final model integrates 12 clinically accessible variables: microvascular invasion (MVI) status, receipt of an HBsAg-positive graft, history of ALD, exceeding Milan criteria, detectable HBV DNA, tumor size > 5 cm, the model for end-stage liver disease (MELD) score, recipient HBsAg level, TC, serum HDL, very low-density lipoprotein (VLDL), and triglyceride (TG) level (
Fig. 3(a)).
To elucidate the contribution of clinical features, an explainable machine learning approach, SHAP, was employed. The risk of HBV reactivation was assessed using survSHAP values, where lower values indicated a higher likelihood of reactivation and vice versa. At the global level, the five most influential predictors were MVI, receipt of an HBsAg-positive graft, HDL, MELD score, and ALD (
Figs. 3(a) and
(b)). Furthermore, metabolic parameters played a crucial role in predicting HBV reactivation, with HDL, TG, VLDL, and TC identified as key determinants. Notably, lower serum HDL levels, alongside elevated TG, VLDL, and TC levels, correlated with an increased risk of HBV reactivation (
Fig. 3(b)). Partial dependence survival analysis demonstrated that recipients with a history of ALD, receipt of an HBsAg-positive graft, MVI status, exceeding Milan criteria, detectable HBV DNA, tumor size > 5 cm, higher recipient HBsAg levels, reduced serum HDL, and elevated serum VLDL, TG, TC, and MELD scores exhibited a progressively increasing cumulative hazard ratio for HBV reactivation over time (
Fig. 3(c)).
3.4. Impact of the AMBRI score on prognosis among liver recipients
The prognostic assessment encompassed the entire dataset of 453 liver transplant recipients from both the training and validation cohorts. Stratification into eight subgroups identified distinct AMBRI thresholds at 45 and 90 (Fig. S8 in Appendix A). Based on these thresholds, the cohort was categorized into three risk groups: low risk (AMBRI < 45, 72.1%), intermediate risk (AMBRI 45–90, 23.0%), and high risk (AMBRI ≥ 90, 4.9%) (Fig. S8).
The clinical characteristics of these subgroups are detailed in
Table 2. Recipients in the intermediate- and high-risk groups exhibited more aggressive disease phenotypes, including significantly larger tumor sizes (
P < 0.001), a higher number of tumors (
P < 0.001), increased MVI incidence (
P < 0.001), and a greater likelihood of exceeding Milan criteria (
P < 0.001). Additionally, these groups displayed elevated MELD scores (
P < 0.001) and higher AFP levels (
P < 0.001). In terms of HBV-related parameters, intermediate- and high-risk recipients more frequently received HBsAg-positive grafts (
P < 0.001), had detectable pre-transplant HBV DNA (
P < 0.001), and exhibited higher HBsAg levels (
P = 0.001). Moreover, intraoperative blood loss was significantly greater in the high-risk group (
P = 0.009), reflecting the increased malignancy and complexity of their tumors. Postoperative outcomes further underscored the prognostic relevance of AMBRI. Intermediate- and high-risk recipients demonstrated a significantly higher incidence of HBV reactivation (
P < 0.001) and metastatic progression, including lung (
P < 0.001), liver (
P < 0.001), and bone (
P < 0.001) metastases. No significant differences in post-transplant alcohol relapse or consumption were observed among the three subgroups (
P = 0.907).
Correlation analysis revealed a significant negative association between AMBRI and key prognostic endpoints, including time to HBV reactivation (R = –0.45, P < 0.001), OS (R = –0.31, P < 0.001), and tumor RFS (R = –0.4, P < 0.001), underscoring its potential as a prognostic biomarker (Fig. S9(a) in Appendix A). Sensitivity analysis, excluding survivors with follow-up durations under five years, yielded consistent results, with stronger negative correlations observed for time to HBV reactivation (R = –0.71, P < 0.001), OS (R = –0.4, P < 0.001), and tumor RFS (R = –0.4, P < 0.001) (Fig. S9(b) in Appendix A).
The prognostic implications across the three AMBRI-defined risk groups were systematically analyzed. The five-year HBV reactivation-free survival rate was highest in the low-risk group (96.2%), markedly lower in the intermediate-risk group (52.9%), and dropped to less than 12.7% in the high-risk group (
Fig. 4(a), Fig. S4). A similar trend was observed for OS, with five-year survival rates of 79.3% in the low-risk group, declining to 34.8% in the intermediate-risk group, and less than 42.2% in the high-risk group (Figs. S4, S10(a), and S10(b) in Appendix A). Tumor RFS followed a comparable pattern, with rates of 73.2%, 27.0%, and less than 18.2% in the low-, intermediate-, and high-risk groups, respectively (Figs. S4, S10(c), and S10(d) in Appendix A), highlighting the pronounced impact of AMBRI stratification on clinical outcomes.
Compared to the low-risk group, the intermediate-risk group exhibited a significantly higher incidence of HBV reactivation (HR = 16, 95% confidence interval (95%CI) = 8.2–30,
Fig. 4(b)), lower OS probability (HR = 3.8, 95%CI = 2.6–5.5, Fig. S10(b)), and reduced tumor RFS probability (HR = 4.3, 95%CI = 3.1–5.9, Fig. S10(d)). The high-risk group demonstrated an even greater increase in HBV reactivation risk (HR = 46, 95%CI = 22–96,
Fig. 4(b)). However, despite significantly worse OS and RFS compared to the low-risk group, OS probability (HR = 4.6, 95%CI = 2.4–8.8, Fig. S10(b)) and tumor RFS probability (HR = 5.6, 95%CI = 3.3–9.4, Fig. S10(d)) in the high-risk group were comparable to those in the intermediate-risk group. Analysis of ALD prevalence across risk groups revealed a significantly higher proportion of intermediate- and high-risk patients in AB-HCC compared to B-HCC (
P < 0.001,
Fig. 4(c)).
To minimize the potential confounding effect of post-transplant alcohol consumption on AMBRI scoring, a sensitivity analysis was conducted, excluding cases with documented alcohol relapse or intake (n = 21). The results remained consistent (Figs. S11(a)–(f) in Appendix A), with the association between pre-transplant ALD and the increased proportion of intermediate- and high-risk groups persisting (Fig. S11(g) in Appendix A), reinforcing the strong correlation between ALD and AMBRI classification.
For clinical application, a user-friendly web-based predictive tool was developed†, as illustrated in Fig. S12 in Appendix A. Upon entering the 12 required clinical variables in the upper panel, the predicted HBV reactivation risk is displayed in the lower panel. For instance, an AMBRI score of 112.13, classified as high-risk, corresponded to an HBV reactivation rate of 41.5%, substantially exceeding the baseline HBV reactivation rate of 16.5% observed in previous studies.
4. Discussion
HBV reactivation is a critical event that can precipitate liver failure and mortality
[13]. This study represents the first to identify ALD as a significant risk factor for HBV reactivation in liver transplant recipients with HBV-related HCC (entire cohort: 23.9% vs 14.4%,
P = 0.028; PSM cohort: 23.9% vs 13.7%,
P = 0.028, comparing ALD and non-ALD recipients, respectively). Subgroup analysis consistently demonstrated poor outcomes following HBV reactivation, with alcohol use further exacerbating OS (beyond 12 months; landmark
P = 0.032) and tumor RFS (log-rank
P = 0.026) among recipients without HBV reactivation. Notably, the subgroup comprising HBV-related HCC patients with ALD but without post-transplant HBV reactivation exhibited an overlooked survival risk, despite being previously considered a favorable prognostic group
[7].
Several factors could influence the study findings, particularly patient compliance. A multidisciplinary assessment incorporating psychosocial evaluation, rather than the conventional six-month abstinence rule, was utilized to determine liver transplant eligibility for ALD patients
[14]. Rigorous selection criteria ensured that post-transplant alcohol relapse or intake rates did not significantly differ between the AB-HCC and B-HCC groups (pre-matching: 7.1% vs 3.8%,
P = 0.243; post-matching: 7.1% vs 4.4%,
P = 0.441). Additionally, the duration of antiviral therapy varies across clinical guidelines
[13],
[15],
[16],
[17],
[18], and in this study, a standardized first-line regimen was employed. Pre-transplant HBV prophylaxis with entecavir or tenofovir was combined with lifelong low-dose hepatitis B immunoglobulin administration to ensure sustained viral suppression post-transplant. A uniform immunosuppressive protocol and close clinical follow-up further reinforced treatment adherence. To mitigate potential confounders, sensitivity analyses were conducted, excluding recipients who failed abstinence or received HBsAg-positive grafts—both of which have been implicated in HBV reactivation risk. The results remained consistent, affirming ALD as an independent risk factor for HBV reactivation, as well as for poorer OS and tumor RFS.
Machine learning, often criticized as a “black box” due to its opacity in decision-making, poses challenges for clinical adoption, as physicians may hesitate to rely on models lacking interpretability. This study addresses this limitation by employing the SHAP approach, which elucidates both global model behavior and local predictions. Through SHAP-based interpretation, our 12-feature clinical model, incorporating metabolic parameters, identified ALD as an independent risk factor for HBV reactivation, underscoring the need for intensified management in liver transplant recipients with a history of HBV- and ALD-related HCC. The predictive model exhibited superior performance, achieving an AUC of 0.914 in internal validation and 0.884 in external validation, significantly surpassing the previous Cox model (AUC = 0.78). Sensitivity reached 0.921, with a specificity of 0.839, demonstrating robust predictive accuracy. To enhance clinical applicability, the model was integrated into a user-friendly web platform, facilitating real-time risk assessment and expanding accessibility for liver transplant specialists.
An AMBRI-based stratification system was developed to classify liver recipients with HBV background according to HBV reactivation risk. Beyond accurately identifying high-risk individuals, AMBRI also effectively differentiated surgical outcomes, including OS and tumor RFS. Recipients in the intermediate- and high-risk groups exhibited significantly shorter OS and tumor RFS, alongside more aggressive clinical features such as increased MVI incidence, elevated MELD scores, and higher AFP levels. These findings highlight the need for early interventions in patients with elevated AMBRI scores. Further analysis revealed a strong correlation between ALD and AMBRI stratification, providing a basis for refining risk prediction in ALD-affected recipients. For patients with pre-transplant ALD and HBV comorbidity, an intensified follow-up protocol is recommended: monthly assessments within the first postoperative year, followed by quarterly monitoring during years 1–2, and biannual evaluations thereafter. The observed stabilization in HBV reactivation-free survival and tumor RFS beyond two years suggests this as a critical milestone in disease progression. Essential follow-up assessments should include HBV-related biomarkers, AFP, and abdominal computed tomography (CT) imaging. Additionally, an escalation in HBV antiviral therapy dosage may be warranted, pending further clinical validation.
This study has several limitations. A primary constraint is its retrospective design, which inherently introduces potential biases, such as selection bias in grouping patients. To mitigate known confounders, PSM was performed to balance influential variables (e.g., sex). However, the study remains vulnerable to the influence of unidentified non-oncological factors, including patient selection variability, differences in perioperative protocols, surgical techniques, donor alcohol consumption, and baseline recipient medical conditions (e.g., metabolic dysfunction-associated fatty liver disease). These factors are inherent to the multicenter registry-based approach. Additionally, in clinical practice, liver recipients with a history of HCC are increasingly treated with immunotherapies (e.g., anti-PD-1 or PD-L1 therapies), an emerging area of interest within the transplant community. However, our database lacked detailed records on preoperative treatments. Therefore, the impact of preoperative therapies, such as targeted treatment, which has been linked to post-transplant HBV reactivation, cannot be disregarded
[19]. Furthermore, the optimal timing for pre-transplant immunotherapy in HBV-related HCC patients, whether with or without a history of ALD, remains undetermined.
In conclusion, liver transplant recipients with a history of both ALD and HBV-related HCC demonstrate worse post-transplant outcomes, including reduced OS, RFS, and HBV reactivation-free survival, compared to those with a single HBV history. These findings underscore the importance of tailored risk assessment for individuals with high-risk ALD history to optimize clinical management and outcomes.
CRediT authorship contribution statement
Renyi Su: Writing – original draft, Methodology, Formal analysis, Data curation, Conceptualization. Huigang Li: Writing – review & editing, Conceptualization. Xuanyu Zhang: Writing – review & editing, Data curation. Linping Cao: Writing – review & editing, Data curation. Zhe Yang: Writing – review & editing, Data curation. Jinyan Chen: Data curation. Shusen Zheng: Conceptualization. Xiao Xu: Writing – review & editing, Supervision, Conceptualization. Di Lu: Writing – review & editing, Supervision, Conceptualization. Xuyong Wei: Writing – review & editing, Supervision, Conceptualization.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was supported by the National Natural Science Foundation of China (92159202), the National Science and Technology Major Project of China (2017ZX10203205), the Fundamental Research Funds for the Central Universities (2022ZFJH003), the Major Research Plan of Key Research and Development Project of Zhejiang Province (2024C03149 and 2023C03046), the National Key Research and Development Program of China (2021YFF1200404 and 2021YFA1201200), the Zhejiang Provincial Natural Science Foundation of China (LQ24C050005 and LQ22H160052), and the Medical and Health Technology Program Project of Zhejiang Province (2021434810).
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.eng.2025.03.016.
PDF
(1998KB)
