Baishideng Publishing Group Inc, 7041 Koll Center Parkway, Suite 160, Pleasanton, CA 94566, USA
Share the Article
Huang JC, Wang HX, Zhao Y, Li XL, He Q, Lang R, Zhou L. Nomogram integrating prognostic immune-inflammatory-nutritional score with clinicopathological parameters for predicting post-transplantation outcomes in hepatocellular carcinoma. World J Gastroenterol 2027; 33(1): 122971 [DOI: 10.3748/wjg.122971]
World J Gastroenterol. Jan 7, 2027; 33(1): 122971 Published online Jan 7, 2027. doi: 10.3748/wjg.122971
Nomogram integrating prognostic immune-inflammatory-nutritional score with clinicopathological parameters for predicting post-transplantation outcomes in hepatocellular carcinoma
Jin-Can Huang, Han-Xuan Wang, Yang Zhao, Xian-Liang Li, Qiang He, Ren Lang, Lin Zhou, Division of Hepatobiliary and Pancreaticosplenic Surgery, Department of General Surgery, Beijing Chao-Yang Hospital, Capital Medical University, Beijing 100020, China
Jin-Can Huang, Department of Hepatobiliary Surgery, The Affiliated Lianyungang Hospital of Xuzhou Medical University, Lianyungang 222000, Jiangsu Province, China
Co-first authors: Jin-Can Huang and Han-Xuan Wang.
Co-corresponding authors: Ren Lang and Lin Zhou.
Author contributions: Huang JC and Wang HX contributed equally to this manuscript and shared the first authorship; Zhou L and Lang R were the co-corresponding authors of this research; Huang JC and Zhou L conceptualized and designed the research study; Huang JC and Wang HX performed the formal analysis, wrote the original draft; Zhao Y collected the data; Li XL and He Q provided resources; Lang R and Zhou L supervised the project and administered the study; Zhou L acquired funding for the study; Lang R and Zhou L reviewed and edited the manuscript; all authors have read and approved the final version of the manuscript.
AI contribution statement: AI tools (ChatGPT) were used for language polishing during manuscript preparation. No AI tool was involved in the generation of research data, interpretation of results, or formulation of conclusions. All AI-generated outputs were critically reviewed and revised by the authors.
Supported by the National Natural Science Foundation of China, No. 82370665.
Institutional review board statement: This study was conducted in accordance with the Declaration of Helsinki. The study was reviewed and approved by the Ethics Committee of Beijing Chao-Yang Hospital, Capital Medical University (approval No. 2020-D-303).
Informed consent statement: Given the retrospective design, the requirement for written informed consent was waived by Ethics Committee of Beijing Chao-Yang Hospital, Capital Medical University.
Conflict-of-interest statement: The authors declare that they have no conflict of interest.
STROBE statement: The authors have read the STROBE Statement—a checklist of items, and the manuscript was prepared and revised according to the STROBE Statement-a checklist of items.
Data sharing statement: The datasets used during the current study are available from the corresponding author (richard_fmmu@126.com) on reasonable request.
Corresponding author: Lin Zhou, MD, Doctor, Division of Hepatobiliary and Pancreaticosplenic Surgery, Department of General Surgery, Beijing Chao-Yang Hospital, Capital Medical University, No. 8 Gongtinan Road, Chaoyang District, Beijing 100020, China. richard_fmmu@126.com
Received: May 7, 2026 Revised: July 3, 2026 Accepted: July 29, 2026 Published online: January 7, 2027 Processing time: 189 Days and 20.5 Hours
Abstract
BACKGROUND
Nutritional, immune, and inflammatory status of patients are closely associated with tumor progression and prognosis. However, comprehensive prognostic indicators that integrates nutritional, immune, and inflammatory status remain limited in hepatocellular carcinoma (HCC) patients after liver transplantation (LT).
AIM
To develop a composite index prognostic immune-inflammatory-nutritional (PIIN) score and evaluating its predictive value for post-LT prognosis in HCC patients.
METHODS
This study adopted a retrospective cohort study design and included 269 HCC patients underwent LT. Nutritional, immune, and inflammatory indexes were integrated to construct the PIIN score using least absolute shrinkage and selection operator regression. A predictive nomogram was subsequently established by incorporating the PIIN score and clinicopathological variables, which was intended for perioperative or postoperative risk stratification. Its performance was evaluated using calibration curves, time-dependent receiver operating characteristic analysis, the concordance index, and decision curve analysis.
RESULTS
Neutrophil-to-lymphocyte ratio, prognostic nutritional index, platelet-to-lymphocyte ratio, systemic immune-inflammation index and systemic inflammation response index were identified to construct PIIN score. Patients with high PIIN score had significantly worse overall survival (OS) and disease-free survival (DFS) than those with low PIIN score in both training and validation cohorts after stratified according to the training-set-derived cutoff value of the PIIN score. Multivariable analysis confirmed PIIN score as an independent prognostic factor for OS [hazard ratio (HR) = 3.277, 95% confidence interval (CI): 2.270-4.731; P < 0.001] and DFS (HR = 3.007, 95%CI: 2.099-4.307; P < 0.001) in both training and validation cohort. A PIIN-based nomogram incorporating PIIN score, tumor number, tumor size, alpha fetoprotein and differentiation was developed, showing favorable discrimination, calibration, and clinical utility, with better predictive performance than conventional criteria.
CONCLUSION
PIIN score may help predict post-transplant outcomes in HCC patients. A PIIN-based model incorporating PIIN score and clinicopathological indexes may serve as a potentially useful tool for risk stratification.
Core Tip: Although nutritional, immune and inflammatory status affect hepatocellular carcinoma (HCC) prognosis, composite indicators reflecting these factors remain unreported. We retrospectively enrolled 269 HCC patients undergoing liver transplantation and constructed a prognostic immune-inflammatory-nutritional (PIIN) score using least absolute shrinkage and selection operator regression on routine blood test-derived indicators. A PIIN-based prediction model incorporating PIIN score, tumor number, tumor size, alpha fetoprotein and differentiation was also established. A high PIIN score independently predicted worse overall and disease-free survival, showed favorable discrimination, calibration, and clinical utility superior to conventional criteria, and may serve as a practical tool for evaluating post-transplant prognosis in HCC.
Citation: Huang JC, Wang HX, Zhao Y, Li XL, He Q, Lang R, Zhou L. Nomogram integrating prognostic immune-inflammatory-nutritional score with clinicopathological parameters for predicting post-transplantation outcomes in hepatocellular carcinoma. World J Gastroenterol 2027; 33(1): 122971
Hepatocellular carcinoma (HCC) continues to be a significant global public health issue, exerting a considerable disease burden internationally. Despite a gradual decline in HCC incidence in recent years attributable to antiviral therapy for hepatitis B virus (HBV), HCC remained the 4th and 7th most diagnosed cancer in Chinese males and females[1]. In addition, HCC is also a lethal disease, ranking as the 5th and 7th leading cause of cancer-related mortality[2]. Currently, surgical treatment remains the central component of comprehensive treatment for HCC, while liver transplantation (LT) enables HCC patients with multiple tumor sites and high tumor burden who are not suitable for liver resection to have a chance for radical operation, with confirmed 5-year OS rates of approximately 70%-90%[3-5]. However, previous research also found that patients with advanced tumor stage and macrovascular invasion tended to have poorer clinical outcomes, indicating that the survival advantage of LT is not universal across patients[6]. Considering the substantial procedural risks and financial burden, clearly predicting post-transplantation clinical outcomes and making optimal surgical decisions is beneficial for both patients and clinicians.
Researchers have identified a series of indexes, such as tumor size, histological grade, and alpha-fetoprotein (AFP) levels, as effective predictors for post-LT prognosis in HCC patients[7,8]. Since chronic liver inflammation caused by HBV infection has been shown to promote HCC by suppressing immune surveillance, and malnutrition may increase the risk of recurrence and mortality in HCC, researchers have focused on the roles of various inflammatory and nutritional factors in predicting postoperative prognosis[9-11]. To date, a series of indexes such as the neutrophil-to-lymphocyte ratio (NLR), lymphocyte-to-monocyte ratio (LMR), systemic inflammation response index (SIRI), and prognostic nutritional index (PNI) have been confirmed as effective prognostic biomarkers in HCC patients after LT[12-15]. However, the predictive accuracy of these markers in isolation was unsatisfactory due to either low specificity or low sensitivity, limiting their clinical utility. To address this issue, prior researchers attempted to establish prediction models by incorporating various inflammatory indexes and achieved significantly improved predictive value in HCC patients undergoing liver resection[16]. Thus, we believed that integrating these biomarkers into a unified prognostic model could provide a more accurate and comprehensive evaluation of postoperative survival in HCC patients after LT.
This study aimed to develop a composite index, termed prognostic immune-inflammatory-nutritional (PIIN) score, incorporating immunological, inflammatory, and nutritional biomarkers, and to evaluate its predictive value in HCC patients after LT.
MATERIALS AND METHODS
Patient selection
This study adopted a retrospective cohort study design. Patients who were preoperatively diagnosed with primary HCC for the first time and received curative-intent LT from January 2012 to June 2023 were retrospectively enrolled. Based on the inclusion and exclusion criteria, eligible patients were randomly assigned to a training cohort and a validation cohort at a ratio of 7:3. The inclusion criteria were as follows: (1) Regular follow-up after LT; and (2) Histopathological confirmation of HCC. The exclusion criteria were as follows: (1) Had concurrent acute infection; (2) Had missing continuous data with a missing rate ≥ 20%; (3) Had missing categorical data; (4) Had incomplete follow-up data; and (5) Suffered perioperative death. The study was reviewed and approved by the Ethics Committee of Beijing Chao-Yang Hospital, Capital Medical University (approval No. 2020-D-303). Given the retrospective design, the requirement for written informed consent was waived by the Ethics Committee of our hospital.
Data collection and follow-up
Patients’ demographic characteristics and laboratory test results were collected from the electronic medical record system of our hospital. For all participants, venous blood was collected within 3 days before surgery. A complete blood count was performed for each patient, comprising neutrophil count (NEUT), platelet count (PLT), lymphocyte count (LYC), monocyte count (MONO), hemoglobin, albumin (ALB), total bilirubin, and other related parameters. We calculated immune, inflammatory, and nutritional indexes using standard formulas: NLR = NEUT/LYC; PLR = PLT/LYC; LMR = LYC/MONO; Systemic inflammatory index (SII) = PLT × NEUT/LYC; SIRI = MONO × NEUT/LYC; PNI = serum ALB (g/L) + 5 × LYC (109/L); ALB-bilirubin score (ALBI) = [log10 total bilirubin (μmol/L) × 0.66] + [ALB (g/L) × -0.0852]. In the regression analysis, these markers were retained as continuous variables. For visualization purposes only, the PIIN score was dichotomized using a training-set-derived threshold based on receiver operating characteristic (ROC) analysis and the Youden index, and this prespecified threshold was then applied to the validation cohort.
We followed all patients until death or June 2025. Follow-up was performed by phone calls and outpatient and inpatient visits. Patients were assessed at 3-month intervals throughout the first two postoperative years. After that, follow-up was performed every six months. When serum AFP levels increased or imaging findings suggested possible liver or lung lesions, further diagnostic evaluation was performed. Tumor recurrence was defined based on radiological evidence and, when necessary, histopathological confirmation. Overall survival (OS) was defined as the duration from operation to death from any cause or the last follow-up. Disease-free survival (DFS) was defined as the interval from operation to tumor recurrence. Patients who died without documented recurrence were censored at death, and patients alive without recurrence were censored at the last follow-up.
Construction of the PIIN score
The PIIN score was developed from seven prespecified immune-, inflammation-, and nutrition-related biomarkers (ALBI, PNI, PLR, NLR, LMR, SII, and SIRI) using least absolute shrinkage and selection operator regression (LASSO)-Cox regression in the training cohort. The resulting non-zero coefficients were combined to generate a fixed PIIN score formula. In the subsequent prognostic modeling stage, the PIIN score was entered as a single continuous composite predictor together with a limited set of clinically relevant clinicopathological variables, therefore, the hazard ratios (HRs) reported for the PIIN score correspond to each 1-unit increase in the continuous PIIN score. Clinicopathological variables were prespecified based on clinical relevance and data availability, and variables showing P < 0.05 in univariable analysis were further considered in multivariable modeling. To avoid redundancy and multicollinearity, the individual biomarkers included in the PIIN score were not separately entered into the nomogram models. Model robustness was further assessed by proportional hazards testing and bootstrap-based internal validation.
Prognostic significance of the PIIN score
The predictive efficacy of the PIIN score was evaluated using time-dependent ROC curve analysis to evaluate its predictive accuracy for 1-year, 3-year, and 5-year OS in both the training and validation sets. The area under the ROC curve (AUC) was calculated to quantify discriminatory capacity. For Kaplan-Meier visualization, patients in training and validation cohorts were categorized into high-risk and low-risk groups according to the optimal cutoff of PIIN score derived from the training cohort using ROC analysis and the Youden index. Kaplan-Meier curves were generated, and survival differences between groups were assessed using the log-rank test.
Development and validation of the nomogram
The proportional hazards assumption of the final Cox models was assessed using Schoenfeld residual-based tests. Univariate Cox regression analysis was performed in the training cohort to identify clinicopathological variables associated with OS and DFS, with the PIIN score entered as a continuous variable. Variables with P < 0.05 in the univariate analysis were subsequently entered into multivariable Cox regression models to determine independent prognostic factors. Based on the final multivariable models, two nomograms were developed to estimate the probabilities of 1-year, 3-year, and 5-year OS and DFS, respectively.
The performance of the nomograms was evaluated in terms of discrimination, calibration, and clinical utility. Discrimination was assessed using the concordance index (C-index) with 95% confidence intervals (CIs) and time-dependent AUC values. Pairwise comparisons of time-dependent AUCs were performed at 1 year, 3 years, and 5 years between the established nomogram and conventional prediction models to compare model discrimination. Specifically, the established model for OS was compared with the Milan and University of California, San Francisco (UCSF) criteria, while the established model for DFS was compared with the Milan and Risk Estimation of Tumor Recurrence After Transplant (RETREAT) criteria[17]. Calibration was evaluated using calibration curves. Internal validation was conducted in the training cohort using 1000 bootstrap resamples to quantify model optimism and obtain optimism-corrected estimates of model performance, including the optimism-corrected C-index, bootstrap-corrected time-dependent AUCs at 1 year, 3 years, and 5 years, the optimism-corrected calibration slope, and bootstrap-corrected calibration curves. Decision curve analysis (DCA) was performed to compare the clinical utility of the nomograms with established criteria, including the Milan and UCSF criteria for OS and the Milan and RETREAT criteria for DFS[17]. The predictive performance of the nomograms was further evaluated in the validation cohort.
Statistical analysis
Continuous data are presented as mean ± SD for normally distributed variables or median (Q1, Q3) for non-normally distributed variables. Continuous data conforming to a normal distribution were compared using the t test, whereas non-normally distributed data were compared using the rank-sum test. Categorical variables were compared using the χ2 test or Fisher’s exact test. Variables with a missing rate < 20% were imputed using the multiple imputation method, while those with a missing rate ≥ 20% were excluded from further analysis. Patients with missing postoperative outcomes were also excluded. Data exceeding the range of (Q1 - 1.5 interquartile range, Q3 + 1.5 interquartile range) were identified as outliers. Outliers of normally distributed variables were replaced by the mean value, while outliers of non-normally distributed variables were replaced by median of that variable in current cohort. A log-rank test was employed to compare survival outcomes using Kaplan-Meier curves. Univariable and multivariable Cox proportional hazards regression models were used to find independent prognostic variables. Statistical significance was set at P < 0.05. Sensitivity analysis excluding patients with Child-Pugh C liver function was performed to ensure the robustness of findings. Nomograms were developed using the independently significant prognostic factors. Predictive performance was assessed using the C-index, ROC analysis, calibration plots, and DCA. All statistical analysis were performed using R (4.3.2).
RESULTS
Baseline clinicopathological characteristics
A total of 269 patients were ultimately recruited, including 82.5% males and 17.5% females. The flowchart of patients selection is shown in Supplementary Figure 1. The cohort had a median age of 53 years (22-74 years) and a median follow-up duration of 46 months. Baseline demographics, clinicopathological variables, and laboratory profiles for the training set (n = 188) and the validation cohort (n = 81) are summarized in Table 1. Regarding liver function status, 55 patients (20.4%) were classified as Child-Pugh A, 124 (46.1%) as Child-Pugh B, and 90 (33.5%) as Child-Pugh C. During follow-up, 80 OS events and 71 DFS events were observed in the training cohort, whereas 35 OS events and 30 DFS events occurred in the validation cohort. There were no discernible differences (P > 0.05) in these baseline characteristics between the two cohorts. Kaplan-Meier survival analysis showed that lower ALBI, NLR, PLR, SII, and SIRI, together with higher LMR and PNI, were significantly correlated with better OS (Figure 1). Consistently, Kaplan-Meier analysis for DFS showed similar associations (P < 0.05) (Figure 2).
Figure 1 Kaplan-Meier curves for overall survival categorized by immune, inflammatory, and nutritional indexes.
A: Kaplan-Meier curves of overall survival (OS) stratified by albumin-bilirubin grade; B: Kaplan-Meier curves of OS stratified by prognostic nutritional index; C: Kaplan-Meier curves of OS stratified by lymphocyte-to-monocyte ratio; D: Kaplan-Meier curves of OS stratified by systemic immune-inflammation index; E: Kaplan-Meier curves of OS stratified by neutrophil-to-lymphocyte ratio; F: Kaplan-Meier curves of OS stratified by platelet-to-lymphocyte ratio; G: Kaplan-Meier curves of OS stratified by systemic inflammation response index. ALBI: Albumin-bilirubin index; PNI: Prognostic nutritional index; LMR: Lymphocyte-to-monocyte ratio; SII: Systemic immune-inflammation index; NLR: Neutrophil-to-lymphocyte ratio; PLR: Platelet-to-lymphocyte ratio; SIRI: Systemic inflammation response index.
Figure 3A illustrates the relationships among the seven immune, inflammation, and nutrition related biomarkers. We used LASSO regression to further analyze these seven candidate indexes (Figure 3B and C). This analysis identified five biomarkers with non-zero coefficients, which were NLR, PNI, PLR, SII, and SIRI. All were selected as components of the PIIN score. The PIIN score was calculated using a fixed formula determined by LASSO regression. PIIN score = 0.0004 × PLR + 0.0198 × NLR - 0.0520 × PNI + 0.0003 × SII + 0.0990 × SIRI. To assess the discriminative ability of the PIIN score, ROC analysis was performed. In the training set, the AUCs for predicting 1-year, 3-year, and 5-year post-transplant OS were 0.681, 0.689, and 0.742, respectively. In the validation set, the respective AUCs were 0.743, 0.796, and 0.712 (Figure 3D and E).
Figure 3 Development of the prognostic immune-inflammatory-nutritional score using a least absolute shrinkage and selection operator regression model.
A: Heatmap of pairwise connections across indicators related to immunity, inflammation, and nutrition; B: Coefficient path plot for seven candidate variables in least absolute shrinkage and selection operator regression; C: Cross-validation curve of least absolute shrinkage and selection operator regression; D: Receiver characteristic curves of prognostic immune-inflammatory-nutritional score in predicting postoperative 1-year, 3-year, and 5-year overall survival in the training set; E: Receiver characteristic curves of prognostic immune-inflammatory-nutritional score in predicting postoperative 1-year, 3-year, and 5-year overall survival in the validation sets. LMR: Lymphocyte-to-monocyte ratio; PNI: Prognostic nutritional index; PLR: Platelet-to-lymphocyte ratio; SII: Systemic immune-inflammation index; ALBI: Albumin-bilirubin index; NLR: Neutrophil-to-lymphocyte ratio; SIRI: Systemic inflammation response index; AUC: Area under the receiver characteristic curve.
Prognostic implications of the PIIN score
According to the ROC analysis, the numerical cutoff value of the PIIN score was -1.235. Patients were categorized into high- and low-risk groups based on this optimal cutoff of PIIN score as the threshold. Kaplan-Meier analysis demonstrated significantly worse OS and DFS in the high-risk group compared with the low-risk group, and these disparities were consistently evident in both the training and validation sets (Figure 4A-D).
Figure 4 Prognostic significance of the prognostic immune-inflammatory-nutritional score.
A: Kaplan-Meier curve of overall survival (OS) in the training set stratified by prognostic immune-inflammatory-nutritional (PIIN) score; B: Kaplan-Meier curve of disease-free survival (DFS) in the training set stratified by PIIN score; C: Kaplan-Meier curve of OS in the validation set stratified by PIIN score; D: Kaplan-Meier curve of DFS in the validation set stratified by PIIN score; E: Forest plot of the multivariable Cox regression for OS in the training set; F: Forest plot of the multivariable Cox regression for DFS in the training set; G: Forest plot of the multivariable Cox regression for OS in the validation set; H: Forest plot of the multivariable Cox regression for DFS in the validation set. PIIN: Prognostic immune-inflammatory-nutritional; AFP: Alpha fetoprotein; MVI: Microvascular invasion; HR: Hazard ratio; CI: Confidence interval.
As summarized in Tables 2 and 3, univariable Cox proportional hazards analysis revealed a significant association between the PIIN score and both OS and DFS in the training set (P < 0.001). After adjustment for established clinicopathological covariates in multivariable models, the PIIN score persisted as an independent predictive factor in the training set for OS (HR = 3.277, 95%CI: 2.270-4.731, P < 0.001) and DFS (HR = 3.007, 95%CI: 2.099-4.307, P < 0.001) (Figure 4E and F). These findings were replicated in the validation set, where the PIIN score likewise retained prognostic significance (P < 0.001) (Figure 4G and H). In a sensitivity analysis excluding the 90 patients with Child-Pugh C, the PIIN score remained independently associated with both OS and DFS (Supplementary Tables 1 and 2).
Table 2 Cox regression analysis for overall survival in the training set.
Development and validation of the PIIN-based prediction model
Schoenfeld residual-based tests showed that the global proportional hazards assumption was not significantly violated in any of the final OS or DFS models. Detailed variable-specific and global test results are presented in Supplementary Table 3. The final OS and DFS nomograms (Figure 5A and B) incorporated the PIIN score, tumor number, tumor size, AFP, and tumor differentiation. Model discrimination was first evaluated using time-dependent ROC analysis. In the training cohort, the AUCs of the OS nomogram at 1 year, 3 years, and 5 years were 0.871, 0.835, and 0.890, respectively, whereas those of the DFS nomogram were 0.825, 0.847, and 0.880 (Figure 5C and D). In the validation cohort, the corresponding AUCs for OS were 0.796, 0.787, and 0.784, while those for DFS were 0.797, 0.802, and 0.783 (Figure 5E and F). The PIIN-based prediction model also showed favorable discriminative ability in terms of the C-index. The C-indexes for OS and DFS were 0.810 (95%CI: 0.762-0.858) and 0.806 (95%CI: 0.760-0.853) in the training cohort, and 0.760 (95%CI: 0.671-0.848) and 0.764 (95%CI: 0.678-0.849) in the validation cohort, respectively, which were numerically higher than those of the Milan and UCSF criteria. Formal pairwise comparisons of time-dependent AUCs further showed that, for OS prediction, the nomogram achieved significantly higher AUCs than the Milan and UCSF criteria at all evaluated time points in both the training and validation cohorts (Supplementary Table 4). For DFS prediction, the nomogram likewise showed significantly higher AUCs than the Milan criteria at all evaluated time points in both cohorts (Supplementary Table 4). In comparison with the RETREAT score, the nomogram showed more modest gains, with statistically significant differences observed only at selected time points, whereas the remaining comparisons were not statistically significant (Supplementary Table 4).
Figure 5 Development and validation of the prognostic immune-inflammatory-nutritional score-based nomograms.
A: Nomogram incorporating the prognostic immune-inflammatory-nutritional (PIIN) score and clinicopathological variables to predict overall survival (OS) in the training set; B: Nomogram incorporating the PIIN score and clinicopathological variables to predict disease-free survival (DFS) in the training set; C: Receiver operating characteristic curves (ROC) of the PIIN-based nomogram in predicting postoperative 1-year, 3-year, and 5-year OS in the training set; D: ROC of the PIIN-based nomogram in predicting postoperative 1-year, 3-year, and 5-year DFS in the training set; E: ROC of the PIIN-based nomogram in predicting postoperative 1-year, 3-year, and 5-year OS in the validation set; F: ROC of the PIIN-based nomogram in predicting postoperative 1-year, 3-year, and 5-year DFS in the validation set. AFP: Alpha fetoprotein; PIIN: Prognostic immune-inflammatory-nutritional; OS: Overall survival; DFS: Disease-free survival; AUC: Area under the receiver characteristic curve.
Bootstrap internal validation further confirmed the robustness of the nomograms, with only mild overfitting observed. After optimism correction, the C-index was 0.803 for the OS nomogram and 0.799 for the DFS nomogram, while the optimism-corrected calibration slopes were 0.925 and 0.918, respectively. Moreover, the bootstrap-corrected time-dependent AUCs of the OS nomogram at 1 year, 3 years, and 5 years were 0.873, 0.850, and 0.877, respectively, and the corresponding values for the DFS nomogram were 0.860, 0.855, and 0.872. Calibration curves demonstrated good agreement between the predicted and observed probabilities for both OS and DFS across all evaluated time points in the training and validation cohorts (Figure 6), and the bootstrap-corrected calibration curves showed similarly good agreement (Supplementary Figure 2). DCA was performed to compare the net clinical benefit of the PIIN-based prediction model with that of conventional criteria. In the decision curves, the PIIN-based model exhibited favorable performance, with favorable net benefit across a wide range of threshold probabilities compared with the Milan criteria, UCSF criteria, and RETREAT score (Figure 7).
Figure 6 Calibration of the prognostic immune-inflammatory-nutritional score-based nomograms for overall survival and disease-free survival.
A: Calibration curve of predicted and observed 1-year overall survival (OS) probabilities in the training cohort; B: Calibration curve of predicted and observed 3-year OS probabilities in the training cohort; C: Calibration curve of predicted and observed 5-year OS probabilities in the training cohort; D: Calibration curve of predicted and observed 1-year disease-free survival (DFS) probabilities in the training cohort; E: Calibration curve of predicted and observed 3-year DFS probabilities in the training cohort; F: Calibration curve of predicted and observed 5-year DFS probabilities in the training cohort; G: Calibration curve of predicted and observed 1-year OS probabilities in the validation cohort; H: Calibration curve of predicted and observed 3-year OS probabilities in the validation cohort; I: Calibration curve of predicted and observed 5-year OS probabilities in the validation cohort; J: Calibration curve of predicted and observed 1-year DFS probabilities in the validation cohort; K: Calibration curve of predicted and observed 3-year DFS probabilities in the validation cohort; L: Calibration curve of predicted and observed 5-year DFS probabilities in the validation cohort. OS: Overall survival; DFS: Disease-free survival.
Figure 7 Decision curve analysis of the prognostic immune-inflammatory-nutritional score based on prediction models for overall survival and disease-free survival.
A: Comparison of decision curve analysis (DCA) curve of prognostic immune-inflammatory-nutritional (PIIN)-based prediction model, Milan criteria and University of California, San Francisco (UCSF) criteria for 1-year overall survival (OS) in the training set; B: Comparison of DCA curve of PIIN-based prediction model, Milan criteria and University of UCSF criteria for 3-year OS in the training set; C: Comparison of DCA curve of PIIN-based prediction model, Milan criteria and UCSF criteria for 5-year OS in the training set; D: Comparison of DCA curve of PIIN-based prediction model, Milan criteria and Risk Estimation of Tumor Recurrence After Transplant (RETREAT) score for 1-year disease-free survival (DFS) in the training set; E: Comparison of DCA curve of PIIN-based prediction model, Milan criteria and RETREAT score for 3-year DFS in the training set; F: Comparison of DCA curve of PIIN-based prediction model, Milan criteria and RETREAT score for 5-year DFS in the training set; G: Comparison of DCA curve of PIIN-based prediction model, Milan criteria and UCSF criteria for 1-year OS in the validation set; H: Comparison of DCA curve of PIIN-based prediction model, Milan criteria and University of UCSF criteria for 3-year OS in the validation set; I: Comparison of DCA curve of PIIN-based prediction model, Milan criteria and UCSF criteria for 5-year OS in the validation set; J: Comparison of DCA curve of PIIN-based prediction model, Milan criteria and RETREAT score for 1-year DFS in the validation set; K: Comparison of DCA curve of PIIN-based prediction model, Milan criteria and RETREAT score for 3-year DFS in the validation set; L: Comparison of DCA curve of PIIN-based prediction model, Milan criteria and RETREAT score for 5-year DFS in the validation set. UCSF: University of California, San Francisco; RETREAT: Risk Estimation of Tumor Recurrence After Transplant.
DISCUSSION
In this retrospective study, we first proposed a novel inflammatory-nutritional composite index, the PIIN score, and confirmed its prognostic value in HCC patients undergoing LT. Currently, the Milan criteria and UCSF criteria remain the most widely adopted and effective standards for recipient selection in HCC patients. However, their strict morphological limits may exclude patients who could benefit from LT. Because tumor differentiation is usually obtained after surgery, the proposed nomograms should be regarded primarily as perioperative or postoperative tools for risk stratification rather than purely preoperative selection tools. Our proposed PIIN score may potentially provide additional prognostic value to complement conventional models, indicating its potential value in assisting the prediction of clinical outcomes and clinical decision-making. However, it must be emphasized that the present findings should be interpreted as exploratory and hypothesis-generating. Although the PIIN score may be a potential early prognostic indicator, its clinical utility requires further prospective and external validation before any clinical application can be considered.
Several recurrence prediction models have been developed for HCC patients after LT, including models based on tumor burden, AFP, explant pathological features, and dynamic tumor behavior, such as the RETREAT score, the UCLA/Agopian nomogram, and the RELAPSE score[17-19]. In addition, biomarker-oriented models have been explored, including a tumor necrosis factor receptor 2-based nomogram for post-transplant lung metastasis[20]. These studies have improved individualized recurrence-risk estimation after LT, but many models mainly emphasize tumor morphology, serum tumor markers, or pathological features. The added value of the present PIIN-based model is that it incorporates a readily available systemic host-status indicator reflecting inflammation, immune balance, and nutrition together with established clinicopathological variables. Therefore, it may provide complementary prognostic information beyond tumor-centered criteria alone.
Accumulating evidence suggests that inflammation is a key contributor to tumor progression, post-treatment recurrence, and poor survival among HCC patients[21,22]. Peripheral blood-based biomarkers reflecting immune and inflammatory balance have individually been confirmed to be associate with prognosis in HCC patients after LT, especially those that were included in PIIN score, partly explaining reason for the prognostic value of PIIN score[13,14,23]. Moreover, according to previous researches, increasing NLR, PLR, SII and SIRI were all correlated with increasing risk of poor post-LT outcomes[14,23-25]. Consistent with previous research, an increase in these indexes was associated with a higher PIIN score, further validating the reliability of our proposed risk scoring system. From a biological perspective, abnormal inflammatory biomarkers likely reflect a protumorigenic systemic milieu characterized by excessive inflammatory activation and impaired antitumor immunity. Neutrophil- and platelet-driven inflammatory responses may suppress cytotoxic lymphocyte function and promote tumor angiogenesis and metastasis, thereby influencing the prognosis of HCC patients after LT[26,27].
Apart from inflammation, post-LT outcome is also closely associated with nutritional status and blood-derived nutritional indexes have been confirmed as effective prognostic biomarkers[28-31]. PNI, a widely adopted nutrition-related index based on laboratory tests, has also been confirmed as an effective indicator of post-transplant outcomes in HCC[15]. The 5-year OS and DFS rates in high-PNI group reached 88.5% and 94.7%, respectively, significantly higher than those in low-PNI group. In the proposed PIIN score, elevated PNI can also significantly decrease the PIIN score, indicating a lower risk of unfavorable clinical outcomes, consistent with previous research. According to previous researches, low ALB levels were correlated with elevated AFP, larger tumor size, and increased nodule number, suggesting that decreased PNI may be a sign of more aggressive tumor biological behavior[32]. Moreover, it was reported that decreased PNI was associated with a higher risk of developing postoperative complications, delaying the bridging to subsequent adjuvant therapy[33,34]. These findings may partly explain the predictive value of PNI for post-LT prognosis in HCC patients, but the exact underlying mechanism requires further research to fully elucidate.
To further improve prognostic accuracy, we established PIIN-based prediction models for OS and DFS by incorporating the PIIN score with statistically significant clinicopathological factors, achieving satisfactory predictive accuracy for 1-year, 3-year, and 5-year OS and DFS. DCA of models in training and validation cohort further showed a satisfactory net benefit when predicting post-LT outcomes, demonstrating their potential clinical utility. Currently, although the Milan and UCSF criteria remain the gold standards for LT recipient selection in HCC patients, their prognostic efficacy for post-LT prognosis is limited, especially in patients exceeding these criteria, since their prediction relies on preoperative morphological parameters without considering tumor biological behavior and systemic inflammation. In contrast, our PIIN-based model incorporates systemic immune-inflammation-nutrition markers, tumor markers, and histopathological features, and therefore may provide additional prognostic information beyond morphological criteria alone. Accordingly, we consider that the PIIN-based model may serve as a useful complement to the Milan and UCSF criteria, although further validation is still required to determine its incremental clinical value. In the future, it may serve as a potent tool for post-transplant risk stratification and clinical decision assistance for HCC patients. However, it should be noted that the proposed PIIN-based model has not yet been validated through prospective, multicenter studies, and its definitive clinical utility remains to be further elucidated.
In the present study, neither the Child-Pugh classification nor the model for end-stage liver disease (MELD) score showed a significant association with OS or DFS in the Cox regression analyses. This finding should be interpreted in the context of the different biological dimensions captured by these scores. Child-Pugh and MELD primarily reflect hepatic functional reserve and short-term liver-related mortality risk, whereas the PIIN score integrates systemic inflammation, immune status, and nutritional status. Therefore, the PIIN score may provide prognostic information that is complementary to conventional liver function-based scores, particularly for post-transplant oncological outcomes.
There are several limitations in this study that should be acknowledged. First, this is a retrospective study conducted at a single center with a relatively small cohort size. The restricted number of sample size and limited number of positive events in the validation cohort constrain the statistical precision of the C-index and may increase the risk of optimism. Although bootstrap resampling was used to verify the internal robustness of the model, future large-scale, multicenter prospective studies are necessary to confirm its generalizability and clinical value. Second, the extended patient enrollment period may have introduced chronological bias. Over this time span, continuous evolutions in patient perioperative care, bridging treatments, pathological evaluations, and post-LT surveillance might have confounded the observed associations. Third, owing to the retrospective design, several transplant-related variables were unavailable for the present analysis. Data on donor type and exact waiting times were not systematically documented and were therefore excluded from the final analysis. In addition, pre-transplant treatment was highly heterogeneous in timing, course, and regimen among the included patients, precluding its meaningful incorporation into the analysis. These unmeasured factors may have introduced potential bias. Finally, because the immune-inflammatory-nutritional status dynamically changes after LT, the PIIN score may fluctuate over the follow-up period. Further investigations integrating longitudinal biomarker measurements and molecular immune profiling are warranted to fully elucidate the clinical utility of the proposed PIIN and PIIN-based models.
CONCLUSION
In summary, the proposed PIIN score, which constitutes of NLR, PLR, SII, SIRI, and PNI, may serve as a potentially effective composite biomarker for survival outcomes prediction in HCC patients following LT. Integrating the PIIN score with other prognostic clinical and pathological data may help refine prognostic stratification for post-LT OS and DFS in HCC patients.
Shimamura T, Akamatsu N, Fujiyoshi M, Kawaguchi A, Morita S, Kawasaki S, Uemoto S, Kokudo N, Hasegawa K, Ohdan H, Egawa H, Furukawa H, Todo S; Japanese Liver Transplantation Society. Expanded living-donor liver transplantation criteria for patients with hepatocellular carcinoma based on the Japanese nationwide survey: the 5-5-500 rule - a retrospective study.Transpl Int. 2019;32:356-368.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 131][Cited by in RCA: 124][Article Influence: 17.7][Reference Citation Analysis (3)]
Tran BV, Moris D, Markovic D, Zaribafzadeh H, Henao R, Lai Q, Florman SS, Tabrizian P, Haydel B, Ruiz RM, Klintmalm GB, Lee DD, Taner CB, Hoteit M, Levine MH, Cillo U, Vitale A, Verna EC, Halazun KJ, Tevar AD, Humar A, Chapman WC, Vachharajani N, Aucejo F, Lerut J, Ciccarelli O, Nguyen MH, Melcher ML, Viveiros A, Schaefer B, Hoppe-Lotichius M, Mittler J, Nydam TL, Markmann JF, Rossi M, Mobley C, Ghobrial M, Langnas AN, Carney CA, Berumen J, Schnickel GT, Sudan DL, Hong JC, Rana A, Jones CM, Fishbein TM, Busuttil RW, Barbas AS, Agopian VG. Development and validation of a REcurrent Liver cAncer Prediction ScorE (RELAPSE) following liver transplantation in patients with hepatocellular carcinoma: Analysis of the US Multicenter HCC Transplant Consortium.Liver Transpl. 2023;29:683-697.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 43][Cited by in RCA: 43][Article Influence: 14.3][Reference Citation Analysis (4)]
Li H, Lin Z, Zhuo J, Yang M, Shen W, Hu Z, Ding Y, Chen H, He C, Yang X, Dong S, Wei X, Sun B, Zheng S, Lang R, Lu D, Xu X. TNFR2 is a potent prognostic biomarker for post-transplant lung metastasis in patients with hepatocellular carcinoma.Chin J Cancer Res. 2023;35:66-80.
[RCA] [PubMed] [DOI] [Full Text][Cited by in RCA: 16][Reference Citation Analysis (0)]
Lin ZY, Lu D, Wu SD, Hu ZH, Yang XD, Liu P, Mao SQ, Shan YY, Chen H, He CY, Li HG, Yang XY, Zhuo JY, Shen W, Wang K, Dong SY, Chen JY, Wei XY, Zhuang L, Cai JZ, Lu CD, Zheng SS, Xu X. Body composition predicts prognosis of patients with hepatocellular carcinoma after liver transplantation: A multicenter study.Hepatobiliary Pancreat Dis Int. 2025;S1499-3872(25)00181.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 1][Cited by in RCA: 2][Article Influence: 2.0][Reference Citation Analysis (1)]
