Zheng ZH, Xiao R, Liu N, He BY, Yu W, Wu ZS, Chen MS, Mei J, Hu DD, Guo RP. CRAPT-M: An effective prognostic predictor for hepatocellular carcinoma patients treated with locoregional-bevacizumab-immunotherapy. World J Gastroenterol 2026; 32(25): 119588 [DOI: 10.3748/wjg.119588]
Corresponding Author of This Article
Rong-Ping Guo, Department of Liver Surgery, Sun Yat-sen University Cancer Center, No. 651 Dongfeng Road East, Guangzhou 510060, Guangdong Province, China. guorp@sysucc.org.cn
Research Domain of This Article
Gastroenterology & Hepatology
Article-Type of This Article
research-article
Open-Access Policy of This Article
This article is an open-access article which was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Baishideng Publishing Group Inc, 7041 Koll Center Parkway, Suite 160, Pleasanton, CA 94566, USA
Share the Article
Zheng ZH, Xiao R, Liu N, He BY, Yu W, Wu ZS, Chen MS, Mei J, Hu DD, Guo RP. CRAPT-M: An effective prognostic predictor for hepatocellular carcinoma patients treated with locoregional-bevacizumab-immunotherapy. World J Gastroenterol 2026; 32(25): 119588 [DOI: 10.3748/wjg.119588]
Ze-Hao Zheng, Rui Xiao, Na Liu, Ben-Yi He, Wei Yu, Min-Shan Chen, Jie Mei, Dan-Dan Hu, Rong-Ping Guo, Department of Liver Surgery, Sun Yat-sen University Cancer Center, Guangzhou 510060, Guangdong Province, China
Zhong-Shi Wu, Department of General Surgery, Clifford Hospital, Jinan University, Guangzhou 511495, Guangdong Province, China
Rong-Ping Guo, State Key Laboratory of Oncology in South China, Sun Yat-sen University Cancer Center, Guangzhou 510060, Guangdong Province, China
Rong-Ping Guo, Collaborative Innovation Center for Cancer Medicine, Sun Yat-sen University Cancer Center, Guangzhou 510060, Guangdong Province, China
Co-corresponding authors: Dan-Dan Hu and Rong-Ping Guo.
Author contributions: Zheng ZH, Xiao R, and Liu N designed the study and drafted the manuscript; Zheng ZH, Xiao R, Liu N, and He BY participated in data collection and analysis; Zheng ZH and Xiao R made significant contributions to this study and should be considered co-first authors; Yu W helped draw the figures; Guo RP, Chen MS, and Hu DD provided financial support; Hu DD and Guo RP supervised the study; Hu DD and Guo RP provided significant support in academic decision-making and content guidance and they contribute equally to this study as co-corresponding authors; all the authors have read and approved the final version of the manuscript.
Supported by the National Science and Technology Major Project of China, No. 2024ZD0520402; and Science and Technology Planning Project of Guangzhou, No. 2023A03J0601.
Institutional review board statement: This study was approved by the Ethics Committee of Sun Yat-sen University Cancer Center (Approval No. B2025-237-01).
Informed consent statement: The informed consent was waived by the Institutional Review Board.
Conflict-of-interest statement: The authors have no conflicts of interest to declare.
Data sharing statement: The datasets of the current study are available from the corresponding author on reasonable request.
Corresponding author: Rong-Ping Guo, Department of Liver Surgery, Sun Yat-sen University Cancer Center, No. 651 Dongfeng Road East, Guangzhou 510060, Guangdong Province, China. guorp@sysucc.org.cn
Received: February 2, 2026 Revised: February 19, 2026 Accepted: March 13, 2026 Published online: July 7, 2026 Processing time: 150 Days and 23.7 Hours
Abstract
BACKGROUND
The CRAPT-M model, which comprises C-reactive protein, albumin, protein induced by vitamin K absence or antagonist-II, total bilirubin, and macrovascular invasion, is recognized to serve as a prognostic indicator in patients with hepatocellular carcinoma (HCC) receiving atezolizumab plus bevacizumab. However, its ability to predict survival in HCC patients receiving locoregional-bevacizumab-immunotherapy (L-Bev-IO) has not been fully elucidated.
AIM
To evaluate the performance of the CRAPT-M model in predicting the efficacy of first-line L-Bev-IO in HCC patients.
METHODS
This study enrolled HCC patients receiving L-Bev-IO between 2020 and 2023. According to the CRAPT-M model, patients were stratified into low (≤ 4), intermediate (5-12), or high (≥ 13) risk groups. The primary endpoint was defined as overall survival (OS), and secondary endpoints included progression-free survival (PFS) and tumor response rate.
RESULTS
Median OS was 13.4 months (95%CI: 11.3-16.8) in the high-risk group and was not reached in the low- and intermediate-risk groups. Median PFS was 17.5 months (95%CI: 11.3-not reached), 15.6 months (95%CI: 13.8-21.1), and 8.2 months (95%CI: 7.0-10.0) in the low-, intermediate-, and high-risk groups, respectively (P < 0.01). The C-index of the CRAPT-M model was 0.637 for OS and 0.593 for PFS. The Simple-CRAPT-M model ultimately retained effective risk stratification and strong predictive performance compared with the original CRAPT-M model.
CONCLUSION
The CRAPT-M model was an independent prognostic indicator in HCC patients receiving L-Bev-IO and demonstrated favorable predictive performance. In addition, the Simple-CRAPT-M model may facilitate risk stratification, enabling physicians to rapidly identify HCC patients likely to gain therapeutic benefit from L-Bev-IO.
Core Tip: This study validated the prognostic value of the CRAPT-M model in patients with hepatocellular carcinoma receiving locoregional-bevacizumab-immunotherapy. The model effectively stratified overall survival and progression-free survival not only in the locoregional plus atezolizumab plus bevacizumab group but also in the locoregional plus sintilimab plus bevacizumab group. In addition, we developed a Simple-CRAPT-M model based on the number of risk factors, which may provide a more convenient tool for rapid risk stratification in clinical practice.
Citation: Zheng ZH, Xiao R, Liu N, He BY, Yu W, Wu ZS, Chen MS, Mei J, Hu DD, Guo RP. CRAPT-M: An effective prognostic predictor for hepatocellular carcinoma patients treated with locoregional-bevacizumab-immunotherapy. World J Gastroenterol 2026; 32(25): 119588
Liver cancer is one of the most lethal and prevalent malignancies worldwide, representing a substantial global health burden[1]. Hepatocellular carcinoma (HCC), constituting over 70% of all primary liver malignancies, is the dominant histopathological subtype of liver cancer[2]. Due to its insidious onset and the lack of effective early screening, 60%-70% of HCC cases in China are diagnosed at advanced disease stage, for which curative therapeutic approaches are no longer applicable[3]. Over the past decade, immunotherapies have improved survival in patients with advanced HCC. In the IMbrave150 study, atezolizumab combined with bevacizumab delivered a significant improvement in both overall survival (OS) and progression-free survival (PFS) in comparison with sorafenib, leading to its adoption as a first-line systemic therapy[4]. Similarly, the ORIENT-32 study showed that bevacizumab combined with sintilimab significantly improved survival relative to sorafenib and this combination is recommended as a primary treatment option for HCC in China[5]. Nevertheless, neither the atezolizumab + bevacizumab regimen nor the sintilimab + bevacizumab regimen has achieved an objective response rate (ORR) exceeding 30%.
Transarterial chemoembolization (TACE) is a well-proven effective locoregional treatment modality for advanced HCC. In recent years, hepatic arterial infusion chemotherapy (HAIC), another transarterial therapy, has exhibited superior locoregional disease control in HCC patients with high tumor burden[6]. Several single-arm clinical studies have indicated that triple therapy combining HAIC or TACE with bevacizumab and atezolizumab or sintilimab can achieve considerable ORR and disease control rate (DCR) in HCC patients[7-9]. Moreover, a phase III trial demonstrated that adding HAIC to sorafenib improved response rates and prolonged survival in HCC with macrovascular invasion (MVI)[10]. Collectively, these findings suggest that combining locoregional and systemic therapies may provide greater survival benefit than systemic therapy alone. However, owing to tumor heterogeneity, the patient population most likely to benefit from transarterial therapy combined with systemic treatment remains uncertain.
Consequently, early identification of patients likely to benefit from these therapies is critical to advancing precision medicine. Nam et al’s study[11] developed and validated the CRAPT-M model—comprising C-reactive protein (CRP) ≥ 1.0 mg/dL, albumin (ALB) < 3.5 g/dL, protein induced by vitamin K absence or antagonist-II (PIVKA-II) ≥ 1500 mAU/mL, total bilirubin (TBIL) ≥ 1.0 mg/dL, and MVI—as a set of biomarkers to predict outcomes in patients receiving atezolizumab + bevacizumab therapy. However, it remains unclear whether the CRAPT-M model can effectively stratify prognosis in patients with HCC receiving transarterial therapies combined with atezolizumab/sintilimab and bevacizumab [locoregional-bevacizumab-immunotherapy (L-Bev-IO)].
This study aimed to evaluate and validate the prognostic value of the CRAPT-M model in HCC patients receiving L-Bev-IO. In addition, we refined the model using real-world survival data and clinical considerations to provide evidence to guide therapeutic decision-making in this population.
MATERIALS AND METHODS
Study design and participants
From 2020 to 2023, we retrospectively reviewed 443 HCC patients receiving L-Bev-IO as initial treatment. The study was approved by the Ethics Committee of Sun Yat-sen University Cancer Center (Approval No. B2025-237-01) and was conducted in full compliance with the Declaration of Helsinki and the STROCSS criteria[12].
Inclusion criteria: (1) HCC confirmed by imaging examinations or pathological biopsy; (2) Initial treatment with HAIC or TACE; (3) Bevacizumab-immunotherapy administered either concurrently with HAIC or TACE or after achieving disease control with HAIC or TACE; (4) Eastern Cooperative Oncology Group performance status ≤ 2; (5) At least 1 month of follow-up; (6) Complete baseline data; and (7) Child-Pugh class A or B liver function.
Exclusion criteria: (1) Child-Pugh C; (2) History of other malignancies; and (3) Bevacizumab-immunotherapy initiated after progression following HAIC or TACE.
Definition of CRAPT-M and CRAFITY
The CRAPT-M and the CRAFITY scores were defined according to Nam et al[11] and Scheiner et al[13], respectively. The CRAPT-M model comprises five clinical risk factors: CRP, ALB, TBIL, PIVKA-II, and MVI. The cutoff values for the CRAPT-M score are listed in Tables 1 and 2.
Table 1 The risk factors and corresponding scores.
The treatment strategy was formulated by experienced experts following discussion by the multidisciplinary tumor team. All the patients in the present study signed the informed consent before treatment. Patients received intravenous injections of 1200 mg atezolizumab[4] or 200 mg sintilimab[5], alongside 15 mg/kg bevacizumab, once every 3 weeks. The systemic therapy was administered until treatment abandonment, unacceptable toxicity, disease progression, or death. The procedures of HAIC[10] and TACE[14] followed our previously published protocols. Briefly, TACE was conducted using a mixture of 50 mg epirubicin and 50 mg lobaplatin with lipiodol, whereas HAIC was based on the mFOLFOX regimen (oxaliplatin 85 mg/m2 and leucovorin 400 mg/m2 on day 1, plus fluorouracil 400 mg/m2 bolus followed by 2400 mg/m2 as a 24-hour infusion).
Data collection and study outcomes
The laboratory blood test data and clinical data before the initiation of L-Bev-IO were retrieved from medical system of our hospital. OS, the primary endpoint, was defined as the interval from bevacizumab-immunotherapy (Bev-IO) treatment initiation to cancer-related death. PFS, the secondary endpoint, was defined as the time from Bev-IO treatment initiation until disease progression, cancer-related death. Initial tumor status and therapeutic response assessment were performed independently by two independent radiologists, using contrast-enhanced magnetic resonance imaging or computed tomography, with evaluations conducted in strict accordance with both the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 and the modified RECIST (mRECIST) criteria[15].
Statistical analysis
Variables were categorized as categorical or continuous based on their distribution and clinical relevance. The classification criteria were as follows. Categorical variables: (1) Demographic characteristics: Sex (male/female) and age (< 65 years vs ≥ 65 years); (2) Clinical and tumor-related characteristics: HBV infection (yes/no); tumor size (< 7 cm vs ≥ 7 cm); tumor number (single vs multiple); extrahepatic metastasis (yes/no); Barcelona Clinic Liver Cancer (BCLC) stage (A/B/C); and Child-Pugh stage (A or B); and (3) Treatment-related variables: Transarterial therapy type (HAIC vs TACE) and systemic therapy type (atezolizumab + bevacizumab vs sintilimab + bevacizumab). Categorical variables were summarized as n (%). The Pearson χ2 test or Fisher’s exact test was used for inter-group comparisons. Continuous variables were presented by median (interquartile range) and analyzed using Student’s t test when the variables distribute normally, and a nonparametric test when they were not. Survival analysis was conducted using the Kaplan-Meier method, and the survival benefit was compared via Log-rank test. The R package compareC was used to compare the C-indices among the three models, and the timeROC package was applied to compare the time-dependent area under the curves (AUCs). All statistics analyses were conducted in R software (v 4.5.1) and statistically significant was set as P < 0.05.
RESULTS
Basic characteristics
We finally enrolled 443 HCC patients who received L-Bev-IO at our hospital (Figure 1). Among them, 375 (94.7%) patients were < 65 years of age, 396 (89.4%) were male, and 416 (91.6%) had hepatitis B virus infection. At diagnosis, 311 patients (70.2%) had a maximum tumor diameter > 7.0 cm, and 306 (69.1%) had multiple lesions. Moreover, 34 patients (7.7%) were classified as BCLC A, 95 (21.4%) as BCLC B, and 314 (70.9%) as BCLC C. Most patients (n = 384, 86.7%) had Child-Pugh class A liver function. HAIC was the predominant transarterial therapy (n = 378, 85.3%). For systemic therapy, 194 patients (43.8%) received atezolizumab + bevacizumab and 249 (56.2%) received sintilimab + bevacizumab. Based on the CRAPT-M model, 58, 162, and 223 patients were stratified into the low-, intermediate-, and high-risk groups, respectively. As shown in Table 3, the CRAPT-M risk group was associated with sex, tumor size, extrahepatic metastasis, and Child-Pugh class.
The median follow-up time was 26.3 months (95%CI: 25.4-27.5) in the entire cohort. Median OS was 21.8 months (95%CI: 19.1-31.3), and median PFS was 11.7 months (95%CI: 10.0-13.8). During follow-up, 220 patients died and 285 experienced tumor progression.
The CRAPT-M model comprises CRP, ALB, TBIL, PIVKA-II, and MVI. We found that these five factors were significantly associated with OS (Supplementary Figure 1). For PFS, a similar pattern was observed for most factors, except MVI. We then compared survival across CRAPT-M risk groups. Patients in the CRAPT-M high-risk group had the worst OS (P < 0.001; Figure 2A) and PFS (P < 0.001; Figure 2B). Median OS was not reached in the low- and intermediate-risk groups, whereas it was 13.4 months (95%CI: 11.3-16.8) in the high-risk group. Median PFS was 17.5 months (95%CI: 11.3-not reached) in the low-risk group, 15.6 months (95%CI: 13.8-21.1) in the intermediate-risk group, and 8.2 months (95%CI: 7.0-10.0) in the high-risk group.
Figure 2 Kaplan-Meier curves for overall survival and progression-free survival of patients treated with locoregional-bevacizumab-immunotherapy.
A-D: Overall survival (A and C) and progression-free survival (B and D) in patients stratified into different groups according to CRAPT-M model (A and B) and CRAFITY model (C and D); E: Sankey diagram showing clinical outcomes stratified by the CRAPT-M model.
Consistent with Nam et al[11], we also evaluated the CRAFITY score and found that stratification by this score showed significant discriminatory ability for OS and PFS. Median OS was not reached in patients with CRAFITY scores of 0 or 1, whereas it was 16.8 months (95%CI: 11.9-19.0) in those with a score of 2 (Figure 2C). Median PFS was 15.6 months (95%CI: 12.8-not reached), 13.7 months (95%CI: 11.5-16.0), and 8.4 months (95%CI: 7.0-10.7) for CRAFITY scores of 0, 1, and 2, respectively (Figure 2D). To assess the robustness of these findings, we conducted landmark analyses at 6 months and 12 months. In the 6-month landmark analysis, between-group differences in OS (Supplementary Figure 2A and B) and PFS (Supplementary Figure 3A and B) for the CRAPT-M and CRAFITY subgroups remained consistent with those observed in the overall cohort. In the 12-month landmark analysis, a separation in OS across CRAPT-M subgroups was still evident (Supplementary Figure 2C), although the differences did not reach statistical significance. Similarly, there was no statistically significant difference in OS between groups with different CRAFITY scores in the 12-month landmark analysis (Supplementary Figure 2D). Additionally, no statistical significance was shown in PFS among different CRAPT-M and CRAFITY subgroups in the 12-month landmark analysis (Supplementary Figure 3C and D). Consistent with these survival results, the Sankey diagram showed markedly divergent clinical outcomes across CRAPT-M risk strata: Most patients in the low- and intermediate-risk groups remained alive, whereas the majority of patients in the high-risk group died during follow-up (Figure 2E).
We further stratified HCC patients into an atezolizumab + bevacizumab group (n = 194) and a sintilimab + bevacizumab group (n = 249). Atezolizumab + bevacizumab group and sintilimab + bevacizumab group had similar OS (Supplementary Figure 1F) and PFS (Supplementary Figure 1L). The CRAPT-M model effectively stratified both OS and PFS in each treatment group (P < 0.05; Supplementary Figure 4A and B). Similarly, the CRAFITY score significantly discriminated OS and PFS in the sintilimab + bevacizumab group (P < 0.05; Supplementary Figure 4C and D). In contrast, in the atezolizumab + bevacizumab group, the CRAFITY score was not significantly associated with OS or PFS (P = 0.11; Supplementary Figure 4C and D).
Notably, substantial overlap in both OS and PFS was observed between the low- and intermediate-risk groups defined by the CRAPT-M model in both the atezolizumab + bevacizumab and sintilimab + bevacizumab cohorts, suggesting limited utility for distinguishing these two subgroups among patients receiving L-Bev-IO.
Treatment efficacy and the CRAPT-M model
Because the survival curves were similar between the CRAPT-M low- and intermediate-risk groups, we compared treatment response by mRECIST between the high-risk group and the combined low/intermediate-risk group. No statistically significant difference in the distribution of best overall tumor response between the CRAPT-M low-/intermediate-risk and high-risk groups, whether evaluated by RECIST 1.1 or mRECIST criteria (Table 4). For HCC patients in the overall cohort, the modified ORR (mORR) was 58.0%, and the corresponding DCR reached 85.8% in the final efficacy analysis. The mORR was 52.9% in the high-risk group and 63.2% in the low/intermediate-risk group (P = 0.036). The DCR was 82.1% in the high-risk group and 89.5% in the low/intermediate-risk group (P = 0.034).
To identify independent prognostic factors for survival, we constructed multivariable Cox regression models that included the CRAPT-M risk group, the CRAFITY score, and other clinical covariates. We found that the CRAPT-M high-risk group—but not the CRAFITY score—was an independent biomarker for both OS (HR = 2.52; 95%CI: 1.34-4.72; P < 0.01) and PFS (HR = 1.87; 95%CI: 1.13-3.09; P = 0.01; Table 5).
Table 5 Multivariate Cox regression analyses of the prognostic factors for overall survival and progression-free survival.
To compare prognostic performance of the CRAPT-M model and the CRAFITY score, we calculated C-indices and time-dependent ROC curves. After bootstrap optimism correction, the calibrated C-indices for CRAPT-M were 0.637 for OS and 0.593 for PFS, while those for CRAFITY were 0.601 for OS and 0.568 for PFS. The C-indices between the 2 models were compared, and the results demonstrated that the C-index of CRAPT-M for predicting OS was superior to that of CRAFITY (P = 0.045). The time-dependent ROC curves indicated that CRAPT-M had a better capacity to estimate OS than CRAFITY at 12 and 24 months (Figure 3A, P = 0.007; Figure 3B, P = 0.045). Notably, the CRAPT-M model exhibited a statistically significant advantage over CRAFITY in predicting 6-month PFS (Figure 3C, P = 0.01). However, the difference between the two models for 12-month PFS did not reach statistical significance (Figure 3D, P = 0.15). As shown in Figure 3E and F, this trend persisted throughout follow-up for both OS and PFS.
Figure 3 Predictive performance of the CRAPT-M, Simple-CRAPT-M and CRAFITY model in patients treated with locoregional-bevacizumab-immunotherapy.
A and B: Receiver operating characteristic (ROC) curves of the CRAPT-M, Simple-CRAPT-M and CRAFITY model for overall survival (OS) at 12 months (A) and 24 months (B); C and D: ROC curves of the CRAPT-M, Simple-CRAPT-M and CRAFITY model for progression-free survival (PFS) at 6 months (C) and 12 months (D); E and F: Time-dependent ROC analysis of the CRAPT-M, Simple-CRAPT-M, CRAFITY model and other factors for OS (E) and PFS (F) throughout follow-up. The X-axis represents the follow-up time (months) from the initiation of locoregional-bevacizumab-immunotherapy therapy. ROC: Receiver operating characteristic; OS: Overall survival; PFS: Progression-free survival; AUC: Area under the curve; AUROC: Area under the receiver operating characteristic; CRP: C-reactive protein; TBIL: Total bilirubin; ALB: Albumin; PIVKA-II: Protein induced by vitamin K absence or antagonist-II; AFP: Alpha fetoprotein.
In addition, we performed calibration analyses and decision curve analysis (DCA). For calibration, CRAPT-M (Supplementary Figure 5) showed predictive accuracy for 12- and 24-month OS and 6- and 12-month PFS that was comparable to that of the CRAFITY score (Supplementary Figure 6). However, DCA showed that CRAPT-M provided better net clinical benefit (Supplementary Figure 7), suggesting superior utility for informing clinical decision-making.
Exploratory evaluation of Simple-CRAPT-M
Based on the results described above and to facilitate clinical application, we performed an exploratory revision of the CRAPT-M risk stratification. In the original definition, the intermediate-risk group (score 5-12) encompassed patients with up to two risk factors. In our revised classification, patients with 0-2 risk factors were regarded as low risk, whereas those with 3-5 risk factors were stratified into high-risk group. This revised Simple-CRAPT-M model effectively stratified patients by both OS and PFS (P < 0.05; Figure 4A and B). In the 6-month landmark analysis, the Simple-CRAPT-M subgroups maintained significant between-group differences in OS (Supplementary Figure 2E) and PFS (Supplementary Figure 3E), consistent with the results in the primary cohort. In the 12-month landmark analysis, separation in OS and PFS across the Simple-CRAPT-M risk strata remained apparent, although the between-group differences did not reach statistical significance (Supplementary Figures 2F and 3F). Notably, the Simple-CRAPT-M model also effectively stratified OS and PFS in both the atezolizumab + bevacizumab and sintilimab + bevacizumab subgroups, see Figure 4C and D. After bootstrap optimism correction, the calibrated C-indices for the Simple-CRAPT-M model were 0.631 for OS and 0.594 for PFS, which were noninferior to those of the original CRAPT-M model (both P > 0.05). Time-dependent ROC analysis further showed that the AUCs for 12- and 24-month OS (0.679 and 0.659, respectively) and 6- and 12-month PFS (0.637 and 0.627, respectively) were comparable to those of the original model (Figure 3E and F; both P > 0.05). In addition, calibration plots (Supplementary Figure 8) and DCA (Supplementary Figure 7) indicated similar calibration and net clinical benefit for the Simple-CRAPT-M model compared with the original CRAPT-M model.
Figure 4 Kaplan-Meier curves for overall survival and progression-free survival of patients treated with locoregional-bevacizumab-immunotherapy.
A and B: Overall survival (OS; A) and progression-free survival (PFS; B) in patients stratified into different groups according to the Simple-CRAPT-M model; C and D: Subgroup analysis showing OS (C) and PFS (D) in patients receiving different bevacizumab-based immune therapies and stratified into different groups according to the Simple-CRAPT-M model.
DISCUSSION
It is well established that atezolizumab + bevacizumab or sintilimab + bevacizumab provides superior survival benefit in comparison with sorafenib in HCC patients[4,5]. Notably, locoregional therapy and immunotherapy may have synergistic effects, potentially enhancing clinical benefit[16]. However, substantial tumor heterogeneity and suboptimal treatment efficacy highlight the need to identify, early in the treatment course, the subgroup of HCC patients most likely to gain favorable therapeutic outcomes from these regimens. Nam et al[11] developed a novel risk prediction model composed of five clinical factors that effectively stratified survival in HCC patients treated with atezolizumab + bevacizumab. Accordingly, we assessed the prognostic performance of the CRAPT-M model in 443 HCC patients receiving L-Bev-IO in the present study. We found that the CRAPT-M model effectively predicted prognosis in this population. In the 6-month landmark analysis of OS and PFS, the CRAPT-M model maintained discriminative ability. In multivariable analysis, the CRAPT-M high-risk group was an independent risk prognostic indicator for both OS and PFS. Furthermore, patients classified as low/intermediate risk by the CRAPT-M model had higher ORR and DCR than those with high-risk. Overall, CRAPT-M model demonstrated discriminative ability comparable to that of the CRAFITY score and other clinical indicators, with satisfactory calibration and greater net clinical benefit for predicting OS and PFS. We also performed an exploratory analysis in which we recategorized the CRAPT-M model into a Simple-CRAPT-M model based on the number of risk factors. The simplified model retained effective risk stratification and predictive performance comparable to the original CRAPT-M.
The CRAPT-M model showed prognostic value in patients receiving L-Bev-IO, likely because it captures three key domains relevant to treatment outcomes: Systemic inflammation, hepatic functional reserve, and tumor burden. CRP, an acute-phase protein, reflects the systemic inflammatory status. Elevated CRP levels are associated with inferior survival following immune checkpoints inhibitors across multiple cancer types[17]. Moreover, CRP is also a biomarker of HCC patients treated with atezolizumab + bevacizumab[13,18] or durvalumab + tremelimumab[19]. Elevated CRP levels may impair CD8+ T-cell responses[20] and are significantly associated with increased infiltration of immunosuppressive cells[21,22]. Beyond reflecting hepatic synthetic function, ALB is also an indicator of overall nutritional status. Hypoalbuminemia has been consistently related with poor survivals in HCC patients[23]. Jiang et al[24] revealed that HCC patients with hypoalbuminemia or malnutrition responded poorly to immunotherapy and had inferior survival. Similarly, Nakanishi et al[25] found among non-small cell lung cancer patients treated with PD-1 inhibitors, those with a low ALB-to-globulin ratio, had significantly worse OS and PFS. Mechanistically, impaired hepatic reserve may reduce tolerance to intensive multimodal combination therapy, while malnutrition-associated immune dysfunction and chronic inflammation may further compromise outcomes[26]. Bilirubin, a core indicator of hepatic excretory function, directly affects treatment safety and feasibility. A meta-analysis by Mishra et al[27] reported that low ALB and high bilirubin levels predict poorer outcomes after TACE. Similarly, our previous study validated the prognostic utility of ALB and bilirubin in HCC patients receiving locoregional immunotherapy[28]. In the present study, patients received L-Bev-IO, an intensive regimen with substantial hepatic burden. Patients with hypoalbuminemia and elevated bilirubin often have poorer overall condition and limited hepatic functional reserve, which may predispose them to treatment-related liver injury and severe adverse events. These complications can lead to treatment discontinuation and, consequently, unfavorable prognosis. PIVKA-II reflects not only tumor burden but also aggressive tumor biology. Serum PIVKA-II is a useful indicator for monitoring treatment response in patients receiving transarterial therapies or immunotherapy[29-31]. Beyond its role in response assessment, PIVKA-II may participate in tumor angiogenesis and could reduce the effectiveness of VEGF-targeted agents[32]. MVI is another critical prognostic factor in HCC. Evidence indicates that MVI is associated with inferior outcomes not only after transarterial therapies such as HAIC[33,34] or TACE[35], but also with bevacizumab plus immunotherapy[36,37]. Collectively, these factors help explain why the CRAPT-M model has prognostic value in HCC patients receiving L-Bev-IO.
Some findings in the present study differed from those reported by Nam et al[11]. First, we observed longer median OS, even in patients with BCLC stage C disease. Second, the CRAPT-M model showed limited discrimination for PFS between the low- and intermediate-risk groups in the overall cohort and in treatment-specific subgroups (atezolizumab or sintilimab). The observed differences may be partly attributable to the incorporation of transarterial therapies, which are commonly used for HCC in China and Japan. Transarterial therapy enables localized drug delivery to tumor sites, thereby facilitating local control and tumor shrinkage, which have been associated with improved outcomes. Evidence also suggests that reducing tumor burden may enhance the efficacy of immunotherapy. In addition, transarterial therapies have been reported to favorably modulate the tumor immune microenvironment. HAIC could activate the T and B cells in a functional state and promote the formation of tertiary lymphoid structures[38]. Meanwhile, TACE could effectively induce immunogenic cell death[39]. Together, these effects may increase responsiveness to immune checkpoint inhibitors. Notably, a phase III trial from our center showed that HAIC combined with sorafenib provided a significant OS advantage over sorafenib alone in patients with HCC and MVI[10]. Consequently, our findings suggest that harboring up to two risk factors in the CRAPT-M model did not adversely impact outcomes for patients undergoing L-Bev-IO. We therefore simplified the CRAPTM model by merging the original low- and intermediate-risk groups. Specifically, patients with two or fewer risk factors were classified as low risk, enabling rapid risk stratification by simply counting risk factors and thereby facilitating clinical decision-making. Importantly, the simplified model was noninferior to the original model in terms of stratification ability and predictive performance.
Several limitations should be acknowledged in the present study. To begin with, the single-center retrospective design carries a potential risk of selection bias. We attempted to mitigate this risk by consecutively enrolling all eligible patients receiving L-Bev-IO through systematic screening of electronic medical records. Second, because 91.6% of the cohort had HBV-related HCC and all patients were Chinese, the generalizability of our findings may be limited in non-Chinese populations or in patients with other etiologies (e.g., nonalcoholic fatty liver disease or alcohol-associated liver disease). Third, all patients in this study received bevacizumab as the antiangiogenic agent. In contrast, lenvatinib is more widely used as a first-line antiangiogenic therapy in East Asia, including China. Finally, it is important to emphasize that the Simple-CRAPT-M model is an exploratory simplification derived from observations in a single cohort, and its generalizability requires independent external validation. Therefore, large-scale, multicenter, prospective studies in diverse populations with varying etiologies and treatment regimens are needed to further confirmed these results.
CONCLUSION
This present study validated the CRAPT-M model as an independent prognostic indicator in HCC patients receiving L-Bev-IO. In this cohort, it demonstrated favorable predictive performance compared with the CRAFITY score and other conventional clinical indicators. In addition, we proposed the Simple-CRAPT-M model, a simplified risk stratification tool based solely on the number of risk factors, which could help clinicians more rapidly identify HCC patients who are more likely to gain therapeutic benefit from L-Bev-IO. However, given the lack of external validation and the predominance of HBV-related HCC in our study population, the generalizability of these results to other HCC etiologies and to alternative targeted-immunotherapy regimens warrants further investigation.
ACKNOWLEDGEMENTS
We thank all those who provided excellent technical support and assistance during the study.
Finn RS, Qin S, Ikeda M, Galle PR, Ducreux M, Kim TY, Kudo M, Breder V, Merle P, Kaseb AO, Li D, Verret W, Xu DZ, Hernandez S, Liu J, Huang C, Mulla S, Wang Y, Lim HY, Zhu AX, Cheng AL; IMbrave150 Investigators. Atezolizumab plus Bevacizumab in Unresectable Hepatocellular Carcinoma.N Engl J Med. 2020;382:1894-1905.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 6170][Cited by in RCA: 5669][Article Influence: 944.8][Reference Citation Analysis (8)]
Ren Z, Xu J, Bai Y, Xu A, Cang S, Du C, Li Q, Lu Y, Chen Y, Guo Y, Chen Z, Liu B, Jia W, Wu J, Wang J, Shao G, Zhang B, Shan Y, Meng Z, Wu J, Gu S, Yang W, Liu C, Shi X, Gao Z, Yin T, Cui J, Huang M, Xing B, Mao Y, Teng G, Qin Y, Wang J, Xia F, Yin G, Yang Y, Chen M, Wang Y, Zhou H, Fan J; ORIENT-32 study group. Sintilimab plus a bevacizumab biosimilar (IBI305) versus sorafenib in unresectable hepatocellular carcinoma (ORIENT-32): a randomised, open-label, phase 2-3 study.Lancet Oncol. 2021;22:977-990.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 918][Cited by in RCA: 850][Article Influence: 170.0][Reference Citation Analysis (4)]
Liu D, Mu H, Liu C, Zhang W, Cui Y, Wu Q, Zhu X, Fang F, Zhang W, Xing W, Li Q, Song T, Lu W, Li H. Sintilimab, bevacizumab biosimilar, and HAIC for unresectable hepatocellular carcinoma conversion therapy: a prospective, single-arm phase II trial.Neoplasma. 2023;70:811-818.
[RCA] [PubMed] [DOI] [Full Text][Cited by in RCA: 19][Reference Citation Analysis (0)]
He M, Li Q, Zou R, Shen J, Fang W, Tan G, Zhou Y, Wu X, Xu L, Wei W, Le Y, Zhou Z, Zhao M, Guo Y, Guo R, Chen M, Shi M. Sorafenib Plus Hepatic Arterial Infusion of Oxaliplatin, Fluorouracil, and Leucovorin vs Sorafenib Alone for Hepatocellular Carcinoma With Portal Vein Invasion: A Randomized Clinical Trial.JAMA Oncol. 2019;5:953-960.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 465][Cited by in RCA: 444][Article Influence: 63.4][Reference Citation Analysis (6)]
Nam H, Kim DY, Kim DY, Kim JH, Kim CW, Lee J, Yang K, Han JW, Sung PS, Yoon SK, Cho HS, Yang H, Bae SH, Lee SK, Kwon JH, Nam SW, Lee A, Song DS, Chang UI, Kim SH, Song MJ, Lee HL, Kim HY, Lee SW, Jang JW. Development and validation of a risk prediction model for patients with hepatocellular carcinoma receiving atezolizumab-bevacizumab.Hepatology. 2026;83:1098-1110.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 1][Cited by in RCA: 3][Article Influence: 3.0][Reference Citation Analysis (0)]
Scheiner B, Pomej K, Kirstein MM, Hucke F, Finkelmeier F, Waidmann O, Himmelsbach V, Schulze K, von Felden J, Fründt TW, Stadler M, Heinzl H, Shmanko K, Spahn S, Radu P, Siebenhüner AR, Mertens JC, Rahbari NN, Kütting F, Waldschmidt DT, Ebert MP, Teufel A, De Dosso S, Pinato DJ, Pressiani T, Meischl T, Balcar L, Müller C, Mandorfer M, Reiberger T, Trauner M, Personeni N, Rimassa L, Bitzer M, Trojan J, Weinmann A, Wege H, Dufour JF, Peck-Radosavljevic M, Vogel A, Pinter M. Prognosis of patients with hepatocellular carcinoma treated with immunotherapy - development and validation of the CRAFITY score.J Hepatol. 2022;76:353-363.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 247][Cited by in RCA: 241][Article Influence: 60.3][Reference Citation Analysis (5)]
Kuwano A, Yada M, Tanaka K, Takahira J, Suzuki H, Ohishi Y, Motomura K. Tumor-infiltrating CD8(+) T-cell numbers and serum C-reactive protein levels as a prognostic biomarker in hepatocellular carcinoma patients receiving atezolizumab plus bevacizumab.Hepatol Res. 2025;55:600-610.
[RCA] [PubMed] [DOI] [Full Text][Cited by in RCA: 1][Reference Citation Analysis (0)]
Hatanaka T, Yata Y, Hiraoka A, Tada T, Hirooka M, Kariyama K, Tani J, Atsukawa M, Takaguchi K, Itobayashi E, Kakizaki S, Fukunishi S, Tsuji K, Ishikawa T, Tajiri K, Toyoda H, Koshiyama Y, Ogawa C, Nishikawa H, Nishimura T, Kawata K, Kosaka H, Matsui K, Naganuma A, Tanaka H, Ohama H, Kuroda H, Matono T, Aoki T, Ochi H, Imai M, Nakamura S, Kanayama Y, Tanaka K, Tada F, Yoshida O, Nouso K, Morishita A, Tsutsui A, Nagano T, Itokawa N, Okubo T, Arai T, Enomoto H, Kaibori M, Hiasa Y, Kudo M, Kumada T; Real‐life Practice Experts for HCC (RELPEC) Study Group, HCC 48 Group (hepatocellular carcinoma experts from 48 clinics in Japan). Clinical Impact of Serum CRP Levels on Advanced HCC Treated With Durvalumab and Tremelimumab: A Multicentre Study.Liver Int. 2025;45:e70192.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 2][Cited by in RCA: 6][Article Influence: 6.0][Reference Citation Analysis (0)]
Nakanishi Y, Masuda T, Yamaguchi K, Sakamoto S, Horimasu Y, Mimae T, Nakashima T, Miyamoto S, Tsutani Y, Iwamoto H, Fujitaka K, Miyata Y, Hamada H, Okada M, Hattori N. Albumin-globulin ratio is a predictive biomarker of antitumor effect of anti-PD-1 antibody in patients with non-small cell lung cancer.Int J Clin Oncol. 2020;25:74-81.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 12][Cited by in RCA: 27][Article Influence: 3.9][Reference Citation Analysis (1)]
Mishra G, Majeed A, Dev A, Eslick GD, Pinato DJ, Izumoto H, Hiraoka A, Huo TI, Liu PH, Johnson PJ, Roberts SK. Clinical Utility of Albumin Bilirubin Grade as a Prognostic Marker in Patients with Hepatocellular Carcinoma Undergoing Transarterial Chemoembolization: a Systematic Review and Meta-analysis.J Gastrointest Cancer. 2023;54:420-432.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 12][Cited by in RCA: 21][Article Influence: 7.0][Reference Citation Analysis (1)]
Pinato DJ, Murray SM, Forner A, Kaneko T, Fessas P, Toniutto P, Mínguez B, Cacciato V, Avellini C, Diaz A, Boyton RJ, Altmann DM, Goldin RD, Akarca AU, Marafioti T, Mauri FA, Casagrande E, Grillo F, Giannini E, Bhoori S, Mazzaferro V. Trans-arterial chemoembolization as a loco-regional inducer of immunogenic cell death in hepatocellular carcinoma: implications for immunotherapy.J Immunother Cancer. 2021;9:e003311.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 166][Cited by in RCA: 170][Article Influence: 34.0][Reference Citation Analysis (9)]
Footnotes
Peer review: Externally peer reviewed.
Peer-review model: Single blind
Specialty type: Gastroenterology and hepatology
Country of origin: China
Peer-review report’s classification
Scientific quality: Grade B, Grade B, Grade B, Grade C
Novelty: Grade B, Grade B, Grade B, Grade B
Creativity or innovation: Grade B, Grade C, Grade C, Grade C
Scientific significance: Grade B, Grade C, Grade C, Grade C
P-Reviewer: Malmir I, PhD, Post Doctoral Researcher, United States; Wang CL, MD, PhD, China S-Editor: Lin C L-Editor: A P-Editor: Zheng XM
