Prognosis of Patients with Hepatocellular Carcinoma Treated with Transarterial Chemoembolization(MC-hccAI 001): Development and Validation of the ALFP Score

Introduction

Hepatocellular carcinoma (HCC) is the predominant form of liver cancer and poses a significant global health burden. It ranks as the seventh most common cancer and the second leading cause of cancer-related deaths worldwide.^1,2 While surgical resection and liver transplantation offer potential curative options.³ HCC is often diagnosed at advanced stages, such as BCLC stage B or advanced BCLC stage.^4,5 According to the Barcelona Clinic Liver Cancer (BCLC) staging system, transarterial chemoembolization (TACE) serves as the primary treatment for patients with intermediate-stage HCC, characterized by asymptomatic, large, or multifocal tumors without macrovascular invasion or extrahepatic metastasis.⁶ However, the heterogeneity within the intermediate stage, including variations in liver function and tumor characteristics, introduces challenges in predicting patient outcomes and response to TACE,^7,8 Consequently, the median survival time for patients treated with TACE has been reported to range from 3.4 to 54.2 months.^8–10 It is important to note that not all patients with intermediate-stage HCC experience similar benefits from TACE. Therefore, predicting the survival of HCC patients undergoing TACE is crucial for making informed clinical decisions but remains a formidable challenge at present.

The liver plays a crucial role in synthesizing most coagulation factors, including factors VII, X, V, and II.¹¹ Prothrombin Time (PT) is a commonly used test that assesses the tissue factor (TF) and coagulation pathways involving factors VII, X, V, II, and fibrinogen.¹² Abnormalities in coagulation factors are observed in patients with HCC, such as Prothrombin induced by vitamin K absence or antagonist-II (PIVKA-II).¹³ PIVKA-II leads to prolonged Prothrombin Time and an increased risk of bleeding. Its diagnostic value in HCC patients undergoing surgery and curative ablation has been confirmed,^13,14 although its clinical utilization remains limited. On the other hand, PT is a common and easily accessible test, however, its role in assessing the prognosis of HCC patients has not been thoroughly investigated. Some studies indicate that PT may serve as a reliable predictor of patient prognosis in the context of surgical treatment.^15–17 Furthermore, PT has been utilized to develop new scoring systems for evaluating liver function and patient survival.^18,19 In conclusion, PT shows promise as a prognostic factor for HCC. Since most HCC patients are diagnosed at advanced stages and often undergo TACE or systemic therapy according to guidelines,⁶ this patient population requires an effective and straightforward clinical scoring system. Based on the findings of our study, we have developed a novel scoring system called the albumin-bilirubin grade, α-fetoprotein, and prothrombin time (ALFP) score. By combining PT with other established prognostic markers such as α-fetoprotein (AFP) and the albumin-bilirubin grade (ALBI), we aim to enhance the accuracy and reliability of prognostic assessment in HCC patients. This comprehensive approach takes into account the liver function, tumor markers, and coagulation status, providing a more comprehensive evaluation of patient prognosis.

Methods and Patients

Patient Selection

In this retrospective study, we reviewed a cohort of 2020 consecutive HCC patients diagnosed at Sun Yat-sen University Cancer Center (SYSUCC) between January 2007 and May 2012.^20–23 In this study, no formal sample capacity calculation was performed. All data available from patients were considered for this study. 2020 patients with histologically or radiologically diagnosed HCC who received transarterial chemoembolization (TACE) were brought into this study. Among these 2020 patients, we excluded 215 patients whose baseline AFP, PT, and Aspartate transaminase (AST) data were missing. As a result, the remaining 1805 patients were included in this study, and we randomly divided patients 7:3 into a training set (n = 1264) and a validation set (n = 541) (Figure 1).

Figure 1 Flow diagram of the patients with hepatocellular carcinoma. Abbreviations: TACE, transcatheter arterial chemoembolization; HCC, hepatocellular carcinoma; AFP, alpha-fetoprotein; PT, Prothrombin Time; AST, aspartate transaminase.

We collected the baseline characteristics of the patients, including gender, age, diameter of the main tumor, biochemical parameters, coagulation profile, and liver function. The liver function was defined using the albumin–bilirubin (ALBI) grade.²⁴ These two cohorts represent patient populations treated with a similar therapeutic strategy, making them comparable in terms of patient and tumor characteristics.

Statement of Ethics

The study protocol (2017-FXY-129) was approved by the Department of Clinical Research at Sun Yat-sen University Cancer Center. Additionally, the Hospital Ethics Committee at the four medical centers granted approval for this study, which exempted the requirement for written informed consent due to its retrospective nature.

Statistical Analysis Methods

In this retrospective study, our cohort was randomly divided into two groups to ensure the accuracy and practicality of the score. We used the chi-squared or Fisher’s exact test to compare categorical data and the Mann–Whitney U-test to compare continuous variables. The prognostic factors were investigated using the Cox proportional hazards regression model. Variables with a p-value < 0.05 in the univariable analysis were considered for multivariable analysis. Variables with a p-value < 0.05 in the multivariable analysis were identified as prognostic factors. Based on the results of the multivariate analysis in the training set, we established a new and simple score. To validate the score and avoid overoptimistic results, we applied it to the validation set. Overall survival (OS) was calculated from the start of TACE until the date of death from any cause, while patients who were still alive were censored at the date of last contact. Survival curves were generated using the Kaplan-Meier method and compared using the Log rank test.

P values of <0.05 were considered statistically significant. All analyses were performed using Empower (R) (www.empowerstats.com, X&Y Solutions, Boston, Massachusetts, USA).

Results

Patient Characteristics (Training and Validation Sets)

A total of 1805 patients were included in this study. The patient characteristics of the training and validation sets are presented in Table 1. In the training set, the median age was 52.6 years, and 69.8% of the patients were male. Among them, 836 patients (66.1%) had a tumor diameter greater than 5 cm, and 777 patients (61.5%) had intrahepatic lesions larger than 3. Serum AFP levels ≥ 100 ng/mL were observed in 737 patients (58.3%), while 256 patients (20.3%) had a Prothrombin time > 13.1 s. According to the ALBI score, grades 1, 2, and 3 were represented by 309 patients (24.4%), 856 patients (67.7%), and 99 patients (7.8%), respectively. The patient characteristics in the validation set were similar to those in the training set, with no significant differences observed (p > 0.05, Table 1).

Table 1 Baseline Characteristics of Patients in the Training and Validation Sets (N=1805)

The Development of the ALFP Score in the Training Set

The multivariate analysis revealed that baseline ALBI grade 2 or 3 and AFP ≥ 100 ng/mL and PT 13.1 s were associated OS (ALBI grade 2 or 3, hazard ratio [HR] 1.301, 95% CI 1.073–1.577, p = 0.007: AFP ≥ 100 ng/mL, HR 1.328, 95% CI 1.133–1.557, p < 0.001: PT> 13.1 s HR 1.262, 95% CI 1.047–1.521, p=0.015: Table 2). These factors, ALBI grade, AFP, and PT, emerged as independent prognostic factors, and their hazard ratios were similar in the multivariate analysis. Based on this result, we built an easy score named the albumin–bilirubin grade and α-fetoprotein and prothrombin time score (ALFP score). We assigned 1 point for ALBI grade 2 or 3 and 1 point for baseline AFP ≥ 100 ng/mL, and 1 point for baseline PT> 13.1 s. The grading of the score is as follows: ALFP score 0 points (ALBI grade 1 and AFP < 100 ng/mL and PT <13.1s), 1 point (either ALBI grade 2 or 3, and AFP ≥ 100 ng/mL, and PT>13.1s), and 2 points (both ALBI grade 2 or 3, and AFP ≥ 100 ng/mL, or PT>13.1s and AFP ≥ 100 ng/mL, or ALBI grade 2 or 3 and PT>13.1s), and 3 points (ALBI grade 2 or 3, and AFP ≥ 100 ng/mL, and PT>13.1s).

Table 2 Results of Univariate and Multivariate Analysis in the Training Set

According to these ALFP scores, the overall Survival (OS) is significantly connected with the score level. The median OS in patients with ALFP scores of 0, 1, 2, and 3 points was 50.1 months (95% CI not applicable [NA]), 33.6 months (95% CI 27.4–39.8), and 20.4 months (95% CI 18.0–22.8); and 12.7 months (95% CI 10.2–15.2), respectively; the difference was statistically significant (p < 0.001: Figure 2A). The 2022 BCLC version stratifies the Intermediate stage HCC into 3 groups,⁶ the first subgroup could be candidates for a liver transplant if they meet the extended Liver Transplant criteria’ according to the criteria of the Institution,²⁵ aim for improving treatment strategy. Based on Milan criteria²⁶ (1 lesion ≤5 cm or 2–3 lesions ≤ 3), we divided patients into two groups by the diameter of tumor (≤5 cm or >5cm) and the No. Of intrahepatic lesions (≤3 or >3), and calculating ALFP score in two sets, we found the OS is strongly connected with the level of score in each group, the median OS in the diameter of tumor ≤5 cm group with ALFP scores of 1, 2 and 3 points were: 54.9 months (95% CI not applicable), 50.0 months (95% CI 38.5–61.5), 23.9 months (95% CI 12.3–35.5), respectively; The difference was statistically significant (p = 0.0077: Figure 3A). The diameter of tumor >5 cm group with ALFP scores of 0, 1, 2 and 3 points were: 31.8 months (95% CI 11.7–51.9), 20.3 months (95% CI 17.4–23.2), 15.6 months (95% CI 13.1–18.1), 10.8 months (95% CI 8.2–13.4) respectively; The difference was statistically significant (p < 0.0001: Figure 3A). The median OS in No. Of intrahepatic lesions ≤ 3 groups with ALFP scores of 1, 2, and 3 points were: 42.7 months (95% CI 27.4–58.0), 34.4 months (95% CI 21.4–47.4), 16.8 months (95% CI 6.8–26.8), respectively; The difference was statistically significant (p = 0.002: Figure 4A). The No. of intrahepatic lesions >3 group with ALFP scores of 0, 1, 2 and 3 points were: 39.5 months (95% CI 23.6–55.4), 22.4 months (95% CI 14.5–30.3), 18.2 months (95% CI 16.0–20.4), 10.8 months (95% CI 7.8–13.8); The difference was statistically significant (p < 0.001: Figure 4A).

Figure 2 The OS according to the ALFP score in the training and validation sets. (A) The median OS according to the ALFP score in the training set was as follows: 0 points (n = 112), 50.1 months (95% CI not applicable); 1 point (n = 484), 33.6 months (95% CI 27.4–39.8); 2 points (n = 540), 20.4 months (95% CI 18.0–22.8); and 3 points (n = 128), 12.7 months (95% CI 10.2–15.2); this amounted to a statistically significant difference (p < 0.001). (B) The median OS according to the ALFP score in the validation set was as follows: 0 points (n = 54), 42.5 months (95% CI 16.1–68.9); 1 point (n = 188), 32.6 months (95% CI 25.1–40.1); 2 points (n = 240), 20.4 months (95% CI 16.4–24.4); and 3 points (n = 59), 13.0 months (95% CI 0.0–27.3); this amounted to a statistically significant difference (p = 0.0019). Abbreviations: CI, confidence interval; ALFP, score albumin–bilirubin grade, α-fetoprotein and prothrombin time score; OS, overall survival.

Figure 3 The OS according to the ALFP score in the training and validation sets stratified by the diameter of tumor. (A) The median OS according to the ALFP score in the training set stratified by the diameter of tumor was as follows: The diameter of tumor ≤5 cm group, 0 points (n = 52), not applicable (95% CI not applicable); 1 point (n = 170), 54.9 months (95% CI not applicable); 2 points (n = 162), 50.0 months (95% CI 38.5–61.5); and 3 points (n = 44), 23.9 months (95% CI 12.3–35.5); this amounted to a statistically significant difference (p = 0.0077). The diameter of tumor >5 cm group, 0 points (n = 60), 31.8 months (95% CI 11.7–51.9); 1 point (n = 314), 20.3 months (95% CI 17.4–23.2); 2 points (n = 378), 15.6 months (95% CI 13.1–18.1); and 3 points (n = 84), 10.8 months (95% CI 8.2–13.4); this amounted not to a statistically significant difference (p < 0.001). (B) The median OS according to the ALFP score in the validation set stratified by the diameter of tumor was as follows: The diameter of tumor ≤5 cm group, 0 points (n = 26), 59.0 months (95% CI 38.8–79.2); 1 point (n = 69), 47.7 months (95% CI not applicable); 2 points (n = 70), 29.6 months (95% CI 19.1–40.1); and 3 points (n = 24), 27.0 months (95% CI 0.0–54.1); this amounted to a statistically significant difference (p = 0.015). The diameter of tumor >5 cm group, 0 points (n = 28), 21.6 months (95% CI 14.5–28.7); 1 point (n = 119), 25.1 months (95% CI 16.5–33.7); 2 points (n = 170), 16.3 months (95% CI 11.8–20.8); and 3 points (n = 35), 9.8 months (95% CI 2.8–16.8); this amounted not to a statistically significant difference (p = 0.092). Abbreviations: CI, confidence interval; ALFP, score albumin–bilirubin grade, α-fetoprotein and prothrombin time score; OS, overall survival.

Figure 4 The OS according to the ALFP score in the training and validation sets stratified by the No. of intrahepatic lesions. (A) The median OS according to the ALFP score in the training set stratified by the No. of intrahepatic lesions was as follows: The No. of intrahepatic lesions ≤3 group, 0 points (n = 54), not applicable (95% CI not applicable); 1 point (n = 191), 42.7 months (95% CI 27.4–58.0); 2 points (n = 201), 34.4 months (95% CI 21.4–47.4); and 3 points (n = 41), 16.8 months (95% CI 6.8–26.8); this amounted to a statistically significant difference (p = 0.002). The No. of intrahepatic lesions >3 group, 0 points (n = 58), 39.5 months (95% CI 23.6–55.4); 1 point (n = 293), 22.4 months (95% CI 14.5–30.3); 2 points (n = 339), 18.2 months (95% CI 16.0–20.4); and 3 points (n = 87), 10.8 months (95% CI 7.8–13.8); this amounted not to a statistically significant difference (p < 0.001). (B) The median OS according to the ALFP score in the validation set stratified by the No. of intrahepatic lesions was as follows: The No. of intrahepatic lesions ≤3 group, 0 points (n = 23), 59.0 months (95% CI 28.2–89.8); 1 point (n = 87), 39.3 months (95% CI 24.3–54.3); 2 points (n = 87), 35.6 months (95% CI 15.3–55.9); and 3 points (n = 17), 28.3 months (95% CI 3.0–53.6); this amounted not to a statistically significant difference (p = 0.35). The No. of intrahepatic lesions >3 group, 0 points (n = 31), 42.5 months (95% CI 7.9–77.1); 1 point (n = 101), 28.0 months (95% CI 11.2–44.8); 2 points (n = 152), 19.2 months (95% CI 15.1–23.3); and 3 points (n = 42), 12.5 months (95% CI 8.3–16.7); this amounted not to a statistically significant difference (p = 0.034). Abbreviations: CI, confidence interval; ALFP, score albumin–bilirubin grade, α-fetoprotein and prothrombin time score; OS, overall survival.

The Validation of the ALFP Score in the Validation Set

We validated our research result in the validation set (Figure 1), Baseline patients’ characteristics are described in Table 1, and there is no significant difference between the two cohorts (p>0.05: Table 1), which is considered comparable. The median OS in patients with ALFP scores of 0, 1, 2, and 3 points were 42.5 months (95% CI 16.1–68.9),32.6 months (95% CI 25.1–40.1), 20.4 months (95% CI 16.4–24.4), 13.0 months (95% CI 0.0–27.3), respectively; the difference was statistically significant (p < 0.0019: Figure 2B). The median OS in the diameter of tumor ≤5 cm group with ALFP scores of 0, 1, 2 and 3 points were: 59.0 months (95% CI 38.8–79.2), 47.7 months (95% CI not applicable), 29.6 months (95% CI 19.1–40.1), 27.0 months (95% CI 0.0–54.1), respectively; The difference was statistically significant (p = 0.015: Figure 3B). The diameter of tumor >5 cm group with ALFP scores of 0, 1, 2 and 3 points were: 21.6 months (95% CI 14.5–28.7), 25.1 months (95% CI 16.5–33.7), 16.3 months (95% CI 11.8–20.8), 9.8 months (95% CI 2.8–16.8); respectively. The difference was insignificant (p =0.092: Figure 3B). The median OS in intrahepatic lesions number ≤ 3 groups with ALFP scores of 0, 1, 2, and 3 points were: 59.0 months (95% CI 28.2–89.8), 39.3 months (95% CI 24.3–54.3), 35.6 months (95% CI 15.3–55.9), 28.3 months (95% CI 3.0–53.6), respectively; The difference was no statistically significant (p = 0.35: Figure 4B). In intrahepatic lesions number >3 groups with ALFP scores of 0, 1, 2 and 3 points were: 42.5 months (95% CI 7.9–77.1), 28.0 months (95% CI 11.2–44.8), 19.2 months (95% CI 15.1–23.3), 12.5 months (95% CI 8.3–16.7), respectively; The difference was statistically significant (p < 0.034: Figure 4B). The performance of ALFP score in validation is similar to the training set, indicate the accuracy and predictive ability of the score.

Discussion

In the current study, we identified baseline PT > 13.1 s, AFP > 100 ng/mL, and ALBI grade 2b/3 as independent unfavorable prognostic factors in HCC patients undergoing TACE. Building upon these findings, we developed a straightforward and practical scoring system called the ALFP score. The ALFP score exhibited good performance in terms of overall survival (OS) curves, both in the training set and in the validation set, providing further confirmation of its predictive ability.

TACE is the recommended therapy for unresectable HCC and has shown improved survival outcomes for patients.^5,27 The prognosis of HCC patients is connected with liver function and tumor burden, and the prediction system should include both factors. Child-Pugh system (CPS) predicts prognosis in patients with chronic liver disease (CLD). It has been extended inclusion in many HCC staging systems, including the Cancer of the Liver Italian Program (CLIP)²⁸ score, the Japan Integrated Staging Score (JIS score),²⁹ and the widely used Barcelona Clinic Liver Cancer (BCLC)³⁰ score. However, CPS is not the most suitable score for assessing liver dysfunction.³¹ To address this limitation, the albumin-bilirubin (ALBI) grade²⁴ was proposed, which calculates a score based on serum albumin and bilirubin levels. The ALBI grade has demonstrated utility and accuracy in multiple external validations,^32–35 and exhibited the highest homogeneity for survival prediction among various liver functional reserve models.³⁶ In the 2022 version of BCLC staging system,⁶ the ALBI grade is recommended as a score for assessing compensated liver function. Previous studies have consistently indicated that the ALBI grade provides more precise and accurate prognostic information than the CPS, particularly in patients with well-preserved liver function. In the present cohort, the multivariate analysis in the training set confirmed that a baseline ALBI grade of 2 or 3 was associated with a poor prognosis (p = 0.007, Table 2). Based on these previous findings, we incorporated the ALBI grade as a favorable prognostic factor in our current study.

Alpha-fetoprotein (AFP) is a glycoprotein that is produced in the fetal yolk sac and liver during the gestational period.³⁷ AFP has been widely used as a biomarker to monitor HCC treatment and guide therapeutic decision-making, as well as for predictive analysis.³⁸ Researchers have observed that AFP-high tumors exhibit enrichment in the VEGF pathway, which promotes tumor progression and metastasis.³⁹ Additionally, in cases where TACE is ineffective, there is an increase in VEGF (vascular endothelial growth factor) and b-FGF (basic fibroblast growth factor) during subsequent treatments, which may explain the association between elevated AFP levels and poor prognosis in HCC patients undergoing TACE.⁴⁰ In this study, AFP levels ≥100 ng/mL were strongly associated with overall survival (OS) in 58% of the patients (p < 0.001, Table 1). Although the precise role of AFP in tumor growth is still not fully understood, it serves as a reliable indicator in HCC patients and is associated with the VEGF pathway. Therefore, AFP levels may guide the development of new treatment strategies in the future. In conclusion, we have incorporated AFP as an unfavorable indicator for HCC patients undergoing TACE.

In 1984, Liebman discovered that Prothrombin induced by vitamin K absence or antagonist-II (PIVKA-II) was significantly elevated in the serum of HCC patients.⁴¹ Numerous studies have since confirmed that PIVKA-II can be used for HCC diagnosis in combination with AFP and is strongly associated with the prognosis of HCC patients who receive curative treatment.^14,42 In the absence of vitamin K, hepatocytes are unable to synthesize normal vitamin K-dependent coagulation factors (VII, X, V, II), resulting in impaired coagulation function and prolonged prothrombin time. In this study, we found that PT > 13.1 s was observed in 20% of both cohorts and was statistically significant in multivariate analysis with regard to overall survival (p < 0.025, Table 1). This finding is consistent with previous studies, which further validates the performance of PT as a prognostic factor. The prognostic factors of liver cancer are complex. AFP has been firmly established as a predictive factor for post-liver transplantation recurrence in large-scale external validation studies.⁴³ ALBI demonstrates superior sensitivity to early liver damage compared to JIS score and CPS.⁴⁴ Researchers achieved promising experimental results when using AFP combined with ALBI score to predict the prognosis of liver cancer patients receiving Atezolizumab plus bevacizumab (AtezoBev) treatment.⁴⁵ To achieve the best predictive efficacy in the treatment of liver cancer, consideration must be given to both tumor-related factors and liver-related factors. Building upon the previous study, we developed a scoring system to enhance prediction accuracy. By incorporating these comprehensive scoring criteria based on common serum laboratory indexes, we are able to categorize HCC patients with TACE into four risk categories. This scoring system provides a more detailed and standardized approach to the treatment of liver cancer.

The multivariate analysis identified AFP > 100 ng/mL, ALBI grade 2b/3, and PT > 13.1 s as independent prognostic factors, with similar hazard ratios (HRs). Based on the ALFP score system, each score was found to be statistically significant in relation to overall survival (OS), with higher scores indicating a poorer outcome. This result was confirmed in the validation set and is consistent with previous studies. In the subgroup analysis, higher scores were associated with lower OS, and these associations were statistically significant. The importance of tumor burden and liver function reserve in HCC prognosis cannot be understated. Patients were grouped based on tumor diameter and the number of intrahepatic lesions, and under the same tumor burden, those with lower scores had better OS. This finding was confirmed in both sets and was particularly statistically significant in the higher tumor burden group (p < 0.0001, Figure 3; p < 0.00047, Figure 4). It indicates that the prognosis of these patients is strongly influenced by liver function, and those in the higher score group did not derive significant benefits from TACE compared to the low score group. However, in the validation set, the group with a tumor diameter >5 cm and the group with ≤3 intrahepatic lesions did not show statistically significant differences (p > 0.05, Figures 3 and 4). This could be attributed to the smaller sample size of the validation set compared to the training set.

Due to the retrospective nature of this study, it is important to acknowledge certain limitations. Firstly, the true survival time of many patients remains unknown due to loss to follow-up, thus necessitating the treatment of the last contact time as a censoring event. Secondly, the enrolled patients do not cover a comprehensive range of hepatitis types. Additionally, there is missing clinical information for some patients. To mitigate these limitations, future prospective studies are planned to complement our findings.

In conclusion, the ALFP score can predict the prognosis of HCC patients receiving TACE and guide the subdivision and precise treatment of intermediate hepatocellular carcinoma.