Skip to main content

Construction and validation of a predictive model for hepatocellular carcinoma based on serum markers

Abstract

Background

Early hepatocellular carcinoma (HCC) detection with non-invasive biomarkers remains an unmet clinical need. We aimed to construct a predictive model based on the pre-diagnostic levels of serum markers to predict the early-stage onset of HCC.

Methods

A total of 339 HCC patients (including 157 patients from Changzhou cohort and 182 patients from Wuxi cohort) were enrolled in our retrospective study. Levels of 25 baseline serum markers were collected. Propensity score matching (PSM) analysis was conducted to balance the distributions of patients’ gender, age, and the surveillance time between HCC group and control group. Then, Receiver operating characteristic (ROC) and Logistic regression analysis were performed to screen the independent predictive variables and construct a non-invasive predictive model. Subsequently, ROC curve and Kaplan–Meier (K–M) curve were used to evaluate the predictive values of the model. Clinical net benefit of the model was demonstrated by decision curve analysis (DCA) and clinical impact curve.

Results

Five independent predictive variables for HCC onset and two general characteristics of patients (age and gender) were incorporated into the score model. ROC and DCA curves showed that the score model had better predictive performance in discrimination and clinical net benefit compared with single variable or other score systems, with the area under the curve (AUC) of 0.890 (95% CI 0.856–0.925) in Changzhou cohort and 0.799 (95% CI 0.751–0.849) in Wuxi cohort. Meanwhile, stratification analysis indicated that the score model had good predictive values for patients with early tumor stage (AJCC stage I) or small tumors (< 2 cm). Moreover, the score of HCC patient began to increase at 30 months before clinical diagnosis and reach a peak at 6 months.

Conclusion

Based on this model, we could optimize the current risk stratification at an early stage and consider further intensive surveillance programs for high-risk patients. It could also help clinicians to evaluate the progression and predict the prognosis of HCC patients.

Peer Review reports

Introduction

Hepatocellular carcinoma (HCC), one of the most fatal malignant tumors, is the third leading cause of cancer-related death in the world (8.2%) [1]. According to the Global Cancer Statistics 2018 [2], the incidence and mortality rate of HCC in China after age standardization were 17.7/100,000 and 16.4/100,000, respectively. More than 70% of the patients are initially diagnosed at intermediate-to-advanced stages, which may be attributed to the insidious onset of HCC. Although medical technology has been improved greatly in recent years, the prognosis of patients with HCC remains unfavorable [3], the 5-years survival rate is only 11.7–14.1% [4]. In contrast, the 5-years survival rates of patients with early-stage who underwent the curative surgical therapies were approximate 69.0–86.2% [5]. As a result, developing a scientific screening method for early identification and timely treatment is critical for improving the prognosis of HCC patients.

Although the routine clinical screening methods for early HCC diagnosis historically involved abdominal ultrasound (US) examination and serum alpha-fetoprotein (AFP) detection [6]. However, the accuracy of the results obtained from the US is significantly influenced by the skills of the image observer, the conditions of instrumentation, and the characteristics of the patient (e.g., obesity, liver texture), with only 39–65% sensitivity for small liver tumors (less than 2 cm) [7]. AFP, a marker commonly used in HCC diagnosis, is positively correlated with the tumor size, but it remains negative in approximately 15–30% of HCC patients [8]. In addition, the AFP level was elevated significantly under certain pathological conditions, such as chronic liver disease, germ cell tumors, and gastric cancer [9, 10]. Recently, it has been reported that AFP-L3 and protein induced by Vitamin K absence or Antagonist-II (PIVKA-II) have high specificity in the diagnosis of HCC (92.9% and 89%, respectively), but the sensitivity of these individual serum markers for early HCC diagnosis is suboptimal [11, 12]. Furthermore, combined indicators such as neutrophil–lymphocyte ratio (NLR) [13], gamma-glutamyl transpeptidase to platelet ratio (GPR) [14], and gamma-glutamyl transpeptidase to lymphocyte ratio (GLR) [15] can be used to predict the prognosis and onset of HCC. Some studies have developed score models based on different variables such as gender, age, AFP levels, and pathological data, which have improved the accuracy of early HCC diagnosis. Nevertheless, some variables incorporated in the model were difficult to collect, which limits the usefulness of these models. Therefore, the models that could accurately predict HCC and conveniently guide clinicians for early HCC diagnosis were urgently needed. In this study, we retrospectively analyzed the levels of pre-diagnostic serum markers in patients with HCC to construct a non-invasive predictive model which could accurately predict the HCC onset and might facilitate early clinical detection and prognostic assessment.

Material and methods

Study subjects

We retrospectively reviewed the clinical data of 157 HCC patients hospitalized in Third People’s Hospital of Changzhou between June 2019 and December 2020, of whom 65 HCC patients had complete prognostic data. The inclusion criteria for our study were as follows: (1) Clinical diagnostic criteria for all HCC were based on the China guidelines (2017 Edition) [16] for Diagnostic and Treatment of Primary Liver Cancer; (2) The clinical data, including general characteristics, the surveillance time, and the levels of serum markers, were collected from all patients; (3) Patients with HCC had no other types of malignant tumors. Meanwhile, 734 patients suffering from chronic hepatitis or liver cirrhosis were enrolled as control. Additionally, clinical data of 305 patients (182 HCC patients and 123 liver cirrhosis patients) hospitalized in the Fifth People's Hospital of Wuxi during the same period were collected as external validation data, of whom 99 HCC patients had complete prognostic data.

Serum variables analysis

The variables used to conduct the study are presented: (1) General clinical characteristics of all patients were collected, including age, gender, and medical history; (2) All the serum variables or combined indicators that could predict the onset of HCC were carefully screened in the PubMed database with the search term ‘serum predict HCC’. Then, 25 variables that could predict the early onset or prognosis of HCC were retrieved, including AFP, AFP-L3, PIVKA-II, Alanine transaminase (ALT), Aspartate transaminase (AST) levels, Albumin (ALB), Total protein (TP), Total bilirubin (TB), Apolipoprotein1, Antithrombin III, Fibrinogen, Neutrophil to lymphocyte ratio (NLR), Gamma-glutamyl transpeptidase (GGT) to platelet ratio (GPR), GGT to lymphocyte ratio (GLR), platelet to lymphocyte ratio (PLR), and bilirubin to albumin ratio (TB/ALB); (3) Predictive models for HCC constructed by other researchers, including GALAD model [17], ALBI score [18], and Fib-4 index [19]; (4) Follow-up data of patients were obtained either through the outpatient clinic or via telephone inquiry every 3 months.

Statistical analysis

IBM SPSS software (version 22) was employed for statistical analysis. Continuous measurement data with normal distribution were expressed as mean ± standard deviation (\({\overline{\text{x}}}\) ± SD) and a comparison of the two groups was done using a t-test. On the other hand, we presented the continuous measurement data with non-normal distribution as median (interquartile range) [M (P25, P75)] and non-parametric U-test (Mann–Whitney U test) was employed to compare the two groups of parameters. Categorical variables were described as percentages and the differences were compared using chi-square test. R (version 3.6.1) software was used for PSM analysis, ROC curve analysis, Logistic regression, and Decision curve analysis. Comparisons between the ROC curves were done using the method of Delong et al. [20]. Moreover, all tests were two-sided, and two-tailed test with p value < 0.05 was considered a significant difference.

Results

Baseline characteristics of patients

Totally 891 patients who met the entry criteria were enrolled in current study, including 157 patients with HCC (HCC group), and 734 patients with chronic hepatitis or liver cirrhosis (control group). Baseline analysis demonstrated that there were significant differences in patients’ general characteristics (age and gender) between the two groups (Table 1). Patients in the HCC group were older than those in the control group (59.66 ± 10.71 versus 54.23 ± 13.66, p < 0.001) and the male gender occupied a higher ratio (83.4% versus 62.5%, p < 0.001). To reduce the influence of these potential confounding factors, we performed PSM analysis with a 1:1 ratio and caliper of 0.05 using nearest neighbor method. After PSM, the final sample size in this study was 154 patients in each group, and most of the HCC patients were early stage (AJCC TNM Stage I–IV, 48.7%, 19.48%, 17.53%, 14.29%, respectively). The general characteristics (age, gender, and surveillance time) of the two groups were significantly balanced (p > 0.01), suggesting the PSM analysis improved the comparability of the two groups. However, statistical analysis of the matched data revealed that there were still differences in some serum variates between the two groups, such as AFP, AFP-L3, PIVKAII, etc. (p < 0.01).

Table 1 Baseline characteristics of patients in Changzhou cohort

Construction of a predictive model for early onset of HCC based on the screened serum markers

ROC curve analysis was applied to evaluate the predictive efficiency of the serum markers, and univariate logistic regression analysis was used to screen risk factors for predicting HCC. Using the criteria of AUC > 0.55 and p < 0.25, respectively. 15 variates that could predict the onset of HCC were identified, including AFP, AFP-L3, AG, ALB, ALT, AST, GGT, TBIL, LY, PDW, MPV, GPR, TB/ALB, the status of HBsAg and HBeAg (Table 2; Additional file 1: Fig. S1). Multivariate logistic regression analysis revealed that AFP- L3, ALB, ALT, the status of HBsAg, and the status of HBeAg were independent risk factors for the onset of HCC. Finally, a score model for predicting the onset of HCC was constructed based on these independent risk factors and two general characteristics (age and gender) of patients: Score model = 0.009755*AFPL3 − 0.141957*Gender − 0.017824*Age (years) + 0.036840*ALB (g/L) − 0.004326*ALT (U/L) + 0.642490*(status of HBsAg) − 0.832696*(status of HBeAg), (Male equal to 1 and Female equal to 0; the status of HBsAg-positive and HBeAg-positive both equal to 1 and the status of HBsAg-negative and HBeAg-negative both equal to 0).

Table 2 ROC curve analysis and univariate logistic regression analysis of predictors for HCC onset

Evaluation of the risk score model in Changzhou cohort

According to the score model, the score for each patient in Changzhou cohort was calculated. ROC curve of the risk score model showed that the AUC value for predicting the onset of HCC was 0.890, which was significantly higher than that of individual characteristics (0.509–0.689, p < 0.05; Fig. 1A; Additional file 2: Table S1), ALBI score, Fib-4 index, and GALAD model (0.514–0.756, p < 0.05; Fig. 1B; Additional file 2: Table S1). Interestingly, the risk score model also had similar high predictive values in patients with different TNM stages (AUC: 0.875–0.902, Fig. 1C), but no remarkable differences among them (p > 0.05, Additional file 2: Table S2). Meanwhile, decision curve analysis revealed that this score model had a larger clinical net benefit compared with the GALAD model, ALBI score, and Fib-4 index (Fig. 1D). Moreover, clinical impact curves also showed that this risk model had good prediction power for the onset of HCC (Fig. 1E). Furthermore, the Kaplan–Meier curve was performed. For Changzhou cohort, the cumulative incidence of HCC was significantly higher in patients with high-risk score than in those with low-risk score (p < 0.05, Fig. 1F). In addition, the score predictive model showed promising levels of sensitivity (89.4% versus 33.1–82.4%) and NPV (87.3% versus 55.4–73.3%), respectively (Additional file 2: Table S1), even in patients with different tumor stages (sensitivity: 85.2–97.3%; NPV: 97.0–99.1%) (Additional file 2: Table S2). In conclusion, this risk score model based on the pre-diagnostic level of serum markers was a reliable and superior model for predicting the onset of HCC.

Fig. 1
figure 1

Performance of non-invasive predictive model for predicting the onset of HCC in Changzhou cohort. A ROC curve analysis of the risk model and single variable for predicting the onset of HCC. B ROC curve analysis of the risk model and other score systems for predicting the onset of HCC. C ROC curve analysis of the risk model for predicting the onset of HCC in patients with different AJCC TNM stages. D Decision curve analysis demonstrates the clinical net benefit of the risk score model and other score systems for the onset of HCC in Changzhou cohort. E Clinical impact curve of the risk model. F Cumulative event between high risk and low risk groups at indicated time before clinical diagnosis

Evaluation of the model score in HCC progression

To investigate the potential clinical value of this model in the progression of HCC, we calculated the risk score of patients at each clinical surveillance time and visualized the longitudinal changes of risk score for disease progression. The smooth curve showed that the risk score of the HCC group before clinical diagnosis time was greatly elevated than that of the control group. Interestingly, the risk score of the HCC group increased at 30 months before clinical diagnosis and reached a peak at 6 months, suggesting that the score model may be a useful tool for early predicting the risk of HCC (Fig. 2A). Furthermore, similar trends were also observed in patients with different tumor stages (Fig. 2B–D). In addition, according to the best cutoff value for predicting the onset of HCC (0.227), patients were divided into two subgroups (high-risk and low-risk subgroups). The Kaplan–Meier curve analysis revealed that patients in the high-risk subgroup had a lower overall survival (p < 0.05, Additional file 1: Fig. S2A). These results strongly indicated that the risk score model could predict the onset of HCC at early time points and might facilitate the prognostic evaluation of HCC patients.

Fig. 2
figure 2

The longitudinal changes of the risk score for disease progression. AD The smooth curve of risk score in selected datasets (A Changzhou cohort. B Patients with AJCC TNM stage I. C Patients with AJCC TNM stage II. D Patients with AJCC stage III and IV)

Validation of the risk model in Wuxi cohort

Since the risk score model based on the non-invasive serum markers showed reliable and superior predictive ability for HCC development in Changzhou cohort, we further validated the predictive ability of the risk score model in an independent cohort from Wuxi (Additional file 2: Table S3), and the score for every patient was calculated using the same formula. ROC curve analysis was performed for predicting the development of HCC. The AUC value was 0.799 with a sensitivity of 84.6% and specificity of 60.5%, which demonstrated that this score model still had a superior predictive ability (Fig. 3A). Furthermore, the risk model also demonstrated strong predictive power for patients with different tumor stages based on AJCC criteria (AUC, 0.704–0.774, Fig. 3B). Decision curve analysis and clinical impact curve revealed that the risk score model had a good clinical utility for predicting the onset of HCC in Wuxi cohort (Fig. 3C–E). In addition, the score model also had a high predictive value for small tumor (diameter size < 2 cm, AUC, 0.791, Fig. 3F). However, survival analysis showed no difference between two subgroups (p > 0.05, Additional file 1: Fig. S2B). In summary, the risk score model constructed using non-invasive serum markers was a reliable and effective tool for predicting the development of HCC.

Fig. 3
figure 3

Validation of the risk model in Wuxi cohort. A Performance of the risk model for predicting the onset of HCC. B Performance of the risk model for predicting the onset of HCC in patients with different AJCC TNM stages. C Decision curve analysis demonstrates the clinical net benefit of the risk score model for the onset of HCC. DE Clinical impact curve of the risk model. F Predictive value of the risk model for the onset of HCC in patients with small tumor (< 2 cm)

Discussion

The prognosis of the HCC patient remains dismal due to the insidious onset [21]. Effective early screening strategies and curative treatments could significantly improve the prognosis of HCC patients [22]. However, Current approaches and biomarkers have some limitations in the early HCC diagnosis [23], such as poor compliance, low sensitivity, and time consumption. Surveillance of early-stage patients is inadequate. On the other hand, Early and accurate HCC detection with non-invasive biomarkers remains an unmet clinical need. In this study, we screened the pre-diagnostic serum biomarkers related to the development of HCC and constructed a non-invasive risk score model for early detection of HCC. This score model showed satisfactory discriminant function (AUC: 0.890, with 89.4% sensitivity and 71.4% specificity, respectively). Similar predictive value of the model for HCC onset was observed in an independent cohort (AUC: 0.799). Moreover, the predictive efficacy of the risk model showed powerful and satisfactory clinical performance, which was evaluated by the Kaplan–Meier curve, decision curve analysis, and clinical impact curve analysis. Notably, the risk score model still showed satisfactory predictive power in patients with early stage (AJCC Stage I) or those with small tumors (< 2 cm). Based on this model, we could optimize the early risk stratification of patients and evaluate the prognosis of HCC patients.

This model consists of five pre-diagnostic serum biomarkers, and these markers could be grouped into three categories: 1) host factors (age, gender, and ALB); 2) viral activity-related factors (ALT, HBsAg status, and HBeAg status); 3) malignant hepatocytes growth-related factors (AFP-L3). Compared to the GALAD model, the indicators of viral activity-related factors were included. Hepatitis B virus (HBV) infection is a key risk factor for the etiology of HCC, especially in eastern Asia and sub-Saharan Africa [24], nearly 90% of HCC patients in China had a history of HBV infection [1]. Several mechanisms by which Hepatitis B virus progression to HCC were proposed [25]. Nevertheless, HBeAg positivity indicated active replication of HBV virus [26]. and several case–control studies indicated that HBeAg seemed to be a better predictive marker for HBV-related HCC [27]. AFP and AFP-L3 are the most widely used serum markers in HCC diagnosis. Studies have demonstrated that serum AFP levels in HCC patients were elevated with disease progression and correlated with tumor size [28], the sensitivity and specificity of AFP for early detection of HCC were 45.3–62% and 87–93%, respectively [29]. AFP isoform L3 (AFP-L3), a glycoprotein of primary origin in hepatocellular cancer cells, was a marker independent of AFP. The levels of AFP-L3 were significantly associated with the development of HCC, with a sensitivity of 45.9–50.7% and a specificity of 92.9%. Nearly 34.3% of HCC patients with normal AFP levels had abnormal AFP-L3 expression before HCC diagnosis [11]. Recent studies have reported that the serum AFP-L3 could be detected in about 35% of patients with small HCC [30]. In addition, serum level of ALB, an essential indicator of nutrition, reflects the capability to regulate the immune and antioxidant reaction against carcinogenesis [31, 32]. Several studies have reported that the abnormal levels of ALB could independently predict a worse prognosis in patients with HCC [33]. Taken together, these markers, included in our score model, are related to the development and prognosis of the tumor. Therefore, this scoring model we constructed could be applied to numerous objectives for HCC, such as stratification of risk, follow-up care after treatment, and prediction of prognosis.

There are still certain limitations. First, the total number of HCC patients in our study is inadequate. Second, the follow-up data for all patients obtained from hospital records were partially missing, especially for patients with early-stage HCC (AJCC stage I and II). Finally, the genotypes of virus were not incorporated into the score model. Therefore, we will enlarge the patient cohort, long-term follow-up, and multi-center studies in the future to further validate the clinical values of the risk model.

Conclusion

In summary, the score model established in this study had the characteristics of high accuracy, reliable, and easy to use. Patients with high-risk scores should consider more intensive surveillance programs. Prospectively, clinicians could use this risk score model as early as possible to screen high-risk patients, evaluate the progression and predict the prognosis of HCC patients.

Availability of data and materials

All data generated and/or analyzed during this study are available from the corresponding author on reasonable request.

Abbreviations

HCC:

Hepatocellular carcinoma

ROC:

Receiver operating characteristic

K–M curve:

Kaplan–Meier curve

DCA:

Decision curve analysis

AUC:

Area under the curve

95%CI:

95% Confidence interval

AFP:

Alpha-fetoprotein

AFP-L3:

Lens culinaris agglutinin-reactive fraction of AFP

PIVKA-II:

Protein induced by Vitamin K absence or Antagonist-II

ALB:

Albumin

ALP:

Alkaline phosphatase

GGT:

Gamma-glutamyl transpeptidase

TP:

Total protein

WBC:

White blood cell

NEUT:

Neutrophil count

NLR:

Neutrophil–lymphocyte ratio

GPR:

GGT to platelet ratio

GLR:

GGT to lymphocyte ratio

References

  1. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018;68(6):394–424.

    Article  Google Scholar 

  2. Ferlay J, Colombet M, Soerjomataram I, Mathers C, Parkin DM, Pineros M, Znaor A, Bray F. Estimating the global cancer incidence and mortality in 2018: GLOBOCAN sources and methods. Int J Cancer. 2019;144(8):1941–53.

    Article  CAS  Google Scholar 

  3. Kim E, Viatour P. Hepatocellular carcinoma: old friends and new tricks. Exp Mol Med. 2020;52(12):1898–907.

    Article  CAS  Google Scholar 

  4. Allemani C, Matsuda T, Di Carlo V, Harewood R, Matz M, Nikšić M, Bonaventure A, Valkov M, Johnson CJ, Estève J, et al. Global surveillance of trends in cancer survival 2000–14 (CONCORD-3): analysis of individual records for 37 513 025 patients diagnosed with one of 18 cancers from 322 population-based registries in 71 countries. Lancet. 2018;391(10125):1023–75.

    Article  Google Scholar 

  5. Tsilimigras DI, Bagante F, Sahara K, Moris D, Hyer JM, Wu L, Ratti F, Marques HP, Soubrane O, Paredes AZ, et al. Prognosis after resection of barcelona clinic liver cancer (BCLC) stage 0, A, and B hepatocellular carcinoma: a comprehensive assessment of the current BCLC classification. Ann Surg Oncol. 2019;26(11):3693–700.

    Article  Google Scholar 

  6. Chen G, Zhang W, Ben Y. Identification of key regulators of hepatitis C virus-induced hepatocellular carcinoma by integrating whole-genome and transcriptome sequencing data. Front Genet. 2021;12: 741608.

    Article  CAS  Google Scholar 

  7. Harris PS, Hansen RM, Gray ME, Massoud OI, McGuire BM, Shoreibah MG. Hepatocellular carcinoma surveillance: an evidence-based approach. World J Gastroenterol. 2019;25(13):1550–9.

    Article  Google Scholar 

  8. Wang L, Pan L, Yao M, Cai Y, Dong Z, Yao D. Expression of oncofetal antigen glypican-3 associates significantly with poor prognosis in HBV-related hepatocellular carcinoma. Oncotarget. 2016;7(27):42150–8.

    Article  Google Scholar 

  9. Quaglia A. Hepatocellular carcinoma: a review of diagnostic challenges for the pathologist. J Hepatocell Carcinoma. 2018;5:99–108.

    Article  CAS  Google Scholar 

  10. Kim MN, Kim BK, Kim SU, Park JY, Ahn SH, Han KH, Kim DY. Longitudinal assessment of alpha-fetoprotein for early detection of hepatocellular carcinoma in patients with cirrhosis. Scand J Gastroenterol. 2019;54(10):1283–90.

    Article  CAS  Google Scholar 

  11. Yi X, Yu S, Bao Y. Alpha-fetoprotein-L3 in hepatocellular carcinoma: a meta-analysis. Clin Chim Acta. 2013;425:212–20.

    Article  CAS  Google Scholar 

  12. Xing H, Zheng YJ, Han J, Zhang H, Li ZL, Lau WY, Shen F, Yang T. Protein induced by vitamin K absence or antagonist-II versus alpha-fetoprotein in the diagnosis of hepatocellular carcinoma: a systematic review with meta-analysis. Hepatobiliary Pancreat Dis Int. 2018;17(6):487–95.

    Article  Google Scholar 

  13. Hung HC, Lee JC, Cheng CH, Wu TH, Wang YC, Lee CF, Wu TJ, Chou HS, Chan KM, Lee WC. Impact of neutrophil to lymphocyte ratio on survival for hepatocellular carcinoma after curative resection. J Hepatobiliary Pancreat Sci. 2017;24(10):559–69.

    Article  Google Scholar 

  14. Luo D, Li H, Hu J, Zhang M, Zhang S, Wu L, Han B. Development and validation of nomograms based on gamma-glutamyl transpeptidase to platelet ratio for hepatocellular carcinoma patients reveal novel prognostic value and the ratio is negatively correlated with P38MAPK expression. Front Oncol. 2020;10: 548744.

    Article  Google Scholar 

  15. Zhang H, Zhou Y, Li Y, Qin W, Zi Y, Liu Y, Qiu X, Xu H, Liao W, Huang Z. Predictive value of gamma-glutamyl transpeptidase to lymphocyte count ratio in hepatocellular carcinoma patients with microvascular invasion. BMC Cancer. 2020;20(1):132.

    Article  CAS  Google Scholar 

  16. Zhou J, Sun HC, Wang Z, Cong WM, Wang JH, Zeng MS, Yang JM, Bie P, Liu LX, Wen TF, et al. Guidelines for diagnosis and treatment of primary liver cancer in China (2017 Edition). Liver Cancer. 2018;7(3):235–60.

    Article  Google Scholar 

  17. Best J, Bilgi H, Heider D, Schotten C, Manka P, Bedreli S, Gorray M, Ertle J, van Grunsven LA, Dechene A. The GALAD scoring algorithm based on AFP, AFP-L3, and DCP significantly improves detection of BCLC early stage hepatocellular carcinoma. Z Gastroenterol. 2016;54(12):1296–305.

    Article  CAS  Google Scholar 

  18. Johnson PJ, Berhane S, Kagebayashi C, Satomura S, Teng M, Reeves HL, O’Beirne J, Fox R, Skowronska A, Palmer D, et al. Assessment of liver function in patients with hepatocellular carcinoma: a new evidence-based approach-the ALBI grade. J Clin Oncol. 2015;33(6):550–8.

    Article  Google Scholar 

  19. Sterling RK, Lissen E, Clumeck N, Sola R, Correa MC, Montaner J, Sulkowski MS, Torriani FJ, Dieterich DT, Thomas DL, et al. Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection. Hepatology. 2006;43(6):1317–25.

    Article  CAS  Google Scholar 

  20. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988;44(3):837–45.

    Article  CAS  Google Scholar 

  21. Kanwal F, Singal AG. Surveillance for hepatocellular carcinoma: current best practice and future direction. Gastroenterology. 2019;157(1):54–64.

    Article  Google Scholar 

  22. Bruix J, Llovet JM. Major achievements in hepatocellular carcinoma. Lancet. 2009;373(9664):614–6.

    Article  Google Scholar 

  23. Alzamzamy A, Elsayed H, Abd Elraouf M, Eltoukhy H, Megahed T, Aboubakr A. Serum vascular endothelial growth factor as a tumor marker for hepatocellular carcinoma in hepatitis C virus-related cirrhotic patients. World J Gastrointest Oncol. 2021;13(6):600–11.

    Article  Google Scholar 

  24. Sherman M. Hepatocellular carcinoma: epidemiology, surveillance, and diagnosis. Semin Liver Dis. 2010;30(1):3–16.

    Article  CAS  Google Scholar 

  25. Rossner MT. Review: hepatitis B virus X-gene product: a promiscuous transcriptional activator. J Med Virol. 1992;36(2):101–17.

    Article  CAS  Google Scholar 

  26. Luo W, Xin H, Zhao P, Jiang S. Effects of HBeAg status on cellular immune function of patients with Hepatitis B virus/treponema pallidum co-infectio. Pak J Med Sci. 2021;37(7):1871–6.

    Article  Google Scholar 

  27. Yang HI, Lu SN, Liaw YF, You SL, Sun CA, Wang LY, Hsiao CK, Chen PJ, Chen DS, Chen CJ, et al. Hepatitis B e antigen and the risk of hepatocellular carcinoma. N Engl J Med. 2002;347(3):168–74.

    Article  CAS  Google Scholar 

  28. Hong G, Suh KS, Suh SW, Yoo T, Kim H, Park MS, Choi Y, Paeng JC, Yi NJ, Lee KW. Alpha-fetoprotein and (18)F-FDG positron emission tomography predict tumor recurrence better than Milan criteria in living donor liver transplantation. J Hepatol. 2016;64(4):852–9.

    Article  CAS  Google Scholar 

  29. Choi J, Kim GA, Han S, Lee W, Chun S, Lim YS. Longitudinal assessment of three serum biomarkers to detect very early-stage hepatocellular carcinoma. Hepatology. 2019;69(5):1983–94.

    Article  CAS  Google Scholar 

  30. Zhou L, Liu J, Luo F. Serum tumor markers for detection of hepatocellular carcinoma. World J Gastroenterol. 2006;12(8):1175–81.

    Article  CAS  Google Scholar 

  31. Reebye V, Saetrom P, Mintz PJ, Huang KW, Swiderski P, Peng L, Liu C, Liu X, Lindkaer-Jensen S, Zacharoulis D, et al. Novel RNA oligonucleotide improves liver function and inhibits liver carcinogenesis in vivo. Hepatology. 2014;59(1):216–27.

    Article  CAS  Google Scholar 

  32. Garcia-Martinez R, Andreola F, Mehta G, Poulton K, Oria M, Jover M, Soeda J, Macnaughtan J, De Chiara F, Habtesion A, et al. Immunomodulatory and antioxidant function of albumin stabilises the endothelium and improves survival in a rodent model of chronic liver failure. J Hepatol. 2015;62(4):799–806.

    Article  CAS  Google Scholar 

  33. Cai X, Chen Z, Chen J, Ma X, Bai M, Wang T, Chen X, Wu D, Wei L, Li X, et al. Albumin-to-alkaline phosphatase ratio as an independent prognostic factor for overall survival of advanced hepatocellular carcinoma patients without receiving standard anti-cancer therapies. J Cancer. 2018;9(1):189–97.

    Article  Google Scholar 

Download references

Acknowledgements

We would like to thank the department of Liver Oncology of the Changzhou Third People`s Hospital and Wuxi Fifth People`s Hospital for technical assistance.

Funding

This work was supported by the Major Projects of Changzhou Health Commission (ZD202119).

Author information

Authors and Affiliations

Authors

Contributions

LL and XZ contributed to the conception of the study; LZ contributed significantly to draw the figures and write the manuscript; MH and XL acquires the clinical data from hospitalization records; ZH and XL performed the data analyses; GZ and XZ helped with the critical revision. All authors read and approved the final manuscript.

Corresponding authors

Correspondence to Xike Zhou or Longgen Liu.

Ethics declarations

Ethics approval and consent to participate

This study was approved by the ethics committees of Changzhou Third hospital and Wuxi Fifth hospital, respectively (Approval number: 02A-A20220018 and 2021-015-1), and due to the retrospective nature of the study, the need for informed consent was waived by the Ethics Committee Review Board of Changzhou Third hospital and Wuxi Fifth hospital, respectively. A statement to confirm that all methods were carried out in accordance with relevant guidelines and regulations.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1. Figure S1.

The Venn diagram indicated intersection of the serum markers associated with HCC onset (Green, serum markers identified by ROC analysis with AUC > 0.55; blue, serum markers screened by univariate logistic regression analysis with p < 0.25). Figure S2. Kaplan–Meier analysis for overall survival between high risk group and low risk group. A In Changzhou cohort. B In Wuxi cohort.

Additional file 2. Table S1.

Predictive accuracy comparison of the variables for the onset of HCC in Changzhou cohort. Table S2. Predictive accuracy comparison of score model for the onset of HCC in patients with different AJCC TNM stages. Table S3. Baseline characteristics of patients in Wuxi cohort.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Zheng, L., Huang, Z., Li, X. et al. Construction and validation of a predictive model for hepatocellular carcinoma based on serum markers. BMC Gastroenterol 22, 418 (2022). https://doi.org/10.1186/s12876-022-02489-2

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12876-022-02489-2

Keywords