Prognostic nomogram for hepatocellular carcinoma with fibrosis of varying degrees

Background To investigate prognostic value of varying degrees of liver fibrosis in hepatocellular carcinoma (HCC) patients and establish an effective prognostic nomogram. Methods Eligible HCC patients with different degrees of fibrosis between 2004 and 2015 were matched from the Surveillance, Epidemiology, and End Results (SEER) database using propensity score matching (PSM). Prognostic value was evaluated using Kaplan-Meier and Cox hazard analysis. The nomogram on variables selected by multivariate analyses was established and subjected to internal validation. The predictive accuracy of nomograms was tested by concordance index (C-index) and calibration. Results HCC patients with advanced fibrosis/ cirrhosis was correlated with poor survival than those with none/moderate fibrosis in all patients HR: 1.111, 95%CI (1.042-1.185); p=0.001 and in the PSM cohort HR: 1.131, 95%CI (1.032-1.240); p=0.009. Multivariate analysis of propensity-matched cohort revealed that age more than 63, higher fibrosis score, AJCC stage T3-4, distant metastasis (M1), Tumor size ≥1, vascular invasion and elevated AFP level were independent factors. The nomogram integrating these factors offers an effective prognostic prediction for HCC patients (C-index 0.749, 95%CI: 0.735~0.763) relative to AJCC 7th edition (0.727) and TNM (0.73). The calibration plots suggested optimal agreement between nomogram prediction and actual observation. Conclusions Increased fibrosis was an independent risk factor for HCC survival. The prognostic nomogram integrating fibrosis score offers a more accurate prediction.

Increased fibrosis was an independent risk factor for HCC survival. The prognostic nomogram integrating fibrosis score offers a more accurate prediction.
Background Hepatocellular carcinoma (HCC) ranks the fifth most common cancer, with up to 10% of cancer-related mortality worldwide and 70% -100% of cumulative risk of recurrence within 5 years after surgery [1][2][3]. The majority (70%-90%) of HCCs arises with advanced fibrosis and cirrhosis as a consequence of various etiologies [4]. Liver fibrosis is a protective wound healing response to chronic liver damage and becomes fibrous scar and cirrhosis when injury persists [5]. Often leading to hepatic dysfunction, fibrosis and cirrhosis was considered as a premalignant state that increased the HCC risk and development [6]. However, the effect of different fibrosis severities for liver fibrosis on HCC prognosis remains controversial.
Currently, there are two surgical staging systems, namely the Tumor-Node-Metastasis (TNM) from the Liver Cancer Study Group of Japan (LCSGJ) and the American Joint Committee on Cancer (AJCC) staging system. The 7 th edition of AJCC staging system adopting TNM classification is a commonly used staging systems and maintains the prognostic power for HCC patients [7]. Unfortunately, its application is not completely satisfactory for individual patients, especially for patients with advanced HCC [8]. One possible explanation for this imperfection lies in its failure to incorporate some other important characteristics such as age and sex in survival prediction. Nomogram is an effective and convenient statistical tool for prognostic prediction by combining all these factors and it has been developed in a variety of cancer types [9][10][11].
Through this SEER analysis, we aimed to evaluate the effect of different fibrosis degrees on HCC prognosis and to establish for the first time a prognostic nomogram integrating fibrosis score and other independent variables to predict HCC survival.
Our model will provide precise prognostic information for clinicians in clinical practice.

Patients selection and Data extraction
We performed a retrospective cohort analysis by querying the SEER database HCC patients younger than 18 years, those whose survival time and fibrosis score were unavailable, patients with the same ID number, patients diagnosed before 2004, patients with no evidence of primary cancer were excluded. Fibrosis score, also called Ishak score, was classified as F0-4 [code 0, fibrosis scores 0-4 (none to moderate fibrosis)] and F5-6 [code 1, fibrosis scores 5-6 (severe fibrosis or cirrhosis)]. We compared patients with none to moderate fibrosis (F0-4) to patients with severe fibrosis or cirrhosis (F5-6). AFP level was described using code10 (positive/elevated), code 20 (Negative/normal; within normal limits) and others.

Statistical analysis
As previously described [12], 1:1 propensity score matching (PSM) analysis was carried out to minimize the selection bias and balance the baseline covariates between F0-4 and F5-6 group. After matching, covariate balance was tested and no covariates exhibiting a large imbalance indicated the adequacy of the constructed propensity score. Statistical analysis from SEER database was performed using SPSS 19.0 software. Continuous variables were presented as mean ± standard deviation (SD) and was analyzed using unpaired t test. Chi-square or Fisher's exact test was used to compare categorical variables. Survival curves were generated by Kaplan-Meier method (log-rank test). Significant factors affecting survival in univariable analysis were further determined by multivariable Cox regression analysis. The prognostic nomogram based on variables of multivariate analyses was established using the rms package in R project. The predictive accuracy and discriminative ability were determined by using concordance index (C-index) and calibration curve.
A threshold of P < 0.05 was considered statistically significant.

Clinicopathologic characteristics
According to the eligibility criteria, a total of 8119 patients diagnosed with HCC  Figure 1). The covariates between the two groups were well-balanced and showed no significant differences at baseline (p 0.05 for all; Table 1 and figure 2A, 2B).

Predictors of survival among advanced fibrosis/cirrhosis patients
Using multivariate analysis, predictors of survival among advanced fibrosis/cirrhosis patients were performed. As shown in Table 4

Prognostic Nomogram construction and validation
The prognostic nomogram integrated all significant independent factors for survival in multivariate analyses ( Figure 3A). The C-index for survival prediction was 0.749 (95%CI: 0.735~0.763). In addition, the calibration plot for the probability of survival at 1, 3 and 5 year presented an optimal agreement between the prediction by nomogram and actual observation ( Figure 3B). Thirty percent of all cohort was randomly generated by R software to make internal validation. The C-index of internal validation of the nomogram was 0.761 (95%CI 0.736~0.786), calibration curves also showed good agreement between prediction and observation in 1-, 3and 5-year OS ( Figure 3C). These results indicated the nomogram that integrated fibrosis score with other factors of age, AJCC T stage, distant metastasis, tumor size, vascular invasion and AFP level was reliable for survival prediction of HCC patients.

Comparison between nomogram and single independent factors/ conventional staging systems
The larger of the C-index represented a more accurate the predictive model. The Cindex of the prognostic nomogram was larger than the AJCC 7 th edition, the AJCC

Discussion
In this study, we demonstrated that advanced fibrosis/ cirrhosis was independently associated with HCC survival and for the first time established and internally validated a fibrosis score-based nomogram for HCC survival prediction. This prognostic model integrating fibrosis score was reliable, robust and practical for HCC patients.
HCC is closely associated with liver fibrosis and its end-stage, cirrhosis, with about 80-90% of HCC patients having underlying fibrosis and approximately one in three cases with liver cirrhosis will develop HCC in their lifetime. The incidence rate of HCC within 5 years in patients with advanced liver fibrosis/ cirrhosis ranges from 5% to 30% [6,13]. However, the impact of fibrosis score on survival outcome in HCC has yet been reached [14][15][16][17]. Therefore, a better understanding of the role of fibrosis score in HCC prognosis and a nomogram constructed to include fibrosis score may provide more-accurate prognostic prediction for HCC patients.
Fibrosis score is also called Ishak score. AJCC classifies fibrosis scores as none to moderate fibrosis and severe fibrosis or cirrhosis. To eliminate selection bias, we categorized the entire study population into groups with different degrees of fibrosis (none/ moderate fibrosis and advanced fibrosis/ cirrhosis) using the propensity score analyses. No significant difference in factors of age, sex, AJCC TNM stage, AFP level, tumor size and vascular invasion indicated the patients of both groups was well-balanced. Previously, Noda et al found no significantly association between liver fibrosis and OS of HCC patients (p=0.1185) [18]. Also, Suh et al found no significant difference in OS between mild and severe fibrosis (P = 0.267) [14].
However, our results revealed that advanced fibrosis/ cirrhosis provided worse survival than none/ moderate fibrosis. In addition, severe fibrosis/ cirrhosis was an independent risk factor for OS. This finding was consistent with a recent metaanalysis by Zhang et al and the study by Toyoda et al [19,20]. The discrepancies may be partly attributed to the different criteria of enrolling patients. Noda et al evaluated the effect of different degrees of fibrosis on the prognosis of non-viral HCC patients [18]. Suh et al investigated the influence of liver fibrosis on prognosis only for HCC with Child-Pugh grade A and a single HCC <5 cm [14]. Appropriately 70% HCC patients were afflicted with hepatitis virus infection [6]. Child-Pugh class B and HCV positivity were also associated with the poor prognosis of HCC [21]. Our study synthetically explored the relationship between fibrosis score and clinical prognosis of HCC.
In the multivariate analyses, besides fibrosis score, age more than 63, AJCC stage T3-4, distant metastasis (M1), tumor size ≥1cm, minor and major vascular invasion, and elevated AFP level were identified as independent predictors for poor prognosis of HCC, which was in agreements with previous studies [8, [22][23][24][25]. Consistency with previous studies [25,26], sex and lymph node metastasis failed to stratify survival in our study. In addition, we used multivariate analysis to identify predictors of survival among advanced fibrosis/cirrhosis patients and the results showed that advanced fibrosis/cirrhosis patients whose age older than 63-year old demonstrated worse survival. In advanced fibrosis/cirrhosis patients, AJCC stage T3-4, distant metastasis (M1), Tumor size ≥1cm, major vascular invasion and positive AFP level were associated with worse survival. Subsequently, we developed a nomogram that incorporated the predictors to predict the patients' survival. The C-index of our nomogram and internally validated nomogram was 0.749 and 0.761, which was higher than AJCC 7 th edition (entire cohort: 0.727, validation cohort: 0.732) and AJCC TNM staging system (entire cohort: 0.730, validation cohort: 0.742). The calibration plot for the 1-, 3-, or 5-year probability of survival showed an optimal agreement between the prediction by the nomogram and actual observation in both the entire cohort and validation cohort.
Our study has several limitations that may affect the results to some extent. Firstly, the nomogram was constructed based on data from the SEER databases, which lacked the record of HCC etiology (e.g. viral status), liver function index, Child-Pugh class or severity of portal hypertension and performance status. Therefore, the predictive accuracy of our nomogram failed to be compared with Barcelona Clinic Liver Cancer (BCLC) scoring system [27]. Secondly, large amount of data was unclear. This represented a strong limitation to conclusion on survival and the established nomogram. Thirdly, it is well-known that HBV suppression with nucleos(t)ide analogues and HCV eradication with direct-acting antivirals (DAAs) or with pegylated interferon have a significant impact on hepatic decompensation and ultimately on survival of patients with HCC. For the same reasons, another critical point is the lack of data on HCC treatment received by patients, that could significantly affect survival. Fourthly, although PSM was applied to minimize the potential bias, the retrospective nature of this study made it hard to avoid other biases caused by some confounding factors. A further validation from prospective or randomized control trials were required. Finally, this analysis was performed based on a database and validated internally. Therefore, externally validations with more patients from other institutions or study groups were warranted.

Conclusions
Our study proposed that HCC patients with advanced fibrosis/cirrhosis had a poor outcome than those with lower degrees of fibrosis. We established a reliable and superior nomogram based on fibrosis score and other independent risk factors to predict the prognosis of patients with HCC.

Ethics approval and consent to participate
Not applicable.

Consent for publication
Not applicable.