Validation of the PAGE-B score to predict hepatocellular carcinoma risk in caucasian chronic hepatitis B patients on treatment

Pinar Gokcen, Fatih Guzelbulut, Gupse Adali, Ayca Gokce Degirmenci Salturk, Oguzhan Ozturk, OzgurBahadir,Emine Kanatsiz,Mevlut Kivak, Kamil Ozdil, Hamdi Levent Doganay

Abstract

Key Words: Chronic hepatitis B; Hepatocellular carcinoma; PAGE-B score; Surveillance

INTRODUCTION

Hepatocellular carcinoma (HCC) is the 7thmost prevalent among all cancers and ranks 4 th in cancer-related mortality[1 ]. Chronic hepatitis B (CHB) virus infection affects 257 million people worldwide and is one of the most common etiologies of HCC,accounting for 33 % of HCC-related mortality[2 ]. Nucleos(t)ide analogs suppress hepatitis B virus (HBV) replication in most patients; however, the risk of HCC persists even in patients with suppressed viral replication. Treatment options for advancedstage HCC are quite limited, and the 5 -year survival rate is 18 .1 %[3 ]. Therefore,identifying patients who are at high risk for HCC development and detecting tumors at early stages are crucial. Recent guidelines recommend HCC surveillance with ultrasound (USG) twice a year in patients who are at high risk for HCC development[4 ]. However, not all patients with CHB have the same risk for HCC. There is an ongoing need for a scoring system to predict HCC risk that offers an easy application in clinical practice and a high predictive value to perform effective surveillance in high-risk patients and eliminate unnecessary surveillance in low-risk patients.

Various risk scores have been developed to identify CHB patients at high risk for HCC development. However, many of the risk scores for HCC have focused on untreated patients, and they are mostly based on pretreatment risk factors for HCC,such as hepatitis B e antigen (HBeAg) status, serum HBV DNA, alanine aminotransferase (ALT), albumin and bilirubin levels[5 ,6 ]. Most of the baseline virological factors are no longer considered risk factors for HCC, as HBV replication is suppressed early in the majority of patients receiving potent antiviral treatment. Therefore, risk scores that include parameters not easily modified by treatment are needed. Among these,the PAGE-B is a practical risk score that includes platelet count, age and sex and has been validated in various patient populations[7 -9 ]. In this study, we aimed to evaluate the accuracy of the PAGE-B score in predicting HCC risk in Turkish CHB patients on tenofovir disoproxil fumarate (TDF) or entecavir (ETV) therapy.

MATERIALS AND METHODS

Patient population and Follow-up Definitions

Medical records of CHB patients who were on follow-up in hepatology outpatient clinics at Umraniye Training and Research Hospital (Istanbul, Turkey) and Haydarpa?a Numune Training and Research Hospital (Istanbul, Turkey) between January 2007 to December 2018 were retrospectively evaluated. The inclusion criteria were as follows: Age ≥ 16 years, HBsAg positivity for ≥ 6 mo, and treatment with TDF or ETV for at least 12 mo. The exclusion criteria were as follows: Age ＜ 16 years,decompensated cirrhosis, having HCC diagnosis before or during the first 6 mo of therapy, history of liver transplantation, and coinfection with hepatitis C virus,hepatitis D virus or human immunodeficiency virus.

Laboratory tests, including HBeAg, anti-HBe, HBV DNA, aspartate aminotransferase, ALT, albumin, bilirubin and alpha-fetoprotein (AFP) levels, international normalized ratio, and complete blood count at the start of therapy and during followup at 3 -6 mo intervals, were recorded. The results of imaging studies,e.g., USG,triphasic computed tomography (CT) and dynamic contrast-enhanced magnetic resonance imaging (MRI), at the start of therapy and during follow-up were recorded.The presence of comorbidities and liver biopsy results, if available, were also recorded.Virological response was defined as a serum HBV DNA level ＜ 80 IU/mL. Maintained virological response was defined as serum HBV DNA negativity without subsequent positivity. A biochemical response was achieved when the serum ALT level dropped below 42 U/L. Hepatic flare was defined as an elevation of ALT ≥ 2 × upper limit of normal with subsequent HBV DNA positivity in patients with virological response.Liver biopsies were evaluated according to the ISHAK staging system[10 ]. Patients with pretreatment fibrosis scores between 0 and 4 (F0 -4 ) were considered noncirrhotic.Patients with fibrosis scores 5 and 6 (F5 -6 ) or those with radiological (nodular appearance of liver surface, parenchymal thickening, caudate lobe enlargement, portal vein diameter ＞ 13 mm) or endoscopic (varices, portal gastropathy) findings of cirrhosis were considered compensated cirrhotic. Decompensated cirrhosis was defined as the presence of ascites, variceal bleeding or hepatic encephalopathy.Patients underwent HCC surveillance with abdominal USG at 6 -12 mo intervals. In the presence of suspicious lesions on USG, cross-sectional imaging with triphasic CT and/or dynamic contrast-enhanced MRI were performed. A diagnosis of HCC was made following the current guidelines[5 ]. The PAGE-B score included the parameters of platelet count, age, and sex. Scoring was performed as follows: (1 ) For age 16 -29 years, 0 points; 30 -39 years, 2 points; 40 -49 years, 4 points; 50 -59 years, 6 points; 60 -69 years, 8 points; ≥ 70 years, 10 points; (2 ) For female gender, 0 points and for male gender, 6 points; and (3 ) For platelet count (/mm3) ≥ 200000 , 0 points; 100000 -199999 , 6 points; ＜ 100000 , 9 points. The score ranged from 0 -25 points. Based on their PAGE-B scores, patients were classified as ≤ 9 , low risk; 10 -17 , moderate risk; and ≥ 18 , high risk[7 ].

Statistical analysis

Statistical data were analyzed using SPSS v.23 .0 (SPSS Inc., Chicago, IL, USA).Descriptive statistics are presented as the mean ± standard error of the mean for continuous variables. Variables were tested for normality using the Kolmogorov-Smirnov test. Independentt-tests were used to compare parametric variables, and chisquared tests, continuity correction, or Fisher’s exact tests were used to compare categorical variables. The cumulative effect of PAGE-B risk groups on survival was assessed using the log-rank test. Survival rates were computed by Kaplan-Meier survival analysis. Accuracy in predicting HCC occurrence was evaluated using a timedependent area under the receiver operating characteristic (AUROC) curve at all study time points. Univariate and multivariate logistic regression analysis models were used to determine the effects of the variables on the risk of developing HCC. Cirrhosis and platelet count were analyzed separately in logistic regression model as they showed collinearity. Tests were interpreted at a 95 % confidence interval. A P value ≤ 0 .05 wasconsidered statistically significant.

Table 1 Demographic and follow-up characteristics of study population

The study was approved by the local ethics committees of Umraniye Training and Research Hospital and Haydarpasa Numune Training and Research Hospital.

高斯（中國(guó)）提出要在鞏固報(bào)紙印刷的基礎(chǔ)上，大力發(fā)展書(shū)刊印刷和商業(yè)印刷，適時(shí)進(jìn)入包裝印刷領(lǐng)域的戰(zhàn)略，并以此展開(kāi)了一場(chǎng)場(chǎng)以“創(chuàng)新”“綠色”“智能化”等為主題的攻堅(jiān)戰(zhàn)。

RESULTS

A total of 742 patients were enrolled in the study. The mean age was 45 .0 ± 0 .5 (17 -93 )years, and 472 (63 .6 %) patients were male. One hundred and sixty-one patients (21 .7 %)had cirrhosis. Of the total patients, 502 (67 .7 %) received TDF, and 240 (32 .3 %) received ETV. One hundred and sixty-two (21 .8 %) of patients were lamivudine-experienced. At month 12 , 597 patients (85 .4 %) achieved virological response, and 620 patients (85 .9 %)achieved ALT normalization. Twenty-five (3 .4 %) patients had hepatic flare. The mean follow-up time was 54 .7 ± 1 .2 (5 -145 ) mo. The demographic and clinical characteristics and follow-up data of the patients are presented in Table 1 .

During the follow-up period, 26 patients (3 .5 %) developed HCC. Patients who developed HCC were older, male predominant, had lower albumin and platelet levels,and had higher AFP levels than those who did not develop HCC (allP＜ 0 .05 ).Cirrhosis and diabetes mellitus were more common in patients who developed HCC than in those who did not develop HCC (bothP＜ 0 .05 ) (Table 2 ).

In the univariable analysis, older age, male sex, lower platelet levels, presence of diabetes mellitus, presence of cirrhosis, absence of ALT normalization at month 6 , and pretreatment AFP levels were associated with HCC development (allP＜ 0 .05 ). HCC was not detected in any patients with hepatic flare. In the multivariable analysis, older age [odd ratio (OR) = 1 .1 ; 95 % confidence interval (CI): 1 .0 -1 .1 ], male sex (OR = 8 .9 ;95 %CI: 1 .1 -70 .7 ), lower platelet levels (OR = 1 .0 ; 95 %CI: 1 .0 -1 .0 ), presence of cirrhosis(OR = 3 .1 ; 95 %CI: 1 .1 -8 .2 ), and absence of ALT normalization at month 6 (OR = 0 .2 ;95 %CI: 0 .1 -0 .7 ) were associated with HCC occurrence (all P ＜ 0 .05 ) (Table 3 ).

Validation of PAGE-B risk score

The mean PAGE-B score was 11 .1 ± 0 .2 . According to the PAGE-B score, 281 (37 .9 %),341 (46 %) and 120 (16 .2 %) patients had low-risk, moderate-risk and high-risk of HCC development, respectively. Nineteen (6 .8 %), 78 (22 .9 %) and 64 (53 .3 %) patients had cirrhosis in the low-, moderate- and high-risk groups, respectively (P＜ 0 .001 ). One(0 .4 %), 7 (2 .1 %) and 18 (15 %) patients developed HCC in the low-, moderate- and high-risk groups, respectively (P＜ 0 .001 ). For a PAGE-B score cutoff value ≤ 9 , the sensitivity, specificity, positive and negative predictive values for the prediction of HCC were 96 .2 %, 39 .1 %, 5 .4 % and 99 .6 %, respectively. The AUROCs of the PAGE-B score in the prediction of HCC risk at 1 , 3 , 5 and 10 years were 0 .977 , 0 .903 , 0 .903 and 0 .865 , respectively (Figure 1 ). The cumulative HCC incidences at 1 , 3 , 5 and 10 years were 0 %, 0 %, 0 % and 0 .4 %, respectively, in the PAGE-B low-risk group; 0 %, 1 .2 %, 1 .5 %and 2 .1 %, respectively, in the PAGE-B moderate-risk group; and 5 .0 %, 11 .7 %, 12 .5 %,and 15 .0 %, respectively, in the PAGE-B high-risk group (log-rank P ＜ 0 .001 ) (Figure 2 ).

DISCUSSION

The ultimate goal of CHB therapy is to extend the survival of patients by preventing progression to cirrhosis, HCC development and the need for transplantation. This objective has been achieved substantially with the widespread use of TDF and ETV,which have a high genetic barrier to resistance. However, the risk of HCC is not eliminated despite effective antiviral drugs. Various studies have aimed to evaluate the risk of HCC development in various populations using risk scores that include clinical or laboratory parameters. The REACH-B (age, sex, HBsAg status, and HBV DNA concentration) score was the first scoring system that did not include cirrhosis as a parameter, and it was associated with a 5 -year HCC incidence of 2 .6 % in the low-risk group[11 ]. The GAG-HCC (age, sex, HBV DNA, core promoter mutations and cirrhosis) and CU-HCC (age, viral load, bilirubin, albumin, cirrhosis) scores have a negative predictive value of 98 .3 % for 5 -year HCC incidence in treatment-naive Asian patients[12 ]. These scoring systems, which were validated in untreated patients, were less predictive when applied to patients on antiviral treatment[13 ]. For example, a high serum HBV DNA level, which is included in REACH-B, is no longer regarded as a risk factor for HCC with the use of potent antivirals[8 ]. Moreover, detection of core promoter mutations that are included in the GAG-HCC scoring system is not alwayspossible. To solve these problems, novel scoring systems were developed in patients on treatment. The PAGE-B (age, sex and platelet count), CAGE-B (age, presence of baseline cirrhosis), SAGE-B (age, liver stiffness measurements), CAMD (cirrhosis, age,male sex, diabetes mellitus) and HCC-RESCUE (age, sex, cirrhosis) scoring systems all have high negative predictive values for HCC development in their low-risk groups[7 ,14 -16 ]. Among current scoring systems, the PAGE-B is the only one that does not include cirrhosis as a parameter. The presence of cirrhosis is the most important risk factor for HCC development, and the annual risk of HCC in patients with cirrhosis is 2 .5 %-4 %[17 ]. Liver biopsy is the gold standard method for the diagnosis of cirrhosis.However, biopsy is associated with certain disadvantages, such as being an invasive method that can lead to potential complications, requiring tissue samples of an appropriate amount and from an appropriate localization, and producing falsenegative results in the early period. Meanwhile, noninvasive methods that assess fibrosis, such as transient elastography, can produce operator-dependent false-positive results. It should also be emphasized that liver biopsy is not performed in patients with lamivudine, adefovir or telbivudine resistance prior to the start of new antiviral agents with a high genetic barrier to resistance. Therefore, these nucleos(t)ide analogexperienced patients may not have cirrhosis at the start of rescue therapy due to the resolution of cirrhosis after years of therapy. Our cohort also included 162 patients(21 .8 %) who had received lamivudine and developed resistance prior to the start of TDF/ETV treatment. Therefore, definitive confirmation of cirrhosis for all patients is impractical. To that point, the PAGE-B score has an advantage compared to the other scoring systems that include cirrhosis as a parameter. Supporting this, implementing ISHAK stage in the PAGE-B score did not improve the prediction of HCC risk[18 ].

9例左冠狀動(dòng)脈異位起源于肺動(dòng)脈的患兒中,6例手術(shù)病人術(shù)前均做了升主動(dòng)脈造影和左心導(dǎo)管檢查,造影能清楚的顯示冠狀動(dòng)脈起源,評(píng)價(jià)左、右冠狀動(dòng)脈的側(cè)枝情況。具體情況見(jiàn)表2。

Table 3 Risk factors associated with the risk of hepatocellular carcinoma development

Figure 1 Receiver operating characteristic curves of the PAGE-B score for hepatocellular carcinoma development according to years. A:1 st year; B: 3 rd year; C: 5 th year; D: 10 th year.

In the present study, the incidence of HCC was determined to be 3 .5 %, and the cumulative HCC incidences at 5 years in the low-, moderate-, and high-risk groups were 0 %, 1 .5 % and 12 .5 %, respectively. These rates were lower than those in the PAGE-B database (0 %, 3 %, 17 %) but higher than those in Spain’s CIBERHEP database(0 %, 2 .8 %, 5 %), which was used for the validation of the PAGE-B score. All three databases (ours, PAGE-B, CIBERHEP) had similar reliability of the PAGE-B score in the prediction of overall HCC development for a cutoff value ≥ 10 [7 ,9 ]. It was thought that the lower HCC rates in CIBERHEP may be related to the relatively low total number of patients and the number of patients who completed the 5 -year follow-up[9 ]. The present study included 267 (36 %) patients who had completed the 5 -year follow-up, which was similar to the PAGE-B database; however, patients were younger on average than the PAGE-B database (45 vs 52 ). This might be a reason for the lower HCC incidence seen in this study. Additionally, the present study included mainly genotype D patients who are known to have less risk for HCC development,while genotypes A, B, D predominated in the PAGE-B database and genotype B and D in the CIBERHEP database.

In this study, we also evaluated 10 -year HCC incidence. Although we did not find any cases of HCC in the low-risk group during the first 5 years of follow-up, one patient developed HCC at month 80 . This 37 -year-old noncirrhotic male patient was treated with ETV, did not have comorbidities, and had a PAGE-B score of 8 . The study by Brouweret al[18 ] showed that the estimated HCC incidences at 10 years in PAGE-B low-, moderate-, and high-risk groups were ＜ 1 .5 %, 1 .5 %-17 .5 % and ≥ 17 .5 %,respectively, supporting our results (0 .4 %, 2 .1 %, 15 %).

Current international guidelines recommend that patients with cirrhosis should undergo HCC surveillance systematically. However, there is no consensus about noncirrhotic CHB patients[19 -22 ]. Diverging from other guidelines, the European Association for the Study of the Liver suggests that only patients with a PAGE-B score≥ 10 in the noncirrhotic group should be included in screening[19 ]. In the present study, unnecessary tests would be prevented in 281 (37 .9 %) patients with low-risk HCC by following this cost-effective approach.

The present study has some limitations. First, it is a retrospective study, so the effect of treatment non-compliance could not be determined precisely. Second, the patient number was relatively low. Third, only one-third of the patients completed the 5 -year follow-up. Fourth, in Turkey, almost all patients are infected with genotype D virus(32 ), so no interpretation could be made for other genotypes.

CONCLUSION

PAGE-B successfully predicted patients who had a low risk for HCC during treatment with genetically high barrier antivirals. Ease of use without the need for biopsy or an impractical molecular test justifies implementing this score in clinical practice.

ARTICLE HIGHLIGHTS

Research objectives

We aimed to evaluate the accuracy of the PAGE-B scoring system in the prediction of HCC risk in CHB patients receiving entecavir (ETV) or tenofovir disoproxil fumarat therapy.

Research methods

We recruited 742 CHB patients who had been treated with tenofovir disoproxil fumarate or ETV for more than 1 year. Risk groups were determined according to the PAGE-B scores. We evaluated the accuracy of the PAGE-B score in predicting HCC.

Research results

HCC was diagnosed in 26 patients (3 .5 %) during 54 .7 ± 1 .2 mo mean follow up. The cumulative HCC incidences at 5 years were 0 % in the PAGE-B low-risk group; 1 .5 %moderate-risk group; and 12 .5 %, in the high-risk group (log-rank p ＜ 0 .001 ). The AUROCs of the PAGE-B score in the prediction of HCC development at 5 years follow up was 0 .903 .

Research conclusions

PAGE-B had successfully predicted the patients who had a low risk of HCC during treatment with genetically high barrier antivirals.

Research perspectives

PAGE-B is a simple score that does not require biopsy or any impractical molecular test. The efficiency of PAGE-B justifies implementing this score in daily clinical practice.