High Levels of Hepatitis B Surface Antigen Increase Risk of Hepatocellular Carcinoma in Patients With Low HBV Load TAI –

*Division of Gastroenterology, Department of Internal Medicine, Buddhist Tzu Chi General Hospital Taipei Branch, Taipei, Taiwan; Division of Gastroenterology, epartment of Internal Medicine; Graduate Institute of Clinical Medicine; Hepatitis Research Center; Department of Medical Research; and Department of icrobiology, National Taiwan University College of Medicine and National Taiwan University Hospital, Taipei, Taiwan; **School of Medicine, Tzu Chi University, ualien, Taiwan; Faculty of Medicine, Dentistry and Health Sciences, University of Melbourne, Melbourne, VIC, Australia


BACKGROUND & AIMS:
Patients with chronic hepatitis B virus (HBV) infection have a high risk for developing hepatocellular carcinoma (HCC).Patients with lower levels of hepatitis B surface antigen (HBsAg) have higher chances of losing HBsAg than those with high levels.However, little is known about whether higher levels of HBsAg increase risk for HCC.METHODS: We followed 2688 Taiwanese HBsAg-positive patients without evidence of cirrhosis for a mean time period of 14.7 years.In addition to the known risk factors of HCC, we investigated the association between levels of HBsAg and development of HCC.RESULTS: Of the patients followed, 191 developed HCC, with an average annual incidence rate of 0.5%.Baseline levels of HBsAg and HBV were associated with development of HCC, and risk increased with level.Compared to HBsAg level, by receiver operating characteristic curve analysis, HBV DNA level better predicted the development of HCC during 10-year and 15-year periods (both, P Ͻ .001).However, when we evaluated hepatitis B e antigenϪnegative patients with levels of HBV DNA Ͻ2000 IU/mL, factors that determined HCC risk included sex, age, and levels of alanine aminotransferase and HBsAg (Ն1000 IU/mL), but not level of HBV DNA.Multivariate analysis showed that the adjusted hazard ratio for HCC in patients with levels of HBsAg Ն1000 IU/mL versus Ͻ1000 IU/mL was 13.7 (95% confidence interval: 4.8Ϫ39.3).CONCLUSIONS: Among patients infected with HBV genotype B or C, determinants of HCC risk include their sex, age, hepatitis B e antigen status, HBV genotype, and levels of alanine aminotransferase and HBV DNA, but not level of HBsAg.Among hepatitis B e antigen؊negative patients with low viral loads, HCC risk is determined by levels of HBsAg and alanine aminotransferase and age, but not HBV DNA.

Watch this article's video abstract and others at http:// tiny.cc/j026c.
Scan the quick response (QR) code to the left with your mobile device to watch this article's video abstract and others.Don't have a QR code reader?Get one at mobiletag.com/en/download.php.
][7] In particular, patients with serum HBV DNA levels Ն2000 IU/mL at study entry have an increased risk of developing HCC. 5 In contrast, those with HBV DNA levels Ͻ2000 IU/mL are usually designated inactive or low-risk HBV carriers. 3,8,9owever, data from longitudinal studies indicated that these subjects still carry an annual incidence rate of 0.06% for HCC development. 5,10Therefore, identification of factors predictive of HCC other than viral load in these low-risk patients remains mandatory and deserves additional studies.
2][13][14][15][16][17] A lower HBsAg level is shown to be associated with a higher chance of HBsAg loss and lower risk of hepatitis activity in patients with HBV genotype B or C infection. 11,17,18n addition, HBsAg level Ͻ1000 IU/mL was found to convincingly define inactive carrier state in Italian patients with HBV genotype D infection. 16Because a lower HBsAg level usually signifies a better prognosis, it is of clinical interest to know whether a higher HBsAg level would be associated with a higher risk of HCC, especially in the special population of lowly viremic patients.
To address this interesting and important issue, we enrolled a large cohort of 2688 treatment-naïve patients who were diagnosed with chronic HBV infection and received long-term follow-up at the National Taiwan University Hospital.The primary aim of our study was to explore whether HBsAg level could complement HBV DNA level as a predictor of HCC development.

Patient Cohort
Figure 1 shows the inclusion and exclusion process of patients in the Elucidation of Risk Factors for Disease Control or Advancement in Taiwanese Hepatitis B Carriers (ERADICATE-B) study.Part of this cohort (patients enrolled from 1985 to 1995) had been used to investigate the issue of HBsAg loss. 17The enrollment time frame was extended to 2000 in this study.In total, 3947 HBsAgpositive patients aged older than 28 years were consecutively enrolled between 1985 and 2000.All of them had been HBsAgpositive for longer than 6 months and received more than 3 years of regular follow-up at the National Taiwan University Hospital.After excluding patients with evidence of hepatitis C virus (HCV) or hepatitis D virus co-infection, and those without adequate serum samples for analysis, 3489 patients remained.We further excluded 411 patients who were diagnosed with cirrhosis at baseline because this is an indication for antiviral therapy in practice guidelines, 19 -21 and 390 patients who received antiviral therapy either before HCC diagnosis or before the end of follow-up because of the possible modification of HCC risk by treatment. 22Finally, a total of 2688 HBV carriers were included into analysis.A subgroup analysis was also performed on hepatitis B e antigen (HBeAg)-positive patients (n ϭ 523) and HBeAg-negative patients (n ϭ 2165), who were divided into high viral load group (1097 with HBV DNA level Ն2000 IU/mL) and low viral load group (1068 with HBV DNA Ͻ2000 IU/mL).All enrolled patients gave informed consent as approved by the National Taiwan University Hospital Ethical Committee.

Data Collection
Patients were tested for serological markers (HBsAg, HBeAg, anti-HBe, antibodies against hepatitis C virus [anti-HCV], and antibodies against hepatitis D virus), and had liver function tests and ␣-fetoprotein levels at baseline.Throughout the follow-up period, if alanine aminotransferase (ALT) levels were within normal limits, liver function tests and ␣-fetoprotein were assayed every 6 months, and at least every 3 months if ALT levels were elevated.Serum samples collected at each visit were stored at Ϫ20°C until analysis.Serum ␣-fetoprotein and abdominal ultrasonography using a high-resolution and real-time scanner were performed for HCC surveillance every 3 to 6 months from enrollment.

Diagnosis of Cirrhosis and HCC
Cirrhosis was diagnosed by histology or ultrasonographic findings, together with clinical features such as thrombocytopenia, gastroesophageal varices, or ascites. 23For the diagnosis of cirrhosis made via abdominal ultrasound, the findings had to be consistent on at least 2 occasions 6 months apart. 6CC was diagnosed either by histology/cytology or by typical image findings (arterial enhancement and venous wash-out by contrast-enhanced computed tomography or magnetic resonance imaging scanning) in hepatic nodules Ͼ1 cm. 24

Quantification of HBV DNA and HBsAg Levels
Serum samples at enrollment were tested for both HBV DNA and HBsAg levels.HBV DNA level was quantified using the Abbott RealTime HBV assay, 0.2 mL protocol (Abbott Laboratories) with a low detection limit of 15 IU/mL.HBsAg level was quantified using the Architect HBsAg QT (Abbott Laboratories) according to manufacturer's instructions. 11,15The detection range of Architect assay is 0.05 to 250 IU/mL.If the HBsAg level was found to be Ͼ250 IU/mL, samples were diluted to 1:100 to 1:1000 to obtain a reading within the calibration curve range.

Extraction of Viral DNA
Viral DNA in the serum was extracted using commercial kits (QIAamp DNA Blood and Tissue Mini Kit; QIAGEN Inc, Valencia, CA, USA).The extracted DNA was used for HBV genotype determination.

Determination of HBV Genotype
HBV genotype was determined by real-time polymerase chain reaction-based single-tube assay as described previously. 25This method consists of 2 consecutive steps.The first step uses polymerase chain reaction to amplify the region (nt 1261Ϫ1600), and the second step uses melting curve analysis to genotype HBV.

Statistical Analysis
Mean and standard deviation (SD) were calculated for continuous variables and percentages were used for categorical variables.Both HBV DNA levels (IU/mL) and HBsAg levels (IU/mL) were logarithmically transformed for Pearson's correlation analysis.HBV DNA levels were assigned as 15 IU/mL for those with undetectable values.
The clinical follow-up started at the time of enrollment.Person-years were censored on the date of identifying HCC, death, the last date of follow-up, or December 31, 2010, whichever came first.The cumulative incidence of HCC by different variables was derived using the Kaplan-Meier curve analysis, and log-rank test was used to test for the statistical difference.
In order to compare the predictive values of different factors for HCC, we restricted the study population to patients who were followed for at least 10 and 15 years.Receiver operating characteristic (ROC) curve analysis was used to compute the area under the ROC curves for different factors.Their performance in predicting 10-year and 15-year HCC was compared.In addition, we evaluated the relationship between HBsAg level and risk of HCC in HBeAg-negative patients with HBV DNA levels Ͻ2000 IU/mL using the restricted cubic spline regression with different number of knots. 26The reference level of HBsAg was 10 IU/mL (1 log10 IU/mL).The best-fitting cubic spline model was determined according to the values of Akaike information criterion. 27ox proportional hazards regression model was used to calculate the crude and multivariate-adjusted hazard ratios (HR) of HCC.Age, sex, serum HBV DNA, HBsAg, and ALT levels were included as adjusting variables as they are known to be associated with HCC development. 5,28 -30tatistical significance of all tests was defined as P Ͻ .05 by 2-tailed tests.All analyses were performed using Stata statistical software (version 10.0; Stata Corp, College Station, TX).
Relationships between levels of HBsAg and HBV DNA were further investigated in patients with HBV DNA levels Ͻ2000 and Ն2000 IU/mL, respectively.A positive correlation did exist between HBsAg and HBV DNA levels in both groups (Supplementary Figure 1).The correlation was better in the higher HBV DNA group (r 2 ϭ 0.44; P Ͻ .001)than the lower HBV DNA group (r 2 ϭ 0.07; P Ͻ .001).

Follow-Up Results
Our study had 39427.2person-years of follow-up, with a mean (ϮSD) follow-up duration of 14.7 Ϯ 4.3 years (median, 13.9 years; range, 2.5-25.8years).Throughout the follow-up period, 191 patients developed HCC, with an incidence rate of 4.8 cases per 1000 person-years.The mean (ϮSD) duration from enrollment to HCC development was 11.0 Ϯ 4.5 years (median, 10.2 years; range, 2.5-24.5 years).In addition, there were 294 patients who developed liver cirrhosis during the follow-up and 151 (79.1%)HCC patients had recognized liver cirrhosis before or when HCC was diagnosed.

Cumulative Incidence of HCC by HBeAg Status, ALT, HBV DNA, and HBsAg Levels and Other Risk Factors
We first correlated the cumulative incidence of HCC with HBeAg status and levels of ALT, HBV DNA, and HBsAg (Figure 2AϪD).We found that HBeAg positivity and higher levels of ALT and HBV DNA were associated with a higher cumulative incidence of HCC.The risk of HCC was shown to increase when HBV DNA level was Ն2000 IU/mL (Table 2).Regarding HBsAg level, a higher HBsAg level was also associated with an increased HCC risk with a dose-response manner (P ϭ .001).In addition, other common risk factors, including older age, male sex, and genotype C infection, were found to be associated with HCC development (Table 2).

Comparing ALT, HBV DNA, and HBsAg Levels for Predicting HCC
Because ALT, HBV DNA, and HBsAg levels were shown to be associated with HCC development, we evaluated which one was a better predictor.The performance of these 3 factors in predicting HCC development was compared by the subcohorts of 10 years of follow-up (n ϭ 2491; 90 developed HCC) and 15 years of follow-up (n ϭ 1219; 154 developed HCC) because the mean time to HCC was 11.0 years and mean follow-up period of this cohort was 14.7 years.In terms of predicting 10-year HCC, the area under the ROC were 0.70 (95% confidence interval [CI]: 0.65-0.75)for HBV DNA level, 0.75 (95% CI: 0.70 -0.79) for ALT level, and 0.58 (95% CI: 0.52-0.64)for HBsAg level.The ROC curve analysis showed that both ALT and HBV DNA levels were superior to HBsAg level in predicting HCC development within 10 years (Figure 2E) and within 15 years (Figure 2F).
In HBeAg-positive patients, we evaluated these 3 factors in predicting 10-year and 15-year HCC development (Supplementary Table 1).Again, both ALT and HBV DNA levels were shown to be superior to HBsAg level in predicting HCC development.
We also evaluated the predictive accuracy of HBV genotype in patients with positive HBeAg or HBV DNA level Ն2000 IU/mL (Supplementary Table 1).The area under the ROC was 0.66 (95% CI: 0.60 -0.72) and 0.63 (95% CI: 0.59 -0.67) for the prediction of 10-year and 15-year HCC risk, respectively, and were both superior to HBsAg level.When patients were stratified by different HBV genotypes, the predictive accuracy of HBsAg level for HCC development did not significantly differ between genotype B and C patients (Supplementary Table 2).

Different Impact of HBsAg Level on HCC Development Between HBeAg-Negative Patients With Low and High Viral Loads
The risk of HCC among patients with HBV DNA level Ͻ2000 IU/mL was comparable at different ranges of viral load (Table 2).We explored whether HBsAg level could be another risk factor for HCC development in HBeAg-negative patients with a low viral load.The relationship between HBsAg level and HCC risk was first evaluated using the restricted cubic spline regression.The shape of the best-fitting regression spline for HBsAg levels, which was based on the smallest Akaike information criterion value (Supplementary Figure 2), is presented in Figure 3A.This model was derived using the restricted cubic spline regression with 3 knots placing on the 25th, 50th, and 75th percentiles of HBsAg distribution.We found that HCC risk appeared to be similar in those with HBsAg levels Ͻ1000 IU/mL, but gradually increased when the level was Ͼ1000 IU/mL.
To validate this finding, we categorized these patients with HBV DNA level Ͻ2000 IU/mL by the following 3 cutoff levels of HBsAg: 10 IU/mL (Figure 3B), 100 IU/mL (Figure 3C), and 1000 IU/mL (Figure 3D), and analyzed the data using the Kaplan-Meier curve analysis.Compatible with the finding in Figure 3A, the cumulative incidence of HCC was different only when patients were categorized by HBsAg level of 1000 IU/mL (P Ͻ .001),but not 10 or 100 IU/mL.The 10-year cumulative incidence of HCC was 0.2% and 2.2% for HBsAg Ͻ1000 and Ն1000 IU/mL, respectively.In contrast, in HBeAg-negative patients with HBV DNA level Ն2000 IU/mL, the HCC risk was not related to serum HBsAg levels (P ϭ .247; Figure 3E).As we aimed to determine the impact of HBsAg level on disease progression, we focused on the subcohort of HBeAg-negative patients with HBV DNA level Ͻ2000 IU/mL.

Factors Affecting HCC Risk in HBeAg-Negative Patients With Low Viral Load
In HBeAg-negative patients with HBV DNA level Ͻ2000 IU/mL, advanced age, male sex, and elevated ALT level, but not HBV DNA level, were found to be independent risk factors for HCC development (Table 3).As for HBsAg level, compared to patients with HBsAg level Ͻ1000 IU/mL, the HR of HCC was 5.4 (95% CI: 2.1-14.2) for patients with HBsAg level Ն1000 IU/mL using univariate analysis.Further multivariate analysis showed that HBsAg level Ն1000 IU/mL remained as an independent risk factor of HCC with an HR of 13.7 (95% CI: 4.8 -39.3).

Relationships Between HCC Risk and Dynamic Changes of HBV DNA, HBsAg, and ALT Levels in HBeAg-Negative Patients With Low Viral Loads at Baseline
In 1068 HBeAg-negative patients with HBV DNA level Ͻ2000 IU/mL, 980 (91.8%) had available stored serum samples at the third year of follow-up for the determination of HBV DNA and HBsAg levels.Compared to patients with persistently low levels of HBV DNA, HBsAg, or ALT, those with persistently high levels of these 3 factors were at a higher risk of HCC using univariate analysis (Table 4).For example, compared to patients with HBsAg level Ͻ1000 IU/mL at baseline and year 3, the HR of HCC was 8.0 (95% CI: 2.2-27.3)for those with HBsAg level Ն1000 IU/mL at baseline and year 3.In addition,  4).

Discussion
For chronic hepatitis B patients, HCC is a devastating complication.It is important to identify risk factors of HCC in clinical practice. 3In this hospital-based cohort study, we demonstrated that advanced age, male sex, elevated ALT level, genotype C, positive HBeAg, and higher HBV DNA level were associated with HCC development over time.The risk of HCC started to increase when HBV DNA level was Ͼ2000 IU/mL.All these results were consistent with earlier studies, 5,28,30 -33 validating the accuracy of our findings in this large hospital-based cohort.Although both levels of HBV DNA and HBsAg were shown to be associated with HCC development, we found that HBV DNA level had better predictive accuracy than HBsAg level when investigating the overall cohort.However, in patients with HBV DNA level Ͻ2000 IU/mL who had a similar risk of HCC, HBsAg level Ն1000 IU/mL was identified as a new independent risk factor for HCC development.These data suggested that HBsAg level might complement HBV DNA level in predicting HCC development, especially in the lowly viremic HBV carriers.
It is known that there is a positive correlation between HBsAg and HBV DNA levels. 12,13,34The correlation has been shown to be higher at HBeAg-positive phase, lower at HBeAg-negative phase, and lowest at the lowly replicative phase, which was consistent with our results.This discrepancy between levels of HBsAg and HBV DNA at the lowly replicative phase might be caused by accumulation of integrated viral envelope sequences in infected hepatocytes.The HBsAg is mainly derived from the integrated form of HBV DNA rather than the episomal form. 13,35In other words, lowly viremic patients who have high HBsAg    level might harbor more hepatocytes with HBV integration than those who have low HBsAg level.Therefore, the higher risk of HCC in former patients can be attributed to the increased genomic instability as a result of integrated viral sequences, which plays an important role in hepatocarcinogenesis. 36 Another possible explanation for the positive correlation between HCC risk and HBsAg level, but not HBV DNA level, is the narrow range in HBV DNA.
In HBV carriers with HBV DNA level Ͻ2000 IU/mL, the dynamic range of HBsAg is wider than HBV DNA level (0.05 to Ͼ10,000 IU/mL vs 15 to 2000 IU/mL).A widerange factor is prone to provide more power to differentiate patients at different risk.
Previous studies have indicated that a lower HBsAg level is associated with better clinical outcomes, including a higher likelihood of HBsAg loss and lower risk of HBeAg-negative hepatitis. 11,17,18In a recent study comparing the prognosis between inactive HBV carriers (viral load Ͻ2000 IU/mL) and non-HBV plus non-HCV controls, the 10-year cumulative incidence rates of HCC were 0.6% and 0.2%, respectively. 10In our study, the 10-year cumulative incidence rate of HCC in patients with HBV DNA level Ͻ2000 IU/mL plus HBsAg level Ͻ1000 IU/mL was 0.2%, which was similar to the controls. 10These data suggested that, in addition to HBV DNA level Ͻ2000 IU/mL, HBsAg level Ͻ1000 IU/mL

CLINICAL LIVER
can be considered to define low-risk patients who are infected with HBV genotype B or C. Of particular note, a recent study on Italian patients with genotype D virus infection also indicated that HBsAg level Ͻ1000 IU/mL is associated with a sustained viral suppression within a 3-year follow-up. 16These findings lend strong support to our hypothesis that a lower HBsAg level can signify adequate host immune control against HBV infection, leading to the decrease of HCC risk over time.Taking these lines of evidence together, HBsAg level Ͻ1000 IU/mL could be considered as a general criterion to define low-risk or inactive HBV carriers infected with different HBV genotypes.However, this speculation needs to be confirmed in HBV patients infected with genotypes other than B, C, and D. The cutoff HBsAg level reported in this study would become clinically useful in different aspects.First, the recently introduced nomogram is based on the results derived from the REVEAL-HBV study, 30 and its usefulness has been confirmed by an external cohort including patients from Korea and Hong Kong. 37Our data further illustrated that HBsAg Ն1000 IU/mL was another key factor associated with HCC development in patients with HBV DNA level Ͻ2000 IU/mL, and this easy-touse marker should be integrated into the nomogram.Second, when treating HBeAg-negative patients with nucleos(t)ide analogues, there still lacks a reliable marker to predict sustained viral suppression after stopping therapy.HBV DNA level is not useful because most patients have an undetectable HBV DNA level at the end of therapy. 38In addition, although clearance of HBsAg is considered as an ideal end point of HBV therapy, it is very difficult to achieve, especially in Asian HBV patients who acquire the infection early in life.Thus practicing physicians urgently require a good indicator to stop nucleos(t)ide analogue therapy in HBeAg-negative patients.If HBsAg level Ͻ1000 IU/mL could be reliably used to define low-risk or inactive HBV carriers, we can adopt this cutoff level as the intermediate treatment goal when its value is validated in future studies.
This study had several unique features.It is well known that HCC has a low occurrence rate, thus a large cohort with a reasonable follow-up period is mandatory to evaluate this rare event.To fulfill this requirement, we adopted a strict inclusion criterion that patients needed to have regular follow-up for at least 3 years.This allowed us to ensure that included patients adhered to our follow-up program, which is especially important, as we used a nonconcurrent prospective cohort.Consequently, we achieved a large hospital-based cohort with a mean follow-up period of 14.7 years.This cohort offered adequate statistical power to address the rare event of HCC development.Second, the ERADICATE-B study, a hospital-based cohort study, has different characteristics from the REVEAL-HBV study, which is a community-based cohort study.Compared to the REVEAL-HBV study, the ERADI-CATE-B study consisted of a higher proportion of patients with elevated ALT level (29% vs 6%) and a lower proportion of patients with HBV DNA levels Ͻ2000 IU/mL (35% vs 56%), just like the patient population in our clinical practice.Of particular note is that although these 2 cohorts have different composition, both reveal the importance of HBV DNA level in predicting HCC risk.These consistent findings validate the impact of HBV DNA on HCC development unequivocally.Third, serial changes in levels of ALT and HBV DNA have been shown to affect HCC development. 39In our lowly viremic cohort, the impact of dynamics of HBV DNA, HBsAg, and ALT levels on HCC was also investigated.We found that in patients with low baseline levels of HBV DNA, HBsAg, or ALT, their HCC risk increased in parallel with increased levels of HBV DNA, HBsAg, or ALT at year 3.These data suggested that dynamic data could improve the predictive accuracy of baseline factors.Taking these lines of evidence together, even for patients traditionally regarded as the low-risk group based on baseline data, a regular follow-up with repeated measurements of these risk factors is required because the risk of HCC can vary over time.
Our study had a few limitations.First, we excluded patients who had cirrhosis at enrollment because these patients require antiviral therapy under current treatment guidelines. 19 -21Second, this study is a hospital-based cohort, and it is inevitable that patients had a higher chance to receive antiviral therapy, which has been shown to decrease HCC development. 22To avoid this interference, we only enrolled patients before 2000, when nucleos(t)ide analogues were not widely available in Taiwan.In addition, we excluded patients who received antiviral therapy either before HCC diagnosis or before the end of follow-up.In other words, this is a homogenous cohort that is free from treatment and ensures that the derived results are not biased.
In summary, in a cohort of 2688 patients infected with HBV genotypes B or C, determinants of HCC risk include their sex, age, HBeAg status, HBV genotype, and levels of ALT and HBV DNA, but not level of HBsAg.Among HBeAg-negative patients with low viral loads, age, baseline levels, and dynamic changes of HBsAg and ALT predict HCC development.Therefore, combination of HBV DNA level Ͻ2000 IU/mL and HBsAg level Ͻ1000 IU/mL can be considered as essential criteria to define the minimal-risk HBV carriers.

Figure 2 .
Figure 2. Cumulative incidence of HCC in a cohort of 2688 HBsAg-positive patients was associated with (A) serum HBeAg status, (B) ALT level, (C) HBV DNA level, and (D) HBsAg level at study entry.A better prediction of serum HBV and ALT levels than HBsAg level for HCC within (E) 10 years of follow-up (n ϭ 2491) and (F) 15 years of follow-up (n ϭ 1219) by ROC curve analysis.

Figure 3 .
Figure 3.In 1068 HBeAg-negative patients with HBV DNA level Ͻ2000 IU/mL.(A) HR of HCC in relation to HBsAg levels was analyzed using the restricted cubic spline regression with 3 knots on the 25 th , 50 th , and 75 th percentiles.Cumulative incidence of HCC was analyzed by cutoff HBsAg levels of (B) 10 IU/mL, (C) 100 IU/mL, and (D) 1000 IU/mL in HBeAg-negative patients with HBV DNA level Ͻ2000 IU/mL and (E) by HBsAg level of 1000 IU/mL in HBeAg-negative patients with HBV DNA levels Ն2000 IU/mL.

Table 1 .
Demographic Data of 2688 HBV Carriers a Only determined in patients with either positive HBeAg or HBV DNA level Ն2000 IU/mL.

Table 2 .
Univariate Analysis of Factors Associated With HCC in 2688 HBV Carriers by Cox Proportional Hazards Regression Model a Only determined in patients with either positive HBeAg or HBV DNA level Ն2000 IU/mL.CLINICAL LIVER

Table 3 .
Univariate and Multivariate Analysis of Factors Associated With HCC in 1068 HBeAg-Negative Patients With HBV DNA Level Ͻ2000 IU/mL by Cox Proportional Hazards Regression Model

Table 4 .
HCC Risk in 980 Patients With Baseline Serum HBV DNA Ͻ2000 IU/mL: Categorized by Serum Levels of HBV DNA, HBsAg, and ALT at Baseline and 3-Year Follow-Up a Adjusting variables include age, sex, and dynamics of HBV DNA, HBsAg, and ALT levels.