Potential conflict of interest: Nothing to report.
After hepatitis B e antigen (HBeAg) seroconversion, hepatitis B may become inactive or progress to HBeAg-negative hepatitis with persistent or intermittent alanine aminotransferase (ALT) elevation. The aim of this study was to prospectively identify factors predictive of the clinical course in HBeAg-negative chronic hepatitis B (CHB). Patients were stratified by ALT and HBeAg status and followed every 3 months for up to 5 years. Kaplan-Meier and Cox regression analysis using the change from normal ALT to elevated ALT as endpoints were performed to determine factors associated with ALT elevation/normalization. Seventy-four HBeAg-negative and 32 HBeAg-positive patients were prospectively evaluated. For HBeAg-negative patients, hepatitis B virus (HBV) DNA was predictive of future ALT. Only 1 patient with normal ALT and an HBV DNA value lower than 10,000 copies/mL developed an elevated ALT within the subsequent year, whereas 67% with an HBV DNA value greater than 100,000 copies/mL had a rise in ALT above normal within 1 year. Patients with a previous history of ALT elevation and longer follow-up at all levels of HBV DNA were more likely to experience ALT elevations. For HBeAg-negative patients with elevated ALT and all HBeAg-positive patients, HBV DNA did not predict future ALT. Other viral and host factors were not predictive of future ALT. Conclusion: HBeAg-negative CHB has a fluctuating course. HBV DNA values lower than 10,000 copies/mL predict persistently normal ALT for at least 1 year. Patients with HBV DNA values between 10,000 and 100,000 copies/mL can safely be followed at 6 monthly intervals, whereas HBV DNA values greater than 100,000 copies/mL are highly predictive of future ALT elevation and should prompt regular follow-up. (HEPATOLOGY 2007.)
During the natural history of chronic hepatitis B virus (HBV) infection, hepatitis B e antigen (HBeAg) seroconversion is usually characterized by a marked reduction in viral replication, normalization of serum aminotransferases, and long-term remission of hepatic inflammation on liver biopsy.1, 2 When this pattern of disease is sustained over a long term, patients are often described as having inactive HBV and have a very good prognosis.3, 4 On the basis of this observation, HBeAg seroconversion became the goal of treatment trials and was believed to define the end of active disease from HBV infection.
However, in the late 1980s, reports from the Mediterranean documented active viral replication associated with ongoing hepatic inflammation despite persistent loss of HBeAg.5, 6 Further investigation revealed that the majority of such patients had a G-to-A mutation at position 1896 in the precore (PC) open reading frame (ORF).7 This PC mutation results in a premature stop codon at codon 28, thus preventing the production of HBeAg without affecting the core-coding region. Subsequently, mutations further upstream in the basal core promoter (BCP) region were identified, the most common of which is the doublet T-for-A substitution at position 1762 and A-for-G substitution at nucleotide 1764.8 Both result in a selective down-regulation of PC messenger RNA leading to a transcriptional decrease in HBeAg production without affecting core protein expression or HBV DNA replication.9 In addition to chronic ongoing active hepatitis, cases of fulminant hepatic failure have been described with these various mutants.10–12 Although initially thought to be rare, HBeAg-negative chronic hepatitis B (CHB) is being increasingly recognized worldwide.13, 14
CHB often follows a fluctuating course, with periods of active hepatitis interspersed with quiescent disease. However, identifying active hepatitis may be challenging because the majority of flares are clinically silent.15 Hence, close follow-up is necessary to recognize disease activity in order to institute and/or adjust antiviral therapy. In contrast, truly inactive individuals have persistently quiescent disease with an excellent long-term prognosis.3 Unfortunately, recognizing that a given patient is inactive rather than simply having temporarily quiescent HBeAg-negative CHB requires close follow-up of all hepatitis B carriers.
We prospectively followed a cohort of HBeAg-negative patients every 3 months for up to 5 years to determine if any factors were predictive of future disease activity or defined the inactive state. In addition to serial measurements of HBV DNA, HBeAg, and alanine aminotransferase (ALT), we evaluated the importance of HBV genotype, HBV PC, and BCP mutations and demographic and historical data. The HBeAg-negative patients were compared to a group of HBeAg-positive patients to contrast the differing natural histories of these 2 stages of HBV infection.
ALT, alanine aminotransferase; BCP, basic core promoter; CHB, chronic hepatitis B; HBeAg, hepatitis B e antigen; HBsAg, hepatitis B surface antigen; HBV, hepatitis B virus; NS, not significant; ORF, open reading frame; PC, precore; PCR, polymerase chain reaction; ULN, upper limit of normal.
Patients and Methods
All patients visiting the Toronto Western Hospital Liver Clinic between January 2001 and August 2003 with positive serum hepatitis B surface antigen (HBsAg) for at least 6 months were approached for entry into the study. Both new patients and patients who were already being followed in the clinic were offered inclusion into the study. Patients coinfected with hepatitis C virus, hepatitis delta virus, human immunodeficiency virus, and other known causes of liver disease were excluded, as were patients receiving treatment for HBV. The study was approved by the institutional review board (the University Health Network Committee for Research on Human Subjects), and informed written consent was obtained from each patient.
Three groups of patients were evaluated. Group I included patients who were HBeAg-negative with elevated ALT (>40 IU/L) at entry, and group II included HBeAg-negative patients with normal ALT (≤40 IU/L) at baseline. To compare the differences between HBeAg-positive and HBeAg-negative HBV infections, group III included all HBeAg-positive patients. Recruitment continued until 35 patients were enrolled in each group. Patients were censored at the initiation of antiviral therapy for HBV.
Patients were followed every 3 months for up to 5 years. At each visit, ALT, HBeAg, and HBV DNA were measured. HBV DNA isolated from serum from the initial study visit was also used for genotyping and mutational analysis. Demographic, historical, and liver biopsy information were collected for all patients. Cirrhosis was defined by the usual histological criteria or on the basis of clinical evidence of portal hypertension (thrombocytopenia, splenomegaly, or esophageal varices on endoscopy).
HBsAg and HBeAg were measured with standard enzyme-linked immunosorbent assays (Abbott Diagnostics). Patients were screened for hepatitis C virus, hepatitis delta virus, and human immunodeficiency virus antibodies by a third-generation enzyme-linked immunosorbent assay (Abbott Diagnostics). Serum was stored at −20°C until it was analyzed. Because we were primarily interested in defining the level of DNA at which active hepatitis occurs, we chose the Roche polymerase chain reaction (PCR) assay with an accurate range of 500-200,000 copies/mL (Amplicor HBV monitor test, Roche Diagnostic Systems). For some (199/276, or 72%), but not all patients with high levels of serum HBV DNA, serial dilutions were performed for accurate quantification.
DNA extraction from serum samples was performed with the Qiagen blood and tissue kit per the manufacturer's recommendations. Nested PCR for the nearly complete genome of HBV was conducted as described.16 These amplicons were used for genotype determination.
In addition, for the sequencing of the core promoter and PC coding region, a nested PCR targeting the positive strand of the HBV genome and encompassing these regions was used. The first round of PCR used the primer pair 5′-AAGCTCTTACATAAGAGGACTCTTGGACT-3′ and 5′-GATAAGATAGGGGCATTTGGTGGTC-3′; the second round of PCR used the primer pair 5′-ACCGGCCTCGAGGAATACTTCAAAGACTGT-3′ and 5′-TAAGCTGGAGGAGTGCGAATCCACA-3′. Each PCR reaction was performed in a volume of 50 μL containing 5 μL of 10× PCR buffer II (Applied Biosystems), 5 μL of 25 mM MgCl2 (Applied Biosystems), 0.5 μL of AmpliTaq Gold (Applied Biosystems), 1 μL of a 10mM deoxynucleotide triphosphate mix (Pharmacia), and 25 pmol of each primer; each reaction was overlaid with 50 μL of mineral oil. For the second round of the nested PCR, 5 μL of the first reaction was transferred into a PCR mix prepared as described previously, except for an adjustment of the buffer and MgCl2 to take into account the salts transferred with the 5-μL aliquot of the first reaction. PCRs were conducted on a Robocycler 40 thermal cycler (Stratagene) with the following parameters: a first cycle consisting of 95°C for 10 minutes, 55°C for 1 minute, and 72°C for 2 minutes 30 seconds followed by 35 cycles of 95°C for 1 minute, 55°C for 1 minute, and 72°C for 2 minutes 30 seconds.
Amplicons were used directly for automated DNA sequencing performed at the DNA sequencing facility of the Centre for Applied Genomics at the Hospital for Sick Children (Toronto, Canada). Sequences were edited with the programs GeneRunner (version 3.05, Hastings Software) and Genedoc (version 2.3, Nicholas KB 1997; http://www.psc.edu/biomed/genedoc).
Genotyping was performed with the nearly complete length amplicons as a substrate for sequencing with the sequencing primer 5′-GTGGTGGACTTCCAATTTTC-3′, which targets a segment in the surface antigen coding region.17 Sequence segments of 500 base pairs starting 40 nucleotides downstream from the sequencing primer were obtained for each clinical sample. This variable region of the HBV genome has been shown to contain enough genetic information to allow for genotype determination.17 Sequences were aligned with sequences of prototype strains of known genotypes with the program ClustalX for Windows (version 1.8),18 and phylogenetic trees were generated with the program Treeview for Windows (version 1.5.2).19 Genotypes were assigned by the clustering of the sequence with prototype sequences. The GenBank accession numbers for the prototype sequences used were X02763 (genotype A), D00329 (genotype B), X01587 (genotype C), J02203 (genotype D), X75657 (genotype E), X75658 (genotype F), and AF160501 (genotype G).
Mutations in the core promoter and PC coding regions were detected through the sequencing of the amplicons obtained from the positive-strand PCR described previously. The sequencing primer 5′-GCACAGAATAGCTTGCCTGAG-3′ was used. Sequences were edited and compared with prototype sequences with the program Genedoc (version 2.3).
Baseline laboratory and demographic data were compared between groups. The Student t test was used for continuous data, and the χ2 or Fisher's exact test was used for categorical data. ALT elevation was defined as a change from normal ALT (ALT ≤ 40 IU/L) to elevated ALT (ALT > 40 IU/L), and normalization was defined as a change from elevated ALT (ALT > 40 IU/L) to normal ALT (ALT ≤ 40 IU/L) from one visit to the next. The Cox proportional hazards model was used to identify factors predictive of future ALT elevation or normalization. To further evaluate more significant changes in ALT, the analysis was repeated with flares of ALT to 3 times the upper limit of normal (ULN). To account for the fact that patients entered the analysis repeatedly, a covariate for the number of previous ALT elevations or normalizations was included in all models. Any events found to be significant by univariate analysis were evaluated in a multivariable regression model. With ALT elevation and normalization as endpoints, Kaplan-Meier curves were constructed for various levels of serum HBV DNA. Because the entry point into the study was arbitrary in terms of the course of disease, all ALT and HBV DNA data were treated equally. If patients reached an endpoint (elevation or normalization), they were followed until they reached a second endpoint. For example, if a patient with normal ALT had an ALT elevation, he or she was then followed until ALT normalization occurred. Similarly, after ALT normalization, this patient would again be followed for subsequent ALT elevation. The time to event for both Kaplan-Meier and Cox regression analyses was counted on the basis of the time at which the patient became at risk for elevation or normalization. If a patient had an ALT elevation or had an elevated ALT at the baseline, he or she would not be considered at risk for an ALT elevation unless there was a subsequent ALT normalization (that is, ALT had to normalize in order for a second elevation to occur). The survival time was counted as the interval between the time at which the patient became at risk (for example, ALT normalization) and the time of the event (for example, ALT elevation). Consequently, the number at risk below the Kaplan-Meier curves reflects the duration of persistently normal or elevated ALT rather than the duration of clinical follow-up. All analyses were performed on both HBeAg-negative and HBeAg-positive patients.
A total of 124 HBsAg-positive patients were recruited for this prospective study. Of these, 18 patients were excluded: 4 refused to continue participation (1 HBeAg-positive and 3 HBeAg-negative), 1 died of multifocal hepatocellular carcinoma (HBeAg-positive), 5 started antiviral therapy after the initial study visit (2 HBeAg-positive and 3 HBeAg-negative), and 8 did not return for follow-up beyond their entry visit (8 HBeAg-negative). Consequently, there were 37 HBeAg-negative patients with an elevated ALT at the baseline (>40 IU), 37 HBeAg-negative patients with normal baseline ALT (≤40 IU/L), and 32 HBeAg-positive patients (10 with normal ALT and 22 with elevated ALT).
There was a male predominance in all 3 groups, and the majority of the patients were Asian. A greater proportion of HBeAg-negative patients were Caucasian (P < 0.05), and HBeAg-negative patients were older than HBeAg-positive patients (P < 0.0001). Genotypes B and C were more common in all groups. Genotype D was more common among HBeAg-negative patients than HBeAg-positive patients. Seven patients (5 HBeAg-negative and 2 HBeAg-positive) were not genotyped because of consistently undetectable HBV DNA (Table 1). Patients were followed for a median of 2.99 years (range: 0.5-4.9 years).
Table 1. Baseline Characteristics of the Patients
P < 0.05 for the comparison between all HBeAg-negative patients and all HBeAg-positive patients.
P < 0.05 for the comparison between HBeAg-negative patients with high ALT and HBeAg-negative patients with normal ALT at the baseline.
Classical G1896A PC mutations creating a stop codon at codon 28 were seen in 32 (30.2%) patients. Eight additional patients (7 HBeAg-negative and 1 HBeAg-positive) had nonclassical mutations in the PC region expected to abrogate HBeAg production, such as stop codons at other positions or insertions/deletions changing the ORF. PC mutations were more common in HBeAg-negative patients than in HBeAg-positive patients (P < 0.05) but did not differ between HBeAg-negative patients with elevated or normal baseline ALT (Table 1).
BCP mutations were seen in 45 (42.4%) patients but were no more common among HBeAg-negative patients than HBeAg-positive patients [P = not significant (NS)]. Approximately 20% of HBeAg-negative patients carried the virus with both PC and BCP mutations versus only 6.25% of the HBeAg-positive patients.
Sequencing was done directly from amplicons without subcloning in order to identify the consensus sequence. Presumably, HBeAg-positive patients harboring mutations expected to block HBeAg production had mixed populations of virions. In support of this, 13 of 14 HBeAg-positive patients with PC and/or BCP mutations had HBV DNA above 200,000 copies/mL at all visits, making them more likely to have mixed viral populations.
PC mutations were much more common in patients with genotypes D (71.4%) and B (52.9%) than in those with genotypes A (20%) and C (21.6%; P < 0.05); none of the PC mutations in genotype A isolates consisted of a stop codon at codon 28 (G1896A). The opposite pattern was seen with BCP mutations, with higher rates in genotype C (72.9%) and A (50%) infections than in genotype B (20.6%) and D (35.7%) infections. Of the 7 patients with genotype B and a BCP mutation, 6 also had a PC mutation. Likewise, 7 of the 8 patients with genotype C with a PC mutation had a coexistent BCP mutation.
The viral load was not predictive of the presence of PC or BCP mutations. Of the 16 HBeAg-negative patients with HBV DNA levels greater than 200,000 copies/mL throughout the study, 8 (50%) harbored BCP mutations, 7 (43.8%) had PC mutations, and 3 (18.8%) had both mutations present (P = NS). Notably, 3 HBeAg-negative patients with a high viral load had neither PC nor BCP mutations. The breakdown was similar for the 40 HBeAg-negative patients with HBV DNA levels persistently below 200,000 copies/mL: 19 (47.5%) had BCP mutations, 17 (42.5%) had PC mutations, and 10 (25%) had both mutations (P = NS).
A total of 61 (57.5%) patients underwent a liver biopsy within 1 year of study entry or during the study, 18 of whom had cirrhosis. Six other patients had radiographic evidence of cirrhosis but did not have a liver biopsy. Of the 24 patients with cirrhosis, 22 were HBeAg-negative (10 had normal ALT and 12 had elevated ALT), 10 (41.6%) had PC mutations, 14 (58.3%) had BCP mutations, and 7 (29.2%) had both mutations present. Both PC and BCP mutations were more common in patients with cirrhosis, but this was not statistically significant. Division by genotype was similar between patients with and without cirrhosis. In the patients without cirrhosis, fibrosis varied: F0, 9 patients (14.8%); F1, 16 patients (26.2%); F2, 14 patients (23.0%); and F3, 4 patients (6.6%). Eight of the 10 patients with cirrhosis and normal ALT at baseline experienced an ALT elevation, and 5 were promptly started on therapy. One patient refused treatment, and the other 2 had mild ALT elevations (<50 IU/L) with normalization at the next visit. Of those with elevated ALT at the baseline, 9 of 12 underwent a liver biopsy during the study and were started on therapy after the diagnosis of cirrhosis with elevated ALT was confirmed.
For the 48 documented ALT elevations, the mean ALT was 98.1 IU/L with a range of 41-674 IU/L. Ten patients had an initial ALT rise to greater than 3 times ULN, and an additional 6 patients had a peak ALT above this level during follow-up. More patients likely would have reached this level of ALT elevation; however, patients were censored at the time of starting antiviral therapy. By logistic regression, an increase in baseline ALT and baseline HBV DNA above all thresholds was associated with reaching a peak ALT greater than 3 times ULN (Table 2).
Table 2. Logistic Regression for Factors Associated with the Peak ALT Elevation to Greater than 3 Times ULN
Odds Ratio (95% Confidence Interval)
HBV DNA (log copies/mL)
HBV DNA and ALT
Although higher HBV DNA levels were generally associated with higher ALT values in HBeAg-negative patients, the correlation was poor (r2 = 0.072), and no absolute threshold of HBV DNA was identified that correlated with an ALT elevation (Fig. 1). There was no correlation between ALT and HBV DNA levels in HBeAg-positive patients. A total of 14 HBeAg-negative patients with an HBV DNA level below 30,000 copies/mL had a corresponding elevated serum ALT. In 5 of these patients, this discrepancy signified a rising HBV DNA titer with values above 100,000 copies/mL at the next visit. Two patients had declining HBV DNA levels, with the previous value above 100,000 copies/mL and subsequent low HBV DNA and normal ALT. In the remaining 7 patients, the discrepant pattern persisted, and 5 underwent a liver biopsy, which revealed mild active hepatitis B in 3 patients and nonalcoholic fatty liver disease in 2. Similarly, 12 HBeAg-negative patients had HBV DNA levels greater than 200,000 copies/mL with a corresponding ALT value in the normal range. Again, the discrepant findings were most commonly related to rising or falling viral titers, so that in 5 of these 12 patients, an ALT elevation with high HBV DNA was noted at the next visit, whereas 4 subsequently had low HBV DNA levels (<50,000 copies/mL) with persistence of normal ALT. In the other 3 patients, this pattern of normal ALT and high HBV DNA persisted; however, a liver biopsy was not performed, given the normal aminotransferase values.
A total of 24 HBeAg-negative patients had persistently normal ALT throughout follow-up. HBV DNA for these patients ranged from undetectable to 2 × 107 copies/mL, including 8 (33.3%) patients with HBV DNA values greater than 10,000 copies/mL, 7 (29.2%) with HBV DNA greater than 50,000 copies/mL, and 2 with HBV DNA greater than 200,000 copies/mL. Nine HBeAg-negative patients had an elevated ALT throughout follow-up (0.3-3.7 years) with corresponding HBV DNA levels ranging from 0-5 × 107 copies/mL. Five of these patients were started on therapy, and 4 underwent a liver biopsy documenting nonalcoholic steatohepatitis in 2 and very mild HBV-induced hepatitis in the other 2. In contrast, 7 HBeAg-positive patients had persistently normal ALT values, and all had HBV DNA levels above 1 × 107 copies/mL, which demonstrated their immunotolerant state. Nine HBeAg-positive patients had persistently elevated ALT values with HBV DNA levels ranging from 2 × 105 to 6 × 109 copies/mL. Five patients were started on therapy, and 4 had mild liver biopsy findings leading to a watchful-waiting approach.
To evaluate whether HBV DNA at a given time was predictive of future liver disease, Kaplan-Meier curves were constructed for various threshold levels of HBV DNA. Because the study entry date was arbitrary, all study visits were evaluated equally. The time interval from a visit with a normal ALT to a visit with an elevated ALT was used for Kaplan-Meier and Cox regression analysis. A similar analysis was performed for changes from elevated to normal ALT values. Previous literature identified threshold values of HBV DNA ranging from 10,000 copies/mL to the commonly accepted value of 100,000 copies/mL as correlating with active liver disease. Therefore, we evaluated the chance of elevation or normalization at threshold values of HBV DNA of 10,000, 30,000, 50,000, 100,000, and 200,000 copies/mL (Figs. 2–5 and Table 3).
Table 3. Likelihood of Future ALT Elevation/Normalization at Given HBV DNA Threshold Values for HBeAg-Negative and HBeAg-Positive Patients
HBV DNA Threshold (copies/mL)
Percentage with ALT Elevation Within Next Year
Percentage with ALT Normalization Within Next Year
P < 0.05 for the comparison of the likelihood of elevation/normalization at HBV DNA values above and below the given threshold.
For HBeAg-negative patients with a normal ALT, the corresponding HBV DNA level was helpful in predicting the likelihood of future ALT elevation, particularly in the first year of follow-up. Only 1 patient (2.9%) with a normal ALT and HBV DNA below 10,000 copies/mL developed an elevated ALT during the next year of follow-up, whereas 43.0% of patients with HBV DNA greater than 10,000 copies/mL with a normal ALT experienced an ALT elevation during this period (P = 0.0003). However, with longer follow-up, although patients with higher HBV DNA were still more likely to develop an elevated ALT, an increasing proportion of those with low HBV DNA also had a rise in ALT. By the end of the follow-up, 77.6% of those with HBV DNA above 10,000 copies/mL and normal ALT at baseline had a rise in ALT, but 37.6% of those with HBV DNA below 10,000 copies/mL had also experienced an ALT elevation (P = 0.0003); highlighting the importance of continued follow-up even in patients with low HBV DNA at baseline (Fig. 2A). HBV DNA threshold values of 30,000 or 50,000 copies/mL were less discriminating than a value of 10,000 copies/mL. At 1 year of follow-up, more patients with normal ALT and HBV DNA levels above either threshold developed an ALT elevation than those with HBV DNA below the threshold (P = 0.039); however, with longer follow-up, the discriminating value of 30,000 or 50,000 copies/mL disappeared (P = 0.093; Fig. 2B,C). A threshold value of 100,000 copies/mL was able to discriminate between those who would experience an elevation of ALT and those who would not, particularly in the first year of follow-up; however, the data are limited because only a small number of patients had an HBV DNA level above 100,000 copies/mL with a corresponding normal ALT. A threshold value of 200,000 copies/mL did not discriminate well for future ALT elevations.
HBeAg-Negative Patients with Elevated ALT.
The HBV DNA level was not helpful in predicting future ALT normalization. Patients with elevated ALT were equally likely to normalize their ALT whether their corresponding HBV DNA value was above or below all HBV DNA thresholds tested. With longer follow-up, the majority of patients did normalize ALT; however, HBV DNA was not predictive of this event (Fig. 3 and Table 3).
All HBeAg Positive Patients.
HBV DNA was not predictive of future ALT for HBeAg-positive patients, regardless of baseline ALT (Figs. 4 and 5 and Table 3). Notably, however, HBeAg-positive patients with normal ALT were more likely to develop an ALT elevation if HBV DNA levels were below, rather than above, thresholds of 100,000 or 200,000 copies/mL (Table 3). Four HBeAg-positive patients lost HBeAg during follow-up. Three had seroconversion flares with a mean ALT elevation of 168 (122-254) and an associated decline in HBV DNA. With longer follow-up, more patients with ALT normalization and declining HBV DNA may have ultimately reached this endpoint.
Risk Factors for ALT Elevation and Normalization During Follow-Up.
To evaluate if factors other than HBV DNA were predictive of ALT elevation or normalization, Cox proportional hazards regression was performed. A total of 33 (40.7%) patients had a rise in ALT of the 81 who had at least 1 visit after a normal ALT during follow-up. Six patients had more than 1 ALT elevation, including 2 with 4 separate elevations during follow-up. Higher baseline HBV DNA and ALT and a history of a rise in ALT were predictive of future ALT elevation. Other clinical and virologic predictors were not of predictive value (Table 4). Regression results differed significantly between HBeAg-negative and HBeAg-positive patients.
Table 4. Cox Proportional Hazards Regression for Factors Associated with the ALT Elevation
Multivariable Cox proportional hazards regression with hazard ratios for the baseline ALT and previous ALT elevations adjusted for an HBV DNA threshold of 10,000 copies/mL, the most predictive threshold value. The hazard ratios for the HBV DNA thresholds are adjusted for the baseline ALT and previous ALT elevations.
Log HBV DNA is adjusted for the baseline ALT and previous ALT elevations in the multivariable model.
Cox regression examining only HBeAg-negative patients revealed that HBV DNA is more important for predicting future ALT elevation in this subgroup than for the cohort as a whole. HBV DNA above each threshold and as a continuous variable was predictive of a future rise in ALT. Similar to the group as a whole, higher baseline ALT and the number of previous ALT elevations were both predictive of future ALT elevation (Table 4). By multivariable analysis, an HBV DNA level above 10,000 copies and the number of prior ALT elevations were significant. At higher threshold levels, ALT became more important. Similar to the Kaplan-Meier data, the regression results suggest that those with HBV DNA below 10,000 copies/mL have quiescent disease, but once above this threshold, prediction becomes more difficult, and more regular follow-up is required. Patients with dynamic ALT values are likely to continue to have a fluctuating course, and HBV DNA and ALT values trending upward likely herald a future rise in ALT.
For HBeAg-positive patients, no factors were predictive of future ALT elevation (Table 4). Similar to the results from the Kaplan-Meier curves, by Cox regression there was a trend toward lower HBV DNA levels predicting ALT elevations, although this was not statistically significant. This may suggest that as immunotolerant HBeAg-positive patients enter the immune clearance phase, HBV DNA falls prior to ALT elevation.
Looking at ALT normalization, older age and a history of previous ALT normalization were predictive of future ALT normalization in HBeAg-negative patients. Only older age was predictive of ALT normalization in HBeAg-positive patients (Table 5). HBV DNA levels were not predictive of ALT normalization in either group of patients.
Table 5. Cox Proportional Hazards Regression for Factors Associated with the ALT Normalization
Multivariable Cox proportional hazards regression with hazard ratios for each factor adjusted for the other covariates significant according to univariate analysis.
Number of previous ALT normalization
Number of previous ALT normalization
HBeAg-negative hepatitis B follows a fluctuating course making prediction of the outcome following HBeAg seroconversion very difficult. If patients are presumed to be inactive and are followed only intermittently, asymptomatic flares may be missed, and this potentially allows the silent progression of liver disease. Conversely, if all patients are followed closely, resources will be used to follow a large number of patients with quiescent disease. This study provides data on the prediction of future active liver disease in HBeAg-negative patients and demonstrates the need for serial testing of HBV DNA to determine the appropriate follow-up and the need for antiviral therapy. The fluctuating levels suggest that decisions should not be made based on a single measurement.
Many previous reports have attempted to define an HBV DNA threshold that corresponds to the presence of active liver disease. The most widely quoted value comes from the National Institutes of Health workshop on the management of HBV, at which it was arbitrarily proposed that an HBV DNA level of 100,000 copies/mL be used to distinguish active HBV infection from inactive HBV infection.20 Manesis et al.21 subsequently suggested that a more conservative value of 30,000 copies/mL would be more accurate to define the inactive carrier state; however, a cutoff of 100,000 copies/mL misclassified only 1.6% more patients than a threshold of 30,000 copies/mL in their study (P = NS). Most recently, guidelines have been proposed using even lower thresholds of 10,000 copies/mL, and they have been followed by an update using 2000 IU/L to define inactive disease.22, 23 Using similar methods, other groups have looked at cross-sectional or retrospective longitudinal data to try to correlate the HBV DNA level with active liver disease. No definite threshold value has been identified, but most groups suggest that if a threshold exists, it lies somewhere between 10,000 and 100,000 copies/mL.24–27
Similarly, in this prospective study, we found that it was not possible to identify a single value of HBV DNA above which active hepatitis was present and below which disease was quiescent. Given the complex host-virus interactions, it is not surprising that a precise viral load threshold cannot be identified that is accurate in all patients at all times.25 Rather than try to define CHB by a specific level of HBV DNA, we chose to look at the predictive value of HBV DNA and other factors for future disease activity.
In HBeAg-negative patients with normal ALT, the most important factor predictive of future quiescent disease was the corresponding HBV DNA titer. Neither host factors (age, gender, and ethnicity) nor viral factors (genotype and presence of PC or BCP mutations) were helpful in predicting future ALT. A low HBV DNA titer was particularly helpful for defining the early course of future disease, with only 1 patient (3.3%) with HBV DNA below 10,000 copies/mL developing active disease in the subsequent year of follow-up. An HBV DNA threshold of 10,000 copies/mL was more discriminating for predicting future ALT than higher cutoff values. With a modest increase to 30,000 copies/mL, 18.7% of those below the threshold had an ALT elevation by 1 year. However, with longer follow-up, an increasing number of patients at every threshold of HBV DNA experienced ALT elevations. Even among patients with HBV DNA below 10,000 copies/mL, 38% had a rise in ALT by 4 years of follow-up. This highlights the very dynamic nature of HBeAg-negative CHB and reinforces the need for prolonged, albeit intermittent, virologic and biochemical follow-up in all patients.
Similarly, the prediction of future active disease was also associated with the corresponding HBV DNA titer, the baseline ALT, and a history of disease activity. In HBeAg-negative patients with normal ALT, if the HBV DNA was above 100,000 copies/mL, 67% developed an ALT elevation during the next year. Hence, such patients require frequent follow-up. Given the high likelihood of a future ALT elevation, it could be argued that such patients should be considered for liver biopsy and even antiviral therapy despite normal liver enzymes. For HBeAg-negative patients with normal ALT and HBV DNA from 10,000-100,000 copies/mL, the prediction of the future outcome was more difficult. By 6 months of follow-up, only 10.7% of patients with HBV DNA values in this range had a rise in ALT versus almost 20% by 1 year, and this suggests that 6 months may be a reasonable interval for screening such individuals. Within this range, thresholds of 30,000 and 50,000 copies/mL were of similar value for the prediction of future ALT elevation. The strength of our data is that each ALT value and HBV DNA level were treated equally in the analysis, and the prediction rules are based on any given ALT value and its corresponding DNA titer. This method of analysis makes the results more clinically useful because it models what is done in practice, in that decisions about future care and follow-up are based largely on results from the current clinic visit. Hence, if a patient presents with normal ALT and an HBV DNA value from 10-100,000 copies/mL and at the subsequent visit in 6 months has an HBV DNA lower than 10,000 copies/mL, he or she can then be followed on the basis of the second value, that is, annually.
Although the HBV DNA level was predictive of future ALT elevation in HBeAg-negative patients, it was not helpful for the prediction of ALT normalization. In general, HBeAg-negative patients with lower HBV DNA levels were more likely to normalize their ALT at subsequent follow-up; however, this was not statistically significant. Older age was associated with ALT normalization, as was a previous history of ALT normalization. The observation that a history of ALT elevation and/or normalization predicts a dynamic disease course in the future reinforces the need for serial follow-up to determine a patient's pattern of disease. In agreement with previous reports, we found no correlation between HBV DNA and ALT in HBeAg-positive patients.25, 28 In addition, no factors, including HBV DNA, were predictive of future ALT in HBeAg-positive patients; however, some trends may reflect differences in the biology of HBeAg-positive and HBeAg-negative hepatitis. Both ALT elevation and normalization were more common among HBeAg-positive patients with lower HBV DNA values. Patients in the immunotolerant phase of HBV infection typically have high viral levels with normal ALT and normal liver histology.29 As they develop immune recognition of the virus, viral levels fall, an ALT flare may occur, and eventually HBeAg seroconversion ensues. Lower HBV DNA levels associated with elevated ALT may represent immune clearance, whereas the finding of lower DNA predicting ALT normalization likely represents patients that are developing immune control of the virus and nearing HBeAg clearance, a state associated with normal ALT and histology.30 Longer follow-up would be required to document this progression.
We demonstrated that it is possible to genotype patients by PCR and sequencing with a single set of primers. Genotyping is usually performed by restriction site analysis but involves a cumbersome set of enzymes and patterns and may be very sensitive to mutations at the restriction sites. With the advances in automated DNA sequencing, the approach that we used may be more straightforward. Amplicons were sequenced without subcloning. The emergence of HBV quasispecies with a gradual rise to dominance of PC mutants has been documented previously and likely accounts for the 6.25% of HBeAg-positive patients with PC mutations.8, 31 In addition, 34% of HBeAg-positive patients harbored a mutation in the BCP region. Although these patients likely had mixed virus populations, this likely also reflects the fact that relative down-regulation of the PC protein by a mutation in the promoter is not as important as a direct block to translation by the introduction of a stop codon in the gene ORF.8, 32–35 Neither PC nor BCP mutations were associated with more active liver disease or higher HBV DNA levels. Most previous studies have found similar results, although 1 group noted less active disease in patients with PC mutations, whereas another noted higher ALT in patients with BCP mutations.25, 26, 28, 36, 37
There are some important limitations to our study. First, we used ALT as a marker of disease activity and performed biopsies only if clinically indicated. Although it would be ideal to have liver biopsy data to truly identify active liver disease, repeated biopsies in such a short time would be unacceptable. Most patients with high ALT and low HBV DNA levels underwent a liver biopsy, and this allowed us to make accurate determinations of their degree of liver disease and its association with hepatitis B. Because we used HBV DNA to predict future ALT and therefore the need for close follow-up, we chose to use a low threshold for ALT elevation. It is preferable that some patients with inactive disease are followed too closely than the alternative of missing patients with persistently mildly active disease. This is particularly important given recent data from large studies in Hong Kong and Korea showing that patients with mild ALT elevations and even those with ALT values in the normal range (0.5-1.0 times ULN) are at increased risk for subsequent long-term complications in comparison with those with ALT values less than 20 IU/L.38, 39 Whether antiviral therapy will prevent complications in this setting is still unknown, and so before treatment is instituted in patients with normal or mild elevations in ALT, serial monitoring with a possible liver biopsy may be required to better assess the extent of liver disease.
In summary, although a definitive threshold of HBV DNA above which liver disease activity is universal was not identified, we found that HBV DNA serves as a useful predictor of future ALT elevation in HBeAg-negative patients with normal ALT at the baseline. The pattern of disease of the given individual is also important for the prediction of future disease activity; however, this can be discerned only over time. Our data suggest that HBeAg-negative patients with normal ALT and HBV DNA lower than 10,000 copies/mL can safely be followed annually or possibly up to every 2 years; however, it is important to continue to follow such patients over the long term because in this study, albeit from a tertiary referral center, 38% followed for up to 4 years were subsequently observed to have a rise in ALT. For patients with normal ALT and HBV DNA between 10,000 and 100,000 copies/mL, 6 monthly visits will identify the vast majority of ALT elevations. Patients with normal ALT and HBV DNA above 100,000 copies/mL are at high risk for a flare of hepatitis and should be followed closely with consideration of liver biopsy and possibly antiviral therapy. Other viral and host factors are not predictive of future disease in HBeAg-negative patients, and no factors, including HBV DNA, are predictive of future outcome in HBeAg-positive HBV infection.
We thank Tamara Arenovich for her help and guidance with the statistical analysis.