Jian-Wu Yu, Department of infectious diseases, Second Affiliated Hospital, Harbin Medical University, 246 Xuefu Road, Nangang District, Harbin 150086, China. Email: email@example.com
Background and Aims: Chronic hepatitis B virus (HBV) infection is a major global health issue, and the prognosis of patients with HBV-associated acute-on-chronic hepatic failure (ACLF) is extremely poor. In this study, the efficacy of lamivudine was investigated in patients with ACLF. The effects of HBV DNA load and its related factors on the prognosis were also further explored.
Methods: A matched retrospective cohort study using data on ACLF patients derived from our hospital database was conducted. One hundred and thirty patients receiving lamivudine were selected into the lamivudine treatment group with another 130 without lamivudine treatment studied as control. They were matched for sex, age and imaging finding with the lamivudine treatment group. All the patients were followed up for 3 months and the survival rates were compared. The influential factors on the mortality were studied by the Cox proportional hazards model.
Results: The cumulative survival rates of patients in the lamivudine group were higher than those of the control group (χ2 = 9.50, P = 0.0021). The mortality of patients in the high virus load group (71/95, 74.7%) was higher than that of those in the low virus load group (15/29, 51.7%) (χ2 = 5.536, P = 0.019). For patients with a Model for End-Stage Liver Disease (MELD) score of 20–30 by week 4, the mortality of those with HBV DNA that was undetectable or declined for more than 2 log10 (2/12, 16.7%; 18/40, 45.0%) was lower than that of those with a less than 2 log10 decline (18/23, 78.3%) (χ2 = 10.106, P = 0.001). In the Cox proportional hazards model, for patients with a MELD score of 20–30, treatment method (P = 0.002), pretreatment HBV DNA load (P = 0.007) and decline of HBV DNA load during therapy (P = 0.003) were independent predictors; for those with a MELD score of above 30, MELD score (P = 0.008) was the only independent predictor.
Conclusion: Lamivudine can significantly decrease the 3-month mortality of patients with a MELD score of 20–30, and a low pretreatment viral load and rapid decline of HBV DNA load are good predictors for the outcome of the treatment.
Acute-on-chronic liver failure (ACLF) is a clinical syndrome where the major liver functions, particularly detoxification, synthetic functions and metabolic regulation, are impaired to different degrees, and may result in life-threatening complications such as hepatic encephalopathy, ascites, jaundice, cholestasis, bleeding and hepatorenal syndrome (HRS).1,2 In China, as a result of the high prevalence of hepatitis B virus (HBV), chronic HBV infection is the most common cause of liver failure. Chronic HBV infection can lead to hepatic failure with a mortality of up to 90%. The poor prognosis of untreated patients with ACLF is partly related to the severity of the disease (high Model for End-Stage Liver Disease [MELD] score) and the presence of active viral replication (high HBV DNA level).3 The precise mechanisms of liver injury from ACLF caused by HBV infection and factors contributing to the progression of liver failure remain unknown. Viral factors are emphasized in the pathogenesis of HBV-associated severe hepatitis, which has been demonstrated by the efficacy of antiviral therapy using nucleoside analogs.4 Early antiviral treatment attenuates the clinical and biochemical impairment can lead to a fast healing and promote complete recovery.
Lamivudine, an L-nucleoside analog, at a daily dose of 100 mg is effective in suppressing HBV DNA with alanine aminotransferase (ALT) normalization and histological improvement in both hepatitis B e-antigen (HBeAg)-positive and HBeAg-negative patients.5 Continuous treatment with lamivudine can delay clinical progression in patients with chronic hepatitis B and advanced fibrosis or cirrhosis by significantly reducing the incidence of hepatic decompensation and the risk of hepatocellular carcinoma.6,7 Lamivudine may prevent the progression of severe hepatitis B to liver failure by decreasing HBV DNA load and reducing inflammatory reaction and improving liver function.8,9 Previous studies about lamivudine therapy in hepatic decompensation during acute exacerbation of chronic hepatitis B in Hong Kong and Taiwan have found that patients with low platelet counts and high serum bilirubin carry the highest risk of mortality.10,11
Data regarding the mortality rate and efficacy of lamivudine in the subgroup of patients with ACLF induced by hepatitis B infection are poor. A logical hypothesis is that rapid reduction in the HBV DNA levels through the use of lamivudine can result in a less intense host response against the HBV and decrease the mortality of these patients. Yu et al.12,13 have reported that MELD score is related to the prognosis of the patients with acute-on-chronic hepatitis.
In this study, we used the MELD scoring system to predict the 3-month prognosis of patients with ACLF after lamivudine treatment and studied the predictive factors. To collect more convincing evidences of lamivudine treatment on the survival of patients with ACLF, a well-designed matched retrospective cohort study method was utilized to control the bias of patient selection between the two treatments.
Selection of patients
Acute-on-chronic hepatitis B liver failure, as defined by the APASL Working Party, is acute hepatic insult manifesting as jaundice (serum bilirubin ≥ 5 mg/dL) and coagulopathy (international normalized ratio [INR]≥ 1.5 or prothrombin activity < 40%), complicated within 4 weeks by ascites and/or encephalopathy in a patient previously diagnosed or undiagnosed chronic liver disease.14 Other inclusion criteria included: (i) the presence of hepatitis B surface antigen (HBsAg) in the serum for at least 6 months; (ii) evidence of active viral replication as documented by measurable HBV DNA in the serum (≥ 1 × 104 copies/mL); and (iii) flare of hepatitis, defined as serum ALT more than five times the upper limit of normal. The exclusion criteria included the following: (i) superinfection or co-infection with hepatitis A, C, D, E, Epstein–Barr virus, cytomegalovirus and HIV; (ii) a previous course of any antiviral, immunomodulator or cytotoxic/immunosuppressive therapy for chronic hepatitis within at least the preceding 12 months; (iii) evidence of decompensated liver disease before enrolment; (iv) hepatocellular carcinoma diagnosed by ultrasonography or computed tomography; (v) coexistence of any other serious medical illness and other liver diseases such as autoimmune hepatitis, alcoholic liver disease, drug hepatitis or Wilson's disease; and (vi) the malignant jaundice induced by obstructive jaundice and hemolytic jaundice and prolonged prothrombin time induced by blood system disease.
To compare lamivudine therapy with the classical therapy methods, a matched retrospective cohort study using data on ACLF patients derived from our database was conducted. The database included 780 ACLF patients who had received lamivudine treatment or not from January 2001 to December 2008 in the Department of Infectious Diseases, Second Affiliated Hospital, Harbin Medical University, China.
One hundred and thirty consecutive hospitalized patients with acute-on-chronic hepatitis B liver failure who fulfilled the inclusion and exclusion criteria were enrolled in this study. There were 117 patients with chronic hepatitis and 13 patients with compensated liver cirrhosis (Child–Pugh score < 6). Patients were treated with lamivudine (GlaxoSmithKline, Brentford, UK) 100 mg daily. Lamivudine treatment was started on day 1 to day 5 on admission (median day 3).
One hundred and thirty patients did not receive lamivudine in the historical control cohort selected from our database. With SAS ver. 8.2 software (SAS Institute, Cary, NC, USA), patients in the control group were matched for sex, age and imaging finding (cirrhosis or not) with the lamivudine treatment group. The match ratio was 1:1. There were 117 patients with chronic hepatitis and 13 patients with compensated liver cirrhosis. All patients of the historical control group met the aforementioned inclusion and exclusion criteria.
The protocol of our study conformed to the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the Clinical Research Ethics Committee of the Harbin Medical University. All the patients or their relatives in the lamivudine treatment group gave written informed consent before enrolment.
The level of serum creatinine, INR for prothrombin time and the level of serum total bilirubin of each ACLF patient on admission were recorded. The MELD score was calculated according to the original formula proposed by the Mayo Clinic group: 3.8 × loge (bilirubin [mg/dL]) + 11.2 × loge (INR) + 9.6 × loge (creatinine [mg/dL]) + 6.4 × (etiology: 0 if cholestatic or alcoholic, 1 otherwise).
Measurement of HBeAg and HBV DNA and HBV genotyping
Hepatitis B e-antigen and antibody to HBeAg were detected by a qualitative HBeAg enzyme immunoassay (Abbott Laboratories, Chicago, IL, USA). Serum HBV DNA was measured by polymerase chain reaction (PCR) assay (Amplicor HBV Monitor Test; Roche Diagnostics, Mannheim, Germany). The detection limit was 1 × 103 copies/mL. HBV DNA levels of the patients were evaluated before and after the 4-week treatment. HBV genotyping was determined by PCR using type-specific primers.
The tyrosine, methionine, aspartate, aspartate (YMDD) motif mutant was detected by PCR assay and restriction fragment length polymorphism (RFLP) assay at baseline and 3-month follow up. With the use of PCR and RFLP assay for mixed viral-genotype populations (wild-type and mutant virus), the lower limit of detection for differentiating between the two viral genotypes has been determined to be 5% of the viral population. The assay has a lower limit of detection of approximately 1 × 104 copies of viral DNA per mL of serum.
The patients were questioned about adverse events. All the adverse events, regardless of their possible association with lamivudine, were recorded.
Analysis of mortality of patients
The 260 patients with ACLF were followed up for at least 3 months. The outcome (recovery, bridging to liver transplantation, or death) of each patient was recorded. For all patients who died, the exact date and cause of death were documented.
Analysis of mortality of different groups was done using the χ2-test. A survival curve method with log–rank test was performed to analyze the influential factors associated with cumulative survival rates of ACLF patients. For all the analyses, P < 0.05 was considered statistically significant.
Cox proportional hazards models were used to estimate the relative risk for 3-month mortality of the patients. Variables included age, sex, treatment method, pretreatment HBV DNA load, HBeAg status, the decline of HBV DNA load during therapy and MELD score. All analyses were performed using SPSS ver. 10.0 statistical software package (SPSS, Chicago, IL, USA).
Clinical and biochemical parameters and MELD scores of the patients between lamivudine treatment and control groups
The MELD scores of all the patients were over 20. They were divided into two groups according to the MELD score: 20–30; and over 30. The baseline characteristics of the treated and control groups are summarized in Table 1. These two groups were matched by age, sex and imaging finding (cirrhosis or not) and there were no significant differences in other baseline clinical and virological characteristics.
Table 1. Comparison of clinical, biochemical and virological characteristics of patients in the lamivudine treatment and control groups
Total bilirubin (µmol/L)
MELD scores 20–30
44.5 ± 3.6
428.5 ± 79.8
410.8 ± 50.6
70.5 ± 18.7
2.2 ± 0.4
25.1 ± 4.8
45.2 ± 3.7
451.5 ± 80.8
405.6 ± 58.5
73.7 ± 19.2
2.1 ± 0.6
24.9 ± 4.2
MELD scores > 30
44.1 ± 3.8
538.4 ± 88.7
511.7 ± 69.6
119.7 ± 31.7
3.6 ± 0.3
37.5 ± 3.6
44.4 ± 3.7
549.7 ± 85.3
502.6 ± 57.6
118.2 ± 32.3
3.5 ± 0.4
37.0 ± 3.9
HBV DNA (copies/mL)
HBV genotype B/C
Compensated liver cirrhosis
Platelet count × 109/mL
≥ 1 × 105
< 1 × 105
Compared with the control group, P > 0.05.
ALT, alanine aminotransferase; HBeAg, hepatitis B e antigen; HBV, hepatitis B virus; INR, international normalized ratio; MELD, Model for End-Stage Liver Disease.
MELD scores 20–30
143 ± 24
142 ± 32
MELD scores > 30
139 ± 34
138 ± 37
Mortality in patients with and without lamivudine treatment
Nine patients who bridged to liver transplantation were excluded from the analysis. During the 3-month follow up, 195 patients died. The causes of death were all related to liver disease (Table 2). The mortality (50.7%, 38/75) of the patients in the lamivudine treatment group with a MELD score of 20–30 was lower than that (75.7%, 56/74) of the control group (χ2 = 10.033, P = 0.002). The mortality of patients with a MELD score higher than 30 was 98.0% (48/49) in the lamivudine treatment group and 100.0% (53/53) in the control group, showing no significant difference between the two groups (χ2 = 1.092, P = 0.296).
Table 2. Comparison of the mortality of patients in lamivudine treatment and control group
Causes of death
HRS, hepatorenal syndrome; MELD, Model for end-stage liver disease.
MELD scores 20–30
MELD scores > 30
Effect of HBeAg status and pretreatment HBV DNA load on the mortality of the patients in the lamivudine treatment group
There was no significant difference in mortality between HBeAg-positive patients (58/84, 69.0%) and HBeAg-negative patients (28/40, 70.0%) (χ2 = 0.012, P = 0.914). Patients in lamivudine treatment group were divided into: high virus load group (HBV DNA ≥ 1 × 105 copies/mL) and low virus load group (HBV DNA < 1 × 105 copies/mL) according to the pretreatment HBV DNA level. The mortality of patients in the high virus load group (71/95, 74.7%) was higher than that of those in the low virus load group (15/29, 51.7%) (χ2 = 5.536, P = 0.019). A similar result was seen in HBeAg-positive patients (high virus load group 51/69, 73.9% vs low virus load group 7/15, 46.7%; χ2 = 4.280, P = 0.039). For HBeAg-negative patients, there was no significant difference in mortality between the high virus load group (20/26, 76.9%) and low virus load group (8/14, 57.1%) (χ2 = 1.695, P = 0.193) (Table 3).
Table 3. Effect of pretreatment HBV DNA load on the mortality of patients in lamivudine treatment group
Pretreatment HBV DNA load
χ2 = 4.280, P = 0.039;
χ2 = 1.695, P = 0.193;
χ2 = 5.536, P = 0.019.
HBeAg, hepatitis B e antigen; HBV, hepatitis B virus.
Effect of the decline in HBV DNA load on the mortality of the patients in the lamivudine treatment group
The relationship between the decline of HBV DNA load and the mortality of patients was discussed (Table 4). For patients with a MELD score of 20–30, by week 4, the mortality of those with HBV DNA that was undetectable or declined for more than 2 log10 (2/12, 16.7%; 18/40, 45.0%) was lower than that of those with a less than 2 log10 decline (18/23, 78.3%) (χ2 = 10.106, P = 0.001). A similar result was seen in HBeAg-positive patients (HBV DNA undetectable or declined for more than 2 log10 1/9, 11.1%; 11/25, 44.0% vs less than 2 log10 decline 13/16, 81.3%; χ2 = 9.191, P = 0.002). For HBeAg-negative patients, no significant difference was seen in mortality (χ2 = 3.365, P = 0.339).
Table 4. Relationship of the decline of HBV DNA load and the mortality of patients in lamivudine treatment group by week 4
HBV DNA load decline (log10 copies/mL)
MELD score 20–30, χ2 = 9.191, P = 0.002;
χ2 = 3.365, P = 0.339;
χ2 = 10.106, P = 0.001;
MELD score > 30, †χ2 = 1.664, P = 0.197;
χ2 = 0.843, P = 0.284;
χ2 = 1.758, P = 0.185.
HBeAg, hepatitis B e antigen; HBV, hepatitis B virus; MELD, Model for End-Stage Liver Disease.
For the patients with a MELD score higher than 30, by week 4 there was no significant difference in mortality between the HBV DNA undetectable group (5/5, 100.0%), more than 2 log10 decline group (26/26, 100.0%) and less than 2 log10 decline group (17/18, 94.4%) (χ2 = 1.758, P = 0.185). Similar results were seen in HBeAg-positive patients (χ2 = 1.664, P = 0.197) and HBeAg-negative patients (χ2 = 0.843, P = 0.284).
Prognostic factors associated with 3-month mortality in Cox proportional hazards models
In Cox proportional hazards models, MELD score (P = 0.017), treatment method (P = 0.009), pretreatment HBV DNA load (P = 0.006) and the decline of HBV DNA load during therapy (P = 0.013) were independent predictors of 3-month mortality in all patients (Table 5). Of them all, treatment method (P = 0.002), pretreatment HBV DNA load (P = 0.007) and decline of HBV DNA load during therapy (P = 0.003) were independent predictors of 3-month mortality in patients with a MELD score of 20–30. Conversely, MELD score (P = 0.008) was the only independent predictor of 3-month mortality in patients with a MELD score over 30.
Table 5. Prognostic factors associated with 3-month mortality in multivariate Cox proportional hazards models
Regression coefficient (B)
95% confidence interval
HBV, hepatitis B virus; MELD, Model for End-Stage Liver Disease.
Pretreatment HBV load
Rapid decline of viral load
MELD score 20–30
Pretreatment HBV load
Rapid decline of viral load
MELD score > 30
Pretreatment HBV load
Rapid decline of viral load
Effects of treatment method, pretreatment HBV DNA load and the decline of HBV DNA load during therapy on 3-month survival
The effects of treatment method, pretreatment HBV DNA load and the decline of HBV DNA load during therapy on 3-month survival are shown in Figure 1.
The cumulative survival rates of patients in the lamivudine group (n = 124) were higher than those of the patients in the control group (n = 127) (χ2 = 9.50, P = 0.0021). A similar result was seen in patients with a MELD score of 20–30 (lamivudine group, n = 75; control group, n = 74) (χ2 = 8.85, P = 0.0029). For those with a MELD score over 30, there was no significant difference (lamivudine group, n = 49; control group, n = 53) (χ2 = 0.16, P = 0.6898).
The cumulative survival rates of patients in the high pretreatment HBV DNA load group (n = 197) were lower than those of patients in the low pretreatment HBV DNA load group (n = 54) (χ2 = 32.74, P < 0.001). A similar result was seen in patients with a MELD score of 20–30 (high HBV DNA load group, n = 106; low HBV DNA load group, n = 43) (χ2 = 16.20, P = 0.001). For patients with a MELD score over 30, there was no significant difference (high HBV DNA load group, n = 91; low HBV DNA load group, n = 11) (χ2 = 0.92, P = 0.3375).
The cumulative survival rates of patients in the HBV DNA load ‘rapid-decline’ group (n = 172) were higher than those of patients in the ‘slow-decline’ group (n = 79) (χ2 = 3.99, P = 0.0471). A similar result was seen in patients with a MELD score of 20–30 (‘rapid-decline’ group, n = 105; ‘slow-decline’ group, n = 44) (χ2 = 5.79, P = 0.0161). For patients with a MELD score over 30, there was no significant difference (‘rapid-decline’ group, n = 67; ‘slow-decline’ group, n = 35) (χ2 = 1.38, P = 0.2395).
HBeAg seroconversion and YMDD mutations in lamivudine treatment group
At follow-up week 12, there were 56 survivors (38 in the lamivudine treatment group, 18 in the control group). In the lamivudine treatment group there were 12 HBeAg-negative patients, and seroconversion of HBeAg occurred in 11 of 26 patients, in which one patient lost hepatitis B HBsAg; whereas in the control group there were nine HBeAg-negative patients, and seroconversion of HBeAg occurred in two of nine patients, and none of the patients lost HBsAg. No patients showed evidence of YMDD mutations at baseline. No clinical evidence of drug-resistant mutants was detected during the 3-month lamivudine treatment in the survivors.
No serious adverse event that could be attributed to lamivudine occurred, and all the patients tolerated the therapy without dose modification or early discontinuation. No pancreatitis, neuropathy or renal impairment occurred in these patients.
Acute-on-chronic hepatic failure is a serious condition with varied etiology and manifestations, as well as high mortality. Among the infectious etiologies, HBV infection is one of the major causes of ACLF in Asia. The pathogenesis of ACLF caused by HBV remains incompletely understood. A ‘two-hit’ hypothesis may be proposed to explain the pathogenesis of acute-on-chronic hepatitis B liver failure. The first hit is considered to be a primary injury caused directly or indirectly by HBV, and the other is a cytokine-cored secondary lesion. HBV replication is one of the key factors causing the progression of severe liver damage to liver failure. Long-term follow-up studies have demonstrated the close relationship between disease severity and viral factors.15 Early antiviral treatment shortens and improves the symptomatic phase of infection and allows a ready clinical and biochemical improvement.
Lamivudine, an oral cytosine nucleoside analog clinically used for the treatment of chronic HBV infection, which can produce marked viral suppression, reduction of hepatic necroinflammatory activity, histological improvement of liver fibrosis16 and improved liver function,17 even in patients with decompensation.18 Lamivudine may be useful in treating patients with fulminant hepatic failure due to exacerbation of chronic hepatitis B.11,19 However, the experience with lamivudine for the treatment of patients with ACLF induced by HBV is limited. Wang et al.20 conducted a large retrospective study of 1036 patients with HBV-associated hepatic failure which demonstrated that the percentage of patients that recovered or had improved outcomes was significantly higher in those who received lamivudine therapy compared with those who did not. They also found that the outcome would be better if the patients were treated early with nucleoside analog.
Our study showed that lamivudine treatment significantly decreased the mortality of patients with a MELD score of 20–30, but had no effect on patients with a MELD score of more than 30. We found that in the group with a MELD score of 20–30, factors including treatment method, pretreatment HBV DNA load and the decline of HBV DNA load during therapy may influence mortality. For patients with MELD scores above 30, MELD score was the only independent predictor. Patients with MELD scores over 30 were at the late stage of ACLF, and factors other than viral replication had a great impact on the prognosis. The livers of these patients had already undergone massive or sub-massive hepatic necrosis. Suppressing viral replication with lamivudine at this late stage was unlikely to be effective, as the main determinants for recovery were liver regeneration and the rapid cessation of ongoing necroinflammation. This could be the reason why some patients' conditions deteriorated even though the replication of HBV had been suppressed by lamivudine.
Our study has suggested that the prognosis of patients with ACLF may be related to the pretreatment HBV DNA load and the decline of HBV DNA load during therapy. We found that the mortality of the patients with high HBV DNA load was higher than that of patients with low HBV DNA load, which may be due to high HBV DNA load patients failing to eradicate HBV at an early stage of liver failure, and a continuously stimulated immune system clearing HBV causes progressive liver damage. Liver failure in patients with low HBV DNA level may be due to the excess of immune reaction. Our study also suggested that for patients with a MELD of score 20–30 (at the early and middle stage of liver failure), by week 4, the mortality with a HBV DNA load decline of more than 2 log10 was lower than that of a less than 2 log10 decline. The decreased mortality may be related to the marked reduction of HBV DNA level by lamivudine relieving inflammatory reaction and improving liver function. Therefore, for the patients with a MELD score of 20–30, an early and effective antiviral therapy based on the combination therapy (including artificial liver support system and liver transplantation) could achieve a better therapeutic outcome. More potent antiviral drugs such as entecavir and tenofovir are now available. It is conceivable that these drugs might be even better, especially in reaching a rapid decrease in viral load and a faster recovery, for the patients with a MELD score of 20–30 who can not achieve a 2 log10 HBV DNA decline.
This study has proved that there is no significant difference in mortality between HBeAg-positive and -negative patients treated with lamivudine. HBeAg status before treatment has little effect on mortality in the lamivudine treatment group. HBV DNA load is more valuable than HBeAg status in predicting the prognosis of patients. In this study, we have confirmed that pretreatment HBV DNA load and the decline of HBV DNA load during therapy are not associated with the mortality of HBeAg-negative patients. It may be related to the late stage of chronic hepatitis, severe necrosis and fibrosis of liver, mutation of pre-core and basic core promoter, low HBV DNA load and cases limitation.
Although the prognosis of HBV-associated ACLF is being transformed by the development of antiviral therapy, it should be noticed that HBV is controlled but never eliminated, and drug resistance remains a major issue. Long-term lamivudine treatment is associated with the emergence of lamivudine-resistant mutants, which has occasionally been associated with severe, and even fatal, flares of hepatitis.21 In this study, the patients with lamivudine treatment were followed up for only 12 weeks. We did not detect any clinical evidence of drug-resistant mutants during the 3-month lamivudine treatment in the survivors. The method of PCR-RFLP is incapable of detection of YMDD mutants under 1 × 104 copies/mL, which limits its sensitivity. The adverse effects of YMDD mutations may be overcome by the addition of adefovir dipivoxil.
In conclusion, our study suggests that lamivudine treatment can significantly decrease the mortality of patients with a MELD score of 20–30, but have no effect on patients with a MELD score higher than 30. For ACLF patients with a MELD score of 20–30, a low pretreatment viral load and rapid decline of HBV DNA load are good predictors for the survival of patients with lamivudine treatment. A significant weakness in the present study is the use of a retrospective control cohort leading to a possible sources of bias. Our report suggests the need for a prospective, randomized, double-blind, placebo-controlled trial of lamivudine in patients with ACLF.