Meta-analysis: anti-viral therapy of hepatitis B virus-associated glomerulonephritis


Dr F. Fabrizi, Divisione di Nefrologia e Dialisi, Ospedale Maggiore, IRCCS, Pad. Croff, Via Commenda 15, Milan, 20122, Italy.




Hepatitis B virus-associated glomerulonephritis is an infrequent complication of chronic hepatitis B virus (HBV) with significant morbidity. A causal association between hepatitis B virus infection and the development of glomerulonephritis remains controversial. Also, the optimal therapy is undefined although several approaches have been made.


To evaluate the efficacy and safety of anti-viral therapy (interferon or lamivudine) in HBV-associated glomerulonephritis by a systematic review and meta-analysis of clinical trials.


The primary outcome was clinical response (as a measure of efficacy); the secondary outcomes were drop-out rate (as a measure of tolerability), and virological response. We used the random effects model of DerSimonian and Laird, with heterogeneity, sensitivity and meta-regression analyses.


We identified six clinical trials (84 unique patients); three had controlled design. The overall estimate for proteinuria remission was 65.2% (95% confidence intervals: 52.7–75.9%), Q-test for heterogeneity = 7.731, P = 0.172, I2 = 35.327. The overall estimate for hepatitis B e antigen clearance was 62.0% (95% confidence intervals: 50.5–72.2%). The overall estimate for drop-out rate was 12.7% (95% confidence intervals: 6.4–23.6%). Meta-regression analysis showed a significant link between hepatitis B e antigen clearance and logit rate of proteinuria remission after interferon therapy [coefficient –2.585 (S.E. 1.089), P = 0.017].


Remission of the nephrotic syndrome is accompanied by clearance of HBV replication, supporting the role of the virus in the pathogenesis of the disease.


Glomerulonephritis (GN) is an uncommon but well-described complication of chronic hepatitis B.1, 2 Epidemiological studies have shown that chronic carriage of hepatitis B virus (HBV) in some individuals (particularly children) leads to the development of nephrotic syndrome with a strong male predominance, the commonest histological type being membranous nephropathy.1, 2 The isolation of immune complexes in the kidney suggests that the pathogenesis of the disease may have an immune-complex basis. The pathogenetic mechanisms by which certain individuals with chronic HBV infection develop GN is unknown, and numerous in vitro investigations have supported a role for hepatitis B surface antigen (HBsAg) by various mechanisms.3–10 Although the natural history of the disease tends often to remission with preservation of renal function, there is considerable morbidity and significant mortality.1, 2

Various therapeutic approaches have been used for HBV-associated GN including corticosteroids,11–13 adenine arabinoside,14 thymic estracts,14 acyclovir,15 interferon (IFN),1 and lamivudine.1 Only anti-viral therapy (IFN or lamivudine)1 proved to be effective in clearing HBV antigens and abrogating proteinuria. However, the literature about anti-viral treatment of HBV-associated GN consists predominantly of small clinical trials and multiple case reports, and its efficacy remains poorly established. The primary goal of our study was to synthesize the available evidence on tolerability and efficacy of anti-viral therapy in patients with HBV-associated GN by performing a systematic review of the literature with a meta-analysis of clinical trials.

Patients and methods

Search strategy and data extraction

Electronic searches of the National Library of Medicine's MEDLINE database, Current Contents, Cochrane Library and manual searches of selected specialty journals were performed to identify all pertinent literature. It had been previously demonstrated that an electronic search alone may not be sensitive enough.16 We searched MEDLINE (PubMed and OVID Technologies), EMBASE (OVID Technologies), Current Contents (Institute for Scientific Information) and the Cochrane Library (Update Software). The key words ‘hepatitis B virus’, ‘glomerulonephritis’, ‘interferon’, ‘nephrotic syndrome’, ‘lamivudine’, and their synonyms and related terms, were used. Reference lists from qualitative topic reviews and published clinical trials were also searched. All articles were identified by a search from June 1980 to December 2005. Data extraction was conducted independently by two investigators (F.F. and D.D.) and consensus was achieved for all data. Studies were compared to eliminate duplicate reports for the same patients, which included contact with investigators when necessary. Eligibility and exclusion criteria were prespecified.

Criteria for inclusion

To be included in this systematic review, a clinical trial had to fulfil a set of criteria. It had to be published as a full paper; report the results of primary anti-viral therapy; and use the virological (VR) and clinical (CR) response as a clinical end-point. The decision as to the inclusion or exclusion of clinical trials was not related to results. We included controlled clinical trials (CCTs), case–control and cohort studies that provided data on anti-viral treatment (IFN or lamivudine) of HBV-related GN.

Ineligible studies

Studies were excluded if they reported inadequate data on measures of response. Trials that were only published as abstracts or as interim reports were excluded; letters and review articles were not considered for this analysis. We excluded clinical trials concerning patients who had functioning renal grafts. Trials that involved previously treated patients were excluded.


The primary outcome of interest in this systematic review was CR as a measure of efficacy. CR was defined as disappearance of proteinuria (<0.3 g/day) after anti-viral therapy. Secondary end-point included drop-out rate as a measure of tolerability, and VR. VR was defined as clearance of hepatitis B e antigen (HBeAg) from serum after anti-viral therapy. Sustained CR and VR were defined as remission of proteinuria and HBeAg clearance occurring over the follow-up period (at least 6 months after completion of IFN therapy), respectively. These definitions are standards currently used in the scientific literature.

Statistical methods

Outcomes were analysed on an intention-to-treat basis, i.e. all patients included in these studies were considered for the calculation of the response rate, while patients without the end-point were considered as failures. When not given in the publication, the response rate according to the intention-to-treat method was calculated by the data abstractors (F.F. and V.D.). Quantitative, pooled, summary estimates of the VR, CR and drop-out rate across individual studies were generated using the random-effects model of Der Simonian and Laird.17 The Cochrane Q-test for heterogeneity was performed per each assessed outcome and potential sources of heterogeneity were discussed where appropriate. A value <0.10 was considered indicative of statistically significant heterogeneity.18 In addition, the consistency of effects across studies was evaluated by I2. To explore the potential effect of several patient or trial characteristics on the summary estimates, a meta-regression analysis was performed. The dependent variable was the observed logit event rate from each trial for the outcome of interest (proteinuria remission). Weights were assigned according to the estimated variance of logit event rate. The residual between trial variance was estimated by an unrestricted maximum likelihood method using an iterative procedure.19 The following covariates were included in the meta-regression analysis: age, gender, reference year, type of glomerular disease, continent, type of anti-viral therapy, HBeAg clearance. A sensitivity analysis using a fixed-effect model was also performed to assess the consistency of results. Publication bias was assessed by the Begg and Mazumadar20 adjusted rank-correlation test and by the Egger et al.21 regression asymmetry test for publication bias. The 5% significance level was used for alpha risk. Every estimate was given with its 95% confidence intervals (95% CI). All the statistical analyses were performed by STATA 8.0 (Stata Corporation, College Station, TX, USA).


Literature review

Our electronic and manual searches identified 291 reports; 121 (41.5%) reports were selected for full text review; the complete list of these reports is available on request. There was 100% concordance between reviewers with respect to final inclusion and exclusion of studies reviewed based on the predefined inclusion and exclusion criteria.

Clinical trials

Six clinical trials fulfilled our inclusion criteria.22–27 One report was excluded as it concerned the same series already included in this meta-analysis.28 Shown in Table 1 are some salient demographic and clinical characteristics of subjects enrolled in the included clinical trials. All trials were published in English language from 1991 to 2005. Many studies (4/6 = 67%) were from centres in Asia. The mean age of subject cohorts ranged from 6.2 to 40.1 years. The gender distribution ranged from 75% to 100% male. Based on the data provided in Tables 2 and 3, it appears that all patients showed persistent HBsAg carriage with active replication (HBsAg positive/HBeAg positive) at the beginning of the study; the majority having nephrotic syndrome related to membranous glomerulopathy.

Table 1.   Baseline characteristics of clinical trials included in the analysis
AuthorsCountryPatients, nStudy design
  1. R, retrospective; P, prospective; Co, cohort study; CCT, clinical controlled study.

Lai et al.22Hong Kong5Co, P
Conjeevaram et al.24US15Co, R
Chung et al.25Korea8Co, P
Bhimma et al.26South Africa24CCT
Tang et al.27China10CCT
Table 2.   Baseline characteristics of clinical trials: clinical and histologic data
AuthorsMale, nMean age, yearsMN, nMPGN, n
  1. MN, membranous glomerulonephritis; MPGN, membranoproliferative glomerulonephritis.

Lai et al.22100% (5/5)27.2 ± 6.2100% (5/5)0
Lin2375% (15/20)6.2 ± 2.4100% (20/20)0
Conjeevaram et al.2486% (13/15)3966.6% (10/15)26.6% (4/15)
Chung et al.25100% (8/8)40.125% (2/8)50% (4/8)
Bhimma et al.2689.4% (17/19)8.779.2% (19/24)4.2% (1/24)
Tang et al.2770% (7/10)48.3 ± 12.8100% (10/10)0
Table 3.   Baseline characteristics of clinical trials: clinical and virologic data
AuthorsHBeAg, nCreatinine, clearanceProteinuria, g/dayNephrotic syndrome, n
  1. HBeAg, hepatitis B e antigen; NA, information needed not available.

Lai et al.22100% (5/5)87 (36–130)3.5 ± 1.5100% (5/5)
Lin23100% (20/20)116.4 ± 7.64.1 ± 0.8NA
Conjeevaram et al.24100% (15/15)86 (29–127)7.0 (1.0–11.5)NA
Chung et al.25100% (8/8)105.2 (76–141)5 (1.1–18.1)62.5% (5/8)
Bhimma et al.26100% (24/24)77.8 ± 8.1NA100% (24/24)
Tang et al.2770% (7/10)NA4.9 ± 2.4100% (10/10)

Table 4 provides details of the study design, including dose and duration of anti-viral therapy, and duration of follow-up after anti-viral therapy. Recombinant IFN alpha 2b (r-IFNα 2b) by subcutaneous route was given in most (5/6 = 83.3%) clinical trials,22–26 and lamivudine therapy in one study.27

Table 4.   Design of clinical trials
AuthorsIFN dose, MUIFN duration, monthsFollow-up, monthsAnti-viral therapy, type
  1. * Lamivudine dose and lamivudine duration of therapy.

  2. IFN, interferon; NA, information needed not available.

Lai et al.223 × 3/week412r-IFNα 2b
Lin235 × 3/week1212r-IFNα 2b
Conjeevaram et al.245 × 7/week443.2r-IFNα 2b
Chung et al.253 × 3/week66 (6–16)r-IFNα 2b
Bhimma et al.2610 × 3/week46r-IFNα 2b
Tang et al.27100 mg/day*49.2 ± 16.5*NALamivudine

Data on the efficacy of treatment are summarized in Table 5. Six (7.3%) of 82 patients treated with anti-viral therapy discontinued treatment. The reason for discontinuation was myalgia (n = 1) and high cost of IFN (n = 5).

Table 5.   Outcomes of clinical trials
AuthorsHBsAg clearanceSustained HBsAg clearanceHBeAg clearanceSustained HBeAg clearance
  1. HBsAg, hepatitis B surface antigen; HBeAg, hepatitis B e antigen; NA, information needed not available.

Lai et al.220020% (1/5)20% (1/5)
Lin2360% (12/20)55% (11/20)80% (16/20)80% (16/20)
Conjeevaram et al.2433.3% (5/15)33.3% (5/15)66.6% (10/15)53.3% (8/15)
Chung et al.250062.5% (5/8)37.5% (3/8)
Bhimma et al.260058.3% (14/24)41.6% (10/24)
Tang et al.270NA50% (5/10)NA

Data on baseline HBV DNA levels have not been provided in all clinical trials. The response of HBV DNA to treatment was not considered as end-point of this meta-analysis because the information was incomplete. No information on liver histology was given (Table 6).

Table 6.   Outcomes of clinical trials
AuthorsProteinuria, remissionProteinuria, Sustained remissionDrop- out rate
  1. NA, information needed not available.

Lai et al.2280% (4/5)20% (2/5)0
Lin23100% (20/20)100% (20/20)0
Conjeevaram et al.2466.6% (10/15)53.3% (8/15)0
Chung et al.2537.5% (3/8)25% (2/8)20% (1/5)
Bhimma et al.2662.5% (15/24)41.6% (10/24)20.8% (5/24)
Tang et al.2770% (7/10)NA0

Primary analysis

The overall estimate for HBsAg clearance was 16.4% (95% CI: 4.7–44.1%), Q-test for heterogeneity 16.787, P = 0.005, I2 70.375. The overall estimate for drop-out rate was 12.7% (95% CI: 6.4–23.6%). Q-test for heterogeneity 4.743, P = 0.448, I2 0.00. The overall estimate for HBeAg clearance was 62.0% (95% CI: 50.5–72.2%). Q-test for heterogeneity 6.263, P = 0.281, I2 20.168. As shown in Table 7, the overall estimate for proteinuria remission was 65.2% (95% CI: 52.7–75.9%), Q-test for heterogeneity 7.731, P = 0.172, I2 35.327.

Table 7.   Proteinuria remission after anti-viral therapy: overall estimate and 95% CI (random effects model)
AuthorsEstimateWeights95% CI
Lai et al.220.85.630.309; 0.973
Lin230.9763.440.713; 0.999
Conjeevaram et al.240.66723.430.406; 0.854
Chung et al.250.37513.210.125; 0.715
Bhimma et al.260.62539.520.422; 0.792
Tang et al.270.714.780.376; 0.900
Overall point estimate0.668100.000.499; 0.802

Sensitivity analysis

We made a sensitivity analysis by including only clinical trials concerning IFN administration. The overall estimate for sustained HBsAg clearance was 19.8% (95% CI: 5.9–49.1%), Q-test for heterogeneity 12.346, P = 0.015, I2 67.6. The overall estimate for sustained HBeAg clearance was 50.6% (95% CI: 31.7–69.4%), Q-test for heterogeneity 8.850, P = 0.065, I2 54.805. The overall estimate for sustained proteinuria remission was 49.8% (95% CI: 33.3–60.1%), Q-test for heterogeneity was 9.216, P = 0.056, I2 56.596.

As a sensitivity analysis, we calculated effect estimates according to the geographical distribution of clinical trials; the overall estimate for proteinuria remission in reports from Asia was 73.6% (95% CI: 39.5–92.3%). Q-test for heterogeneity 14.223, P = 0.043, I2 43.4. The sensitivity analyses by the fixed-effects model yielded very similar results to the random model (data not shown).

Meta-regression analysis

Meta-regression analysis showed no link between logit rate of proteinuria remission and several covariates (Table 8). The association between logit rate of proteinuria remission and frequency of membranous glomerulopathy approached the statistical significance (P = 0.049). Meta-regression analysis showed a significant link between HBeAg clearance and logit rate of sustained proteinuria remission in the subgroup of clinical trials based on IFN [coefficient –2.585 (S.E. 1.089), P = 0.01767].

Table 8.   Meta-regression analysis (dependent variable: logit-rate of Proteinuria remission)
  1. GNM, membranous glomerulopathy; SN, nephrotic syndrome.

GNM rate2.3241.18200.0492
HBeAg clearance0.641051.400570.647
Reference year−0.0440.0900.6237
SN rate3.16452.15090.1421

Publication bias

Egger and Begg tests for publication bias showed that the risk for having missed trials was acceptable low; for proteinuria remission and HBeAg clearance the P-values ranged from 0.133 to 0.226. The funnel plot by precision and logit-rate of proteinuria remission is shown in Figure 1.

Figure 1.

 Funnel plot of precision by logit event rate.


Our meta-analysis showed that most patients with HBV-related GN were successfully treated by anti-viral therapy- the overall estimate for remission of nephrotic syndrome was more than 60%. In addition, the overall estimate for sustained remission of proteinuria induced by IFN therapy was 50%. This response is higher than that typically reported in patients with chronic hepatitis B and normal renal function. In the general population, IFN therapy has a beneficial effect in 30–40% of patients with chronic hepatitis B who respond by clearing HBeAg from serum.29, 30 The reasons for the high rate of response in patients with GN remain unclear. Some patients with HBV-related GN have chronic renal insufficiency at diagnosis and IFN could help to restore the cell-mediated immunity depressed by uraemia. Another possibility is that patients with extrahepatic manifestations of hepatitis B have an enhanced immunologic responsivity, i.e. they appear to produce very high levels of immune complexes. This greater degree of immunologic reactivity could also predispose these patients to a better response to IFN therapy. Also, the natural history of HBV-associated nephrotic syndrome shows gradual improvement in many cases, particularly in children.31–33 The virological features of HBV (HBV genotype or viral load), and the duration of HBV infection could play also a role.

The overall estimate for drop-out rate was 12.7% (95% CI: 6.4–23.6%). The majority (5/6 = 83%) of the patients requiring early termination of therapy did not show serious adverse events, but stopped anti-viral therapy due to financial reasons. Thus, tolerance to anti-viral therapy was quite satisfactory.

Our meta-analysis presents some shortcomings. As with all meta-analyses, this study has the potential limitation of publication bias. Negative trials are sometimes less likely to be published, and we have tried to obtain data from as many sources as possible. Secondly, the number of patients enrolled in our study was small. HBV is an infrequent cause of GN in developed countries where the HBV rate in the general population is low, but it is likely more frequent in those countries where HBV diffusion is higher. No histological end-points were provided; histological data at the end of treatment or in follow-up are particularly difficult to obtain. Finally, the limited number of clinical trials hampers the development of definitive conclusions by meta-regression.19

Remission of nephrotic syndrome by anti-viral therapy in patients with HBV-associated GN was accompanied by clearance of HBV markers in most cases; our meta-regression demonstrated a significant link between logit rate of sustained proteinuria remission and HBeAg clearance in the subset of IFN-based trials. These findings clearly support a role of HBV in the pathogenesis of this nephropathy. It has been claimed by several authors that HBV and GN may both occur with increased frequency in a certain subset of patients but be pathogenetically unrelated.1, 2, 25 We cannot exclude the possibility that some patients included in our systematic review may have spontaneously seroconverted without anti-viral treatment, but the frequency of proteinuria remission and loss of HBeAg was much higher in study than control groups in the CCTs included in this meta-analysis. Spontaneous clearance of HBV antigens (particularly HBeAg) leading to proteinuria abrogation has been already observed.34 Numerous case-reports have given emphasis on the role of IFN 35, 36 or lamivudine37–40 in remitting nephrotic syndrome with clearance of HBV antigens in HBV-related GN. Resolution of HBV-associated GN in association with lowered circulating HBV surface antigen levels has been noted after orthotopic liver transplantation.41 The impact of HBV vaccination on the incidence of HBV-linked membranous nephropathy has been recently emphasized.42, 43 Thus, our meta-analysis offers an additional piece of evidence supporting the immunopathogenetic activity of HBsAg in selected cases of GN with nephrotic syndrome.

We found by meta-regression a direct relationship between logit rate of proteinuria remission and frequency of membranous nephropathy; this finding needs to be confirmed in larger series.

As reported above, only one clinical trial concerned lamivudine therapy; thus, it is difficult to provide a direct comparison between IFN and lamivudine on efficacy and safety of anti-viral therapy in HBV-related GN. IFN-based therapy has generally a shorter duration whereas Tang et al. have not discontinued lamivudine. Issues on cost and lamivudine resistance need to be evaluated further.

In conclusion, this meta-analysis showed the efficacy and safety of anti-viral agents (IFN or lamivudine) in the treatment of HBV-associated GN, especially membranous glomerulopathy. Remission of nephrotic syndrome was accompanied by clearance of HBV replication supporting the role of the virus in the pathogenesis of the disease. Additional clinical trials aimed to address the optimal therapy of HBV-related GN are in progress.


This study was supported in part by the grant ‘Project Glomerulonephritis’ in memory of Pippo Neglia.