Trials of thymosin treatment in chronic hepatitis B virus infection have been small and the results have been inconsistent.
Trials of thymosin treatment in chronic hepatitis B virus infection have been small and the results have been inconsistent.
To conduct a meta-analysis to evaluate the efficacy of thymosin treatment in chronic hepatitis B virus infection.
Randomized controlled trials comparing thymosin for over 24 weeks vs. placebo (or usual care) in the treatment of chronic hepatitis B virus infection were identified through MEDLINE, EMBASE and the Cochrane Register of Clinical Trials. Biochemical (normalization of transaminases) and virological (loss of hepatitis B virus DNA and hepatitis B e antigen) responses were analysed using the intention-to-treat method. The odds ratio was used to measure the magnitude of the efficacy.
Five trials (353 patients) were identified. The odds ratio (95% confidence interval) of the virological response of thymosin over placebo at the end of treatment, 6 months post-treatment and 12 months post-treatment were 0.56 (0.2–1.52), 1.67 (0.83–3.37) and 2.67 (1.25–5.68), respectively. There was an increasing trend of the virological response with time since the cessation of thymosin treatment (P=0.02). There was no difference in the biochemical response between the thymosin and placebo groups at the end of treatment, 6 months post-treatment and 12 months post-treatment.
Thymosin is effective in suppressing viral replication in chronic hepatitis B virus infection, but the effect is delayed until 12 months after the cessation of treatment.
Patients suffering from chronic hepatitis B virus (HBV) infection generally have a weak cytotoxic T-cell response against the virus. Treatment with interferon-α, an immunomodulating agent, is effective in clearing the virus in up to one-third of chronic HBV-infected patients who have evidence of active viral replication and hepatitis.1 The unsatisfactory efficacy and associated adverse effects have largely limited the use of interferon. Thymosin-α1 is a 28-amino acid polypeptide isolated from thymosin fraction 5, a bovine thymus extract containing a number of immunologically active peptides.2 It is an immunomodulating agent and may enhance the clearance of HBV, although the exact mechanism is not entirely clear. In vitro studies have suggested that thymosin-α1 can influence T-cell maturation and antigen recognition, the stimulation of interferon and cytokine production, and the activity of natural killer cell-mediated cytotoxicity.3, 4 Randomized controlled clinical trials have been conducted to evaluate the efficacy of thymosin-α1 in chronic hepatitis B. These trials were usually small in size and their results were inconsistent. Although it seems that thymosin has fewer and milder adverse effects compared with interferon-α, its clinical efficacy remains inconclusive. Thus, we conducted this meta-analysis of randomized controlled clinical trials to assess the evidence obtained on the efficacy of thymosin treatment in chronic HBV infection.
All English articles were retrieved by using searches of MEDLINE, EMBASE and the Cochrane Controlled Trial Registry.5 Additional studies were identified by scrutiny of the reference lists of trial publications and review articles and by writing to principal investigators of identified eligible trials.
The selection of papers and data extraction using the same data extraction form were conducted independently by two investigators (HLYC and JLT). Basic information obtained from each eligible trial included the number of patients randomized into each compared group at the outset of the trial, the treatment regime, duration of follow-up and the treatment outcomes at the end of treatment and/or during the post-treatment follow-up period. Articles were examined to eliminate duplicate reports of the same trial, and uncertainties in the data were clarified by contacting the principal investigators through writing when necessary.
Prospective, randomized controlled trials comparing thymosin vs. placebo or usual care in the treatment of chronic HBV infection were considered for analysis. Patients were HBV DNA-positive by non-polymerase chain reaction assays and had elevated alanine transaminase levels. Studies were included in the meta-analysis if they had a minimum treatment duration of 24 weeks and reported end-of-treatment and/or sustained (6 and/or ≥ 12 months post-treatment) virological and/or biochemical responses. Trials using other anti-viral agents, such as interferon, as controls, and those including patients suffering from other forms of viral hepatitis (hepatitis C or hepatitis D) were excluded.
Ten potentially eligible randomized trials using thymosin in the treatment of chronic HBV infection were identified. Five were excluded: three compared thymosin vs. interferon,6–8 and two9, 10 were duplicate publications of the same trial.11 Five trials were included in the analysis.11–15
Virological response was defined as the disappearance of HBV DNA in the serum by non-polymerase chain reaction assays plus the loss of hepatitis B e antigen (HBeAg) if the studied patients were HBeAg-positive before treatment, and the disappearance of HBV DNA if the studied patients were HBeAg-negative before treatment. Biochemical response was defined as the normalization (or near-normalization in one trial) of the alanine transaminase levels. We analysed the outcome at the end of treatment, 6 months post-treatment and 12 months (or longer) post-treatment.
We established standards for the assessment of the methodological quality of the trials: clearly specified inclusion (1) and exclusion (2) criteria; concealment of randomization (3); baseline equivalence of treatment groups (4); blindness (5); and completeness of follow-up (6). We considered the baseline characteristics of the treatment groups to be equivalent if there were no statistically significant differences for factors that might affect the treatment response (gender ratio, liver enzymes, HBeAg status and HBV DNA levels). We defined complete follow-up as 90% or greater with no preferential loss to follow-up in one group over the other. A point was given for each satisfied standard according to the description contained within the text of each trial. A quality score was generated by summing the standards, and ranged from 0 (no standard satisfied) to 6 (all standards satisfied).
Biochemical and virological responses were analysed separately using the intention-to-treat method. We used the ratio of the odds of the main outcomes in the thymosin-treated group over that in the control group as the measure of efficacy. The 95% confidence interval (95% CI) for the combined odds ratio (OR) is also provided. The fixed-effect method suggested by DerSimonian and Laird was used to combine the results and the related significance test to assess the heterogeneity.16 The combined result was an average OR weighted according to the standard error of the OR of the trial. If a single control group was used for comparison with two treatment groups of different duration of thymosin treatment, the standard error was computed by using half of the patients in the control group. We arbitrarily assigned 0.5 to groups in which no outcome event was observed so that the standard error could be computed. When the heterogeneity test was statistically significant (P=0.10 or lower), the random-effect method was also applied to see how much the conclusion would change as compared with that derived from the fixed-effect method. The methods proposed by Egger et al. and Tang and Liu were used to detect publication bias.17, 18 We used the hierarchical regression suggested by McIntosh to assess the association of the effectiveness and the frequency of end-point events in the control group.19
Table 1 shows the characteristics of the five trials included in the meta-analysis, with a total of 353 patients. Seven valid comparisons were available as two trials had two treatment arms with a different duration of thymosin treatment vs. a single control group. Two trials compared thymosin with placebo in a double-blind manner and three compared thymosin with usual care. At entry, all patients were HBV DNA-positive by non-polymerase chain reaction assays and had elevated alanine transaminase levels. Three trials studied HBeAg-positive patients, one trial studied HBeAg-negative (and anti-HBe-positive) patients, and one included both HBeAg-positive and HBeAg-negative patients. Pre-treatment liver biopsy to confirm chronic hepatitis or early cirrhosis was conducted in three trials. None of the trials included patients with decompensated liver disease or complication of portal hypertension. Four trials used thymosin-α1 at a dose of 900 μg/m2 or 1.6 mg (twice weekly subcutaneous injection), and the remaining trial used both thymosin-α1 and thymosin fraction 5 (at 90 μg/m2 twice weekly subcutaneous injection). Among the seven comparisons, thymosin treatment was continued for 6 months in five comparisons and 12 months or longer in two comparisons. One comparison had outcome data only at the end of treatment, three had main outcome data up to 6 months post-treatment and three had data up to 12 months or longer post-treatment.
The quality scores were comparable among the four fully published trials, with a score of 4–5 out of a maximum of 6 (Table 1). These four trials had clearly stated inclusion and exclusion criteria as well as comparable baseline characteristics and follow-up among the treatment groups; however, only two were conducted in a double-blind manner. One trial was published in abstract form without a clearly stated methodology, and it received a quality score of 1 out of 6. None of the trials in this meta-analysis had clearly mentioned the concealment of allocation in the randomization process.
The virological response at the end of treatment, 6 months post-treatment and 12 months post-treatment is shown in Figure 1. The numbers of comparisons available for analysis were five at the end of treatment, five at 6 months post-treatment and three at 12 months post-treatment. The numbers of patients included in the analysis were 267, 291 and 207, respectively, at the three time points. The heterogeneity test indicated that the variation of trial-specific ORs was statistically significant at 6 and 12 months post-treatment (P=0.0277 and P=0.0550, respectively), but not at the end of treatment (P=0.4337).
The combined OR was 0.56 (95% CI, 0.20, 1.52) at the end of treatment and 1.67 (95% CI, 0.83, 3.37) at 6 months post-treatment (Figure 1). Neither combined OR was statistically significant. In addition, the conclusion at 6 months post-treatment was largely unchanged when the random-effect method was used (OR=1.86; 95% CI, 0.58, 5.99). These results suggest that there is a lack of sufficient evidence to indicate that thymosin may affect the clearance of HBV replication markers at the end of treatment and 6 months post-treatment.
At 12 months or longer post-treatment, all three comparisons showed a virological response in favour of thymosin and the OR was statistically significant in one trial. The combined OR was 2.67 (95% CI, 1.25, 5.68) and remained largely the same and statistically significant when the random-effect model was used to combine the results (OR=2.70; 95% CI, 1.18, 6.19).
The favourable effect of thymosin on the virological response increased with an increase in time since the cessation of thymosin treatment (Figure 2). The combined ORs were 0.56, 1.67 and 2.67 at the end of treatment, 6 months post-treatment and 12 months post-treatment, respectively. A trend using a simple regression method weighted according to the total number of patients was statistically significant (P=0.0201).
The biochemical response at the end of treatment, 6 months post-treatment and 12 months post-treatment is shown in Figure 3. The numbers of comparisons available for analysis were four at the end of treatment, three at 6 months post-treatment and two at 12 months post-treatment. The numbers of patients included in the analysis were 257, 209 and 108, respectively, at the three time points. The heterogeneity test indicated that the variation of trial-specific ORs was statistically insignificant at all three time points (P > 0.10).
The combined ORs for the biochemical response were 0.86 (95% CI, 0.44, 1.68) at the end of treatment, 0.86 (95% CI, 0.45, 1.64) at 6 months post-treatment and 1.20 (95% CI, 0.51, 2.83) at 12 months post-treatment. None of the ORs at the three time points was statistically significant, suggesting a lack of evidence that thymosin has an effect on alanine transaminase normalization.
Apart from local discomfort at the injection site, no systemic or constitutional symptoms or biochemical abnormalities as a result of thymosin treatment were reported in these trials.
This meta-analysis showed that thymosin (mainly thymosin-α1) treatment for 6–12 months was not associated with any significant biochemical or virological response at the end of treatment. There was an increasing trend of virological response after the cessation of thymosin treatment, and the benefit of thymosin treatment became statistically significant at ≥ 12 months after thymosin had been stopped. The reason for the delayed virological response to thymosin is uncertain, although it has been postulated that thymosin may have an immunomodulatory effect that induces permanently elevated T-helper cell levels.12 However, in this meta-analysis, there was no apparent improvement in terms of alanine transaminase response.
There was obvious heterogeneity in the virological response in the included trials. This variation in the estimate of efficacy may partly be attributable to the difference in the inclusion criteria (particularly HBeAg status of the patients), the definition of response to thymosin treatment and the methods used to detect the virological markers. HBeAg-positive patients were studied in four of the five trials, in which the virological response was generally defined as HBeAg seroconversion plus the loss of HBV DNA. However, one trial used a very stringent virological end-point: HBV DNA-negative on two occasions at least 3 months apart during the follow-up period and HBV DNA-negative and HBeAg seroconversion at the end of follow-up.14 Another trial used only HBeAg seroconversion as the virological response and only 79% of the HBeAg-seroconverted patients had negative HBV DNA.13 Two trials in this meta-analysis studied HBeAg-negative patients.11, 15 In HBeAg-negative patients, the meaning of virological response is different from that in HBeAg-positive patients, as it only requires the loss of HBV DNA at the time of assessment (with no HBeAg seroconversion). Finally, the different HBV DNA assays used in the different trials may also have caused additional variability in the sensitivity of HBV DNA detection and thus in the estimate of efficacy.
The selective publication of small positive trials is a great threat to the validity of meta-analyses.17 We used the data obtained at 6 months on HBV DNA to perform a regression analysis to assess the possibility of publication bias in the meta-analysis. We found insufficient evidence for a tendency for selective publication of small positive trials in this meta-analysis (P > 0.10). Based on the same data, the hierarchical regression analysis showed no statistical linear relation between the OR and the rate of outcome event in the control group (P > 0.10).
In conclusion, we found that thymosin has a cumulative effect after the cessation of treatment on the virological clearance in chronic HBV infection. It seemed to take as long as 12 months after the cessation of treatment for the anti-viral effect to become apparent. It is likely that our results reflect the response of HBeAg-positive patients, as HBeAg-positive patients were studied in four of the five included trials (three trials included HBeAg-positive patients and one included both HBeAg-positive and HBeAg-negative patients). As only two trials included HBeAg-negative patients (one trial included HBeAg-negative patients and one included both HBeAg-positive and HBeAg-negative patients), whether the efficacy of thymosin treatment differs from that in HBeAg-positive patients requires further research. As thymosin treatment is relatively free from adverse effects, future research is warranted to study the efficacy of combination therapy of thymosin with interferon or anti-viral agents in the treatment of chronic HBV infection.