Nine RCTs were used for the primary meta-analysis, which included a total of 789 patients. There were 62 uncontrolled and preliminary studies that met the inclusion criteria and provided additional data for the sensitivity analyses.
The method used to randomise the patients was computer in 4 studies and random numbers table in one study; the method was not stated in the remaining 4 studies. Allocation concealment was not stated in any of the studies and blinding was only stated in one study, where it was single.
A biochemical ETR could be evaluated in 7 trials with a total of 752 patients. Of these, 6 trials showed a significant improvement among patients treated with combination therapy in comparison with IFN monotherapy. The overall weighted response was 39.4% with a common odds ratio (OR) of 5.5 (95% CI: 3.7, 8.0) in favour of combination therapy.
A virologic ETR could be assessed in 7 trials with a total of 766 patients. Only 4 of these trials showed a significant improvement among the patients treated with combination therapy. The overall weighted virologic ETR was 23.1% with a common OR of 4.9 (95% CI: 2.9, 8.1) in favour of combination therapy.
A SBR could be assessed in 7 trials with a total of 700 patients. Only 2 of these trials showed a significant improvement among patients treated with combination therapy. The overall weighted SBR was 15.2% with a common OR of 3.8 (95% CI: 2.2, 6.7) in favour of combination therapy.
A SVR could be assessed in 8 trials with a total of 726 patients. Only 2 of these trials showed a significant improvement among patients treated with combination therapy. The overall weighted SVR was 13.2% with a common OR of 4.9 (95% CI: 2.1, 11.2) in favour of combination therapy. Of the 8 trials, 2 studies had a calculated risk difference of zero. The remaining 6 trials had NNTs that ranged between 7 and 28. The pooled risk difference for all 8 trials was 7% (95% CI: 2, 13) with a corresponding NNT of 14 (95% CI: 8, 50).
Overall, the response among the uncontrolled trials was higher than those from the randomised studies. For a treatment duration of 24 weeks, the overall weighted SVR of the combination treatment arms in uncontrolled trials and preliminary controlled trials was estimated to be 14.6% (95% CI: 11, 19.3). When adding the results of the published RCTs, the overall SVR was 14.0% (95% CI: 11, 18), suggesting that the eventual publication of the preliminary studies is unlikely to significantly influence the estimates in the primary analysis.
A number of viral-, patient- and treatment-related characteristics have been associated with a response in the individual clinical trials of IFN monotherapy or combination therapy. Favourable response characteristics include a lower baseline HCV-RNA titre, the absence of cirrhosis, infection with non-type 1 genotypes, non-black race, and treatment for at least 48 weeks. However, the influence of these predictors has varied among the different trials, and has not been well established among IFN nonresponders treated with combination therapy. Thus, the effect of these variables on the study end points in the published controlled trials where data were available was explored. Preliminary trials were used to examine the influence of 48 weeks of combination therapy, because the longest duration of treatment in the published controlled trials was 24 weeks. The SVRs of the combination treatment arms were determined after excluding studies containing the highest or lowest proportion of patients with various covariates. The overall weighted SVR for studies involving a treatment duration of 48 weeks was increased in comparison with the primary analysis: 21.3% (95% CI: 16.7, 26.9) versus 13.2% (95% CI: 10, 17.3). In contrast, the overall weighted SVR for trials that included more than 50% of patients with genotype 1 infection was decreased compared with the primary analysis: 7.9% (95% CI: 3.9, 15.5) versus 13.2% (95% CI: 10, 17.3). These findings suggest a possible advantage for longer treatment duration and for patients with non-type 1 genotypes. However, there were insufficient data to formally test these hypotheses. Minimal differences were detected when the sensitivity analysis was limited to trials with a mean baseline HCV-RNA level greater than three million copies/mL or trials using a dose of ribavirin greater than 1,000 mg/day.
The authors state that there was no significant heterogeneity among the trials within each meta-analysis, although the Q statistic was not reported.