Interferon and Ribavirin vs Interferon Alone in the Re-treatment of Chronic Hepatitis C Previously Nonresponsive to Interferon: Title and subTitle BreakA Meta-analysis of Randomized Trials FREE

Hepatitis C virus (HCV) infection represents a significant public health problem, affecting nearly 4 million people in the United States and more than 150 million worldwide.^{1 - 2} Approximately 70% of those with HCV will develop chronic hepatitis, and up to 20% will progress to cirrhosis.^{1 ,3 - 5} Patients with chronic hepatitis C leading to cirrhosis are at increased risk for hepatocellular carcinoma.⁶ Given its prevalence and capacity for chronicity, HCV is now the leading cause of chronic liver disease and the most common indication for liver transplantation in the United States.⁷

Until recently, therapy for chronic hepatitis C was limited to interferon alone. While 40% to 60% of patients treated with interferon will respond with loss of detectable serum HCV RNA levels and normalization of serum alanine aminotransferase (ALT) concentration, the majority relapse once treatment is stopped. Thus, only 15% to 20% of patients treated with interferon demonstrated a treatment effect 6 to 12 months after therapy.^{4 ,8}

Beginning in 1993, pilot and small-scale study results suggested there was a potential benefit of interferon and ribavirin combination therapy vs interferon monotherapy.^{9 - 11} These findings were confirmed in 1998 by 3 large, placebo-controlled trials.^{12 - 14} With a combined total of 1844 subjects, these studies demonstrated that for patients naive to interferon therapy, combination therapy was superior to interferon alone. A placebo-controlled study of nearly 350 patients showed the same was true for relapsers after interferon treatment.¹⁵

The effectiveness of combination therapy for nonresponders to prior interferon monotherapy is less clear. Studies thus far, most with fewer than 100 subjects, show conflicting results. The safety of combination re-treatment of nonresponders also is a concern. While several narrative reviews have addressed the topic of lack of response,^{16 - 17} to our knowledge, there has been no systematic review to evaluate combination therapy for nonresponders. The objective of this meta-analysis was to assess the efficacy and safety of combination therapy vs interferon monotherapy for treatment of chronic hepatitis C in nonresponders to prior interferon monotherapy.

The literature search was performed with MEDLINE (January 1966-December 1999) and the Science Citation Index, using the keywords "hepatitis C AND interferon AND ribavirin" and the publication type "clinical trial." A manual search of references from publications and a manual review of major internal medicine, gastroenterology, hepatology, and infectious diseases journals (January 1999-August 1999) also were conducted. In addition, the input of experts in investigational clinical hepatology was sought.

To be included in the meta-analysis, a study had to provide original data, to be published as an article, to be carried out in humans with clinically and histologically confirmed chronic hepatitis C, to compare interferon monotherapy with combination therapy in nonresponders to prior treatment with interferon, and to have randomized allocation to treatment group. The end points of interest were virological response, defined as absence of HCV RNA levels in serum, and biochemical response, defined as normalization of serum ALT concentration. These markers of disease status reflect the infectious process (virological quantification) and liver damage (biochemical quantification). Though neither is a direct clinical manifestation of disease, both are standard surrogates that correlate with long-term prognosis.^{6 ,18 - 19} Studies that did not report data for these end points at the end of a follow-up period were excluded. Articles were considered for inclusion regardless of language of publication.

Each study was evaluated separately by 2 investigators, each of whom was blinded as to the other's evaluation. Once completed, any disagreements were resolved through discussion. Recorded information included study characteristics (location, year, blinding, dose, and schedule), patient characteristics (percentage male, risk factors for hepatitis C, and percentage with cirrhosis), and outcomes. The recorded virological and biochemical outcomes were the total number of subjects and the number of responders at the end of follow-up. The recorded histological outcomes were the total number of subjects with pretreatment and posttreatment biopsy specimens, the number of subjects with posttreatment biopsy specimens showing improvement, and the modified Knodell histologic activity index (HAI) score before and after treatment.²⁰ Data on treatment groups other than those of interest were excluded.

The number of patients with a reduction in dose due to adverse events and the number withdrawn due to adverse events (or total number withdrawn from study, if not specified as due to adverse events) were recorded for each treatment group. The number of symptoms reported for each group was recorded by symptom category whenever possible (influenza-like, gastrointestinal, psychiatric, respiratory, and dermatological). Most studies did not report the number of adverse events and symptoms for nonresponders by treatment group. These data were obtained by contacting the authors of those studies.

We designed a quality evaluation questionnaire with sections on study population, bias and confounding, and description of therapy based on suggested randomized controlled clinical trial quality assessment questions.^{21 - 22} Questions regarding outcomes reported, follow-up time, and crossover were added for completeness. For each of the 17 questions, studies were assigned a score of 0 (inappropriate), 1 (fair), or 2 (appropriate).

Funnel plots were used to determine whether a publication bias existed against small studies reporting negative results.²³

The 24-week follow-up response rates based on virological and biochemical measurements were used for quantitative comparison. The measure of effect used was risk difference, the difference in the proportion of subjects with a particular outcome comparing combination therapy with monotherapy. Heterogeneity was assessed using L'Abbe plots and Q-statistics. Fixed effect models, using the inverse variance method, and random effect models, using the DerSimonian and Laird approach,²⁴ were used. When the results of the models differed, the results of the random effects model were reported. Results were presented as the estimated overall risk difference along with 95% confidence intervals (CIs). Pooled response rates for individual treatment groups were calculated as the total number of responses divided by the total number of patients, and these were reported along with their exact binomial CIs. Sensitivity analyses and meta-regressions were performed with study characteristics to evaluate potential sources of heterogeneity. Stratified analyses were performed using the variables found to be statistically significant in the meta-regressions.

The data regarding the discontinuation of treatment were analyzed using the methods described for the biochemical and virological outcomes. The data regarding adverse effects for each treatment group were compared using Poisson regression.

All analyses were based on intention-to-treat, used a 2-sided P = .05 significance level, and were conducted with the "meta," "metan," "labbe," and "funnel" subroutines in STATA 5.0.^{25 - 26}

Of the 50 articles that were identified, 13 met the inclusion criteria.^{27 - 39} Two studies by Brillanti et al^{27 - 28} reported on different stages of the same study. Thus 12 studies, published in English from 1995 through 1999, were included (Table 1). Study size ranged from 14 to 303 subjects. Studies published since 1998 were larger than earlier ones, with 5 reported as being multicenter.^{29 ,33 - 34 ,36 ,39} None of the studies was placebo-controlled. Quality scores ranged from 20 to 28 with a mean of 24.3 (95% CI, 22.8-25.7). Funnel plots revealed that most smaller studies found no significant difference between interferon monotherapy and combination therapy.

Interferon nonresponders were defined in most studies as patients who did not achieve a normalized serum ALT concentration with prior interferon therapy. Several studies also required the failure to achieve an undetectable serum level of HCV RNA in the definition of interferon nonresponders. Six of the studies recruited "interferon-relapsers" in addition to nonresponders.^{28 ,31 ,33 ,35 - 37} A total of 941 subjects, all nonresponders, were included in our meta-analysis.

Response to interferon monotherapy is known to correlate with dose and length of treatment, with higher dose (6 million units [MU]) and longer treatment (48 weeks) yielding higher response rates.⁴⁰ It is not known if the same relationship exists when ribavirin is combined with interferon therapy. The dose of interferon varied across studies; 6 used interferon, 3 MU, for both treatment groups^{27 ,30 ,32 ,34 ,37 - 38} while 6 incorporated a higher dose in at least 1 treatment group (Table 1).^{29 ,31 ,33 ,35 - 36 ,39} All studies used 24-week treatment groups, with the exception of Scotto et al³² and Pol et al,³⁹ which had only 8 weeks of combination therapy.

While different types of interferon appear comparable when used alone,¹⁸ it is not known whether differences in efficacy will become manifest when these interferons are combined with ribavirin. All studies used interferon alfa, either a human leukocyte-derived product (interferon alfa-n or -n3) containing a mixture of interferons or a recombinant product (interferon alfa-2a or -2b) containing a single interferon. Dose of ribavirin also varied (600-1200 mg/d).

Virological response 24 weeks after therapy correlates with favorable long-term (5-10 years) histological and clinical outcomes.^{19 ,41} All studies used at least 24 weeks of follow-up. Barbaro et al³⁶ did not report virological and biochemical outcomes separately but instead used a combined measure. Because it was unclear whether this combined outcome better estimated virological or biochemical response, this study's results were applied to both analyses. Salmeron et al³⁷ included both nonresponders and relapsers, did not provide virological results separately for nonresponders, and therefore was not included in the virological analysis.

Patient characteristics may influence response to treatment. Patients infected via transfusion have been found to have more advanced disease than patients infected via intravenous drug use.⁴² The presence of cirrhosis at the time of treatment may be associated with lower response to therapy,⁴³ although recent evidence has made that association less clear.¹² Viral genotype is a strong predictor of response to interferon, with genotype 1 being associated with lower virological response rates.^{44 - 46} Within studies, treatment groups were generally comparable with respect to these factors as well as to age and sex.

Virological Response. The 11 studies reporting virological outcomes were statistically heterogeneous (P = .006). The heterogeneity was quantitative not qualitative, that is, mainly because of differing estimates of the magnitude of benefit, not the direction of effect. Risk differences ranged from –9% to 20%, with an overall risk difference of 7% (95% CI, 2%-13%; P = .01) (Figure 1). The pooled response rate was 14% (62/456; 95% CI, 11%-17%) for combination therapy and 2% (10/443; 95% CI, 1%-4%) for monotherapy. Excluding the studies by Scotto et al³² and Pol et al³⁹ increased the overall risk difference to 9% (95% CI, 2%-15%; P = .008). These studies were excluded from all meta-regressions.

Univariate meta-regressions were performed with respect to interferon type (alfa-2a/2b vs alfa-n/n3), interferon dosage (3 MU 3 times a week in both treatment groups vs >3 MU 3 times a week in either treatment group), ribavirin dosage (600-800 mg/d vs 1000-1200 mg/d), and quality score. Interferon alfa-2a/2b and higher-dose ribavirin were associated with a significantly higher risk difference for virological response and were included in stratified analyses (Table 2). Stratification showed that studies comparing ribavirin and interferon alfa-2a/alfa-2b combination therapy vs interferon alfa-2a/alfa-2b monotherapy had an overall risk difference of 16% (95% CI, 11%-21%; P<.001). The studies that compared higher-dose ribavirin (1000-1200 mg/d) and interferon combination therapy vs interferon monotherapy, regardless of type of interferon, had an overall risk difference of 13% (95% CI, 8%-18%; P<.001). Because all studies using interferon alfa-2a/alfa-2b also used higher-dose ribavirin (1000-1200 mg/d), it was not possible to assess whether the interferon alfa-2a/alfa-2b, the higher-dose ribavirin, or the combination thereof, led to these response rates.

Biochemical Response. The results based on biochemical outcome were similar to those based on virological outcome. Risk differences ranged from –9% to 30% with an overall risk difference of 8% (95% CI, 1%-15%; P = .03). The pooled response rates were higher: 21% (101/478; 95% CI, 18%-25%) for combination therapy and 6% (28/463; 95% CI, 4%-9%) for monotherapy.

Histological Response. Three studies provided histological data amenable to analysis. Sostegni et al³⁰ found a significantly lower HAI score after treatment in the combination therapy group (n = 12) compared with the interferon alone group (n = 8). Neither Andreone et al³⁴ nor Pol et al³⁹ found such a difference in HAI score, although Pol et al did note a significantly higher percentage of subjects in the combination group (n = 41) had an improvement in the Knodell score compared with the interferon group (n = 40) (69.2% vs 35.9%; P<.02).

The 9 studies reporting withdrawal due to adverse events were statistically homogeneous (P>.20).^{28 - 30 ,32 ,34 - 36 ,38 - 39} The risk difference ranged from 0% to 8%, with an overall risk difference of 4% (95% CI, 1%-7%; P = .01). The pooled withdrawal rate was 9% (39/431; 95% CI, 7%-12%) for combination therapy and 4% (18/418; 95% CI, 3%-7%) for monotherapy. Univariate meta-regressions showed no association between higher doses of interferon or ribavirin and the risk of treatment discontinuation.

Reduction in the dose of interferon did not occur in any of the studies. In the studies by Andreone et al³⁴ and Sostegni et al,³⁰ the dose of ribavirin was reduced in 4% and 24% of patients, respectively. In 2 other studies (Bell et al³³ and Bellobuono et al³¹ ), the dose of ribavirin was reduced for 30% and 17% of patients, respectively. These 2 studies included relapsers as well as nonresponders, and dose reductions were not reported separately.

The interstudy variations in reporting of symptoms did not allow for a rigorous comparison of symptoms by category. For all studies, the total number of adverse events reported in the combination therapy group was greater than in the monotherapy group (Table 3). No treatment-related deaths were reported. Using Poisson regression, the estimated rate of adverse events in the combination therapy group was 1.31 (95% CI, 1.14-1.51; P<.001) times that of the monotherapy group.

The overall risk difference for nonresponders (7%) indicates that combination therapy is more efficacious than interferon alone. However, studies were heterogeneous, with some studies suggesting no difference between the therapies.

We identified several factors that may account for this heterogeneity with clinically important implications. Studies that used interferon alfa-n or -n3 had a lower-risk difference (4%) than those using interferon alfa-2a or -2b (16%), indicating that the interaction between ribavirin and recombinant interferon may not extend to other interferons, such as natural interferon. A previous meta-analysis did not find such a difference between types of interferon; however, that study focused exclusively on initial treatment with interferon alone.⁴⁰ Preliminary results from a recent study by Lau et al,⁴⁷ looking at antiviral synergy between interferon and ribavirin in an HCV surrogate system, did find differences in antiviral activities among several types of interferon. Our results suggest that anticipated novel interferons cannot be presumed to have the same efficacy in combination with ribavirin as do current products. Our findings support the need for clinical trials to compare new interferons with current recombinant types before the effectiveness for novel interferons in combination with ribavirin can be asserted.

We also found that ribavirin dose may contribute to the efficacy of combination therapy. In studies using ribavirin, 1000 to 1200 mg/d, combination therapy was more efficacious than interferon alone. Studies using lower-dose ribavirin (600-800 mg/d) showed no difference between the 2 therapies. To our knowledge, there have been no previous reports of such dose-dependent response. Indeed, preliminary results of an investigation carried out in relapsers and nonresponders suggested that ribavirin, 600 mg/d, may be sufficient.⁴⁸ Given their risk for resistance to treatment, however, nonresponders may need higher-dose ribavirin with interferon for synergy to occur. Furthermore, these patients may need longer treatment; the inclusion of studies using less than 24 weeks of combination therapy slightly lowered the risk difference.

Two of the 3 studies with complete data on histological response found a significant difference in histological features, favoring combination therapy. The incorporation of these treatment outcomes in future studies is warranted.

While our analysis of safety suggests a less favorable safety profile for combination therapy, it resulted in no treatment mortality. Furthermore, symptoms tended to be transient; authors generally noted that adverse events resolved with withdrawal of treatment. Finally, the absence of placebo control arguably contributed to symptom reporting, dose reductions, and treatment withdrawal, as evidenced by the lack of an association between treatment withdrawal and ribavirin dose.

Our analysis benefited from its sample size, inclusion of multiple measures of efficacy and safety, and exploration of factors potentially contributing to heterogeneity. While our literature search was extensive, we did not identify any non-English language publications or unpublished studies that met our inclusion criteria. This raises the issue of publication bias against studies with negative results. Yet, one third of the studies included here, and particularly those of smaller size, found no difference between treatments. Occasionally, we were limited by availability of data, despite our efforts to contact the authors directly. For instance, differences in reporting of viral load and disease duration hampered quantitative analyses of those factors, while the reporting of combined relapser and nonresponder data precluded the inclusion of 1 study (Salmeron et al³⁷ ) in the virological analysis.

Sample size and study design hindered a multivariate analysis of heterogeneity; we were limited to a univariate analysis that was subject to confounding. For instance, it was not possible to control for ribavirin dose when analyzing the effect seen with interferon alfa-2a/2b and vice versa. Thus, we cannot determine the extent to which each of these 2 factors individually contributed to the risk difference observed, although stratification (Table 2) suggests that each may have an effect. Although we did not identify any significant differences in baseline patient characteristics, there may have been unreported factors contributing to the observed risk difference. We cannot exclude, for example, the possibility that disease severity was greater in studies using interferon alfa-n/n3 or lower-dose ribavirin, although the available data on prevalence of cirrhosis do not fit such a pattern. Compliance was not routinely assessed in these studies. How variations in compliance could explain the superior efficacy of a higher-dose ribavirin regimen (presumably more complicated) is not clear.

In conclusion, our results suggest that combination therapy incorporating recombinant interferon and higher-dose ribavirin (1000-1200 mg/d), for at least 24 weeks, is more effective than interferon alone in the re-treatment of interferon nonresponders. However, the response rates for nonresponders are still low, regardless of interferon type or ribavirin dose. While combination therapy is tolerable in most cases, the fact that the vast majority will not benefit highlights the urgent and ongoing need for more effective therapies for the treatment of chronic hepatitis C.