Types of studies

We will include randomised clinical trials irrespective of language or publication status.

Types of participants

Patients with chronic hepatitis C will be included. The diagnosis could be based on presence of hepatitis C virus RNA for more than six months, or presence of hepatitis C virus RNA plus elevated transaminases for more than six months, or histological evidence of chronic hepatitis. Based on previous antiviral treatment, the included patients will be classified as treatment‐naive (not previously treated), relapsers (patients with a transient response to previous treatment), or non‐responders (patients without response to previous treatment). Patients who have undergone liver transplantation and patients with human immunodeficiency virus will be excluded.

Types of interventions

This review will include randomised clinical trials comparing pegylated interferon alpha 2a or pegylated interferon alpha 2b plus ribavirin versus non‐pegylated interferon plus ribavirin for patients with chronic hepatitis C. Co‐interventions will be permitted if received by all trial groups and applied equally. Trials will be included regardless of the dose or the duration of the interventions.

Types of outcome measures

• Serum (or plasma) sustained virological response (SVR): number of patients with undetectable hepatitis C virus RNA in serum by sensitive test six months after the end of treatment.

• Liver‐related morbidity plus all‐cause mortality: number of patients who developed cirrhosis, ascites, variceal bleeding, hepatic encephalopathy, hepatocellular carcinoma, or died.

• Adverse events: number and type of adverse events defined as patients with any untoward medical occurrence not necessarily having a causal relationship with the treatment. We will report on adverse events that lead to treatment discontinuation and those that have not led to treatment discontinuation separately. Serious adverse events are defined according to the International Conference on Harmonisation (ICH) Guidelines (ICH 1997) as any event that leads to death, is life‐threatening, requires in‐patient hospitalisation or prolongation of existing hospitalisation, results in persistent or significant disability, and any important medical event, which may have jeopardised the patient or requires intervention to prevent it. All other adverse events will be considered non‐serious.

Selection of studies

We will list the identified trials, and two of the authors will independently asses their fulfilment of the inclusion criteria. We will list the excluded trials with the reason for exclusion. Disagreements will be resolved by discussion and arbitrated with a third author.

Data extraction and management

A standardised template of data collection form will be used to extract data regarding source ID, eligibility, methods, participants, interventions, outcomes, and results from either published reports or by contacting the authors. Two of the authors will extract all the data independently. Disagreements will be resolved by discussion and arbitrated with a third author.

Assessment of risk of bias in included studies

Methodological quality is defined as the confidence that the design and the report of the randomised clinical trial would restrict bias in the comparison of the intervention (Moher 1998). According to empirical evidence (Schulz 1995; Moher 1998; Kjaergard 2001; Wood 2008), the methodological quality of the trials, hence risk of bias, will be assessed based on the defined below sequence generation, allocation concealment, blinding of participants, personnel, and outcome assessors, incomplete outcome data, selective outcome reporting, and other sources of bias (Gluud 2009).

Sequence generation

• Low risk of bias (the methods used is either adequate (eg, computer generated random numbers, table of random numbers) or unlikely to introduce confounding).

• Uncertain risk of bias (there is insufficient information to assess whether the method used is likely to introduce confounding).

• High risk of bias (the method used (eg, quasi‐randomised trials) is inadequate and likely to introduce confounding).

Allocation concealment

• Low risk of bias (the method used (eg, central allocation) is unlikely to induce bias on the final observed effect).

• Uncertain risk of bias (there is insufficient information to assess whether the method used is likely to induce bias on the estimate of effect).

• High risk of bias (the method used (eg, open random allocation schedule) is likely to induce bias on the final observed effect).

Blinding of outcome assessors

• Low risk of bias (blinding was performed adequately, or the outcome measurement is not likely to be influenced by lack of blinding).

• Uncertain risk of bias (there is insufficient information to assess whether the type of blinding used is likely to induce bias on the estimate of effect).

• High risk of bias (no blinding or incomplete blinding, and the outcome or the outcome measurement is likely to be influenced by lack of blinding).

Incomplete outcome data

• Low risk of bias (the underlying reasons for missingness are unlikely to make treatment effects departure from plausible values, or proper methods have been employed to handle missing data).

• Uncertain risk of bias (there is insufficient information to assess whether the missing data mechanism in combination with the method used to handle missing data is likely to induce bias on the estimate of effect).

• High risk of bias (the crude estimate of effects (eg, complete case estimate) will clearly be biased due to the underlying reasons for missingness, and the methods used to handle missing data are unsatisfactory).

Selective outcome reporting

• Low risk of bias (the trial protocol is available and all of the trial's pre‐specified outcomes that are of interest in the review have been reported or similar).

• Uncertain risk of bias (there is insufficient information to assess whether the magnitude and direction of the observed effect is related to selective outcome reporting).

• High risk of bias (not all of the trial's pre‐specified primary outcomes have been reported or similar).

Other sources of bias

• Low risk of bias (the trial appears to be free of other sources of bias (eg, baseline imbalance, early stopping, and vested of interest bias)).

• Uncertain risk of bias (there is insufficient information to assess whether other sources of bias are present).

• High risk of bias (it is likely that potential sources of bias related to specific design used, early termination due to some data‐dependent process, lack of sample size or power calculation, or other bias risks are present).

All the above bias risk components will be assessed independently by two authors. Disagreements will be resolved by discussion and arbitrated with a third author.

Measures of treatment effect

Dichotomous data

Dichotomous data will be expressed as risk ratio (RR) and risk difference (RD) with 95% confidence intervals (CI). Furthermore, the number needed to treat to obtain a beneficial effect (NNTB) will be derived from the RD in meta‐analyses where the 95% confidence interval or the RD does not include zero.

Rare events (morbidity and mortality) will be estimated using the odds ratio as measure of effect.

Unit of analysis issues

Multiple treatments meta‐analysis

We plan to combine the results of this review with another ongoing review (Awad 2009) using a multiple treatments meta‐analysis model in an overview of reviews. We will use the multiple logistic regression approach proposed by Lu and Ades for combining direct and indirect evidence (Lu 2004). The analyses will be carried out both in a frequentist and Bayesian framework. The multiple treatments meta‐analysis model will be extended by including the covariates for meta‐regression. The included regression covariates are described in detail in the subgroup analysis section.

Dealing with missing data

For trials with missing data we will assess the adequacy of the methods used to deal with missingness. We will extract both the summary statistics based on the missing data methods and the crude summary statistics. All meta‐analyses will be carried out using summary statistics derived via missing data methods and, separately, using crude summary statistics.

When patients are lost to follow‐up and missing data methods are not applied, data will be analysed according to the intention‐to‐treat (ITT) principle and the available case analysis. ITT analysis will be performed assuming poor outcome in both groups (ie, dropouts from both the treatment and control groups will be considered as failures, using the total number of patients as the denominator). In the available case analysis patients lost to follow‐up will be excluded from the analysis, and the total number of patients that completed the trial will be used as the denominator.

Assessment of heterogeneity

Heterogeneity will be explored by chi‐squared test, and the quantity of heterogeneity will be measured by I² statistic (Higgins 2002). Sources of heterogeneity will be assessed in meta‐regression and subgroup analysis. Meta‐regression for pairwise comparison meta‐analysis will be carried out only for meta‐analyses including a sufficient number of trials. Sub‐group analyses will be carried out, strictly for exploratory and hypothesis‐ generating purposes. For the meta‐regression model we will consider the covariates as for the subgroup‐analyses.

Assessment of reporting biases

Different types of reporting biases (eg, publication bias, time lag bias, outcome reporting bias, etc.) will be handled following the recommendations of The Cochrane Collaboration (Higgins 2008). For dichotomous outcomes with intervention effects measured as odds ratios, the arcsine test proposed by Rücker 2008 will be used to test for funnel plot asymmetry. Nevertheless, asymmetric funnel plots are not necessarily caused by publication bias, and publication bias does not necessarily cause asymmetry in a funnel plot (Egger 1997).

Data synthesis

For all analyses, we will use both random‐effects (DerSimonian 1986) and fixed‐effect models (DeMets 1987). Due to the underlying assumptional differences, results from the random‐effects model and fixed‐effect model may differ to a non‐ignorable extent. In case such discrepancies are observed, P values will be ignored, and results will be interpreted according to the implications of the subgroup and heterogeneity analyses according to confidence intervals of the two models.

Subgroup analysis and investigation of heterogeneity

The following subgroup analyses will be considered and performed whenever feasible:

• Risk of bias: trials that are assessed to be at a low compared to trials at unclear or high risk of bias.

• Participants: trials with treatment‐naive patients compared to trials with relapsers or non‐responders.

• Genotype: trials with patients infected with different hepatitis C virus genotypes.

• Type of pegylated interferon: trials with pegylated interferon alpha 2a compared to trials with pegylated interferon alpha 2b.

Sensitivity analysis

Suitable sensitivity analysis will be identified during the review process. For example, if zero‐event trials are found, we will employ a number of different continuity corrections according to the sensitivity analysis proposed by Bradburn 2006.