Summary of main results

A major obstacle to treatment and patient compliance is the association of combined antiviral therapy with frequent and sometimes serious adverse events. Timely recognition of adverse events has a crucial role in the continuation of antiviral therapy with a hope for successful virological treatment results. In clinical trials, approximately 10% to 15% of participants discontinue peginterferon and ribavirin therapy because of adverse effects; in clinical practice, the rate of treatment discontinuation has been reported as substantially higher (Sulkowski 2011). The adverse effects of combined antiviral therapy are numerous and can be observed in many organ systems. The most common adverse effects are haematological, dermatological, neurological, immunological, gastrointestinal, pulmonary, cardiovascular, and ocular (Sulkowski 2011). We are lacking data on direct comparisons of different types of interferons. Among the most frequent adverse effects is myelosuppression. Neutrophil count decreases by an average of 34% (Soza 2002), and mean platelet drop is around 42% (Roomer 2010). Regarding haemoglobin levels, more than 50% of participants experienced a decrease in haemoglobin to around 30 g/L. Women are more prone to experiencing a haemoglobin level less than 100 g/L (Sulkowski 2004). Many of those adverse effects usually require only adjustment of antiviral therapy. In some patients, application of growth factors is needed because of haematological disturbances. Psychiatric disturbances and mood alterations associated with long‐term interferon therapy are well described. In some trials (Constant 2005), mood alterations occurred in 32% of participants, and significant depressive symptoms were noted in 21% to 58% of participants (Raison 2005). Retinopathy associated with pegylated interferon therapy has been reported in 21% of participants, but the vast majority of participants recovered without cessation of treatment (Lim 2010). Among the rare adverse events that could be linked to combined antiviral therapy are some potentially fatal disorders that are usually reported as case reports. Pulmonary complications such as pneumonitis and pulmonary hypertension are rare but require early detection and premature cessation of antiviral therapy (Kumar 2002; Dhillon 2010; Slavenburg 2010). In one trial, mortality secondary to pneumonitis was seen exclusively with peginterferon alpha‐2b (Slavenburg 2010).

In this systematic review, we have summarised the available evidence from randomised clinical trials comparing peginterferon alpha versus interferon alpha, both given in combination with ribavirin. Because of the high risk of bias in the trials, our results seem to suggest that the combination of peginterferon plus ribavirin is associated with a moderate or low beneficial effect on sustained virological response when compared with interferon plus ribavirin. Our results show that evidence on liver‐related morbidity plus all‐cause mortality is sparse, and that the meta‐analysis is underpowered to detect any difference in risk of liver‐related morbidity plus all‐cause mortality. However, regarding adverse events, the two regimens seem to differ significantly, with interferon plus ribavirin causing less harm than peginterferon plus ribavirin.

Both peginterferon alpha‐2a and peginterferon alpha‐2b plus ribavirin were superior in terms of sustained virological response when compared with interferon plus ribavirin for participants with chronic hepatitis. Our subgroup analyses demonstrated that peginterferon alpha‐2a was associated with a significantly higher sustained virological response than peginterferon alpha‐2b versus interferon. This indirect observation is in harmony with our previous observations from head‐to‐head trials that peginterferon alpha‐2a leads to sustained virological response in a greater number of participants than is seen with peginterferon alpha‐2b (Awad 2010; Hauser 2014).

Overall completeness and applicability of evidence

The Summary of Findings table that we have prepared reveals that, in generalwe can have only low confidence in the current evidence on harms measured as liver‐related morbidity plus all‐cause mortality, and adverse events or adverse events leading to treatment discontinuation. In addition, we can have very low confidence in the current evidence on treatment benefits measured as sustained virological response (summary of findings Table for the main comparison). All trials had high risks of bias. If we disregarded bias due to lack of blinding and due to vested interests, about half of the trials could be considered trials with a lower risk of bias. Our sensitivity analyses, however, showed no important influence of bias risk on our estimates of intervention effects. This may be due to the fact that we were unable to identify trials with a lower risk of bias. In our study, trials that adequately reported on the methodology of trial performance are large trials that dominate the pooled estimates of effect. Therefore, it is less likely that the pooled estimates are biased. In the meta‐analysis for sustained virological response, no serious inconsistencies were seen across trials, and the meta‐analyses crossed the Lan‐DeMets monitoring boundary, leaving out any random error. The only trial with blinded assessment of sustained virological response showed no significant differences between the two interventions but suffered from the fact that it was very small (Sjögren 2007).

Subgroup analyses for sustained virological response showed a pronounced difference in comparative treatment effects by genotype and type of peginterferon. Data from 16 trials for genotype one yielded RR 1.5, 95% CI 1.3 to 1.7. Data from nine trials for genotype two or three yielded RR 1.1, 95% CI 1.0 to 1.2. Data from nine trials for genotype four yielded RR 1.8, 95% CI 1.3 to 2.4. These differences were substantiated by tests of interaction that yielded statistically significant differences between genotype one and genotype two or three (P value 0.0002) and between genotype two or three and genotype four (P value 0.029). The larger effects of peginterferons versus interferons on sustained virological response were observed in participants infected with genotype one and genotype four, whereas the difference between peginterferon and interferon was least for participants infected with genotype two or three.

Moreover, peginterferon alpha‐2a plus ribavirin seemed significantly more effective than peginterferon alpha‐2b plus ribavirin when compared with interferon plus ribavirin in subgroup analyses. Data from nine trials for peginterferon alpha‐2a yielded RR 1.9, 95% CI 1.5 to 2.4. Data from 18 trials for peginterferon alpha‐2b yielded only RR 1.1, 95% CI 1.1 to 1.2. These differences were substantiated by tests of interaction that yielded statistically significant differences between peginterferon alpha‐2a plus ribavirin and peginterferon alpha‐2b plus ribavirin trials (P < 0.0001). This is in accordance with our previous observations based on head‐to‐head randomised clinical trials comparing peginterferon alpha‐2a versus peginterferon alpha‐2b (Awad 2010; Hauser 2014).

All other tests of interaction regarding risk of bias (lower risk of bias compared with high risk of bias), treatment history, and baseline viral load yielded non‐significant results.

We found three trials that used amantadine cointervention, four trials that used consensus interferon, and one trial that used leucocyte interferon. However, the paucity of evidence precludes any firm conclusions on effects of triple therapy regimens with the use of amantadine, leucocyte interferon, or consensus interferon. Our analyses strengthen the evidence that peginterferon plus ribavirin increases sustained virological response for participants with chronic hepatitis C, irrespective of genotype, previous antiviral status, or viral load, and with or without amantadine as cointervention.

Quality of the evidence

The methodological quality was unclearly reported in a considerable number of included trials. Only 14 of the 27 trials (51.9%) reported adequate randomisation and allocation concealment. Most trial reports described unclearly blinded outcome assessment. These aspects may be essential in minimising the risks of selection, performance, and ascertainment bias in individual trials (Schulz 1995; Moher 1998; Kjaergard 2001; Wood 2008; Lundh 2012; Savovic 2012; Savovic 2012a). However, we found no significant association between methodological quality and trial results (Analysis 3.1).These findings may support the robustness of our results but do not exclude the possibility of bias, especially as we could not assess risk of bias from blinding and vested interests. Industry support is known to significantly bias trial results (Lundh 2012). Regarding blinding, only one trial assessed sustained virological response with blinding and found no significant effect of pegylated interferon versus interferon (Sjögren 2007). However, this trial was small, and the confidence interval was accordingly wide (Sjögren 2007).

Selective outcome reporting was difficult to assess in this review. None of the included trials was adequately registered or had its protocol publicly available before trial completion (Krleza‐Jeric 2005; WHO 2009). It is hoped that the initiation of the World Health Organization International Clinical Trials Registry Platform will facilitate such assessments for future trials (WHO 2009). Another limitation in this review was insufficient design and reporting. Investigators of future trials therefore are well advised to adhere to the guidelines of Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) (Chan 2013) and the Consolidated Standards for Reporting of Trials (CONSORT) (Moher 2012) to improve the quality of trials.

In general, we can have only low confidence in the current evidence on harms measured as morbidity plus mortality and adverse events leading to discontinuation, whereas we can have moderate confidence in the current evidence on treatment benefits measured as sustained virological response (summary of findings Table for the main comparison). In the meta‐analysis for morbidity plus mortality, very serious imprecision was evident because only nine events were included in the meta‐analysis from five trials (Fried 2002; Esmat 2003; Al‐Faleh 2004; Lee 2005; Nevens 2010).

In the meta‐analysis on adverse events, serious discrepancies across trials and lack of precision were noted. The proportions of observed adverse events differ substantially across trials, and the direction of effect is heterogeneous. However, because the event rate is still relatively low across trials, all of the included trials may be subject to considerable random error, thus explaining the apparent heterogeneity in direction of estimates. The observed treatment effects differ in magnitude as well, but most confidence intervals show considerable overlap. In the meta‐analysis for sustained virological response, no serious inconsistencies across studies were evident, and the meta‐analyses showed adequate precision (i.e., the TSA alpha‐spending monitoring boundary for benefit was crossed, favouring pegylated interferon). Nonetheless, we had some concerns regarding bias (which was high in all trials) and indirectness, as all included randomised clinical trials had virological response as the predominant measure of benefit. Many of the trials measured sustained virological response, which is currently the commonly used surrogate outcome measure of benefit. Recent large cohort studies show a positive correlation between the presence of viraemia and mortality (Adeel 2009; Hirofumi 2009). However, it is important to remember that sustained virological response still is only a putative (unvalidated) surrogate outcome for the patient‐relevant intervention effects of antivirals (Gluud 2007; Koretz 2013; Gurusamy 2013). Because randomised clinical trials need to inform clinical practice, clinical outcomes such as risk of liver failure, hepatocellular carcinoma, mortality, and quality of life would be of greater interest to patients and clinicians. Such measures nevertheless require a follow‐up of maybe up to five years. Currently, no randomised clinical trials assessing the two pegylated interferons are of such long duration.

Potential biases in the review process

The included meta‐analyses had a seemingly reasonable mix of small and large studies, yielding fairly consistent results, thus causing little concern about the presence of publication bias.

The strengths of this Cochrane Hepato‐Biliary Group systematic review are that it builds on a peer‐reviewed published protocol, uses extensive searches until recently (Appendix 1), considers risks of systematic errors (‘bias’) (Schulz 1995; Moher 1998; Kjaergard 2001; Wood 2008; Lundh 2012; Savovic 2012; Savovic 2012a), and considers risks of random errors (‘play of chance’) by adjusting the threshold for statistical significance according to the information and strength of evidence present in the cumulative meta‐analysis (Brok 2008; Wetterslev 2008; Brok 2009; Thorlund 2009, Wetterslev 2009; Thorlund 2010). Furthermore, to avoid publication and ascertainment bias, we included trials regardless of publication status and language, and all data were abstracted independently by at least two review authors to avoid information bias. However, our review has potential limitations that may lower our confidence in the estimates of intervention effect. First, we are aware of possible inclusion bias in excluding special patient groups (hepatitis B virus, human immunodeficiency virus, cancer, and liver‐transplanted patients). Future systematic reviews should assess the intervention effects of antivirals in these patient groups. Second, a possible limitation of our review is the lack of availability of full reports of all included trials. Ten of the 27 included trials were available only as abstracts. However, including these abstracts in our systematic review may likely be a strength rather than a limitation. By including abstracts, we are looking at the complete available body of evidence. By excluding abstracts, we would have been looking only at a subset defined through present‐day publication mechanisms, favouring trials that show beneficial effects. This would considerably increase the likelihood of publication bias. Third, in general, adverse events were reported insufficiently. Most of the participants treated with peginterferon alpha or interferon alpha experience one or more adverse effects (e.g., influenza‐like symptoms, depression, neutropenia, thrombocytopenia). The proportions of observed adverse events differed greatly across trials, and the direction of effect was heterogeneous.

Agreements and disagreements with other studies or reviews

By comparing our results with those of the other meta‐analyses, we have found two main differences. Our searches of the literature were much more detailed than searches previously conducted. In previous meta‐analyses, authors usually did not perform grey literature searches and did not include meeting abstracts. Some authors excluded trials that were not published in English or used just two databases (Chander 2002; Zaman 2003), performed only cost‐effectives analysis (Shepherd 2005; Siebert 2005), or evaluated only one genotype (Khuroo 2004; Aljumah 2013). Second, we performed several sensitivity analyses to ensure the robustness of the results of our systematic review; this procedure is lacking in previously published meta‐analyses. Moreover, our number of included trials is much higher than the number included in the meta‐analyses mentioned above. Fourth, we considered risk of bias. Fifth, we also took precautions to interpret the statistical inference according to the strength of the evidence. In this vein, we constructed adjusted thresholds for statistical significance by using TSA.