Antidepressant prophylaxis reduces depression risk but does not improve sustained virological response in hepatitis C patients receiving interferon without depression at baseline: A systematic review and meta-analysis

C hronic hepatitis C virus (HCV) infection is a major cause of liver cirrhosis and hepatocellular carcinoma, and the most com-mon indication for liver transplantation in Europe and the United States (1). Combination therapy with pegylated interferon (peginterferon)-alpha and ribavirin represents standard treatment for chronic HCV infection (2,3). A sustained virological response (SVR) is achieved in 46% to 80% of patients (4-6). Despite this success rate, the challenging side-effect profile of HCV antiviral therapy limits treatment uptake. A key neuropsychiatric side effect of interferon-alpha (IFN- α ) is major depression (7,8). The risk has been reported to range between 10% and 40% (4,5,9-12). In many cases, antidepressants are required, mental health services provided, IFN doses reduced and/or antiviral therapy interrupted. Antidepressants are a mainstay for treating depression associated with IFN- α -based HCV therapy (6,13). Multiple studies have reported success in retaining patients on HCV antiviral therapy with the use of antidepressants in the management of IFN-induced depression (14,15). it remains unclear as to whether this no benefit to SVR was identified (risk ratio 1.08 [95% CI 0.74 to 1.57]; P=0.69; I 2 =58%). COnCLusiOn: This practice is not justified to improve SVR in populations free of active depressive symptoms leading up to HCV antiviral therapy.

to address the effectiveness of this strategy without clearly resolving the issue (1,(18)(19)(20)(21)(22). To address this unresolved question, we conducted a meta-analysis on the use of prophylactic antidepressants in patients initiating IFN-based HCV antiviral treatment. The impact on virological response, SVR rates and measures of mental health status were specifically addressed.

MeThOds search strategy
A review protocol and search strategy was developed to capture articles describing HCV antiviral treatment in which prophylactic antidepressant therapy was used. English language articles from 1946 to July 2012 were included. The MEDLINE, Embase and Cochrane Central databases were searched. Reference lists of selected articles were also screened for eligible reports. The search strategies used are presented in Appendix 1.

eligibility criteria
Inclusion was restricted to randomized controlled trials (RCTs) in which prophylactic antidepressant therapy was started at least two weeks before the initiation of HCV antiviral treatment.

study selection and data extraction
All titles and abstracts of the citations identified by the literature search were independently screened by two investigators (AA-O and CC). Relevant articles were reviewed in their entirety. Each investigator made a recommendation for inclusion or exclusion of single articles and, if discordant, a third investigator (JC) resolved the discrepancy. When two or more articles had overlap of their populations and reported on the same outcomes, only the most inclusive article was considered with supplementary information taken from the overlapping articles.
Using a standardized form, two investigators (AA-O and JC) systematically collected data on the outcomes of interest, population characteristics and several aspects of study setting and methodological design. Virological response to treatment (ie, SVR) and major depression were specifially addressed. The Montgomery-Asberg Depression Rating Scale (MADRS) and Beck Depression Index (BDI) were also evaluated. The MADRS is a 10-item, clinician-administered measure of current depressive symptoms. It provides a measure of depressive symptomatology in patients with chronic medical conditions that is less influenced by physical symptoms and more sensitive to changes in depressive symptoms. The BDI, Second Edition, is a 21-item, self-reported tool used to evaluate depression symptom severity that is well validated and reliable in HCV patients and in IFN recipients (23)(24)(25).
Individual studies were assessed using the Cochrane Collaboration's tool for assessing risk of bias in randomized trials by two investigators (AA-O and JC) (26). Reporting of the following individual components were assessed: description and method of generation of the randomization sequence; method of allocation concealment; method of blinding; report of incomplete outcome data; selective reporting bias; and other biases such as baseline imbalance or early termination due to some data-dependent process.

synthesis and meta-analysis
Where possible, measures of effect were pooled using standard metaanalysis methods. Pooled risk ratios (RR) and mean differences (MD) with associated 95% CIs were for calculated for dichotomous and continuous outcomes, respectively. Studies were pooled using randomeffects models by the generic inverse variance method (27). However, where homogeneity allowed, both fixed-and random-effects results were calculated. Where studies reported medians, effect distribution was assessed and variance estimates were calculated using standard methods (28). Clinical and methodological heterogeneity was investigated and sensitivity analyses performed where necessary. Statistical heterogeneity was measured using the I 2 statistic (29); all meta-analyses were conducted irrespective of statistical heterogeneity. No subgroup analyses were performed. Study outcomes not eligible for inclusion in meta-analyses were reported descriptively across studies.

resuLTs
A total of 400 potentially eligible trials were retrieved through electronic searches; 310 nonduplicate publications were identified (Appendix 1). Twenty-one full-text articles were reviewed, of which 15 were excluded as companion studies or nonrandomized trials. In total, six RCTs met the inclusion criteria and were included in the meta-analysis (Table 1, Figure 1). Antidepressants evaluated included paroxetine (n=2 [20,21]) escitalopram (n=3 [1,18,22]) and citalopram (n=1 [19]). The period of time required to be free of ongoing depression or psychiatric symptoms ranged from two to six months. All but one study (1) excluded patients with past or current mood disorders, bipolar and psychotic psychiatric conditions. All studies excluded individuals with active substance abuse. The time to randomization before HCV antiviral treatment initiation ranged from zero to four weeks. Some studies used dose escalation and intensification strategies while others used a single, fixed dose. Four of six studies administered antidepressants for the entire duration of HCV antiviral therapy. The exceptions included the Diez-Quevedo et al (18) evaluation, in which patients were randomly assigned to escitalopram or placebo for the initial 12 weeks of treatment, and the de Knegt et al (1) assessment, in which citalopram was dosed for 26 weeks. All study participants received one of several IFN formulations administered subcutaneously and daily oral ribavirin. The targeted duration of HCV antiviral therapy was consistent across studies: 48 weeks for genotypes 1 and 4, and 24 weeks for genotypes 2 and 3.

risk of bias in included studies
An assessment of each study against the individual methodological quality criteria described in the method section is provided in the table of risk of bias summary (Appendix 2). All studies were reported as 'randomized', although one article (20) did not describe the method of randomization. There is only one study that had low risk of bias across all domains. The study by Morasco et al (19) did not address incomplete outcome data and the study code was broken to several patients in the de Knegt et al (1) article. The study by Schaefer et al (22) had a baseline imbalance at the time of randomization in which baseline MADRS was higher in the placebo group even though this score normalized at the time of study. Overall, the included trials were assessed to be of reasonable quality.

Prophylactic antidepressant therapy protected against clinical depression
All studies reported information describing the proportion of patients who developed clinical depression during the study period and/or at the end of the trial ( Table 2). DSM-IV criteria were used to identify patients with clinical depression in four studies (18)(19)(20)(21). The Mini-International Neuropsychiatry Interview (MINI) was used to diagnose depression in one study (1). A score of ≥13 on the MADRS was considered to be clinical depression in one study (22).  (22) study reported MD at four, 12 and 24 weeks but not absolute mean MADRS for each study group. To address this, the mean MADRS and SDs at were calculated at weeks 4, 12 and 24 based on the reported baseline MADRS score. Sensitivity analysis was conducted excluding data extracted from Schaefer et al (22) to determine whether this methodology introduced bias in estimate of effect (data not shown). The MD of MADRS score at four, 12 and 24 weeks were not significantly different using this approach. The BDI score was reported in two studies (1,19). The BDI score at weeks 4, 8 and 24 did not differ between groups ( Figure 4B).

disCussiOn
Ultimately, the goal of antidepressant prophylaxis is to maintain patients on full doses of IFN and ribavirin and enable completion of the full duration of HCV antiviral therapy thereby maximizing the likelihood of achieving an SVR. Our meta-analysis, based on six RCTs with a low overall risk of bias (Appendix 2), suggests that this is not achieved, at least not in the populations evaluated. Individuals eligible for participation in these RCTs were characterized as being free from current  or recent clinical depression, or other concurrent active psychiatric conditions. Furthermore, these participants were not taking any mental health medications at the time of enrollment. Our analysis established that antidepressant prophylaxis in advance of initiating HCV antiviral therapy is not beneficial in improving SVR in individuals with stable mental health. This does not resolve the question with regard to individuals with mild depression or other active psychiatric conditions planning to start HCV treatment. It is plausible that this benefit may have a more clinically significant effect in those with borderline depression or active, but stable, concurrent mental health concerns during the lead-up period to HCV therapy. Our analysis indicated that the risk for developing on-treatment depression was reduced in those randomly assigned to receive antidepressant prophylaxis, although the Diez-Quevedo et al (18) study, which has the highest internal validity or lowest risk of bias among the six trials, did not show decreased rate  of on-treatment clinical depression using DSM-IV criteria. Clearly, the severity of these cases and/or the reduced frequency of depression in those receiving prophylactic antidepressants was insufficient to impact SVR. However, this remains an important outcome because it suggests that the burden of psychiatric side effects experienced on treatment may be partially alleviated with prophylactic antidepressants. MADRS and BDI scores are calculated, and continuous scoring systems are used to grade an individual's depressive state at any one time and to follow it over time. Although trending in favour of benefit with antidepressants, both MADRS and BDI scores were similar between randomization groups at weeks 4, 12 and 24 of HCV antiviral therapy. It is key to note that in individual studies, a relatively high baseline MADRS score was associated with a greater protection from on-treatment depression with the use of a prophylactic antidepressant (21). Moreover, de Knegt et al (1) noted a trend toward protection from on-treatment depression in recipients of escitalopram who had a history of depressive symptoms. It would have been of value to assess these scores during the initial month of therapy because the onset of depressive mood symptoms generally begins within the first four weeks of IFN-based HCV therapy. In clinical practice, therapy is often interrupted, doses of IFN are reduced and/or antidepressant therapy with or without additional mental health care is initiated before week 12 in an effort to manage on-treatment depressive symptoms. These measures would collectively serve to diminish any difference in the MADRS and BDI scores between groups at weeks 12 and 24. Unfortunately, the level of detail required to control for these on-treatment depression management factors was not reported in the individual publications.
Antidepressants are not without side-effect and adverse-event risk. However, there was no evidence from individual articles that the use of prophylactic antidepressants resulted in an increased symptomatic burden (19). Importantly, no suicides were reported. Raison et al (21) reported increased dizziness with paroxetine. At the very least, pill burden is increased with this practice, which is often a challenge for individuals receiving HCV antiviral therapy.
Several limitations of the present study are acknowledged. Relatively small, clinically and methodologically heterogeneous studies were evaluated. For continuous measures of effect, assumptions were made regarding normality. However, we do not believe that this altered the results. Only a small number of the many antidepressants currently in use were evaluated in the present study (ie, escitalopram, citalopram, paroxetine). It is plausible that other medications may be of greater value in preventing depression. Specifically, agents with anxiolytic and/or appetite-enhancing properties may be of overall benefit over the course of HCV treatment (30). It is noteworthy that the collective outcomes observed with any one of the three specific antidepressants assessed in the six RCTs included in our meta-analysis did not clearly differ from the overall findings (Figures 2, 3 and 4). Dose escalation was allowed in some but not all studies based on depressive symptomatology. It is possible that suboptimal doses may have been evaluated in some studies. Different IFN formulations were evaluated within and between studies. Theoretically, this could have influenced the risk of on-treatment depression and/or the response to antidepressant prophylaxis. However, several studies suggest that this is not the case (31-33). It is noteworthy that the mean maximum MADRS  (18). Forest plot of the depression risk compared between recipients of prophylactic antidepressants and the placebo group excluding one outlying study (18). Fixedand random-effects analyses are represented