Bupropion: a systematic review and meta-analysis of effectiveness as an antidepressant

Placebo trials

Five of the six open-label placebo trials reported a statistically significant clinical improvement in favour of bupropion [Brown et al. 2007; Fava et al. 2003; Ferguson et al. 1994; Gross et al. 2007; Walker et al. 1993] and one [Tomarken et al. 2004] did not demonstrate between group differences: they are described in Table 2. Of the six double-blinded, placebo-controlled RCTs, two evaluated flexible dosing of bupropion. In the earlier study [Lineberry et al. 1990] (n = 216) bupropion was titrated to a maximum of 100 mg/TID. Intention-to-treat (ITT) analysis demonstrated significantly greater HAM-D, MADRS, and CGI changes for the bupropion group over 6 weeks in both observed case analysis (p < 0.05 for all) and in last-observation-carried-forward (LOCF) analysis (p < 0.01 for all). A total of 54% attained response (⩾50% reduction in HAM-D total scores) on bupropion, significantly more (p = 0.01) than the 34% in the placebo group. In the work by Jefferson and colleagues (n = 270) the majority of patients (59%) received maximum dosage (450 mg/day) bupropion (versus 38% on 300 mg/day) [Jefferson et al. 2006]. Clinician reported response rates (IDS-C-30) were statistically significant for the bupropion group (50% versus 35%, p = 0.009) but self-reported response rates were not (53% versus 45%, p = 0.084) at the 8-week endpoint; remission rates were significantly greater in both clinician (41% versus 27%, respectively, p = 0.01) and self-reports (32% versus 18%, respectively, p = 0.005).

Two RCTs evaluated fixed dosing regimens over 8-week timeframes, with contrasting results. In the most recent study [Koshino et al. 2013] ITT analysis of 564 Japanese and Korean participants (randomized to bupropion 150 or 300 mg/day or placebo in a 1:1:1 ratio) reported no statistical differences between bupropion SR and placebo on MADRS total scores (p = 0.853) response (⩾50% reduction in MADRS) or remission (MADRS ⩽ 11) rates, though the authors note that significant changes in their placebo group by the study end could have reduced between-group differences. Post hoc analysis showed a trend towards response in those diagnosed with severe MDD. The study had notably stringent inclusion criteria to reduce the potential enrolment of patients with mild depression, a factor reported to impact placebo response rates in clinical trials [Posternak et al. 2002]. However, mean MADRS baseline scores (31.8–32.1) did not seem to differ widely with those reported by other authors in this review who used the same measure. Conversely, an earlier study of similar design (n = 362) [Reimherr et al. 1998] reported that bupropion dosed at 150 and 300 mg/day significantly reduced HAM-D and CGI-I total scores (p < 0.05 and p < 0.01, respectively) compared with placebo. Notably, this study had a lower study completion rate (54%) and, interestingly, of the 46% of patients that had prematurely discontinued, the majority were from the placebo group, with withdrawal due to inadequate response or because their condition had deteriorated.

A two-phase trial evaluated the long-term efficacy and weight change (described later) of fixed dose (300 mg/day) bupropion. Weihs and colleagues administered bupropion dosed at 150-300mg/day in an open-label design for 8 weeks (phase I) [Weihs et al. 2002]; those who responded were then recruited for a 44-week double-blind, placebo-controlled randomized fixed-dose (300 mg/day) study (phase II). In this study the primary outcome measure was time to relapse, defined as the prescription of a new antidepressant or the use of electroconvulsive therapy (ECT) and the study did not report additional data on changes in HAM-D scores. In phase II (n = 417) of those patients who relapsed, mean HAM-D scores were 21 (SD 4.4, range 11–30) and a statistically significant difference in favour of bupropion over placebo was demonstrated when comparing the survival curves for the two treatment groups (p = 0.004). In the placebo treatment group median time to relapse was 24 weeks after randomization compared with 44 weeks for the bupropion treatment group. Furthermore, survival estimates demonstrated that 52% of the placebo group would become depressed by the end of the study compared with 37% of the bupropion treatment group (p = 0.004) and that by end of year 1 the odds of placebo group requiring treatment were 1.83 times greater than for those in the bupropion treatment group.

In an inverse two-phase design with a much smaller sample (n = 16) Tomarken and colleagues administered bupropion dosed at 100–300 mg/day in a double-blinded, placebo-controlled randomized design for 6 weeks (phase I); at study endpoint participants from the bupropion group were titrated up to 400 mg/day and participants from the placebo group to 300 mg/day of medication, in an open-label design (phase II) for a further 6 weeks [Tomarken et al. 2004]. The primary aim of this study was evaluating efficacy of bupropion on specific symptom dimensions of depression (which are discussed later); however, the authors also administered the HAM-D. Both bupropion and placebo demonstrated linear declines on the HAM-D during phase I (p < 0.001) though the rate of change was greater for bupropion than placebo (p = 0.04). Although declines in mean scores were replicated in phase II (p = 0.005) no significant differences were observed between groups (p > 0.3).

In summary, bupropion demonstrated efficacy compared with placebo in five out of the six RCTs, and all but one of the open-label trials, with one of the two fixed dosing studies showing no difference. Study sizes in the RCTs were generally reasonable, with the exception of the Tomarken and colleagues cohort; follow-up periods were generally relatively brief, but consistent with typical antidepressant trials.

A subset of five studies reported sufficient data to allow for quantitative analysis using meta-analysis. Meta-analysis for the main effect of bupropion on depression scores as compared with placebo control showed a consistent large effect favouring bupropion (Hedge’s g = 2.02, df = 4, p < 0.001, 95% confidence interval [CI] 2.93–1.11; Figure 2). However, this analysis also revealed high heterogeneity of study findings (p = 0.001, I² = 79.4%, τ² = 0.832), which is reflected in a large prediction interval (PI 5.28 to −1.24). Moreover, both Egger’s (p = 0.043) and Begg’s (p = 0.027) tests for publication bias produced significant results. Thus, the strength of evidence produced by the present meta-analysis needs to be considered with extreme caution.

Bupropion versus SSRI

Ten RCTs identified evaluated the efficacy of bupropion against an SSRI: one compared with escitalopram [Clayton et al. 2006], five with sertraline [Coleman et al. 1999; Croft et al. 1999; Kavoussi et al. 1997; Rush et al. 2001, 2006], two with fluoxetine [Coleman et al. 2001; Feighner et al. 1991] and two with paroxetine [Grunebaum et al. 2012; Kennedy et al. 2006]. Nine trials employed a double-blinded, randomized design: six of which utilized a double dummy approach, which is a technique for retaining blinding when the two drugs cannot be made to appear identical; all participants thus take two treatments, one of which (depending upon the arm they are in) would be a placebo; and five also had a placebo arm. Trials varied in length from 8 to 24 weeks, and number of participants from 74 to 785.

Bupropion versus escitalopram

A large (n = 785) RCT by Clayton and colleagues randomized participants to receive bupropion XL (300–450 mg/day, n = 276), escitalopram (10–20 mg/day, n = 281) or placebo (n = 273) [Clayton et al. 2006]. A primary measure, reported later, was sexual functioning, but compared with placebo both drugs had statistically superior response (p = 0.015 and p < 0.001, respectively) and remission rates (p = 0.018 and p < 0.005, respectively), as measured by the HAM-D, with no difference between the treatment groups. However, separation from placebo was not achieved at the statistical level of 0.05 for bupropion for mean HAM-D total scores in the individual or pooled analysis. This is surprising given a statistically significant response rate was achieved in the pooled analysis for bupropion. The authors attribute this discrepancy to their sample size and large placebo response (53%).

Bupropion versus sertraline

Five studies compared bupropion and sertraline, two with placebo arms [Coleman et al. 1999; Croft et al. 1999]. Three employed a double-blinded, double dummy design [Coleman et al. 1999; Croft et al. 1999; Kavoussi et al. 1997], one a double-blinded design [Rush et al. 2001] and one was not blinded [Rush et al. 2006]. Trials ranged from 8 to 16 weeks and from 248 to 583 patients. The most recent RCT [Rush et al. 2001], which is 14 years old, employed a double-blind, randomized parallel design (n = 248) over a 16-week treatment phase. Mean HAM-D scores reduced for both treatment groups, with high response (bupropion, 66% and sertraline, 74%) and remission (bupropion, 55% and sertraline, 63%) rates reported, with no between-group differences. Interestingly a larger (n = 583) open-label trial by the same group [Rush et al. 2006] recorded significantly lower response rates despite administering bupropion at a similar mean dose; however, there was a broader inclusion of patients with other comorbid somatic or psychiatric disorders, and it is not clear to what degree these factors might have altered the results. Findings from the RCT are in accord with the earlier, similarly designed (n = 241) trial by Kavoussi and colleagues [Kavoussi et al. 1997]. Both groups demonstrated a ⩾50% improvement in HAM-D scores from week 6 to study endpoint (week 16) with no statistical between-group difference on the HAM-D, CGI-S or CGI-I. Both groups also demonstrated ⩾50% study-end improvement in anxiety scores, as measured by the HAM-A. Neither trial had a placebo arm, nor were p values reported.

Two early studies employed more rigorous, and similar, methodological designs. Coleman and colleagues reported mean HAM-D scores in the bupropion, sertraline and placebo groups improved by ⩾50% by week 8, but only bupropion was statistically significantly superior to placebo by this point (p < 0.05) [Coleman et al. 1999]. Croft and colleagues reported the same reduction in depression scores by week 8 (from week 6) (p < 0.05) LOCF analysis for both active treatment groups, with no statistically significant between them [Croft et al. 1999]. A statistically significant HAM-D clinical response was observed in bupropion (66%, p < 0.004) and sertraline groups (68%, p = 0.002) compared with placebo (47%). These two trials, plus work by Reimherr and colleagues [Reimherr et al. 1998] reported that individuals on bupropion showed significantly greater improvement on CGI-S and CGI-I scores. In the Coleman and colleagues study no difference between sertraline and placebo was established at any time for the CGI-S or CGI-I [Coleman et al. 1999]. Croft and colleagues reported that although no statistical difference between the groups was observed both treatments were statistically superior to placebo in CGI-S (bupropion, p = 0.005 and sertraline, p = 0.05) and CGI-I scores (p < 0.01) [Croft et al. 1999].

In summary, all trials that included a placebo arm reported bupropion as being significantly superior to placebo. In the majority of the studies, bupropion and sertraline demonstrated comparable efficacy, with the exception of the Coleman and colleagues [Coleman et al. 1999] cohort where bupropion but not sertraline was superior to placebo.

Bupropion versus fluoxetine

Two RCTs evaluated bupropion and fluoxetine, with similar study designs including placebo arms. The most recent trial was in 2001 [Coleman et al. 2001]; over the 8-week intervention mean HAM-D scores decreased across all groups (total n = 427). No statistical difference in response rates was observed between bupropion (56%), fluoxetine (57%) and placebo (50%); however, for remission (47%, 40%, 32%, respectively), a statistically significant greater rate was seen for bupropion, but not fluoxetine, over placebo (p < 0.05). The earlier trial [Feighner et al. 1991] had a smaller sample (n = 119) and shorter treatment phase (6 weeks); data similarly demonstrated no statistical difference in HAM-D, CGI-S or CGI-I scores, or response rates (bupropion, 62.5% and fluoxetine, 58.3%), with no statistical difference demonstrated between treatment groups.

Bupropion versus paroxetine

Two RCTs compared the efficacy of bupropion with paroxetine, neither utilizing a placebo arm. Grunebaum utilized an initial 8-week treatment phase (bupropion n = 38, paroxetine n = 36) in individuals with MDD and elevated suicidal risk factors, followed by a 16-week continuation phase design for those patients who had initially responded to treatment [Grunebaum et al. 2012]. The primary outcome was suicidal behaviour and ideation, measured on a modified HDRS scores (mHDRS; subtracting the suicide item). A reduction in mHDRS scores was reported for both active treatment groups; however, for each point more severe at mHDRS baseline, symptoms were 0.46 points lower with paroxetine by week 8. These data suggests that patients with more severe global depression symptoms (minus suicidality) improved modestly more with paroxetine when controlling for the suicidal ideation index at baseline. The earlier trial by Kennedy and colleagues had a larger sample (n = 131) and lower mean dosage of medication for both drugs, and reported a statistically significant reduction in HDRS scores for both treatment groups (p < 0.01), with no significant differences between bupropion and paroxetine on the HDRS scores, response or remission rates [Kennedy et al. 2006]. Overall the lack of a placebo arm hinders the ability to determine the absolute efficacy of the drug, though they produced comparable effects in terms of reduction in mean HDRS scores, with the latest study showing a particular advantage for paroxetine in suicidal patients.

Bupropion versus SNRIs

Four trials evaluated bupropion and a SNRI, three of which looked at venlafaxine [Hewett et al. 2009, 2010b; Thase et al. 2006] and one duloxetine [Rosso et al. 2012]. The venlafaxine trials were all double-blinded, double dummy designs, and two contained a placebo arm; sample size varied from 324 to 569 subjects and treatment phase from 8 to 12 weeks. The single trial evaluating bupropion and duloxetine had a small sample (n = 46) and short treatment duration (6 weeks) [Rosso et al. 2012].

Bupropion versus venlafaxine

The trial by Hewett and colleagues (n = 384) (Hewett et al. 2009) reported a statistically significant difference in mean MADRS total scores for bupropion and venlafaxine compared with placebo in the LOCF (p = 0.006 and p = 0.001, respectively) and observed case analyses (p = 0.003 and p < 0.001, respectively), as well as on CGI-I (p < 0.001 and p = 0.009), CGI-S (p = 0.003 and p < 0.001) and HAM-A (p = 0.019 and p < 0.001). A statistically significantly greater proportion of bupropion and venlafaxine patients, compared with placebo, met the criteria for response (p = 0.033 and p < 0.001, respectively) and remission (p = 0.004 and p < 0.001, respectively) at week 8, with no significant differences between the two active treatment groups.

However a larger (n = 390) subsequent trial by the same group [Hewett et al. 2010b] reported no statistical significance in the least squares mean change from baseline MADRS scores, at study endpoint (week 8) for bupropion (180 mg/day) compared with placebo. Given the limitations of the studies analysis protocol, no further comparisons at the 0.05 level between active treatment groups and placebo were deemed appropriate, and further results are therefore purely descriptive. Comparison of patients classified as responders or remitters (according to MADRS and CGI-I criteria) were significant for venlafaxine (p < 0.05) but not bupropion, and this was also the case for CGI-S and HAM-A scores (p < 0.01, venlafaxine). In this latter study, Hewett and colleagues argued that the study enrolled a population that was inherently less responsive to bupropion, and more so to venlafaxine, although they recognized that such an argument was speculative.

Thase and colleagues conducted a randomized, double-blind, non-placebo-controlled study comparing bupropion and venlafaxine in 324 outpatients with MDD, and reported similar response and remission rates for the bupropion and venlafaxine treatment groups [Thase et al. 2006]. Although the study reported that a significant difference in favour of bupropion was observed in the LOCF analysis for percentage of patients categorized under remission (for both HAM-D and CGI-I criteria) no p values were provided to further interpret this, and antidepressant efficacy was a secondary aim of the study.

In summary bupropion up to doses of 450 mg and venlafaxine up to 225 mg showed comparable levels of antidepressant activity in two trials. One trial indicated superiority of venlafaxine over bupropion and placebo, however results should be interpreted as descriptive. In some studies mean severity scores were trending downward at the end of treatment, suggesting there was potential for patients to achieve remission if treatment phases were extended.

Bupropion versus duloxetine

One study evaluated the efficacy of bupropion (300 mg/day) versus duloxetine (120 mg/day) in a small-scale (n = 46) 6-week single-blinded, randomized trial in individuals with MDD resistant to SSRIs [Rosso et al. 2012]. Rosso and colleagues reported that both groups demonstrated a statistically significant reduction in mean HAM-D and CGI-S scores at study endpoint (p < 0.001), however no differences between treatment groups were found at any time point. Although a similar percentage of patients were categorized as responders (60–70%) and remitters (30–40%), there were no statistical differences between the treatment groups or from baseline to study endpoint (week 6). Whilst the results are positive, given the sample’s refractory cohort, and support class change in MDD refractory to SSRIs, interpretation is hindered by the lack of a placebo arm and the small sample size.

Bupropion versus trazadone

One study, a RCT, evaluated the relative efficacy of bupropion (225–450 mg/day) and trazadone (150–400 mg/day), a SARI [Weisler et al. 1994]. After a 1-week placebo lead in, 124 outpatients with mild–moderate depression were randomized to one of the two drugs. At the 6-week follow up there were no significant differences between the compounds, though trazadone showed earlier gains that might have been due to improved sleep. A total of 58% of bupropion and 46% of trazadone patients were rated as much or very much improved by the end, though the study lacked a placebo arm, and only numerical values of improvement were provided.

Bupropion versus TCAs

A double-blind, double dummy RCT [Masco et al. 1994] evaluated bupropion (n = 55) against nortriptyline (n = 50) over a 6-week treatment phase. Improvements on HAM-D scores and response rate (bupropion, 40% versus nortriptyline, 48%) were comparable across the two treatment groups. As with the trazadone study, there were some early benefits for nortriptyline that the authors considered might be due to a disproportionate weighting of the sleep index on the HAM-D.

Bupropion coprescribed with SSRIs

Five reports evaluated the efficacy of adding bupropion to SSRIs. A methodologically unique work by Weissman and colleagues followed the progress of mothers with depression (n=76) randomized to receive bupropion, escitalopram, or their combination over 12 weeks [Weissman et al. 2015]; their children’s (n = 135) wellbeing was independently assessed. There were no between-group differences for the women, all groups showing statistically significant improvements as measured on the HAM-D (p < 0.001); there was an overall remission rate of 67%. However, the effect upon their children varied, and depended upon the mothers’ baseline symptom profiles; mean Children’s Depression Inventory (CDI) scores declined significantly among children whose mothers received escitalopram monotherapy compared with both bupropion monotherapy (p < 0.04) and combination treatment of bupropion and escitalopram (p < 0.001). Subanalysis of this interesting finding showed that those children with mothers with a so-called high ‘negative affectivity’ (which includes the symptoms of guilt, irritability and fear/anxiety) only improved when their mothers were on escitalopram; mothers in this group reported improvements in their abilities to listen to their children, and the children described the mothers as becoming more caring during treatment. Negative affectivity has been linked with serotonergic dysfunction, which may explain bupropion’s lack of impact on this domain.

A single-blind RCT [Gulrez et al. 2012] evaluated 60 outpatients showing a partial response on SSRIs (escitalopram 10–30 mg/day, citalopram 20–60 mg/day, paroxetine 25–75 mg/day and sertraline 50–200 mg/day; all on treatment for 4 weeks). Participants were randomly assigned to have either placebo or bupropion SR (150–300 mg) added to their antidepressant. By the end of week 4, both groups had significantly improved, but the decrease in depression scores (measured on the HDRS, MADRS and ADI) were significantly greater in those coprescribed bupropion.

Three open-label studies evaluated the addition of bupropion to citalopram or escitalopram, a SSRI that is highly selective for the serotonin reuptake transporter. A novel design naturalistic study by Lam and colleagues explored treatment strategies, whether to augment or switch medication, in 61 individuals taking either citalopram or bupropion SR (at a therapeutic dose but showing minimal improvement) for at least 6 weeks [Lam et al. 2004]. Eligible participants, all of whom had failed to respond to at least one previous antidepressant, had, in alternate months, the other medication added to their treatment, or were switched to the other drug. An advantage of such a methodology is that it attempts to address the issue of whether it is the combination of medications (which could be additive or synergistic) or the novel compound that accounts for any improvement. At 6 weeks treatment, the combination condition (n = 32) was superior to a monotherapy switch (n = 29) in terms of clinical change on the SIGH-SAD score (p < 0.04) and the proportion of participants in remission (28% versus 7%, p < 0.05). The combination remission rate was low, but with the caveat that it was measured at 6 weeks treatment. Leuchter and colleagues examined the coprescribing of citalopram (mean dose 18 mg/day) and bupropion SR (mean dose 329 mg/day) over 12 weeks in 51 medication-free outpatients with chronic or recurrent MDD [Leuchter et al. 2008]. Participants were commenced on escitalopram 10 mg/day, with bupropion SR 150 mg/day added at week 1; they were eligible for an increase in either medication by week 4 (and for bupropion also at weeks 6–10) if they failed to show any, or suboptimal, improvement; mean doses by the trial end were 18 and 329 mg/day, respectively. A total of 62% showed a response to treatment, 50% attained remission (QIDS-C16 ⩽ 5). Rates of discontinuation due to side effects were low, at 6%. Similarly, Mohan and colleagues prescribed 10-20 mg/day of escitalopram to 135 participants with MDD in a 12-week trial (those showing ⩽50% decrease in MADRS scores at week 4 being put on the higher dose) [Mohan et al. 2009]; nonresponders at week 12 were then coprescribed 150–300 mg/day (depending upon response) bupropion for an additional 6 weeks. By the end of 12 weeks 60.7% were responders, 58.5% remitters on citalopram monotherapy; 41 participants defined as nonresponders entered the coprescribing phase, and 61.0% attained response, 53.7% remission in the 6-week follow on. Interestingly, subanalysis of symptom changes showed that melancholic features of depression were less responsive to escitalopram monotherapy, but that these appeared to respond well to bupropion augmentation.

Overall, with regards to adding bupropion to SSRIs, there was only one double-blinded RCT, though this did not have a placebo group, and was in a specific cohort of depressed mothers. In this study bupropion and bupropion coprescribing showed results comparable, though not superior to, SSRI prescribing; the data are interesting in terms of the differential outcomes in the children. The data from the single-blinded RCT and open-label trials support superior efficacy of bupropion, though the sample sizes are modest, and methodologically, open-label trials, with their lack of placebo, are open to challenge.

Bupropion coprescribed with non-SSRIs

Two studies compared the efficacy of bupropion in adjunction with a non-SSRI antidepressant, in both instances a SNRI.

A 6-week double-blind placebo-controlled RCT [Fornaro et al. 2014] evaluated bupropion SR (150 or 300 mg/day) or placebo added to duloxetine (60–120 mg/day) in 48 outpatients with MDD (all of whom had failed at least one drug trial) with DSM-IV criteria atypical features (which include increased appetite and food consumption, weight gain, hypersomnia and atypical diurnal variation in mood). No statistically significant differences were seen between groups, with measurements including the Structured Interview Guide for the HAM-D with Atypical Depression Supplement (SIGH-ADS) in addition to the GAF and HAM-D; 26.1% (duloxetine) and 21.7% (placebo) response rates were attained. The authors noted that the presence of a higher number of atypical features, which generally occur in up to a fifth of those with MDD, significantly predicted nonresponse, as has been demonstrated in other studies, and may have affected their outcomes.

Spier evaluated the use of adding bupropion to patients with MDD in a private practice clinic already on venlafaxine or an SSRI in an open-label trial assessing symptom relief and management of drug-induced side effects in monotherapy nonresponders (n = 15) and responders (n = 10) with side effects, respectively [Spier, 1998]. A numerical reduction in average CGI scores was observed after response to second agent had stabilized, scores reducing from 5.2 (4–7) to 2.2 (1–4). A total of 12 of the 15 monotherapy nonresponders showed a response after the addition of bupropion, but only two of the ten prescribed this to manage medication side effects showed any amelioration of their problems. The small sample size, and a very heterogeneous population, inevitably means these results should be treated with caution.

Bupropion as part of multiple medication trials

Six studies compared the efficacy of bupropion in multiple medication trials.

As part of the highly influential STAR*D project, Trivedi and colleagues randomized (but in an unblinded fashion) 851 individuals with MDD not remitting on citalopram (mean dose 55 mg/day, mean duration 11.9 weeks) to additionally receive either bupropion XL (n = 565, dose of up to 400 mg/day) or buspirone (n = 286, dose of up to 60 mg/day); there was no placebo arm [Trivedi et al. 2006a]. Both treatments had similar results on the HRSD-17 remission, QIDS-SR-16 remission, and QIDS-SR-16 response rates; however bupropion had a greater reduction (p < 0.04) and overall score (p < 0.02) in QIDS-SR-16 total scores, and a lower side-effect drop-out rate (p < 0.009) by the 12-week study end. Subsequent analysis by the same group [Bech et al. 2012] focused on what the authors termed the ‘pharmacopsychometric triangle’ in comparing bupropion with buspirone augmentation. Under this model, which composites the domains of antidepressive activity, side effects and quality of life, bupropion was superior in all domains (though not to statistical significance for ‘quality of life’).

In a double-blind RCT Blier and colleagues randomized 105 medication-free participants with MDD to receive either fluoxetine monotherapy (n = 28, dose 20 mg/day) or one of three mirtazapine (30 mg/day) combination groups: with fluoxetine (20 mg/day, n = 25); with venlafaxine (225 mg/day, n = 26); or with bupropion (150 mg/day, n = 26) over a 6-week period [Blier et al. 2010]. Although, similarly to the STAR*D work, there was no placebo group, the monotherapy group allowed for evaluation of the hypothesis that any response or remission might be solely due to having more time on an initial compound. There were no significant differences in inter-group drop-out rates, suggesting that polytherapy was tolerated well; and all of the combination groups showed significantly greater improvement on HAM-D scores compared to fluoxetine monotherapy, producing a number needed to treat (NNT) of 3–5 over monotherapy. However, the combination protocols did not result in any greater rapidity of response and there were no significant between-combination-group differences on the MADRS or CGI. Furthermore, discontinuation of any drug in those who had shown a marked response led to relapse in approximately 40% of patients.

A post-hoc analysis of data from a subgroup of patients enrolled in a large (n = 296) 52-week open-label multicentre study [Clayton et al. 2014] evaluated the addition of aripiprazole to either bupropion (n = 47) or SSRIs/SNRIs (n = 245, data pooled) in individuals with MDD. The primary outcome measurements were the safety and tolerability of these combinations, which are reported elsewhere in this paper. For both groups, LOCF improvements occurred over the course of the 1-year trial; mean changes of −1.4 points in CGI-S scores occurred in the bupropion group and −1.5 points in the SSRI/SNRI group (observed case n = 76). In a 6-week prospective open-label study, DeBattista and colleagues examined the efficacy of bupropion XL (150 or 300 mg/day) added to 25 participants’ existing antidepressants (all with MDD and an inadequate response on their current treatment of ⩾4 weeks), which included SSRIs (fluoxetine, paroxetine and sertraline) and the SNRI venlafaxine [DeBattista et al. 2003]. A statistically significant reduction in HDRS symptoms was demonstrated by the trial end (p < 0.001), with 54% demonstrating a clinical response of ⩾50% symptom reduction. No between drug differences were noted, though the sample size was very small. Bares and colleagues compared the efficacy of antidepressant monotherapy (ADM) with combinational antidepressant (CAD) treatment in 60 inpatients with treatment-resistant depression (TRD) in a randomized 6-week open-label study [Bares et al. 2013]. After a short initial washout period of a day or two, eligible participants were randomly allocated to either 6-week ADM or CAD groups, with responders from both entered into a further 8-week follow-up protocol. The new treatments were chosen by clinicians based upon professional judgement of the patients mental state and past psychiatric histories, excluding medications that had previously failed, or drugs of the same class (with the exception of SSRIs). There were no differences from baseline between ADM and CAD groups as measured by change on MADRS or response rates; whilst bupropion was administered to participants in this study, the total number on this in both groups (and the combinations in the CAD group) are not recorded, nor are the other drugs reported.

Interpreting these findings, the methodologically strongest work, by Trivedi and colleagues and Blier and coworkers support the addition of bupropion to a range of existing first-line antidepressants [Trivedi et al. 2006a; Blier et al. 2010]. However, there is little to delineate treatment options further, and the lack of placebo arms is notable.

Older adults

Three studies evaluated the use of bupropion in older adults (⩾65 years of age) with MDD. Two were RCTs, evaluating bupropion in comparison with placebo [Hewett et al. 2010a] and with paroxetine [Weihs et al. 2000]. One naturalistic study investigated the use of both bupropion monotherapy and its combination with SSRIs (Steffens et al. 2001). Trials ranged from 6-12 weeks with bupropion doses ranging from 100-400 mg/day. All three studies showed bupropion to be significantly efficacious.

The most recent double-blind RCT, by Hewett and colleagues, demonstrated efficacy for bupropion XR in comparison with placebo (p < 0.05) over 10 weeks in a large sample (n = 418) [Hewett et al. 2010a]; from baseline the median changes in MADRS total score were −15.0 and −11.0 for bupropion XR and placebo, respectively. Weihs and colleagues (n = 100) found significant efficacy for both bupropion SR (n = 48) and paroxetine (n = 52) for the treatment of depression in a double-blinded, double dummy RCT [Weihs et al. 2000]. Mean HAM-D (LOCF) scores were similar at baseline for both treatment groups with a 59% reduction in total score for bupropion SR and a 63% reduction in total score for paroxetine by week 6. LOCF analyses found no significant difference in mean HAM-D scores between the two drugs. Secondary analyses of CGI-S, CGI-I and HAM-A scores showed improvement in both treatment groups at week 6. No significant differences were found between treatment groups in mean CGI-S, CGI-I or HAM-A scores.

In the naturalistic trial, Steffens and colleagues (n = 31) investigated the use of bupropion SR or IR monotherapy (n = 15) and its combination with an SSRI (n = 16):with prescribing based upon clinicians’ choices [Steffens et al. 2001]. For bupropion monotherapy, 61% were classified as responders (MADRS < 15), with 50% achieving partial (CGI = 2) or complete (CGI = 1) remission. Of those receiving bupropion and SSRI combination therapy 81.2% were classified as responders (MADRS < 15) with 56.2% showing partial (CGI = 2) or complete (CGI = 1) remission. Across both types of intervention, 74.2% were classed as responders (defined as a MADRS < 15), with 53.3% achieving a partial (CGI = 2) or complete (CGI = 1) remission.

Of the three studies identified that evaluated the use of bupropion in the treatment of MDD in the elderly all demonstrated significant improvement on depression scores. However, given the general dearth of research in this population and that the current studies were not methodologically strong results should be interpreted with caution.

Bipolar affective disorder

Although defined by pathological mood elevation, depression (so-called bipolar depression) constitutes the majority of illness burden [Lloyd et al. 2011]. Three studies investigated the use of bupropion in this group: one double-blind RCT compared it to the selective α2 adrenergic antagonist idazoxan [Grossman et al. 1999], a single-blinded RCT compared with the anticonvulsant and mood stabilizer topiramate [McIntyre et al. 2002], and a mixed-design study compared bupropion with sertraline and venlafaxine in combination with mood stabilizers [Post et al. 2006]. Trial length varied from 6 to 8 weeks with bupropion doses ranging from 75 to 450 mg/day.

In a small study of 14 individuals with bipolar depression, Grossman and colleagues found a significant reduction in HDRS total score by the end of the 6-week trial for both bupropion and idazoxan [Grossman et al. 1999]. In a slightly larger (n = 36) and longer trial, McIntyre and colleagues reported a significant improvement in HDRS-17 total score from baseline to endpoint for both bupropion SR and topiramate (p < 0.001) [McIntyre et al. 2002]. Both treatment groups demonstrated significant response rates (≥ 50% reduction in HDRS-17); bupropion SR (59%) and topiramate (56%) (p = 0.03 and p = 0.04, respectively), with no significant between-group differences being observed (p = 0.097). This was also reflected in CGI-I total scores, where a significant reduction was observed for both groups at week 8 (p < 0.005), with no significant differences between bupropion and topiramate (p = 0.092).

Post et al. (2006) evaluated the effects of bupropion, sertraline and venlafaxine as adjuncts to mood stabilizers. This 10-week study consisted of a cohort of 174 outpatients with bipolar disorder; 28 of which were treated with open medication and 156 studied in a randomized, double-blind double dummy fashion. Continued medication included lithium (n=68), valproate (n=93), carbamazepine (n=16), lamotrigine (n=8), typical antipsychotics (n=8) and atypical antipsychotics (n=30). Similar response (≥50% improvement in IDS score) and remission rates (IDS score <12) were recorded across all three treatment groups at study end, with no statistically significant differences demonstrated (p = 0.68 and p = 0.29, respectively).

Collectively these studies suggest efficacy of bupropion in the treatment of bipolar depression, although no significant differences were observed between the active comparators. However, interpretation is once again restricted due to the dearth in research within this population and notable methodological issues including a small sample size and absence of a placebo group.

Seasonal affective disorder

Three studies evaluated the use of bupropion in SAD. One prospective study utilized bupropion prophylactically in a double-blinded, randomized design [Modell et al. 2005], and two adopted open-label designs [Dilsaver et al. 1992; Seo et al. 2013]. Studies ranged from 5 to 8 weeks in length, with doses of bupropion ranging from 150 to 400 mg/day.

In their large (n = 1042) multisite RCT of individuals with a diagnosis of SAD but currently asymptomatic, Modell and colleagues randomized participants to receive either bupropion or placebo, and found a significant reduction in recurrence rates by the following Spring in the active group [Modell et al. 2005]. Of the total sample, 16% of patients experienced a recurrence of major depression whilst on bupropion XL, compared with 28% of patients in the placebo group, with an overall relative risk of 0.56. Bupropion XL was favoured over placebo across all three study sites p = 0.081, p = 0.057 and p = 0.001, respectively). Suggesting there is potentially a role for bupropion in preventing recurrence of seasonal major depressive episodes, although further methodologically sound research is required to support this initial data.

Both open-label trials demonstrated a significant reduction in depressive symptoms. The more recent work by Seo and colleagues (n = 52) found a significant reduction in SIGH-SAD total score by 53% from baseline (27.6 ± 6.5) to week 8 (12.2 ± 6.3) (p < 0.001) [Seo et al. 2013]. The earlier, and considerably smaller study (n = 15) by Dilsaver and colleagues found a significant improvement on HAM-D total score from baseline (18.3 ± 6.9) to study end (week 5; 3.2 ± 2.2) with an average reduction of 15.2 (p < 0.0001) [Dilsaver et al. 1992]; there were also significant reductions of modified HAM-D (mHAM-D; modified to include scores on hyperphagia and weight gain) total score from baseline (25.5 ± 6.4) to treatment end (4.1 ± 3.1) with an average reduction of 21.4 (p < 0.0001).

In summary, Modell and colleagues demonstrated the potential prophylactic potential of bupropion in patients with SAD [Modell et al. 2005] although this finding is yet to be replicated. Open-label trials have also demonstrated the efficacy of bupropion in reducing depressive symptomology. However, given the limited number of trials these findings should be considered preliminary.

Dysthymic disorder

A single open-label study evaluated bupropion SR in the treatment of dysthymic disorder, in an 8-week trial consisting of 21 patients [Hellerstein et al. 2001]. A significant improvement in depression scores was demonstrated at study end; HAM-D scores decreased by 73% from 21.71 (SD 5.57) at baseline to 5.90 (SD 3.60) at week 8 (p < 0.001) as did the Cornell Dysthymia Rating Scale (CDRS) total score from baseline (36.33, SD 9.85) to week 8 (12.43, SD 7.90) (p < 0.001). Although in support of bupropion in the treatment of dysthymia, methodologically sound replication is required to validate these preliminary positive findings.

Postpartum depression

One small (n = 8) study evaluated bupropion SR in the treatment of postpartum depression in an 8-week open-label trial [Nonacs et al. 2005]. Overall, patients showed a significant response to bupropion treatment, with a significant reduction in HAM-D by week 4 (p < 0.05), CGI and Kellner depression scores at study end (p < 0.005). Given the small sample and the study design, further validation of findings is required in support of this initial data.

Weight change

Fourteen studies (total n = 8137 participants with MDD), ranging from 6-52 weeks study duration reported bupropion-related weight changes (see Table 3). Overall, the majority concurred that there was significant weight loss in bupropion-treated patients [Coleman et al. 1999; 2001; Croft et al. 1999; 2002; Jefferson et al. 2006; Reimherr et al. 1998; Settle et al. 1999; Weihs et al. 2002], though some studies reported a small weight increase [Blier et al. 2010; Clayton et al. 2014] or no significant change [Hewett et al. 2009, 2010b; Koshino et al. 2013; Thase et al. 2006].

Six double-blinded RCTs reported significant weight loss during bupropion treatment, including two long-term (52-week) follow-up studies. In the work by Weihs and colleagues (the antidepressant efficacy data are reported earlier in this review), there was an initial 8-week open-label phase of 518 participants with recurrent major depression treated with bupropion SR 300 mg/day [Weihs et al. 2002]; following this the responders (n = 471) entered a 44-week double-blind, placebo-controlled phase where they were randomized to continue bupropion or to switch to placebo. At week 52, no fluctuations in weight were found in the placebo group compared with baseline (+0.18 kg; SD 11.0), whereas the bupropion SR group had a mean weight loss of 1.13 kg (SD 15.8); as noted earlier the bupropion cohort had significant improvements in mood over placebo, and one might otherwise expect more illness-driven weight loss in the more depressed placebo group. Interestingly, one bupropion-treated patient discontinued from the study due to weight loss of approximately 7.8 kg or 15.2% of her baseline body weight. In a similarly designed open-label study, Croft et al. (2002) reported a mean weight loss of 1.4 kg by week 8 in the open-label phase; following double-blind randomization at that point, mean weight changes of −1.15 kg were reported for the bupropion group (N = 2101) and +0.02 kg for the placebo arm (n = 213) by week 52 [Croft et al. 2002].

Five briefer 8-week, multicentre, randomized, double-blind, double dummy placebo-controlled studies (totalling 2957 subjects with MDD), also reported weight decreases with bupropion [Coleman et al. 1999, 2001; Croft et al. 1999; Reimherr et al. 1998; Settle et al. 1999]. Across these trials median weight loss ranged from 0.9 to 1.49 kg, compared with losses of 0.5 kg on sertraline and 1.49 kg on fluoxetine, and weight gain of 0.1–0.2 kg in placebo groups. In the studies by Reimherr and colleagues and Settle and coworkers, weight loss was dose dependent, with larger mean decreases on higher dosing regimens [Reimherr et al. 1998; Settle et al. 1999].

Four double-blind RCTs reported minimal weight changes over 8 to 12 weeks of bupropion treatment (n = 2070), though only the work by Koshino and colleagues reported statistical comparisons [Koshino et al. 2013]; they found no significant differences in weight change in a cohort of 569 depressed Asian subjects treated with bupropion 150 mg/day, 300 mg/day or placebo. Thase and colleagues found a mean weight decrease in the bupropion XL group of 0.1 kg compared with a gain of 0.1 kg in the venlafaxine XR group [Thase et al. 2006]. Hewett and colleagues described minimal 8-week weight changes for bupropion, venlafaxine and placebo groups [Hewett et al. 2009, 2010b].

Two studies (n = 358) reported mean weight gains associated with bupropion treatment, though these were studies where it was co-administered with another antidepressant. Blier and colleagues (n = 66) reported significant mean weight increases in the bupropion plus mirtazapine (M = 2.7 kg, SD 2.4), venlafaxine plus mirtazapine (M = 2.2 kg, SD 2.5) and fluoxetine plus mirtazapine groups (M = 3.1 kg, SD 2.5) as compared with fluoxetine monotherapy (p < 0.001) [Blier et al. 2010]. In a 52-week open-label study (n = 292) by Clayton and colleagues mean weight gains of +3.1 kg for bupropion plus aripiprazole, and +2.4 kg for an SSRI/SNRI plus aripiprazole were reported [Clayton et al. 2014].

Sexual functioning

Eleven studies (total participant n = 5582) measured effect of bupropion on sexual functioning in patients with MDD (see Table 3). The majority divided sexual dysfunction into sexual desire disorder, sexual arousal disorder and orgasm dysfunction, while one study reported sexual functioning using a single measure. The consensus across the literature was that bupropion did not differ significantly from placebo in causing sexual dysfunction; whilst it improved sexual functioning in patients experiencing SSRI-induced sexual dysfunction.

Regarding causing sexual dysfunction, Croft and colleagues and Coleman and coworkers both reported that significantly more patients treated with sertraline experienced orgasmic dysfunction than subjects treated with bupropion SR or placebo, with an onset of orgasm dysfunction occurring as early as day 7 in up to a sixth of those on sertraline [Croft et al. 1999; Coleman et al. 1999]. Following a 16-week randomized clinical trial (N = 248), Segraves and colleagues reported that compared with bupropion, more patients in the sertraline group experienced sexual desire disorder (63% of men and 41% of women) and sexual arousal disorder (19% in men, p = 0.02; 12% in women, p = 0.05) [Segraves et al. 2000]; bupropion treatment was related with an increase in sexual satisfaction (from 57% to 79%). Clayton and colleagues combined data from two double-blind, placebo-controlled RCTs comparing bupropion (n = 276) to escitalopram (n = 281) on measures of sexual functioning [Clayton et al. 2006]. They found, in both the individual studies and the pooled dataset, that the incidence of orgasm dysfunction and the incidence of worsened sexual functioning at the end of the treatment (week 8) were statistically significantly lower with bupropion and placebo than with escitalopram (p < 0.05), and not statistically different between bupropion and placebo. Similarly, Hewett and colleagues reported no significant difference between bupropion patients and placebo on the Changes in Sexual Functioning Questionnaire (CSFQ), while venlafaxine treated subjects experienced significant decreases on arousal and orgasm subscales (p < 0.05) in an 8-week double-blind, placebo-controlled RCT (n = 591) [Hewett et al. 2010b]. A double-blind, double dummy RCT [Coleman et al. 2001] compared bupropion with fluoxetine. The overall incidence of sexual desire disorder significantly decreased only in the bupropion group by the week 8 endpoint (p < 0.05). Sexual arousal disorder occurred in significantly more fluoxetine- than bupropion- or placebo-treated patients (p < 0.05), with no difference between bupropion and placebo at any point. The incidence of orgasm dysfunction in those on fluoxetine 20 mg/day was more than twice as high than in those on bupropion 300 mg/day and, interestingly, three times higher than that of those on bupropion 400 mg/day. Subjectively, more patients in the bupropion group were satisfied with their sexual functioning than in the placebo or fluoxetine groups.

The important issue of potential gender differences in sexual functioning was taken into account by two randomized, double-blind trials [Kennedy et al. 2006; Thase et al. 2006] (total n = 473 subjects with MDD but without sexual dysfunction). Thase and colleagues reported that for men, there was a statistically significant difference in sexual functioning in favour of bupropion XL for all comparisons with venlafaxine (p ⩽ 0.048); for women, this was also true at week 5, week 6, and for the average across weeks 5–12 (p ⩽ 0.043). Notably, bupropion XL was also superior to venlafaxine XR on the CSFQ subscales for pleasure, desire/frequency, desire/interest, arousal and orgasm throughout the study period. Following an 8-week treatment period with either bupropion or paroxetine, Kennedy and colleagues found that women (n = 68) in the bupropion group experienced significantly higher levels of Sexual Functioning (Sex FX; total, p < 0.01; desire, p < 0.01; arousal, p < 0.01; overall satisfaction, p < 0.01) [Kennedy et al. 2006]. In men (n = 73), the analysis revealed a significant treatment effect by visit for Sex FX total (p < 0.001), desire (p < 0.001) and overall satisfaction (p < 0.01). The effect was due to a significant decrease in sexual functioning during paroxetine treatment (Sex FX total, p < 0.002; desire, p < 0.005; arousal, p < 0.005; and overall satisfaction, p < 0.057); at week 8 the paroxetine groups displayed a significant deterioration from baseline, on Sex FX total (p < 0.01), desire (p < 0.01), arousal (p < 0.05), orgasm (p < 0.01) and overall satisfaction (p < 0.01) scores, whereas no significant change was observed in any of these measures across visits in men randomised to the bupropion SR group.

Several studies evaluated bupropion’s effectiveness at reducing existing sexual dysfunction: four studies measured changes in sexual functioning when subjects were either switched from their previous SSRI/SNRI to bupropion monotherapy or bupropion was added to their current SSRI treatment plan. Clayton and colleagues reported significant improvements in global CSFQ scores by week 2 in the coprescribing group (SSRI/SNRI+bupropion) and from week 2 to 4 in the monotherapy group (discontinuation from SSRI/SNRI to bupropion monotherapy), with gains over baseline maintained by the 8-week endpoint (p < 0.05) [Clayton et al. 2001]. A similar pattern was observed on the subscales of sexual desire and orgasm, mean scores differing significantly (p < 0.05) from baseline to weeks 4 and 8, indicating improvement following SSRI discontinuation and the start of bupropion monotherapy, but not for coprescribing. In a later, placebo-controlled, double-blinded study (n = 42) by the same group, Clayton and colleagues reported that neither the addition of bupropion nor placebo produced a change in CSFQ total scores after 4 weeks [Clayton et al. 2004]; however, those coprescribed bupropion self-reported improved desire and frequency of sexual activity (p = 0.024). Dobkin and colleagues conducted a 10-week open-label study of 18 ethnic minority depressed women with poor response to current SSRI treatment and hypoactive sexual desire [Dobkin et al. 2006]. Compared with baseline measurement, after 8 weeks of bupropion monotherapy (i.e. week 10), there were significant improvements in desire (p < 0.001), arousal (p < 0.001) and orgasm (p < 0.001). Post-hoc analyses indicated that improvements were observed as early as week 2 (end of cross-taper) for desire scores (p = 0.001) and at week 4 for arousal (p < 0.001) and orgasm scores (p < 0.001). More recently, a 12-week double-blind RCT of 218 women with SSRI-induced sexual dysfunction reported the Female Sexual Function Index total score was higher in those coprescribed bupropion SR than in those randomized to placebo (p = 0.001). In addition, at the end of the trial the mean scores for desire, arousal, lubrication, orgasm, and satisfaction were significantly higher in the bupropion group (p < 0.01) [Safarinejad, 2011].

Other side effects

The majority of studies identified for review described side effects as frequently reported (generally defined as occurring in at least 5% of patients). Overall, following bupropion treatment, side effects experienced were generally of mild to moderate intensity, and the most commonly reported were: dry mouth (5–34.5%), insomnia (1–27.8%), headache (3–34%) and nausea (7–21%). Other rarely reported side effects included restlessness, anxiety, constipation, dizziness, nasopharyngitis and fatigue. Discontinuation rates due to adverse events ranged from 0 to 55%, however, bupropion was not significantly different from other antidepressants and in some cases placebo treatments.