Whilst nicotine replacement is the most widely used pharmacotherapy for smoking cessation, some people prefer a treatment that does not use nicotine. Others require alternative treatments having failed to quit with nicotine replacement. Observations that a history of depression is found more frequently amongst smokers than nonsmokers; that cessation may precipitate depression; that nicotine may have antidepressant effects; and that antidepressants influence the neurotransmitters and receptors involved in nicotine addiction provided a rationale for the study of antidepressant medications for smoking cessation (Benowitz 2000; Kotlyar 2001).
The following antidepressants have been investigated for their effect on smoking behaviour in at least one study:
- the tricyclic antidepressants (TCAs) doxepin, imipramine and nortriptyline
- the monoamine oxidase inhibitors (MAOI) moclobemide and selegiline
- the selective serotonin reuptake inhibitors (SSRIs) fluoxetine, paroxetine and sertraline
- the atypical antidepressants bupropion, tryptophan and venlafaxine
- extracts of Hypericum perforatum L. (St. John's wort)
The focus of this review and meta-analysis is on trials that provide evidence for an effect on long-term smoking cessation. We describe these in the Results section. For pharmacotherapies for which there is still a lack of long-term data, we briefly describe results from excluded short-term trials in the Description of Studies section.
To assess the evidence for the efficacy and safety in assisting long-term smoking cessation of medications with antidepressant properties, including: bupropion, doxepin, fluoxetine, imipramine, moclobemide, nortriptyline, paroxetine, tryptophan, selegiline, sertraline, venlafaxine and hypericum (St John's Wort).
For each medication identified as having been used in a smoking cessation trial we tested the hypothesis that it was more effective than placebo, or an alternative treatment, in achieving long-term smoking cessation.
Criteria for considering studies for this review
Types of studies
For efficacy, we examined randomized trials comparing antidepressant with placebo or with an alternative therapeutic control, or comparing different dosages of an antidepressant that reported six-month or longer follow ups. For safety, we examined data from randomized controlled trials and non-randomized post-marketing surveillance data.
Types of participants
Current smokers, or recent quitters (for trials of relapse prevention).
Types of interventions
Treatment with any medication with antidepressant properties to aid a smoking cessation attempt or to prevent relapse, or to reduce the number of cigarettes smoked and aid subsequent cessation. Trials in which all participants received the same pharmacotherapy regimen but different behavioural support were not included.
Types of outcome measures
Efficacy was measured via a) abstinence from smoking or b) incidence of reducing cigarette consumption to 50% or less of baseline, both assessed at follow up at least six months from start of treatment. Safety was assessed by incidence of serious and other adverse events, and drop-outs due to adverse events.
Search methods for identification of studies
We identified studies from the Tobacco Addiction Group's specialised register. All trials using pharmacotherapy other than nicotine, clonidine or lobeline for smoking cessation were found, and those using medications generally classified as having an antidepressant effect were selected for inclusion in this review. The date of the last search was June 2009. We checked the citation lists of these studies, recent reviews of non-nicotine pharmacotherapy and abstracts from the meetings of the Society for Research on Nicotine and Tobacco. For each medication found from these sources we searched MEDLINE and EMBASE (via Ovid, 9th June 2009) using the medication name and 'smoking' as a free text term. Several studies were located by contacting investigators in the area. We also checked all records of trials of bupropion held on the GlaxoSmithKline Clinical Trials Register (http:ctr.glaxowellcome.co.uk) for unpublished studies.
Data collection and analysis
LS and TL independently extracted data on the number of study participants who had ceased to smoke at final follow up.
In each study we used the strictest available criteria to define cessation, so we extracted figures for sustained abstinence in preference to point prevalence where both were presented. In studies that used biochemical validation of cessation, only those subjects meeting the criteria for biochemically confirmed abstinence were regarded as having stopped smoking. We treated subjects in either group lost to follow up as continuing smokers. As far as possible we used an Intention-to-Treat analysis. Where subjects appeared to have been randomized but were not included in the data presented by the author we noted this in the study description (see 'Characteristics of Included Studies'). Assuming that people lost to follow up are smokers will ensure that actual quit rates are conservative, but may not necessarily lead to conservative relative treatment effects (e.g. risk ratios), if loss to follow up is higher in the control group (Hall 2001). Some studies now use alternative methods to model effects of missing data (Hall 2001; Niaura 2002). Where differential results using alternative models were reported we considered whether the results of the meta-analysis were sensitive to the use of different denominators.
We summarized individual study results as a risk ratio (RR), calculated as: (number of quitters in intervention group/ number randomized to intervention group) / (number of quitters in control group/ number randomized to control group). A risk ratio greater than 1.0 indicates a higher rate of quitting in the treatment group than in the control group. For each type of medication where more than one eligible trial was identified, we performed meta-analysis using a Mantel-Haenszel fixed-effect method to estimate a pooled risk ratio with 95% confidence intervals (Mantel 1959). This is a change from previous review versions that used odds ratios (OR), because ORs can be misleading (Higgins 2008).
To investigate statistical heterogeneity we use the I² statistic, given by the formula [(Q - df)/Q] x 100%, where Q is the chi squared statistic and df is its degrees of freedom (Higgins 2003). This describes the percentage of the variability in effect estimates that is due to heterogeneity rather than sampling error (chance). We used threshold values of 50% and 70% as suggesting moderate and substantial heterogeneity respectively. Although we give a summary statistic, the conclusions that can be drawn from it must be cautious. Where trials are small and few in number the confidence intervals will be wide. The derivation of the summary statistic implicitly assumes that data from all randomized trials are available without any bias due to non-publication of unpromising results or to exclusion of randomized individuals. There is evidence that publication bias occurs in the field of smoking cessation research (Egger 1997), and this issue is discussed further in the Cochrane review of nicotine replacement therapy (NRT) (Stead 2008). Thus, we included unpublished studies or studies found only as abstracts where sufficient detail was available. We contacted authors for further data if necessary.
We have added a subgroup meta-analysis by level of additional support using the same criteria applied in the Cochrane NRT review (Stead 2008); low intensity support was regarded as part of the provision of routine care, so the duration of time spent with the smoker (including assessment for the trial) had to be less than 30 minutes at the initial consultation, with no more than two further assessment and reinforcement visits. We distinguished in the meta-analyses between trials testing an antidepressant as a single pharmacotherapy or as an adjunct to NRT for initial cessation. We also distinguished between cessation trials and those where the intervention addressed relapse prevention or reduction in number of cigarettes smoked. None of the trials located were specifically designed to directly compare antidepressant pharmacotherapy with non-pharmacological therapies.
Adverse events: Tables in the results section summarize the adverse events reported in clinical trials for smoking cessation for medications which have shown evidence of efficacy (bupropion and nortriptyline). For other medications adverse effects are noted in the included studies
. The number of people who have received bupropion and nortriptyline in smoking cessation trials is still relatively small, so there is limited power to estimate accurately the risk of uncommon adverse events. Because the safety of bupropion has been questioned in some countries, we have supplemented trial data with data from observational studies including national post-marketing surveillance schemes where it was possible to estimate a denominator. Nortriptyline is not currently licensed for smoking cessation, so this was not possible for this medication. We did not directly examined safety data for bupropion or nortriptyline when used to treat depression because these studies used higher doses and different populations than those used for smoking cessation.
Description of studies
We identified thirteen additional trials for this update, giving a total of 66 included trials. The new trials were of bupropion (Covey 2007; Croghan 2007; Fossati 2007; Grant 2007; Muramoto 2007; Schmitz 2007; George 2008; McCarthy 2008; Simon 2009), nortriptyline Aveyard 2008) and selegiline, (Biberman 2003; George 2003; Weinberger 2009, not previously covered by this review). Five trials that were previously included based on unpublished results have now been published and the study identifiers are now based on year of publication (Brown 2007; Evins 2007; Piper 2007; Spring 2007; Uyar 2007). The forty-nine trials of bupropion included five (Hays 2001; Hurt 2003; Killen 2006; Covey 2007; Croghan 2007) testing the medication for relapse prevention and one of reduction (Hatsukami 2004). Three of the bupropion trials allowed a direct comparison with nicotine patch therapy (Jorenby 1999; Górecka 2003; Uyar 2007), and three a direct comparison with the nicotine receptor partial agonist varenicline (Gonzales 2006; Jorenby 2006; Nides 2006). Nine trials used nortriptyline including three which also used bupropion (Hall 2002; Wagena 2005; Haggsträm 2006), There were four trials of fluoxetine (Blondal 1999; Niaura 2002; Saules 2004; Spring 2007), three of selegiline (Biberman 2003; George 2003; Weinberger 2009), one of paroxetine (Killen 2000), one of sertraline (Covey 2002), and one of venlafaxine (Cinciripini 2005). These studies excluded smokers with current depression but almost all included smokers with a past history of depression. Further details of the study designs are given in the table 'Characteristics of included studies'.
We list 58 excluded studies. Most of these were short-term or laboratory-based studies. For medications where there is little or no evidence from long-term studies we briefly describe the results of the excluded short-term trials. The reasons for exclusion are given in the table 'Characteristics of excluded studies'. Papers reporting additional outcomes or subgroup analyses from included studies are listed as references under the study identifier. Six further studies are potentially relevant but are either ongoing or do not have sufficient data to assess for inclusion yet.
This antidepressant has both dopaminergic and adrenergic actions, and appears to be an antagonist at the nicotinic acetylcholinergic receptor (Fryer 1999). It may work by blocking nicotine effects, relieving withdrawal (Cryan 2003; West 2008) or reducing depressed mood (Lerman 2002a). It has been licensed as a prescription aid to smoking cessation in many countries. The usual dose for smoking cessation is 150mg once a day for three days increasing to 150 mg twice a day continued for 7 to 12 weeks. The quit attempt is generally initiated a week after starting pharmacotherapy.
The majority of trials were conducted in North America but studies are also included from Europe (Aubin 2004; Dalsgarð 2004; Górecka 2003; Wagena 2005; Zellweger 2005; Fossati 2008); Turkey (Uyar 2007); Brazil (Haggsträm 2006); New Zealand (Holt 2005); Australia (Myles 2004) and two multicontinent studies (Tonnesen 2003; Tonstad 2003). Special populations recruited include smokers with the following conditions; chronic obstructive pulmonary disease (Górecka 2003; Tashkin 2001; Wagena 2005); schizophrenia (Evins 2001; Evins 2005; Evins 2007; George 2002; George 2008); post traumatic stress disorder (Hertzberg 2001); alcoholism (Grant 2007); and cardiovascular disease (Rigotti 2006 - inpatients; Tonstad 2003) as well as hospital inpatients (Simon 2009). Other populations included adolescents (Killen 2004; Muramoto 2007); smokers awaiting surgery (Myles 2004) hospital staff (Dalsgarð 2004); healthcare workers (Zellweger 2005); African-Americans (Ahluwalia 2002), and Maori (Holt 2005). Two studies recruited smokers who had previously failed to quit smoking using bupropion (Gonzales 2001; Selby 2003), and one included smokers who had just failed to quit using NRT (Hurt 2003).
More than half the bupropion studies followed participants for at least 12 months from the start of treatment or the target quit day. Eighteen studies (37%) had only six months follow up (Evins 2001; Hertzberg 2001; Ahluwalia 2002; George 2002; Aubin 2004; Collins 2004; Dalsgarð 2004; Hatsukami 2004; Killen 2004; Myles 2004; Evins 2005; Wagena 2005; Haggsträm 2006; Grant 2007; Muramoto 2007; Uyar 2007;George 2008; Simon 2009). The majority of studies reported an outcome of sustained abstinence. In 12 (24%) only point prevalence rates were given, or the definition of abstinence was unclear (George 2002; Selby 2003; Swan 2003; Killen 2004; Myles 2004; Evins 2005; Grant 2007; Muramoto 2007; Piper 2007; Schmitz 2007; Uyar 2007; George 2008).
Thirty-six trials evaluated bupropion as a single pharmacotherapy to assist initial cessation. Included in separate analyses are trials that tested bupropion as an adjunct to nicotine replacement therapy (Evins 2007; Killen 2004; Jorenby 1999 (part); Simon 2004; Grant 2007; George 2008) and trials making direct comparisons between bupropion and nicotine replacement therapy (Górecka 2003; Jorenby 1999; Uyar 2007), bupropion and nortriptyline (Haggsträm 2006; Hall 2002; Wagena 2005) and between bupropion and varenicline (Gonzales 2006, Jorenby 2006; Nides 2006). Trials testing the extended use of bupropion for relapse prevention (Hays 2001; Hurt 2003; Killen 2006; Covey 2007; Croghan 2007), and its use for assisting in reducing the amount smoked (Hatsukami 2004) are pooled separately.
The five studies that evaluated bupropion SR for relapse prevention each had slightly different designs (Hays 2001; Hurt 2003; Killen 2006; Covey 2007; Croghan 2007). These studies also contribute to a separate Cochrane review on interventions for relapse preventions (Hajek 2009)
One study evaluated bupropion for reducing smoking in people not wanting to make a quit attempt but interested in reducing (Hatsukami 2004). During treatment, if participants decided they wanted to try to quit, they were enrolled in a cessation programme during which they continued to use bupropion and were then followed up for 19 weeks.
Two placebo-controlled trials studied the use of bupropion for smokeless tobacco cessation (Dale 2002; Glover 2002). These trials are excluded from the present review but are covered in the Cochrane review of interventions for smokeless tobacco cessation (Ebbert 2007).
Most of the bupropion trials excluded participants with current depression but not those with a history of depression. One study (Ahluwalia 2002) did include participants who may have been depressed (i.e. several had a Center for Epidemiologic Studies Depression (CES-D) score of over 16). Two studies explicitly excluded participants with a history of major depression (Dalsgarð 2004) or any psychiatric disorder (Collins 2004). Amongst the studies recording the prevalence of a history of depression at baseline the proportion ranged from 6% (Hatsukami 2004) to 44% (Swan 2003), but was typically 20-30%.
Nine published studies of the tricyclic antidepressant nortriptyline are included. Aveyard 2008 is new since the last review. Hall and colleagues conducted three trials, and Prochazka and colleagues two, both in the USA. Two studies were conducted in Brazil (Da Costa 2002; Haggsträm 2006), one in the Netherlands (Wagena 2005) and one in the UK (Aveyard 2008). Seven studies excluded participants with current depression but most of these included people with a history of depression. All studies were placebo controlled and used doses of 75 to 100 mg/day or titrated doses to serum levels recommended for depression during the week prior to the quit date. All studies used a definition of cessation based on a sustained period of abstinence. The three studies by Hall and colleagues, and Aveyard 2008 reported outcomes after 12 months of follow up and the other five had six months of follow up. The design, length of treatment, level of common behavioural support and use of nicotine replacement as common therapy varied as summarised below.
- Hall 1998 used a 2x2 factorial design with nortriptyline versus placebo for 12 weeks crossed with a 10-session cognitive behavioural mood management intervention versus a five session standard health education control. The behavioural arms are collapsed in this meta-analysis.
- Hall 2002 used a 3x2 factorial design with nortriptyline versus bupropion versus placebo for 12 weeks crossed with a five session psychosocial behaviour therapy intervention versus medical management involving only brief advice and counselling. The behavioural arms are collapsed.
- Hall 2004 used a 2x2 factorial design with nortriptyline versus placebo crossed with extended treatment versus brief treatment. All participants received nicotine patch therapy for eight weeks and five sessions of group-based counselling. The 'brief' treatment was 12 weeks of nortriptyline or placebo and weekly individual counselling. The extended treatment was 52 weeks of nortriptyline or placebo and weekly and the monthly individual counselling sessions. Since the brief nortriptyline regimen is similar to that of the other nortriptyline trials, we include the results of brief nortriptyline versus placebo and extended nortriptyline versus placebo separately in the meta-analysis (Hall 2004 Brief; Hall 2004 Extended).
- Prochazka 1998 tested nortriptyline versus placebo for 10 weeks with a behavioural intervention of two group sessions and 12 individual follow-up visits.
- Prochazka 2004 tested nortriptyline versus placebo for 14 weeks as an adjunct to nicotine patch and brief behavioural counselling.
- Da Costa 2002 tested nortriptyline versus placebo for six weeks with a behavioural intervention of six weekly group cognitive therapy sessions.
- Wagena 2005 compared nortriptyline, bupropion or double placebo for 12 weeks with individual counselling and telephone support.
- Haggsträm 2006 tested nortriptyline versus placebo for 60 days with six individual counselling sessions.
- Aveyard 2008 compared nortriptyline to placebo as an adjunct to participants' choice of NRT which could include a combination of products, and four weeks of behavioural support, mainly in a group setting.
Selective Serotonin Reuptake Inhibitors (SSRIs)
No new studies with SSRIs have been completed since the last update of the review.
Four trials with long-term follow up are included. Two studies used fluoxetine as the only pharmacotherapy and had six months follow up; a multicentre trial compared 30 mg daily, 60 mg daily or placebo for 10 weeks (Niaura 2002), and a single-site study used 60 mg or placebo for 12 weeks (Spring 2007). Two trials provided NRT to all participants and compared the addition of fluoxetine and placebo over 12 months follow up; Blondal 1999 used 20 mg/day or placebo for three months as an adjunct to nicotine inhaler; Saules 2004 used 20 or 40 mg/day of placebo for 10 weeks as an adjunct to nicotine patch. Participants in all trials were not currently depressed but may have had a past history of depression. Spring 2007 stratified by history of depression.
We excluded other short-term studies. One study tested 14 weeks of fluoxetine (40 mg/day) or dexfenfluramine (30 mg/day) versus placebo in normal weight women in a study investigating the effects of these medications in controlling post-cessation weight gain (Spring 1995). This study found an apparently lower abstinence rate with fluoxetine (20%) than with placebo (31%) at three months, but the difference was not statistically significant. Pomerleau 1991 and Dalack 1995 were also studies on fluoxetine in smokers attempting to quit, but considered outcomes other than abstinence. Another pharmaceutical company-sponsored multicentre trial was completed but its results were never presented or published. One further trial is known to be underway (Brown 2007b).
One trial with six-month follow up assessed paroxetine (40 mg, 20 mg or placebo) for nine weeks as an adjunct to nicotine patch (Killen 2000).
One trial with six-month follow up assessed sertraline (200 mg/day) for 11 weeks versus placebo in conjunction with six individual counselling sessions. There were 134 participants, all current smokers with a past history of major depression (Covey 2002). One trial that combined sertraline with buspirone was excluded because the specific effect of sertraline could not be evaluated (Carrão 2007).
One short-term study used a crossover design to investigate the effect of the SSRIs citalopram or zimelidine on the smoking behaviour of heavy drinkers who were not attempting to stop smoking. Their cigarette use did not change significantly between active medication and placebo periods (Sellers 1987).
Monoamine oxidase inhibitors
The current review now includes selegiline in this category.
Moclobemide is a reversible monoamine oxidase-A inhibitor. Since smoking acts as a monoamine oxidase-A inhibitor, substituting moclobemide for smoking might help with cessation. This has been tested in one long-term placebo-controlled trial in France (Berlin 1995). Treatment with 400 mg/day began one week before quit day and continued for two months, reducing to 200 mg/day for a further month. No behavioural counselling was provided. Final follow up was at 12 months.
Selegiline is an irreversible, selective monoamine oxidase-B inhibitor at low doses (10 mg/day) and has shown antidepressant activity at higher doses, when it is non-selective (Gaszner 2006). In a short-term study it reduced smoking behaviour under laboratory conditions and reduced craving during two days of attempted abstinence (Houtsmuller 2002). Two long-term trials have been published and one is included based on a conference abstract (Weinberger 2009). All used 10mg/day oral treatment. Two had treatment durations of 9 weeks (George 2003; Weinberger 2009) and one continued therapy for 26 weeks (Biberman 2003). An unpublished study with preliminary short-term data is excluded (Brauer 2000). Three other phase II trials are known to be in progress (Glover (NCT00439413); Killen (NCT00218647); Le Foll (NCT00390923)), two of which are evaluating transdermal delivery.
This selective monoamine oxidase-B inhibitor was evaluated in an eight-week, dose finding, exploratory study (Berlin 2002). The trial was halted early due to liver toxicity observed in trials of the medications for other indications, and lazabemide is not being developed further. Continuous four-week quit rates at the end of treatment, including all drop-outs as treatment failures, were 17% (18/108) for 200 mg/day, 11% (12/108) for 100 mg/day and 9% (10/114) for placebo.
No new studies of other antidepressants have been completed since the last review.
There are no long-term studies of this serotonergic tricyclic antidepressant. It has been evaluated in a single small trial with short-term follow up (Edwards 1989). Treatment was with 150 mg doxepin daily for three weeks prior to quit day and four weeks afterwards. Subjects forfeited a US$135 deposit if they failed to stop smoking for seven days. Two months after cessation, 78% (7/9) of the doxepin group and 10% (1/10) of the placebo group reported abstinence, a statistically significant difference (P < 0.02). However one week post-cessation abstinence rates using stringent validated abstinence criteria failed to show a statistically significant difference. Among withdrawal symptoms, there was a significant group difference only for craving.
There are no long-term studies of this noradrenergic tricyclic antidepressant. One trial (Jacobs 1971) compared imipramine (25 mg 3 times/day) with lobeline, dextroamphetamine, placebo and a no-medication control. Some participants attended group support sessions. After three months, success rates, which included a reduction in smoking to less than 10% of baseline, were 56% (10/18) for imipramine, 40% (6/15) for placebo and 69% (27/39) for the no-medications control. These differences were not statistically significant.
Tryptophan may have antidepressant properties because it increases the level of serotonin. There have been no long-term studies reported. Bowen and colleagues postulated that this serotonin-enhancing action in conjunction with a high carbohydrate (CHO) diet might relieve the negative affect of cigarette withdrawal. Oral l-tryptophan (50mg/kg/day) and instructions to follow a high CHO, low-protein diet were compared with placebo pills and instructions for a low-carbohydrate diet (Bowen 1991). Participants in both groups also received four two-hour weekly multi-component group therapy sessions. Two weeks following the target cessation date 75% (12/16) of the tryptophan and high CHO diet group were abstinent versus 47% (7/15) of the placebo and low CHO diet group. This difference was not statistically significant.
This antidepressant inhibits re-uptake of serotonin and norepinephrine. One trial with 147 participants compared venlafaxine at a dose of up to 225 mg/day with placebo. All participants also received nicotine patches and nine brief individual counselling sessions; follow up was for 12 months (Cinciripini 2005). An unpublished short-term study (Frederick 1997) reported no difference in abstinence rates at eight weeks, and frequent side effects in the treatment group.
Hypericum (St John's Wort)
Extracts of hypericum have antidepressant properties. No studies are eligible to be included in the review. Barnes 2006 compared two doses for smoking cessation in an open randomized study with no placebo control. Quit rates were low and did not differ between dose levels. No participants maintained abstinence for 12 months. One uncontrolled study is excluded (Lawvere 2006). A controlled study has been completed but not yet published (Sood (NCT00405912)).
Risk of bias in included studies
All of the trials used placebo controls except Uyar 2007, and Swan 2003 which compared two different doses of a pharmacotherapy. All the trials were described as randomized, but most failed to report randomization and concealment methods in detail. Thirty-two studies (48%) reported a method of allocation judged adequate to ensure that treatment assignment was concealed at the time of enrollment. All but one of the other trials were categorised unclear because the method of allocation was not described. One bupropion trial (Myles 2004) described the use of a random number table but no mention of a blinded allocation list so was categorised as potentially inadequate. This was a small trial so its inclusion or exclusion did not change the results. Restricting inclusion in the largest meta-analysis (bupropion versus placebo) to studies assessed as using adequate allocation concealment no effect on the results.
The definition of abstinence was not always explicit and biochemical validation of self-reported smoking status was not always used; however all but one of the bupropion studies (Swan 2003) and all but one of the nortriptyline studies (Da Costa 2002) for which details were available did use biochemical verification for most self-reported quitters at some assessment points. In a small number of studies we were able to obtain a sustained outcome that was not given in the published report. Most of the sustained abstinence rates are based on self-reported slip-free abstinence from the start of the third week after the target quit date (TQD) onward and validated at intermediate and final follow ups.
Additional details about the methodology of individual trials are given in the table 'Characteristics of Included Studies'. Consistent with Cochrane methods, we included some trials that have only been published as abstracts, which have limited information on methodological issues (Clarke 2002). For some studies we have obtained additional information from authors, or from the pharmaceutical company funding the study. Use of unpublished data in the meta-analysis is noted in the Included Studies table.
Effects of interventions
(Selected analyses are displayed as Figures in the text. Other analyses are shown in the 'Data & Analyses' section online and full pdf versions of the review)
We distinguish between the subgroup of trials where bupropion was tested as the only pharmacotherapy, and those trials where the effect of bupropion when added to NRT was assessed. One trial (Jorenby 1999) contributed arms to both subgroups.
Compared to placebo control, no other pharmacotherapy
There were 36 trials in which bupropion was the sole pharmacotherapy, with over 11,000 participants. The pooled risk ratio [RR] was 1.69 (95% confidence interval [CI] 1.53 to 1.85) with little evidence of heterogeneity (I² = 11%), see figure 1, analysis 1.1. The estimated effect is smaller when expressed as a risk ratio, but has not really changed since the last version of the review, comparing odds ratios (i.e. OR in 2006; 1.94, OR now; 1.86). The control group quit rates ranged from 0% to 22%, with a weighted average of 9%. Intervention group quit rates ranged from 4% to 43% with a weighted average of 17%.
|Figure 1. Forest plot of comparison: 1 Bupropion. Abstinence at 6m or greater follow-up, outcome: 1.1 Bupropion versus placebo/control. Subgroups by length of follow-up.|
Sensitivity to length of follow-up
Although in this update there was no overall evidence of heterogeneity amongst the 36 trials of bupropion as the only pharmacotherapy we continued to use subgroup analyses to explore potential moderators of treatment effect. Twenty two of these trials had 12-month follow up and fourteen had six months. The estimated RR for the 12-month follow-up group was 1.64 (95% CI 1.46 to 1.84, I² = 34%, analysis 1.1.1) which was not significantly lower than that for trials with only six months (RR 1.81; 95% CI 1.51 to 2.16), I² = 0%, analysis 1.1.2). Much of the difference, and the heterogeneity in the 12-month subgroup, could be attributed to Zellweger 2005.
Sensitivity to clinical setting
In a post hoc subgroup analysis we distinguished between trials that recruited community volunteers and trials that recruited patients in healthcare settings or with specific diagnoses. Whilst the estimated effect was smaller amongst trials that recruited community volunteers than those recruiting in health care settings the confidence intervals overlapped and effects were significant in both groups (analysis 1.2).
Effect of level of behavioural support
Three trials compared bupropion and placebo in factorial designs varying the behavioural support. There was no evidence from any that the efficacy of bupropion differed between lower and higher levels of behavioural support (Hall 2002; McCarthy 2008) or by type of counselling approach used (Schmitz 2007). Two other studies have compared different levels of behavioural support for people prescribed bupropion (Strayer 2004; Swan 2003). These did not include placebo arms so do not provide evidence about within-study interactions between behavioural interventions and pharmacotherapy. We also explored a between-study subgroup analysis of the possible interaction with behavioural support using the classification into low and high intensity used in the Cochrane NRT review (Stead 2008). Low intensity was less than 30 minutes at the initial consultation, with no more than two further assessment and reinforcement visits. Only one of the included trials had such low intensity support (Myles 2004) and it was too small to draw conclusions from. Fossati 2007 (in a primary care setting) and part of McCarthy 2008 had limited behavioural support but in both cases there were more than three visits. We also examined, within the more intensive therapy trials, evidence of a different effect of bupropion versus placebo in eight trials that provided group-based behavioural interventions compared to the majority where individual therapy was provided. We found no evidence of a difference between subgroups (analysis 1.3). (This subgroup analysis was based on the 36 trials contributing to analysis 1.1 but excludes two trials where the level of support could not be classified.)
Effect of dose
In the first multi-dose study (Hurt 1997), cessation rate was linearly related to dose (100 mg versus 150 mg versus 300 mg) through the end of treatment, consistent with pharmacological efficacy, although the difference between 300 mg and 150 mg doses was not significant at long-term follow up. A larger study compared 150 mg and 300 mg daily doses, without a placebo group, and reported similar 12-month point prevalence quit rates for both doses (Swan 2003). A study in adolescents also included 150 mg and 300 mg doses (Muramoto 2007), with higher quit rates in the larger does group. Doses above 300 mg have not been tested. Pooling the three studies and comparing 300 mg versus 150 mg shows no evidence of a significant difference in abstinence (N = 2,042, RR 1.08; 95% CI 0.93 to 1.26, analysis 1.4).
Bupropion & nicotine replacement combined therapy compared to NRT alone
There was substantial heterogeneity in the results of six studies adding bupropion to nicotine patch therapy (Jorenby 1999; Killen 2004; Simon 2004; Evins 2007; Grant 2007; George 2008, I² = 64%). Using a random-effects model to pool the studies did not show evidence of a significant effect of adding bupropion (N = 1,106, RR 1.23; 95% CI 0.67 to 2.26, analysis 1.5). Of the six trials, four recruited people who were potentially hard to treat; adolescents (Killen 2004), smokers with schizophrenia (Evins 2007; George 2008) and smokers in treatment for alcohol dependence (Grant 2007). George 2008 was a small study with no quitters at all in the control group. The significant benefit seen in one trial (Jorenby 1999) may be due in part to the unusually poor results from nicotine patch alone in this study. The confidence intervals around the estimate do not exclude a benefit that would be clinically useful. One relapse prevention study (Croghan 2007) has compared open label nicotine inhaler, bupropion or both combined as initial therapy for cessation. After 12 weeks there was a second phase of randomization, so long term effects cannot be compared.
Bupropion for relapse prevention
Five trials have evaluated extended use of bupropion for preventing relapse in people who have already stooped smoking. Pooling studies suggests the possibility of a small benefit but confidence intervals just include 1 (N = 1,587, RR 1.17; 95% CI 0.99 to 1.39, analysis 1.6). The studies were heterogeneous with respect to the length of initial abstinence, the period of pharmacotherapy and the length of post treatment follow-up. The results are discussed in more detail in a Cochrane review of relapse prevention interventions (Hajek 2009).
Bupropion and depression
There was not sufficient detail in most studies about results in smokers with and without depression to conduct a meta-analysis. In three within-trial analyses, smokers with a past history of depression did not benefit more than those without such a history (Hayford 1999 subgroup analysis of Hurt 1997; Hurt 2002 & Cox 2004 subgroup analyses of Hays 2001; Brown 2007). In the only analysis specifically within the subgroup of smokers with no history of depression, bupropion was effective (Hayford 1999). Bupropion may alleviate some subclinical symptoms of depression during treatment (Ahluwalia 2002; Catley 2005; Lerman 2002a), but although this may facilitate smoking cessation, other mechanisms are probably more important (Catley 2005). In one trial (Collins 2004 reported in Lerman 2004), there was an interaction between nicotine dependence and treatment on post-cessation depression symptoms. Most smokers showed a reduction in depression symptoms during the treatment phase, whether they received bupropion or placebo. The reduction was maintained during follow up. However highly dependent smokers showed a greater reduction in depression scores whilst receiving bupropion, and an increase when treatment ended.
Gender/age differences with bupropion
Too few of the studies have published data on long-term quit rates by gender for it to be possible to conduct a definitive subgroup meta-analysis. A meta-analysis of mainly short-term outcomes and including 12 trials with 4421 participants showed no evidence of a treatment-gender interaction (Scharf 2004). In these trials women were less successful at quitting than men overall, but bupropion was equally beneficial in men and women. A subgroup analysis of long-term data from one study (Jorenby 1999, reported in Smith 2003) did report an interaction such that women appeared to benefit relatively more from medication. A more recent study reported a significant gender by smoking rate by treatment group interaction, such that bupropion seemed to benefit male heavy smokers and female light smokers but not others (Collins 2004). This study also showed an interaction among treatment effect, gender and genotype (Lerman 2002b). At the end of treatment, women with a variant CYP2B6 gene had significantly higher quit rates when treated with bupropion than on placebo. The bupropion treatment effect was not significant for the other three gender/genotype subgroups. A study in smokers with chronic obstructive pulmonary disease (COPD) noted a larger treatment effect for women (ORs 2.7 versus 1.7), although the statistical significance of this interaction was not tested (Tashkin 2001). One study has reported a larger treatment effect for four- to seven-week abstinence in males (Gonzales 2001). This was a study re-treating smokers who had already failed to quit with bupropion. In the Hays 2001 relapse prevention study, there were no significant gender effects (Gonzales 2002a). In summary, gender does not appear to consistently influence the efficacy of bupropion.
Whilst most reports have not indicated any difference in treatment effects between older and younger smokers, subgroup analyses of two trials, Hays 2001 (reported in Hurt 2002), and Hurt 1997 (reported in Dale 2001), found evidence of an interaction, with a larger treatment effect for older smokers. One study in adolescents did not show evidence of an effect for bupropion over nicotine patch alone (Killen 2004).
Bupropion as second treatment
One relapse prevention trial described above (Hurt 2003) also randomized 194 smokers who had not quit successfully using nicotine patch therapy to bupropion or placebo as a second line treatment. Only one person, in the bupropion group, quit and sustained abstinence at six months. This is consistent with the results of other studies, which find low overall success rates in smokers offered further pharmacotherapy soon after treatment failure (eg Gourlay 1995; Tonnesen 1993). In contrast, a subgroup analysis of Jorenby 1999 (reported in Durcan 2002) suggested that bupropion was equally effective in smokers with and without a past history of failure with NRT. In this trial the gap between the previous failed attempt and the second attempt at cessation would have been longer.
Bupropion versus NRT
Three studies allowed a direct comparison between bupropion and nicotine patch (Górecka 2003; Jorenby 1999; Uyar 2007). In the only one that was placebo-controlled, bupropion was significantly more effective than nicotine patch (Jorenby 1999); however, nicotine patch itself was not efficacious in this particular study. The other two smaller studies were open label and had non-significant effects. Because there was slight indication of heterogeneity (I² = 44%) and there was borderline significance using a fixed-effect model we used a random-effects model to estimate the pooled effect, which did not show a significant difference (N = 657, RR 1.26; 95% CI 0.73 to 2.18, analysis 1.7).
Bupropion versus varenicline
In three studies there was a direct comparison between bupropion and varenicline (Gonzales 2006; Jorenby 2006; Nides 2006). The pooled estimate showed a significantly lower rate of quitting with bupropion than varenicline (N = 1622, RR 0.66; 95% CI 0.53 to 0.82, analysis 1.8), with no evidence of heterogeneity. The average quit rate across the bupropion arms was 14% compared to 21% for varenicline.
Bupropion for smoking reduction
One study offered bupropion to smokers not wishing to quit (Hatsukami 2004). There were no significant differences in reduced cigarettes/day, cotinine or cessation (analysis 1.9).
Compared to placebo control, no other pharmacotherapy
Pooling six trials using nortriptyline as the only pharmacotherapy shows evidence of a significant benefit of nortriptyline over placebo (N = 975, RR 2.03; 95% CI 1.48 to 2.78, figure 2, analysis 2.1.1) without evidence of statistical heterogeneity (I² = 7%).
Nortriptyline & nicotine replacement combined therapy compared to NRT alone
Pooling three trials (one with a factorial design entered as two studies) using nortriptyline as an adjunct to nicotine patch therapy does not show evidence of an additional benefit from nortriptyline (N = 1,219, RR 1.29; 95% CI 0.97 to 1.72, figure 2, analysis 2.1.2) with much heterogeneity (I² = 34%).
|Figure 2. Nortriptyline versus placebo, long term abstinence|
Subgroup and sensitivity analyses
There were too few trials of nortriptyline to examine effect of duration of follow up, past depression, or amount of behavioural therapy between subgroups of trials. In one within-study comparison, a past history of depression did not appear to modulate the efficacy of nortriptyline, but subgroup numbers were small (Hall 1998). In two within-study comparisons, the intensity of adjunctive behaviour therapy did not influence the active versus placebo effect (Hall 1998; Hall 2002). In the study by Hall and colleagues of extended treatment (longer duration of both nortriptyline and behaviour therapy) versus brief treatment (similar to other nortriptyline trials), the confidence intervals for nortriptyline versus placebo included 1.0 (i.e. no evidence of an effect) for each treatment. The extended treatment increased absolute rates of abstinence and the relative effect for nortriptyline (RR 1.34 versus 0.62) but this was not statistically significant. Dose-response studies with nortriptyline have not been reported.
Bupropion versus nortriptyline
Three trials included a direct comparison between bupropion and nortriptyline (Haggsträm 2006; Hall 2002; Wagena 2005). In each study the comparison favoured bupropion but none showed significant differences. There was no evidence of heterogeneity. When pooled the difference remains non-significant, but does not exclude a clinically useful difference in favour of bupropion (N = 417, RR 1.30; 95% CI 0.93 to 1.82 analysis 3.1).
Selective Serotonin Reuptake Inhibitors (SSRIs)
Four trials of fluoxetine (Blondal 1999; Niaura 2002; Saules 2004; Spring 2007) and one each of paroxetine (Killen 2000) and sertraline (Covey 2002) were included. The pooled RR for the fluoxetine trials was 0.92 (N = 1,486, 95% CI 0.68 to 1.24, analysis 4.1). Pooling only Niaura 2002 and Spring 2007 that used fluoxetine alone and not an adjunct to NRT did not alter the conclusion that there was no evidence of a clinically important benefit (N = 1,236, RR 0.92; 95% CI 0.65 to 1.30). There was no evidence of benefit from paroxetine (Killen 2000, N = 224, RR 1.08; 95% CI 0.64 to 1.82) or sertraline (Covey 2002, N = 134, RR 0.71; 95% CI 0.30 to 1.64)
Monoamine oxidase inhibitors (MAOIs)
One trial of moclobemide (Berlin 1995) and three of selegiline (Biberman 2003; George 2003; Weinberger 2009) were included. The effect of moclobemide was significant at six-month follow up but was not at the final 12-month follow up (N = 88, RR 1.57 95% CI 0.67 to 3.68, analysis 5.1.1). The selegiline trials were all relatively small and had heterogeneous effects, with the unpublished trial (Weinberger 2009) reporting higher quit rates in the placebo group. When pooled there was no evidence of a significant effect (N = 250, RR 1.45; 95% CI 0.81 to 2.61, I² = 55%, analysis 5.1.2). Pooling all four trials of MAOIs also gave a non significant estimate (RR 1.49; 95% CI 0.92 to 2.41, analysis 5.1). Biberman 2003 also reported significantly reduced ratings of craving for cigarettes.
One trial of befloxatone showed no effect on cessation but data are unpublished (Berlin 2005).
One trial of venlafaxine (Cinciripini 2005) failed to show a significant increase in 12-month quit rates compared to nicotine patch and counselling alone, but confidence intervals do not exclude a clinically useful effect (RR 1.22; 95% CI 0.64 to 2.32, analysis 6.1). Post hoc subgroup analyses suggested that there might be greater evidence for an effect amongst lighter smokers.
We summarize adverse events (AEs) reported in trials of bupropion (analysis 1.10) and nortriptyline (analysis 2.2). In addition, for bupropion we summarize data from national surveillance schemes in the United Kingdom (UK), Australia and Canada (see Appendix 1). Although there are no new data, there have been some new warnings since the last review. Assessing AEs in smoking cessation medications is difficult because any AEs may be due, not to the medication, but to nicotine withdrawal (i.e., physical dependence). In addition, given smokers are more likely to have several medical and psychiatric illnesses, some "new" AEs may be exacerbations of pre-existing illnesses (Hughes 2008).
Adverse events reported for bupropion
The most common side effects are insomnia, occurring in 30% to 40% of patients, dry mouth (10%) and nausea (GlaxoSmithKline; Goldstein 1998). Typical drop-out rates due to adverse events range from 7% to 12%, but in one study 31% of those on 300 mg and 26% on 150 mg discontinued medication (Swan 2003). Early trials of bupropion as a treatment for depression using the immediate release formulation and often doses greater than 300 mg/day suggested it increased the risk of seizures in those with a prior history of alcohol withdrawal, anorexia or head trauma. This led to the development of the slow release (SR) preparation now licensed for smoking cessation. Using this preparation in doses of 300 mg/day or less, and excluding those at risk of seizures, no seizures had been reported in any of the smoking cessation trials until the study in physicians and nurses in Europe (Zellweger 2005). In this study there were two seizures amongst 502 people randomized to bupropion, one of whom had a familial history (data from GlaxoSmithKline). Since then two seizures have been reported in a study in which 126 participants received bupropion (Nides 2006) and one in a study with 329 treated (Gonzales 2006). Two seizures were also reported in an unpublished study with 100 participants prescribed bupropion (Strayer 2004, personal communication). This gives a total of 7 seizures amongst around 8,000 people exposed in clinical trials, so despite the recent reports the overall seizure rate remains less than the rate of 1:1000 given in product safety data. The figure of 1:1000 derives from a large, open, uncontrolled observational safety surveillance study conducted by the manufacturers (Dunner 1998) which examined 3100 adult patients using slow release bupropion for eight weeks for treatment of depression (not smoking cessation). Treatment was extended if necessary to a year, at a maximum dose of 150 mg twice daily. Patients with a history of eating disorder, or a personal or family history of epilepsy were excluded. Three participants (i.e. 1:1000) had a seizure considered to be related to the therapeutic use of bupropion.
Post-marketing surveillance data are available from some countries in which bupropion is licensed only for smoking cessation. Their limitation is that the denominator is not definitely known, and serious adverse events in medical practice are underreported by as much as a factor of 10 (Furberg 2006). However, using number of prescriptions as the denominator, the rate of reported seizures in the United Kingdom and Canada appears to be no higher (and possibly lower) than the rate of 1 in 1000 reported by Dunner et al. In England an observational study provided data on a cohort of 11,753 patients who had been dispensed bupropion (Boshier 2003). Eleven seizures were reported for a rate of 1 in 1000; four of these were associated with a past history of seizure. A second UK study (Hubbard 2005) used a general practice database (The Health Improvement Network) and a self-controlled case series method to estimate the relative incidence of death or seizure in 9329 individuals over a mean (SD) follow up of 1.9 (0.9) years. The self-controlled cases series method involves comparing each individual during a 'high risk' period with him/herself outside this period. The definition of high risk period in this study was 28 days after a prescription for bupropion (a 63 day high risk period was also used to test for robustness of the analysis). The reported death rates (case-series age adjusted estimate) were non-significantly lower during the high risk period (28 days: 0.5, 95% CI 0.12 to 2.05; 63 days: 0.47, 95% CI 0.18 to 1.19) while the seizure rates were non-significantly higher during the same period (28 days: 3.62, 95% CI 0.87 to 15.09; 63 days: 2.38, 95% CI 0.72 to 7.93). The seizures recorded in the first 28 days of treatment occurred on days 5 and 6 in one individual with no previous history of epilepsy. Of note in this study was that 12 people had been prescribed bupropion despite previous diagnoses of seizure.
Allergic reactions have also been reported with bupropion. These include pruritus, hives, angioedema and dyspnoea. Symptoms of this type requiring medical treatment have been reported at a rate of about 1 to 3 per thousand in clinical trials (GlaxoSmithKline), and this is approximately the level at which they are being reported in the national surveillance schemes. There have also been case reports of arthralgia, myalgia, and fever with rash and other symptoms suggestive of delayed hypersensitivity. These symptoms may resemble serum sickness. From the national surveillance schemes it is not possible to calculate the frequency of this outcome. Hypersensitivity reactions are listed as possible rare (occurring at rates less than 1 per 1000) adverse effects in the product data.
In the UK, Australia and France, bupropion is licensed only for smoking cessation. In the UK there were four reported suicides, 78 reports of suicidal ideation and five of suicide attempts/parasuicide between the licensing of bupropion and May 2004 among an estimated 1,000,000 prescriptions (MHRA 2004). In Australia there were 32 reports of suicide/self-injurious ideation from approximately 534,000 prescriptions up to 2004 (TGA 2004). In France there were reports of 22 suicide/attempted suicides and 19 suicidal ideation from 2001 to 2004 amongst approximately 698,000 people exposed (Beyens 2008). In all these populations the risk based on reported events is in the order of 1:10,000.
A review of the safety of bupropion was undertaken by the European Agency for the Evaluation of Medicines for Human Use (EMEA 2002). Suicidal ideation had been observed in six out of a total of 4067 participants in clinical trials for smoking cessation, a rate of 1: 677. The rate of suicidal ideation with bupropion was stated to be low compared with the rates found in the general population but no data were presented. It was also stated that there was neither a pharmacological nor a clinical reason for suspecting bupropion to be causally associated with depression or suicide. The committee concluded that the benefit/risk balance remained favourable, but made recommendations to strengthen warnings on hypersensitivity, and on depression, by advising clinicians to be aware of the possible emergence of significant depressive symptoms in patients undergoing a smoking cessation attempt.
A follow-up of 136 women exposed to bupropion prescribed for smoking cessation or depression during the first trimester of pregnancy suggested that bupropion does not increase the rates of major malformations, but there were significantly more spontaneous abortions (Chun-Fai-Chan 2005). An assessment of potential infant exposure to bupropion and active metabolites in breast milk suggests that the exposure of an infant whose mother was taking a therapeutic dose would be small (Haas 2004). Bupropion is also an inhibitor of CYP2D6 so care is needed when starting or stopping bupropion use in patients taking other medication metabolized by this route (Kotlyar 2005).
Although no patient is reported to have died while taking bupropion in trials for smoking cessation, some have died while taking bupropion prescribed for smoking cessation in routine practice. There has been no formal epidemiological analysis of these deaths, but no national reporting scheme has concluded that bupropion caused these deaths. Bupropion may cause adverse effects in overdose. A review of bupropion-only non-therapeutic exposures reported to the US Toxic Exposure Surveillance System for 1998-1999 identified 3755 exposures to Wellbutrin SR, 2184 to Wellbutrin and 1409 to Zyban (Belson 2002). These non-therapeutic exposures included intentional overdose and unintentional ingestion as well as reports of adverse reactions. Clinical effects related to bupropion exposure developed in 31% of non-therapeutic exposures, with vomiting the most common childhood symptom and tachycardia the most common in teenagers and adults. Six per cent of exposures (19% of symptomatic patients) developed a seizure. Seizures were more common with Wellbutrin exposures (22% of symptomatic patients) compared to bupropion SR (16% of symptomatic) and Zyban (13% of exposures). Moderate or severe outcomes were reported in 17% of Wellbutrin exposures, 12% of Wellbutrin SR exposures and 9% of Zyban exposures. Seventy-eight per cent of the moderate and major effects resolved in less than 24 hours. Five deaths all involved suspected suicides and only one in five involved Zyban or Wellbutrin SR.
In 2003, post-marketing data from studies of SSRIs for depression in adolescents suggested they may increase the risk for suicidal ideation (Hegerl 2006). Based on this finding, the US FDA issued warnings for several clases of antidepressants including bupropion when used for depression in both children and adults (US FDA 2004). In 2009, the US FDA added new warnings about the risk of serious neuropsychiatric symptoms in patients using bupropion for smoking cessation (US FDA 2009a; US FDA 2009b). The FDA stated “these symptoms include changes in behavior, hostility, agitation, depressed mood, suicidal thoughts and behavior, and attempted suicide. The added warnings are based on the continued review of postmarketing adverse event reports for varenicline [a smoking cessation treatment, see Cahill 2008] and bupropion received by the FDA.” There were 46 reports of suicidal ideation and 29 of suicidal behaviour for bupropion to November 27 2007 (US FDA 2009a).
Adverse effects of nortriptyline
The adverse events reported included the well known tricyclic effects of dry mouth, drowsiness, light-headedness and constipation observed in studies treating depression in which doses were often > 150 mg (Khawam 2006). In addition, nortriptyline can be lethal in overdoses. Based on experiences when used to treat depression, nortriptyline would be expected to have the potential for more serious adverse events. In contrast, when used at 75 to 150 mg doses in smokers, drop-out rates in the trials reporting this outcome have ranged from 4% to 12%, with one exception (Wagena 2005). This rate is similar to that for bupropion and NRT. The only serious adverse event in someone treated with nortriptyline was collapse/palpitations thought possibly caused by treatment (Aveyard 2008). Since nortriptyline is not approved for smoking cessation in any country, we are unaware of any post-marketing surveillance data.
Thirty-six trials now provide a large evidence base confirming the benefit from bupropion used as single pharmacotherapy for smoking cessation. There is no statistical heterogeneity evident and the pooled estimate suggests that bupropion increased long term quitting success by relative factor of 1.5 - 1.9. Treatment effects appear to be comparable in a range of populations, settings and types of behavioural support and in smokers with and without a past history of depression. Clear evidence of an additional benefit from adding bupropion to NRT was not demonstrated. The meta-analysis for the updated USPHS clinical practice guideline reported an odds ratio of 1.3 (95% CI 1.0 to 1.8) for a combination versus nicotine patch alone (Fiore 2008 table 6.28). The difference in meta-analytic outcomes may be because the current analysis included several studies of hard-to-treat populations not included in the USPHS analysis. Also, it could be because the Cochrane analysis was a collation of six within-study randomized comparisons whereas the USPS was an across-study comparison of the results the combination arm in three trials to the results of the patch alone arm in 32 studies.
Meta-analysis of the three bupropion trials that compared the recommended dose of 300 mg/day (150 mg twice daily) with a dose of only 150 mg failed to show a significant long-term benefit of the higher dose. Whilst the power of the comparison is not sufficient to establish equivalence, for people with troubling side effects such as insomnia, a reduction in dose to 150mg in the morning would be an alternative to discontinuing pharmacotherapy altogether.
There is still insufficient evidence from head to head trials to prefer bupropion over NRT or vice versa. In indirect, across-study comparison the efficacy seems similar. The choice between these two therapies will depend on patient preferences including a consideration of the risks of adverse events.
In three trials, participants treated with bupropion were significantly less likely to quit than those treated with varenicline, a partial nicotinic agonist. Although this suggests varenicline should be preferred over bupropion as a first line therapy, further study is warranted for several reasons. First, the number of studies is small. Second, the three trials used very similar optimal samples, settings and procedures. Whether superiority for varenicline would occur in a more real-world setting is unclear. Finally, given that both bupropion and varenicline block nicotine receptors and increase dopamine, a biological explanation for superior efficacy for varenicline has not been proposed. The evidence for efficacy of varenicline is covered by another Cochrane review (Cahill 2008).
Further trials of extended therapy with bupropion for individuals who have recently quit bring the number included to five, and the pooled estimate only narrowly excludes 1 (RR 1.17; 95% CI 0.99 to 1.39) but the clinical importance of any effect seems likely to be small. Preventing relapse remains a major challenge.
Nortriptyline has also been shown to assist cessation; there is an adequate evidence base although the number of trials and the total number of participants is much smaller than for bupropion. As with bupropion there is no evidence that the combination of nortriptyline and NRT is more effective than NRT alone.
There are no direct comparisons of nortriptyline with NRT or varenicline. Head to head comparison with bupropion in three trials favour bupropion but do not show a significant difference. The pooled risk ratios of efficacy of nortriptyline and bupropion appear broadly similar. One argument for considering nortriptyline as a first line therapy is its lower cost (Wagena 2005a). The main argument against this is based on the potential for serious adverse effects (Hughes 2005).
Although not widely tested, the efficacy of bupropion and nortriptyline appear to be independent of a past history of depression (Hall 1998; Hayford 1999; Hurt 2002) and post-cessation depression (Catley 2005, reporting an analysis of Ahluwalia 2002). This suggests that their efficacy is not due to a traditional antidepressant effect and that they benefit those with no history of depression. Although the pharmacological mechanism of action of bupropion is still unclear, recent animal studies suggest that it may act as an antagonist at the nicotine receptor (Cryan 2003; Wiley 2002, Young 2002). How nortriptyline increases cessation is unclear.
Although there is considerable research interest in genetic differences that could help predict response to pharmacotherapy (Uhl 2008), there is currently no genetic test that can be used for treatment matching in a clinical setting. There is preliminary evidence that smokers with normal dopamine receptor availability and function might respond better to bupropion (David 2005; Lerman 2006) whilst genotypes that are associated with impaired dopaminergic systems could have relatively better outcomes with NRT (Johnstone 2004). It is also possible that women with particular genotypes may respond differently to bupropion compared with men having the same genotypes (Swan 2005). The rate of metabolism of nicotine has also been suggested as a moderator of treatment effect, with fast metabolisers benefiting from bupropion (Collins 2004, reported in Patterson 2008).
No seizures were reported in the first large studies of bupropion for smoking cessation but more recently four studies (Gonzales 2006; Nides 2006; Strayer 2004; Zellweger 2005) report a total of 7 seizures. Since about 8,000 people have been exposed to bupropion in the cessation studies included in this review, the averaged rate is still less than the 1:1000 estimated risk used in product safety information, although the clustering of seizures in a few small studies is unexpected. Some suicides and deaths while taking bupropion have been reported. Currently, like many other antidepressants and varenicline, bupropion has a warning about the possibility of serious mood and behavioral changes. However, it remains unclear whether these outcomes were caused by bupropion effects.
In studies in depressed patients nortriptyline sometimes caused sedation, constipation, urinary retention and cardiac problems, and when taken as an overdose could be fatal. Based on the rate of significant adverse events when used to treat depression, nortriptyline would be expected to have higher rate of drop-outs. This has not been the case in the relatively small number of subjects receiving nortriptyline in the existing studies (about 500), perhaps because the dose of nortriptyline used (75 to 150 mg) is generally smaller than that used for depression and smokers are not acutely ill. However, given this small sample size, the safety of these doses of nortriptyline for smoking cessation is still unclear.
The six long-term trials of selective serotonin reuptake inhibitors (SSRIs) (fluoxetine, paroxetine and sertraline) and other short-term trials have, somewhat surprisingly, failed to show that this class of antidepressants helps smoking cessation. Some studies have found SSRIs effective in post hoc-determined subgroups (Borrelli 2004; Swan 1999) but this requires verification. The most recent trial (Spring 2007) found that although fluoxetine initially increased cessation amongst smokers with a history of depressive disorder, by the end of the study it impaired cessation regardless of depressive history.
There is no clear evidence of long term efficacy for monoamine oxidase inhibitors. Two early trials of selegiline suggested a possible benefit whilst the most recent trial has not supported this.
Mechanism of action of antidepressants
Whether the efficacy of bupropion and nortriptyline is specific to the unique pharmacology of these medications or whether it would occur in all antidepressants has not been completely resolved. The SSRI antidepressants appear not to be efficacious. This suggests serotonin modulation is not important, leaving the dopaminergic or noradrenergic effects of nortriptyline and bupropion to account for their efficacy. Although the efficacy of bupropion was initially thought to be due to its dopaminergic actions, nortriptyline, which is also effective, has relatively weak dopaminergic activity. In addition, bupropion has as much noradrenergic activity as dopaminergic activity. Another possibility, at least for bupropion, is that it acts as a nicotinic receptor blocker (Warner 2005). Whether the same is true for nortriptyline is not clear (Gambassi 1999). If noradrenergic effects are important in treatments for smoking, then monoamine oxidase inhibitors and other tricyclic antidepressants should be effective; however, only a few small trials of these are available.
Comparison with prior reviews and meta-analyses
The findings of this review are in agreement with the conclusions of other reviews and guidelines (Aubin 2002; Haustein 2003; Hughes 2005; Jorenby 2002; Martinez-Raga 2003; McRobbie 2005; RCP 2000; Tonstad 2002; Tracey 2002; Haustein 2003; Martinez-Raga 2003; West 2000; West 2003). Many national smoking cessation guidelines were last updated six years ago. The US guidelines were updated in 2008 (Fiore 2008) and continue to recommend bupropion as a first line therapy and nortriptyline as a second-line therapy due to possible adverse events. Open uncontrolled trials and observational studies of bupropion have shown real-life quit rates comparable to those found in clinical trials (Holmes 2004; Wilkes 2005; Paluck 2006). Studies of cost-effectiveness also support the utility of bupropion (Bolin 2006; Javitz 2004) and nortriptyline (Hall 2005).
Implications for practice
The existing evidence supports a role for bupropion and nortriptyline in clinical practice. Nicotine replacement therapy has proven efficacy in over 90 studies (Stead 2008) and has a very benign side-effect profile. There is insufficient published evidence to conclude either bupropion or nortriptyline has superior efficacy to NRT or vice versa. The confidence intervals around the efficacy estimates for bupropion, nortriptyline and NRT overlap. Bupropion and nortriptyline are equally effective in smokers with and without a history of depression and their efficacy does not appear to be mediated by improving post-cessation depression. Although the US Guideline (US DHHS 2000) suggests smokers with depression problems should use bupropion rather than NRT, whether smokers with a previous history of depression or mild current depression would do better with antidepressants than NRT has not been tested. Whether bupropion prevents depressive symptoms or relapse to depression better than NRT has also not been studied. Patient preferences, cost, availability and side-effect profile will all need to be taken into account in choosing among medications. Bupropion and nortriptyline may be helpful in those who fail on nicotine replacement therapy. Recent studies (Gonzales 2006; Jorenby 2006; Nides 2006) comparing bupropion with varenicline have shown higher quit rates with varenicline.
All smoking cessation medications can produce clinically significant adverse effects. When people are initially screened for potential adverse effects, however, fewer than 10% of those on antidepressants for smoking cessation stop taking the medications due to adverse effects. Although bupropion use has been associated with deaths in lay public reports, currently there is insufficient evidence to state that bupropion caused these deaths. There has also been concern about antidepressants such as bupropion being associated with psychiatric disorders including suicidal ideation and suicide attempts. Again, is not clear that there is a causal relationship. Smoking cessation may also precipitate depression (Hughes 2007). Also, although nortriptyline is associated with more side effects when used for depression, in the doses used for smoking cessation this may not be true.
Slow release bupropion, under the name Zyban, is licensed for smoking cessation in many parts of the world, including North America, Australia and Europe, but is not available in many other countries. Often, bupropion is available in these countries under the name Wellbutrin SR as a treatment for depression. Nortriptyline is marketed as an antidepressant in many countries but is not currently marketed as a smoking cessation aid in any country. In almost all countries, bupropion and nortriptyline are available only as prescription medications.
Implications for research
More research is needed with different antidepressants to determine which antidepressants or classes of antidepressant are effective in smoking cessation. Determining this could provide insight not only into the mechanism of action of antidepressant efficacy but also into the biological factors controlling nicotine dependence and smoking. Antidepressants of the SSRI category are not effective, which suggests serotonin may not be an important factor. However it is unclear whether dopaminergic, noradrenergic or nicotinic-cholinergic monoaminergic activity or blockage of nicotine receptors is most important for cessation efficacy. Given some suggestive results with selegiline, further trials of this compound and other monoamine oxidase inhibitors (MAOIs), which have mostly adrenergic activity, could be helpful. Similarly, the suggestive findings that genotype might moderate antidepressant treatment efficacy deserves follow-up. Also, it would be helpful to know whether bupropion's efficacy is due to receptor blockade and whether nortriptyline also is a nicotine receptor blocker. Research on the biological and behavioural mediators of the efficacy of bupropion and nortriptyline is needed; e.g. how much of their efficacy is due to craving or withdrawal relief, blocking nicotine reinforcement, or preventing lapses from becoming relapses. Knowledge of whether NRT or antidepressants have more efficacy in decreasing depression post-cessation would help decide whether smokers with a past history of depression should prefer antidepressants over NRT.
The use of antidepressants in combination with nicotine replacement therapy, in smokers who have failed with NRT, in smokers with baseline dysphoria, should be further investigated as initial data suggest the combination may add efficacy. Given the concern by some about deaths and psychiatric disorders from antidepressants used for smoking cessation, continued monitoring is indicated.
Our thanks to Drs Niaura, Borrelli, Spring, Fiore, Hurt, Mizes, Ferry, Schuh, Cinciripini, Hays, Prochazka, Ahluwalia, Mayo, Collins, Novotny, Brown, David & Evins for assistance with additional information or data on studies.
JR Hughes's contribution is supported by Research Scientist Development Award DA-00490 from the National Institute on Drug Abuse.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Index terms
Appendix 1. Suspected adverse events from national reporting schemes
Labelling of graphs
In the graphs for bupropion the comparison is described under the outcome heading rather than the comparison heading
In this review the outcome is always abstinence from smoking at longest follow-up. In order to group comparisons for the same pharmacotherapy and to give an informative summary graph, the decision was taken to use the outcome level for the specific comparisons. Unfortunately it is not possible to alter the fixed headings of 'Comparison' and 'Outcome'.
Lindsay Stead, review author and Managing Editor
Last assessed as up-to-date: 29 July 2009.
Protocol first published: Issue 3, 1997
Review first published: Issue 3, 1997
Contributions of authors
All authors contribute to the text of the review. LS and TL extracted study data.
Declarations of interest
JR Hughes has received consultancy fees from many pharmaceutical companies that provide tobacco related services or products or are developing new products, including Pfizer (the maker of NRTs and varenicline) and GlaxoSmithKline (the makers of bupropion and NRTs).
Sources of support
- No sources of support supplied
- National Institute on Drug Abuse (NIDA), USA.
- NHS Research and Development Programme, UK.
This review was first published as part of the review 'Anxiolytics and antidepressants for smoking cessation'. From Issue 4 2000 the classes of drugs are reviewed separately.
Medical Subject Headings (MeSH)
Anti-Anxiety Agents [adverse effects; *therapeutic use]; Antidepressive Agents [adverse effects; *therapeutic use]; Randomized Controlled Trials as Topic; Smoking [*drug therapy]; Smoking Cessation [*methods]
MeSH check words
* Indicates the major publication for the study