Antidepressants for smoking cessation

Whilst nicotine replacement has become the most widely used pharmacotherapy for smoking cessation, some people prefer a treatment that does not use nicotine. Observations that a history of depression is found more frequently amongst smokers than non smokers, that cessation may precipitate depression, and that nicotine may have antidepressant effects provide a rationale for the use of antidepressant drugs 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 (MAOIs) moclobemide and selegiline
* the selective serotonin reuptake inhibitors (SSRIs) fluoxetine, paroxetine and sertraline
* the atypical antidepressants bupropion, tryptophan and venlafaxine

This meta-analysis complements prior qualitative reviews of the efficacy of the above medications for smoking cessation (Hughes 1994, Henningfield 1998, Benowitz 2000, Covey 2000a, RCP 2000, West 2000, Kotlyar 2001).

SHORT-TERM TRIALS

The focus of this review and meta-analysis is on trials that provide evidence for an effect on long-term smoking cessation and these are described in the Results section. Short-term trials investigating antidepressants have also been completed, and we describe these in this section.

TRICYCLIC ANTIDEPRESSANTS (TCAs)

Doxepin
This serotonergic tricyclic antidepressant has been evaluated in a single small trial with short-term follow-up (Edwards 1989). Treatment was with 150 mg doxepin/day for three weeks prior to quit day and four weeks afterwards. Subjects forfeited a $135 deposit if they failed to stop smoking for 7 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 detect a statistically significant difference. Among withdrawal symptoms, there was a significant group difference only for craving.

Imipramine
One trial (Jacobs 1971) compared the noradrenergic tricyclic antidepressant imipramine (25 mg 3 times/day) with lobeline, dextroamphetamine, placebo and a no drug 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 drug control. These differences were not statistically significant.

Nortriptyline
This noradrenergic TCA has been used in five long-term smoking cessation trials (Hall 1998; Prochazka 1998; Prochazka 2001; Da Costa 2002; Hall 2002) described in the results section.

MONOAMINE OXIDASE INHIBITORS

Moclobemide
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 study reported in the results section (Berlin 1995A).

Selegiline
This reversible monoamine oxidase-B inhibitor has been shown to reduce smoking behaviour under laboratory conditions and to reduce craving during two days of attempted abstinence (Houtsmuller 2002). A trial of selegiline as an adjunct to nicotine patch therapy is in progress, and preliminary short-term results suggest it may promote abstinence (Brauer 2000).

Lazabemide
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 drug 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.

SELECTIVE SEROTONIN REUPTAKE INHIBITORS

Fluoxetine
Spring 1995 tested 14 weeks of fluoxetine (40 mg/day) or dexfenfluramine (30 mg/day) versus placebo in 97 normal weight women in a study investigating the effects of these drugs in controlling post-cessation weight gain. This study found an apparent lower abstinence rate with fluoxetine (20%) than placebo (31%) at three months, but the difference was not statistically significant. Pomerleau 1991 and Dalack 1995 have also reported studies on fluoxetine in smokers attempting to quit, but considered outcomes other than abstinence. Two uncontrolled trials by one group report a decrease in the amount smoked by depressed alcoholic patients not attempting to quit (Cornelius 1997; Cornelius 1999). In another study in non-depressed early stage problem drinkers a dose of 60 mg/day slightly increased the number of cigarettes smoked compared to placebo and to baseline, whilst a dose of 40 mg/day showed no effect (Naranjo 1990). Two studies with long-term cessation outcomes are discussed in the results (Niaura 2002; Blondal 1999). An ongoing trial is assessing fluoxetine (20 or 40 mg/day) as an adjunct to nicotine patch (Schuh 2000). Long-term results for this are not yet available. Another multicentre trial was completed several years ago but never published.

Paroxetine
In the results section we describe one study with long-term follow-up (Killen 2000).

Sertraline
In the results section we describe one study with long-term follow-up (Covey 2002).

Citalopram/zimelidine
One 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 drug and placebo periods (Sellers 1987).

ATYPICAL ANTIDEPRESSANTS

Bupropion
This antidepressant has both dopaminergic and adrenergic actions, and also appears to be an antagonist at the nicotinic acetycholinergic receptor (Fryer 1999). It has been licensed for use in smoking cessation in the USA and the European Union. We have identified 20 studies with long-term follow-up that we report in the results section. The results of nine of these studies are available from abstracts of conference presentations or posters (Ferry 1992; Ferry 1994; Hurt 2001; Swan 2001; Tonstad 2001; Zellweger 2001; Dalsgard 2002; McRobbie 2002; Simon 2002) and 10 have been published in full (Hurt 1997; Jorenby 1999; Evins 2001; Gonzales 2001; Hays 2001; Hertzberg 2001; Tashkin 2001, Ahluwalia 2002; George 2002; Hall 2002). For one trial only the results for participants of European Caucasian ancestry have been published so far (Lerman 2002A).

In another unpublished study, smokers not currently trying to stop smoking were randomized to bupropion or placebo for smoking reduction. Bupropion both increased reduction and produced a higher quit rate at week seven; longer term data have not yet been reported (Rennard 2001).

Two trials have studied the use of bupropion for smokeless tobacco cessation (Dale 2002, Glover 2002). These trials are excluded from the present review but will be included in a forthcoming Cochrane review of interventions for smokeless tobacco cessation.

Venlafaxine
This antidepressant inhibits reuptake of serotonin and norepinephrine. No trials have yet been published in full, but two have been reported in abstracts. One study (Frederick 1997) reported no difference in abstinence rates at eight weeks, and frequent side effects in the treatment group. Twelve month abstinence rates are available for a second trial which we describe in the results section (Cinciripini 1999).

Tryptophan
Tryptophan may have antidepressant properties because it increases the level of serotonin. Bowen and colleagues postulated that this serotonin enhancing action in conjunction with a high-carbohydrate diet might relieve the negative affect of cigarette withdrawal. Oral l-tryptophan (50mg/kg/day) and instructions to follow a high-carbohydrate (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 multicomponent 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.

To assess the evidence for the efficacy in assisting long-term smoking cessation of drugs with antidepressant properties, including: bupropion, doxepin, fluoxetine, imipramine, moclobemide, nortriptyline, paroxetine, tryptophan, selegiline, sertraline and venlafaxine.

For each drug 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.

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 drugs generally classified as having an antidepressant effect were selected for inclusion in this review. The date of the last search was 3rd December 2002. We checked the citation lists of these studies, previous reviews of non-nicotine pharmacotherapy (Benowitz 2000, Hughes 1994, Henningfield 1998, Kotlyar 2001, Covey 2000a, RCP 2000) and abstracts from the meetings of the Society for Research on Nicotine and Tobacco. For each drug found from these sources we searched MEDLINE (Silverplatter to 9/02, PubMed on 3/12/02 limited to 90 day entry date) and EMBASE (to 11/02) using the drug name and 'smoking' as a free text term. Several studies were located by contacting investigators in the area.

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').

We give data on the number of quitters in the treatment and control groups, and an odds ratio with confidence intervals in the Summary of Analyses. For each type of drug where more than one eligible trial was identified, we calculated a meta-analytic estimate of the most likely effect size using the Mantel-Haenszel method. In previous versions of the review we used the Peto method to pool odds ratios (Yusuf 1985). This method may give biased results when the study arms have unequal sample sizes. Since this is now the case for some trials of bupropion we have changed the summary statistic for all the meta-analyses . The Mantel-Haenszel method is now used. In most cases the difference between the two methods is small, and in no case does it change the conclusions. 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) (Silagy 2001). We included unpublished studies or studies found only as abstracts where sufficient detail was available. We contacted authors for further data if necessary.

Adverse events: We summarise the adverse events reported in clinical trials for smoking cessation in tables in the results section for drugs with data from more than one study (bupropion, nortriptyline), and in the included study table for others. The number of people who have received bupropion 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 antidepressants licensed for smoking cessation 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. These data are summarised in a table. We have not included observational data for antidepressants not currently licensed for smoking cessation.

Thirty studies were included: five of nortriptyline, one of moclobemide, two of fluoxetine, one of paroxetine, one of sertraline, 20 of bupropion and one of venlafaxine. One trial compared nortriptyline and bupropion with each other and with placebo. One trial compared nicotine patch and bupropion with each other and with placebo. In all of the studies, pharmacotherapy was begun several days prior to a quit date at which smokers stopped abruptly. Duration of treatment for all medication types was typically 7-12 weeks. Two trials using bupropion for relapse prevention offered six months (Hurt 2001) or 12 months (Hays 2001) of pharmacotherapy.

All of the studies excluded smokers with current depression but many included smokers with a past history of depression. Further details of the study designs are given in the table 'Characteristics of included studies'.

Twenty nine studies were excluded. These include the short-term studies covered in the background section, laboratory based studies, and some subgroup analyses of included studies. The reasons for exclusion are given in the table 'Characteristics of excluded studies'

TRICYCLIC ANTIDEPRESSANTS

Among the tricyclic antidepressants, only nortriptyline has been studied with long-term follow-up. Five studies are included in the meta-analysis. Hall and colleagues conducted two of these. The first (Hall 1998) tested nortriptyline (dose up to 100 mg/day titrated to therapeutic levels for 12 weeks) versus placebo in a 2x2 factorial design which also compared a cognitive behavioural mood management intervention versus standard health education. The second (Hall 2002) compared nortriptyline to placebo and also bupropion in a 3x2 design which also compared medical management alone to a psychosocial behaviour therapy intervention. A third study comparing brief (12 weeks) and extended (52 weeks) nortriptyline and behavioural therapy in addition to nicotine patch versus placebo has not yet been reported in sufficient detail for inclusion (Hall 2002B). Prochazka and colleagues have conducted two studies. The first (Prochazka 1998) used a dose of up to 75 mg/day versus placebo with a behavioural intervention of two group sessions and 12 individual follow-up visits. The second, reported in an abstract (Prochazka 2001) tested nortriptyline versus placebo along with nicotine patch and brief behavioural counselling for all participants. A Brazilian study (Da Costa 2002) used a dose of up to 75 mg/day for six weeks. Two of the studies reported 12 month continuous abstinence rates (Hall 1998; Hall 2002) and three reported six month rates (Prochazka 1998; Prochazka 2001; Da Costa 2002).

MONOAMINE OXIDASE INHIBITORS

Moclobemide has been tested in one placebo controlled trial in France (Berlin 1995A). 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.

SELECTIVE SEROTONIN REUPTAKE INHIBITORS (SSRIs)

Fluoxetine for smoking cessation has been evaluated in a multicentre trial comparing 30 mg/day, 60 mg/day or placebo for 10 weeks with follow-up at six month after the quit date (Niaura 2002). Participants in this trial were not depressed at study entry but may have had a past history of depression. A second trial has studied fluoxetine 20 mg/day vs placebo as an adjunct to nicotine inhaler (Blondal 1999).

There is one published trial using paroxetine (40 mg, 20 mg or placebo) as an adjunct to nicotine patch with six month follow-up (Killen 2000).

One trial (Covey 2002) 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.

ATYPICAL ANTIDEPRESSANTS

Bupropion
Sixteen placebo controlled studies of bupropion for initial cessation include over 5000 patients. Ferry and colleagues evaluated bupropion 300 mg/day versus placebo in two studies (Ferry 1992; Ferry 1994). All participants also attended smoking cessation and relapse prevention meetings. These studies have been reported only as abstracts. We obtained 12 month abstinence rates from the author. A multicentre study (Hurt 1997) evaluated sustained release (SR) bupropion in doses of 100 mg/day, 150 mg/day or 300 mg/day against placebo for seven weeks. The main publication for this study reported point prevalence abstinence rates. Glaxo Wellcome provided continuous abstinence rates at 12 months. A second multicentre study compared a combined treatment of 300 mg bupropion SR plus the nicotine transdermal patch to bupropion alone and patch alone and to placebo in a factorial design (Jorenby 1999). Another multicentre study recruited participants with mild to moderate chronic obstructive pulmonary disease (COPD). Six month abstinence rates have been published; unpublished 12 month rates from an abstract are used in the meta-analysis (Tashkin 2001). One small study with six month follow-up recruited veterans with post-traumatic stress disorder (Hertzberg 2001). One study recruited smokers who had previously failed to quit smoking using bupropion, (Gonzales 2001). Two small studies have evaluated bupropion for smoking cessation in smokers with schizophrenia (Evins 2001; George 2002) with six month follow-up. Three multicountry studies have reported 12 month continuous abstinence rates for bupropion versus placebo in abstracts: Tonstad 2001 recruited smokers in eight countries, McRobbie 2002 recruited smokers with stable cardiovascular disease in 10 countries, Zellweger 2001 recruited doctors and nurses who smoked, from 12 European countries. Dalsgard 2002 recruited Danish hospital employees. Ahluwalia 2002 recruited African American smokers. One paper has reported outcomes for a subgroup of trial participants who were Caucasians of European descent (Lerman 2002A). One study compared bupropion, nortriptyline and placebo crossed with two intensities of counselling, in a factorial design (Hall 2002).

One unpublished study without a placebo control (Swan 2001) compared 300 mg/day and 150 mg/day doses of bupropion, crossed with two levels of behavioural support, in a factorial design. Another unpublished study compared bupropion with placebo as an adjunct to NRT using the nicotine patch (Simon 2002). This is not included in the main analysis but is included in a comparison on bupropion plus NRT versus NRT alone, along with the comparable arms of Jorenby 1999.

Two studies have evaluated bupropion SR for relapse prevention. The first recruited smokers who had remained abstinent during seven weeks of open label bupropion therapy (Hays 2001). In the treatment group bupropion was provided for a further 45 weeks versus a placebo. Follow up continued for a further year after the end of treatment. A second trial reported as an abstract (Hurt 2001) randomized people who successfully quit after eight weeks of nicotine patch therapy to six months of bupropion or placebo. The study reported cessation rates at the end of this period. In a second arm of this study, smokers who failed to quit with the nicotine patch were randomized to eight weeks bupropion or placebo. These participants were only followed up for the period of therapy, so are not included in the results here.

In all of the bupropion, nortriptyline and fluoxetine trials, participants may have had a past history of depression but were not depressed at study entry.

Venlafaxine
One trial, reported in an abstract, compared venlafaxine at a dose of up to 225 mg/day with placebo for 136 smokers also treated with nicotine patch and nine brief individual counselling sessions (Cinciripini 1999).

All of the trials used placebo controls apart from those comparing doses of a pharmacotherapy (Swan 2001). Seventeen reported 12-month and 13 reported six-month outcomes. All the trials were described as randomized, but not all reported randomization and allocation methods in detail. Eight studies (27%) were graded A for describing a method of allocation likely to ensure that treatment assignment was concealed. All other trials were graded B because the method of allocation was not described. 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 for which data were available did use biochemical verification. Where reported, we used prolonged abstinence rates even if this was not the primary outcome. Most of the prolonged abstinence rates are based on self reported slip free abstinence from approximately day 22, or the start of the third week after the target quit date (TQD). In most of the studies, medication dosing began prior to the quit date, but all follow-ups were tied to the quit date and not to the start of medication, except for studies of relapse prevention which tied follow-up to the end of medication.

Additional details about the methodology of individual trials are given in the table 'Characteristics of Included Studies'. A number of the trials have only been published as abstracts, which have limited information on methodological issues. 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.

[Graph numbers identify the graph where the plots of trial results and pooled estimates are displayed]

TRICYCLIC ANTIDEPRESSANTS

Four out of five trials of nortriptyline reported a statistically significant increase in quit rates at long-term follow-up. The pooled odds ratio was 2.80 (95% CI 1.81 - 4.32, Graph 01.01). There was no evidence of heterogeneity between the trial providing nortriptyline as an adjunct to nicotine patch (Prochazka 2001) and the others (Hall 1998; Prochazka 1998; Da Costa 2002; Hall 2002). A past history of depression did not appear to modulate the efficacy of nortriptyline in one study, but subgroup numbers were small (Hall 1998). In two studies the presence or absence of behaviour therapy did not influence the active versus placebo odds ratio (Hall 1998, Hall 2002).

MONOAMINE OXIDASE INHIBITORS

One trial of moclobemide (Berlin 1995A) found a trend toward an effect but no significant difference in abstinence at 12 month follow-up (OR 1.76, 95% CI 0.61 - 5.07, Graph 04.01.01). Nor did the trial detect an effect on withdrawal symptoms.

SELECTIVE SEROTONIN REUPTAKE INHIBITORS (SSRIs) (Graph 04.01)

A multicentre trial of fluoxetine (Niaura 2002) found a small short-term benefit of the higher dose fluoxetine but this was not sustained. Although this report suggested fluoxetine was effective for some post hoc subgroups, quit rates based on the intent-to-treat sample after six months were similar for both dosages and for placebo, and the confidence intervals make it unlikely that a clinically significant benefit is being missed. In a second trial the addition of a low dose (20 mg/day) of fluoxetine did not increase cessation rates over nicotine inhaler alone, although confidence intervals were wide (OR 0.88, 95% CI 0.34 - 2.27) (Blondal 1999).

A trial comparing two dosages of paroxetine with placebo as an adjunct to nicotine patch failed to detect an effect on six month cessation rates (Killen 2000). Sustained abstinence rates did not differ significantly from placebo at six months for 40 mg (OR 1.32, 95% CI 0.62 - 3.28) or 20 mg (OR 0.91, 95% CI 0.41 - 2.00). There were significant differences only at four weeks follow-up.

A trial of sertraline failed to detect an effect on smoking cessation (Covey 2002). Point prevalence abstinence rates did not differ significantly six months after treatment (OR 0.67, 95% CI 0.25 - 1.78 ) or at any earlier assessment.

The pooled estimated odds ratio across all four trials of SSRIs was 0.97 (95% CI 0.71 - 1.32)

ATYPICAL ANTIDEPRESSANTS

Bupropion for cessation induction compared to placebo control
Pooling the results of 10 trials with 12 month abstinence (Ferry 1992; Ferry 1994; Jorenby 1999; Hurt 1997; Gonzales 2001; Tashkin 2001;Tonstad 2001; Zellweger 2001; Hall 2002; McRobbie 2002) and six with six month rates (Evins 2001; Hertzberg 2001; Ahluwalia 2002; Dalsgard 2002; George 2002; Lerman 2002A) gives an estimated odds ratio of 1.97 (95% CI 1.67 - 2.34, Graph 02.01). Whilst the heterogeneity between trial results was not statistically significant, there is increasing variation between trial results as more studies are included. The use of a random effects model to pool the data does not change the conclusion of a significant benefit, and increases the estimated effect because of the increased weight given to small trials, some of which have shown large effects (random effects OR 2.02, 95% CI 1.59 - 2.57). Of the 10 trials with more than 100 participants in each study arm, three had confidence intervals which did not exclude no effect (Hurt 1997; Tashkin 2001; Zellweger 2001). One of these (Tashkin 2001) recruited smokers with chronic obstructive pulmonary disease. Six month quit rates did show a difference just reaching statistical significance but later relapse was higher in the bupropion treated group and at 12 months the difference was smaller and not significant. A second, (Zellweger 2001) recruited health care professionals, doctors and nurses, in European centres. In this study the placebo quit rate was 22%, well above the typical level for placebo recipients in most other studies.

In a secondary display of the data (Graph 02.02) we have grouped the studies by population characteristics; unselected community volunteers; healthcare professionals; primary care patients or hospital outpatients without specific diagnoses; people with cardiovascular disease, COPD, schizophrenia or post traumatic stress disorder. Treatment effects were not significant in all these subgroups, but this is likely to be explained by chance and the small numbers in some subgroups.

Bupropion - effect of dose
In the first multidose study (Hurt 1997), cessation rate was linearly related to dose 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, include on the basis of an abstract but not yet published, compared 300 mg and 150 mg daily doses and reported 12 month point prevalence quit rates (Swan 2001). Pooling the two studies and comparing 300 mg versus 150 mg 12 month quit rates shows no evidence of a significant difference (OR 1.07, 95% CI 0.86 - 1.32, Graph 02.03). The second trial (Swan 2001) also crossed dose with two intensities of counselling, the more intensive of which provided proactive telephone counselling that extended beyond the period of pharmacotherapy. The number of quitters in the lower dose group who received proactive telephone contact increased between three and 12 month follow-up. The use of prolonged abstinence rather than 12 month point prevalence abstinence as the outcome might therefore show a different treatment effect.

Bupropion for relapse prevention
In a trial of bupropion therapy for relapse prevention following cessation induction with bupropion (Hays 2001) an initial benefit from continued therapy was no longer significant one year after the end of therapy (OR 1.16, 95% CI 0.76 - 1.77, Graph 02.04.01). By this time almost three quarters of those who achieved abstinence after the initial seven weeks on bupropion had relapsed. Weight gain was significantly reduced up to one year after treatment. In a second trial in smokers who quit during eight weeks of nicotine patch therapy (Hurt 2001) there was no evidence of significant benefit at the end of six months additional bupropion therapy (OR 1.19, 95% CI 0.61 - 2.32, Graph 02.04.02), with almost three quarters relapsed.

Bupropion and NRT
In one study bupropion was found to be significantly more effective than nicotine patch (OR 2.07, 95% CI 1.22 - 3.53, Graph 02.05) (Jorenby 1999). In this study combined bupropion and nicotine patch also appeared to increase quit rates more than patch alone (OR 2.65, 95% CI 1.58 - 4.45, Graph 02.06), but in a second as yet unpublished study there was no evidence of an incremental benefit for bupropion (OR 0.75, 95% CI 0.31 - 1.39, Graph 02.06) (Simon 2002). There was significant heterogeneity between these two results and thus they could not be combined statistically.

Venlafaxine
One trial of venlafaxine failed to detect a significant increase in 12 month quit rates compared to nicotine patch and counselling alone, but again confidence intervals do not exclude a clinically useful effect. (OR 1.33, 95% CI 0.59 - 3.00, Graph 5.1.7) (Cinciripini 1999).

DIRECT COMPARISON BETWEEN ANTIDEPRESSANTS

One trial has compared bupropion with nortriptyline directly (Hall 2002). There was not a significant difference between the two; the comparison favoured bupropion but the confidence intervals are wide and do not exclude a clinically useful difference in favour of either drug (OR 1.85, 95% CI 0.69 - 5.02, Graph 3.1).

ADVERSE EVENTS

We summarise adverse events reported in trials of bupropion and nortriptyline in tables (see links under Other data heading). In addition, for bupropion we include reports of adverse events from national surveillance schemes in the United Kingdom (with data to July 2002) and Canada (data to April 2001) (see Table 01).

Adverse events reported for bupropion:
The commonest side effects are insomnia, occurring in 30-40% of patients, dry mouth (10%) and nausea (GlaxoSmithKline; Goldstein 1998).Amongst people allocated to bupropion in clinical trials for smoking cessation, the drop out rate due to adverse events ranged from 7% to 12%. Early trials of bupropion as a treatment for depression using the immediate release formulation and often doses > 300 mg/day suggested it increased the risk of seizures. This led to the development of the slow release 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 2001). In this study there were two seizures amongst 502 people randomized to bupropion, one of whom had a familial history (data from GlaxoSmithKline). Based on an observational safety surveillance study, the manufacturers report the risk of seizure with the slow release preparation at a maximum dose of 300 mg/day to be about 1:1000 patients prescribed the drug (Dunner 1998). In this open and uncontrolled study, 3100 adult patients with a diagnosis of depression used bupropion Slow Release for eight weeks, 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 had a seizure considered to be related to the therapeutic use of bupropion. Post-marketing surveillance data are now available from some countries in which bupropion is licensed for smoking cessation. The limitation of these schemes is that the denominator is not definitely known. 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.

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. However, from the national surveillance schemes it is not possible to calculate the frequency of this outcome.

Although no patient died while taking bupropion in trials for smoking cessation, some have died whilst taking bupropion prescribed for smoking cessation in routine practice. Although there is as yet no formal epidemiological analysis of these deaths, no national reporting scheme has concluded that bupropion caused these deaths. Bupropion may cause adverse effects in overdose. A review of bupropion-only exposures reported to the US Toxic Exposure Surveillance System for 1998-1999 identified 3755 exposures to Wellbutrin SR, 2,184 to Wellbutrin and 1409 to Zyban (Belson 2002). These exposures included intentional overdose and unintentional ingestion as well as reports of adverse drug reactions. Clinical effects related to bupropion exposure developed in 31% of patients, with vomiting the most common childhood symptom and tachycardia the most common in teenagers and adults. Six percent (19% of symptomatic patients) developed a seizure. Seizures were more common with Wellbutrin exposures (22% of symptomatic) 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 percent of the moderate and major effects resolved in less than 24 hours. Five deaths all involved suspected suicides and 4/5 involved Wellbutrin.

Adverse effects of nortriptyline:
Drop out rates in the two trials reporting this outcome were 4% and 9%. No serious adverse events were reported in any of the smoking cessation trials. The adverse events reported included the well known tricyclic effects of dry mouth, drowsiness, lightheadedness and constipation.

Bupropion
The results of 16 placebo controlled trials of bupropion confirm that bupropion increases the proportion quitting smoking in a range of populations. There is some heterogeneity between the trial results, but this does not reach statistical significance. Bupropion approximately doubles the odds of long-term cessation. Eight of the studies had confidence intervals that did not exclude an odds ratio of 1, indicating a non significant effect, but five of these had fewer than 100 people in each arm. The largest trial failing to detect a significant effect (Zellweger 2001) was conducted with European healthcare professionals. The high quit rate in the placebo group in this group may have limited the potential for bupropion to increase rates further. Bupropion has now been tested in community volunteers in multiple countries, and in patients with cardiovascular disease. Treatment effects appear to be comparable in a range of populations and settings, and different levels of behavioural therapy. A study in patients with COPD (Tashkin 2001) did not detect a significant treatment effect at 12 month follow-up, but had done so at six months. Overall quit rates were low and this may be a hard to treat group of smokers.

Two bupropion trials that compared the recommended dose of 300 mg/day (150 mg twice daily) with a dose of only 150 mg have failed to detect a long-term benefit of the higher dose although end of treatment rates favour the higher dose. Whilst one trial found bupropion and nicotine patch more effective than patch alone (Jorenby 1999), a second did not detect additional benefit (Simon 2002). One trial showed no significant difference between nortriptyline and bupropion (Hall 2002) but had limited power to establish equivalence. Continued use of bupropion to prevent relapse has not been shown to have a significant long-term benefit.

The efficacy of bupropion appears to be independent of a past history of depression (Hayford 1999) and not due to diminished post-cessation depression (Catley 2002). This suggests the efficacy of bupropion is not due to a traditional antidepressant effect and that bupropion is of benefit to those with no history of depression. Although the mechanism of action of bupropion is still unclear, recent animal studies suggest that it may act as an antagonist at the nicotine receptor (Wiley 2002, Young 2002).

Although some deaths while taking bupropion have been reported, thus far there is insufficient evidence to suggest they were caused by bupropion. Also, no new information has appeared to suggest the rate of seizures is greater than the 1:1000 originally estimated.

Gender/age differences with bupropion
Too few of the studies have published data on long-term quit rates by sex for it to be possible to conduct a definitive subgroup meta-analysis. A recent poster presentation (Gonzales 2002) reported an analysis of short-term outcomes in five trials (Hurt 1997; Jorenby 1999; Tonstad 2001; Zellweger 2001; McRobbie 2002), and long-term outcomes for the last three of these, which are unpublished. Both meta-analysis and logistic regression were used. Neither type of analysis detected evidence that women and men differed in their response to treatment at either time point. Point estimates of effects were slightly larger for women than for men at the end of treatment, but not after 12 months. Overall quit rates were generally lower for women. In contrast to this, other reports suggest a gender difference in treatment response. A secondary analysis of Jorenby 1999 (Smith 2003) reports a gender interaction such that women appeared to benefit relatively more from bupropion than men, whereas on placebo, or nicotine patch alone, their quit rates were lower than men. Another recent study has reported a significant gender X treatment interaction (Collins 2002), and also an interaction between treatment effect, gender and genotype (Lerman 2002A). At the end of treatment women with a variant CYP2B6 gene had significantly higher quit rates when treated with bupropion than on placebo. The treatment effect was not significant for the other three gender/genotype subgroups. Women with the variant low activity gene had low quit rates on placebo, despite intensive group behavioural counselling. Bupropion attenuated this, leading to a significant treatment effect. No significant treatment effect was detected for men, or women with the wild type gene, who all achieved relatively high quit rates with behavioural therapy and placebo. A study in smokers with 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 4-7 week abstinence in males (Gonzales 2001). This was a study re-treating smokers who had already failed to quit with bupropion. No other study reports have noted a gender X treatment group interaction. In the Hays 2001 relapse prevention study, quit rates after the initial treatment period were non significantly lower, and there were no significant gender difference in abstinence rates at any time point during the relapse prevention and follow up phase (Gonzales 2002). Male gender was predictive of success in the Hurt 1997 study but gender was reported to be independent of bupropion dose (Dale 2001).

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.

A sub group 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.

Nortriptyline
Trials of nortriptyline also found that it increases smoking cessation. In one trial this occurred both with and without extra behaviour therapy and in a second it provided additional benefit over nicotine patch therapy. In the trials the efficacy of nortriptyline appeared independent of its antidepressant 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. The most common side effect in the trials was dry mouth. Based on the rate of significant adverse events when nortriptyline and bupropion are used to treat depression, nortriptyline would be expected to have higher rate of dropouts. This has not been the case in the relatively small number of subjects in the existing studies, perhaps because the dose of nortriptyline used (75-150 mg) is much smaller than that used for depression.

Bupropion versus nortriptyline
Based on current evidence bupropion and nortriptyline appear to be equally effective. The single head-to-head comparison of the two drugs did not detect a difference and the confidence intervals for the estimated effects of bupropion (95% CI 1.67 - 2.34) and nortriptyline (95% CI 1.81 - 4.32) in this meta-analysis overlap. Neither bupropion nor nortriptyline has been tested in adolescents, pregnant women or the elderly (>65) or, with the exception of one bupropion study (McRobbie 2002), in those with cardiovascular disease

Other antidepressants
The three long-term trials of selective serotonin reuptake inhibitors (fluoxetine, paroxetine and sertraline) and other short term trials have failed to show this class of antidepressants aids in smoking cessation. Some studies have found SSRIs effective in post hoc determined subgroups but this requires verification. Only one long-term trial of monoamine oxidase inhibitors is available.

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. However, the SSRI antidepressants appear not to be efficacious. This suggests either the dopaminergic or noradrenergic effects of nortriptyline and bupropion are essential for its efficacy. Also, 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. Whether the same is true for nortriptyline is not clear (Gambassi 1999). If noradrenergic effects are essential then MAOI and tricyclic antidepressants should be effective.

The findings of this review are in agreement with the conclusions of other recent reviews and guidelines (Aubin 2002, Benowitz 2000, Covey 2000a; Kotylar 2000, RCP 2000). The updated US and UK clinical practice guidelines (US DHHS 2000; West 2000) recommend the use of bupropion as a first-line medication for smokers trying to quit. The New Zealand prescribing guidelines recommend bupropion as a second-line therapy (Tatley 2001). The US guidelines endorse nortriptyline as a second-line therapy due to possible adverse events. The report of the Royal College of Physicians (RCP 2000) considers nortriptyline, monoamine oxidase inhibitors, and antidepressants with noradrenergic effects as possible therapies, whilst listing selective serotonin reuptake inhibitors as either ineffective or lacking evidence of effectiveness.

Our thanks to Drs Niaura, Borrelli, Spring, Fiore, Hurt, Mizes, Ferry, Schuh, Cinciripini, Hays, Prochazka, Ahluwalia and Mayo for assistance with additional information or data on studies.
JR Hughes is supported by Research Scientist Development Award DA-00490 from the National Institute on Drug Abuse.

JR Hughes has received consultancy fees from many pharmaceutical companies that provide tobacco related services or products, including Pfizer, GlaxoSmithKline (the makers of bupropion and NRTs) and Pharmacia (the maker of NRTs).

This review was first published as part of the review 'Anxiolytics and antidepressants for smoking cessation'. From Issue 4 2000 of the Cochrane Library the classes of drugs are reviewed separately.