Smoking cessation in severe mental ill health: what works? an updated systematic review and meta-analysis

Background People with severe mental ill health are more likely to smoke than those in the general population. It is therefore important that effective smoking cessation strategies are used to help people with severe mental ill health to stop smoking.﻿ This study aims to assess the effectiveness and cost –effectiveness of smoking cessation and reduction strategies in adults with severe mental ill health in both inpatient and outpatient settings. Methods This is an update of a previous systematic review. Electronic databases were searched during September 2016 for randomised controlled trials comparing smoking cessation interventions to each other, usual care, or placebo. Data was extracted on biochemically-verified, self-reported smoking cessation (primary outcome), as well as on smoking reduction, body weight, psychiatric symptom, and adverse events (secondary outcomes). Results We included 26 trials of pharmacological and/or behavioural interventions. Eight trials comparing bupropion to placebo were pooled showing that bupropion improved quit rates significantly in the medium and long term but not the short term (short term RR = 6.42 95% CI 0.82–50.07; medium term RR = 2.93 95% CI 1.61–5.34; long term RR = 3.04 95% CI 1.10–8.42). Five trials comparing varenicline to placebo showed that that the addition of varenicline improved quit rates significantly in the medium term (RR = 4.13 95% CI 1.36–12.53). The results from five trials of specialised smoking cessation programmes were pooled and showed no evidence of benefit in the medium (RR = 1.32 95% CI 0.85–2.06) or long term (RR = 1.33 95% CI 0.85–2.08). There was insufficient data to allowing pooling for all time points for varenicline and trials of specialist smoking cessation programmes. Trials suggest few adverse events although safety data were not always reported. Only one pilot study reported cost effectiveness data. Conclusions Bupropion and varenicline, which have been shown to be effective in the general population, also work for people with severe mental ill health and their use in patients with stable psychiatric conditions. Despite good evidence for the effectiveness of smoking cessation interventions for people with severe mental ill health, the percentage of people with severe mental ill health who smoke remains higher than that for the general population. Electronic supplementary material The online version of this article (doi:10.1186/s12888-017-1419-7) contains supplementary material, which is available to authorized users.


Background
The physical health of people with severe mental ill health (SMI) is poor, with people with a diagnosis of SMI dying 20-25 years earlier than those in the general population [1]. Smoking is one of the most important modifiable risk factors that contributes to this excess mortality [2]. People with SMI tend to smoke more heavily and extract more nicotine from cigarettes than smokers without mental health problems [3], and up to 70% of people with SMI smoke [4].
Whilst the percentage of people who smoke in the general population has been steadily declining, the percentage of people with SMI who smoke has not seen a similar decline [5]. Despite this, when questioned, the percentage of people with SMI who are interested in cutting down or quitting smoking is similar to that of the general population [6]. In 2010 a systematic review was conducted to establish the clinical and cost effectiveness of smoking cessation and reduction strategies for people with SMI to determine the most successful strategies such as the use of pharmacotherapy (e.g. nicotine replacement therapy, varenicline, bupropion) or behavioural interventions [7]. In the United Kingdom, following the publication of guidance issued by the National Institute of Health and Care Excellence (NICE) Guidance PH 48 in 2013 [8], a number of mental health trusts have decided to go smoke free and encourage people with SMI to give up or cut down on their smoking. We have therefore decided to update the 2010 review with the additional inclusion of e-cigarettes as a smoking cessation strategy to provide up to date information on the most effective and cost-effective strategies to help people with SMI cut down or quit smoking.

Objectives
To assess the effectiveness and cost-effectiveness of smoking cessation and reduction strategies in adults with severe mental ill health.

Search strategy
The protocol for this review has been registered on the PROSPERO register of systematic reviews (http://www .crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD 42015029455).
An electronic search strategy based on that used in our previous review, combining search terms for severe mental ill health, smoking cessation and randomised controlled trials, adapted from terms developed by the Cochrane groups for schizophrenia and tobacco addiction was used to search the following database for potentially relevant studies: MEDLINE (PubMed), EMBASE, PsycINFO, CINAHL, Health Management Information Consortium (HMIC) and CENTRAL.
The search strategy was limited to the inception year of each database until September 2016. An example of the search strategy is shown in an additional word file (see Additional file 1).
Searching other resources. Reference lists of all identified studies and existing reviews were checked for additional potentially relevant studies.

Inclusion criteria Types of studies
Randomised controlled trials (RCTs), including clusterrandomised controlled trials, that assess the effects of smoking cessation and reduction interventions in people with severe mental ill health were included. Studies conducted in any country and in either in-patient or out patient settings were eligible for inclusion. Studies that are not published in English were excluded.

Types of participants
Participants were adults aged 18 years and above who had been diagnosed with SMI. We defined SMI as schizophrenia or other psychotic disorders, bipolar disorder and depression with psychotic features. We have not included personality disorder, severe anxiety disorder, post traumatic stress disorder (PTSD), major depression or autism in this review. We have based this classification on diagnoses that would typically be included on a UK primary care SMI register [9]. Diagnosis needed to be made by using International Classification of Disease (ICD10 F20-29 and F30- 31) or Diagnostic and Statistical Manual (DSM IV 295.x, 296.x and 297.x) criteria.
Studies involving participants who had a problem with substance abuse (other than nicotine addiction) without any other mental disorder, or whose participants had learning disability, dementia, other neurocognitive disorders or terminal illness were not included in this review.

Types of interventions
Trials of all types of smoking cessation and reduction strategies, (behavioural or pharmacological as monotherapy or in combination) compared to each other, placebo, usual care or to no intervention were included, including trials of very brief advice. Behavioural interventions include on-toone programmes, group programmes, and telephone counselling. Pharmacotherapy includes products licensed for smoking cessation e.g. nicotine replacement therapy (NRT), varenicline, nortriptyline, and bupropion. Trials in which electronic cigarettes ('e-cigarettes') have been used as a smoking cessation aid were also included. Studies looking at 'implementation of a smoke-free environment' as an intervention were excluded. Behavioral interventions were classed as 'group' or 'individual' therapy.

Types of outcome measure
The primary outcome measure was biochemically verified self-reported smoking cessation. Accepted methods of biochemical verification were expired carbon monoxide (CO level of <10 ppm (p.p.m.), salivary cotinine <15 ng/ml, urinary cotinine <50 ng/ml or serum cotinine <15 ng/ml. All follow-up times were included and categorised as short-term quit if less than or up to four weeks, mid term quit for up to six months, and long-term quit if longer than six months. Participants lost to follow up were treated as 'still smokers'.
The secondary outcomes were: 1. Smoking reduction; as no acceptable standard exists for its measurement, any measure was acceptable as long as it was verified by biochemical assay 2. Change in body weight 3. Change in psychiatric symptoms (any validated symptom scale) 4. Adverse events

Selection of included studies and data extraction
Two authors independently screened 10% of the titles and abstracts of publications identified by the search strategy. Results from this initial screening were compared to check the level of agreement between the two authors over which studies should proceed to full text screening. Both authors were in agreement over which texts should proceed to full text screening therefore one author continued to screen the remaining studies. All studies that were not applicable according to our inclusion criteria were discarded. The full text of the remaining references was obtained.
Two authors independently decided whether the studies meet the inclusion criteria with any disagreements resolved through discussion with a third author.

Data extraction
Two authors independently extracted data from the included studies. Any disagreements were resolved through discussion with a third author where necessary.
Any missing data, relating to the primary outcome only, was sought by contacting the Investigators and/or corresponding authors of primary studies.

Assessment of risk of bias in included studies
The methodological quality of included trials was assessed independently by two reviewers using the Cochrane's tool for assessing risk of bias, [10] which assesses the following domains: 1. Sequence generation (selection bias) 2. Allocation concealment (selection bias) 3. Blinding of participants and personnel (performance bias) 4. Blinding of outcome assessment (detection bias); 5. Incomplete outcome data (attrition bias) 6. Selective outcome reporting (reporting bias) 7. Other potential sources of bias Each of the domains was scored as 'high' , 'low' or 'unclear' risk of bias, following criteria outlined in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions [10].

Data synthesis
A narrative overview of study design features, study populations, outcomes, risk of bias and study results is given.
For smoking cessation data, we present risk ratios with 95% confidence intervals as per our previous review [7]. Where interventions and comparisons were sufficiently similar we conducted a meta-analysis using RevMan (version 5.3, Review Manager (RevMan) [Computer program]. Version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). We performed standard pairwise meta-analysis for every comparison that contained at least two studies and used a random-effects model if studies were statistically heterogeneous as measured by I 2 (I 2 ≥ 50%); otherwise we used a fixed-effect model. Absolute quit rate was taken as the proportion of participants who met criteria for abstinence out of the number randomised to that group.

Unit of analysis issues
The unit of analysis was the individual.

Results
Of the 1312 records identified 106 full texts were screened ( Fig. 1). Of these 28 (based on 26 studies) involving 1978 participants met the inclusion criteria ; 18 more studies than in our previous review. The reasons for ineligibility are shown in Fig. 1, with the most common reason being that the study was not a randomised controlled trial.

Study characteristics
Study characteristics are given in Table 1. No cluster RCTs were identified in this review. The sample size of the studies ranged from five participants [22,37] to 298 participants [18]. The majority of the studies recruited participants who were outpatients (n = 20), one study recruited solely from an inpatient setting [29], and one study recruited from a mixture of inpatients and outpatients [35] the remaining 4 studies did not clearly state whether the participants were inpatients or outpatients.
Sixteen of the studies were conducted in the United States, two in Australia, one in Taiwan, one in England one in the United States, Israel and China and one in the United States and Canada. In four studies the country was not clearly stated.
The majority of the studies recruited participants with schizophrenia or schizoaffective disorder (n = 21), with three studies recruiting participants with bipolar disorder, and two studies included participants with schizophrenia, schizoaffective disorder or bipolar disorder. In eight of the studies it was a study requirement that the participants had stable symptoms, in three studies it was a requirement that participants were on a stable dose of medication and in six studies it was a requirement that participants has stable symptoms and were on a stable dose of medication. Nine studies did not state whether the participants were clinically stable or were on a stable dose of medication.
In just over half of the studies the participants had expressed a willingness to quit smoking (n = 12), in one study participants were excluded if they were planning on quitting in the next 30 days [36] and in the remaining 12 studies participants' views on quitting were not stated. No study stated that it was recruiting participants with no interest in quitting smoking.
Nine of the studies used an intention to treat analysis, one used a per protocol analysis [36] and 16 studies did not report whether or not they used an intention to treat analysis.

Description of the interventions
The included studies covered a range of interventions (Table 1). Nine studies explored the effects of the prescription of bupropion, six studies the prescription of varenicline and one study the prescription of nicotine replacement therapy (NRT). The varenicline studies all followed a standard dosing schedule whereas the dose in the bupropion studies ranged from 150 mg once per day to150 mg twice per day. Five studies explored the effects of a specialist smoking cessation programme for people with  SMI and three studies investigated the effects of contingent reinforcement (i.e., providing people with cash incentives if they had remained abstinent from smoking at defined time points). Of the nine trials (involving 306 participants in total) which explored the effects of bupropion, five tested bupropion plus group therapy versus placebo plus group therapy [13,14,17,22,26], two tested bupropion plus group therapy plus NRT versus placebo plus group therapy plus NRT [19,21] one tested bupropion plus smoking cessation counselling versus placebo plus smoking cessation counselling [30]. The final study employed a factorial design testing contingent plus bupropion versus non-contingent plus bupropion versus contingent plus placebo versus noncontingent plus placebo [24]. Tidey did not report abstinence therefore was not included in the meta-analysis.
The addition of varenicline to a range of interventions in the control arm was tested in six trials (313 participants in total). Of these six trials, four tested varenicline plus smoking cessation counselling versus placebo plus counselling [30,31,35,37], one tested varenicline plus group therapy versus placebo plus group therapy [25], and one tested varenicline versus placebo [27].
Five studies explored the effects of a smoking cessation programme designed for people with SMI (638 participants): two studies compared the smoking cessation programme to usual care [18,33], one explored a specialist programme plus NRT versus a standard smoking programme plus NRT [12], one study compared a specialist programme with medication management [23], and one study compared motivational interviewing with personalised feedback with interactive education with no personalisation [36].
Smoking cessation counselling, whether part of the intervention being tested or part of the control arm, consisted of a range of behaviour change techniques delivered in a variety of formats e.g. face-to-face one-to-one sessions, faceto-face group sessions or one-to-one sessions delivered via telephone. It is important to note that in the trials of varenicline and bupropion, where smoking cessation counselling was delivered, the same programme was delivered in both the medication (varenicline or bupropion) arm of the trial as in the usual care arm of the trial. Therefore it is unlikely that the smoking cessation counselling component of the study had any bearing on the study results. In the majority of the trials the exact content, in terms of the behaviour change techniques employed in the smoking cessation counselling, was insufficiently described.
No studies were identified exploring the effectiveness of very brief advice or the effectiveness of electronic cigarettes. Table 2 Summarises the risk of bias in the included studies. Overall the studies were at high risk or unclear risk of bias aside from Smith 2015 [34] and Smith 2016 [35] which were both at low risk of bias. Overall there was a lack of detail given in the descriptions of key study design features which has led to studies being deemed at an unclear risk of bias. For those studies that were assessed as having an unclear risk of bias the issue may be with the reporting as opposed to actual study conduct. The risk of bias was assessed by two reviewers and there were only few disagreements which were simply resolved by discussion until consensus was reached. Discussion with 3rd reviewer not necessary in any of the instances.

Smoking abstinence
Risk ratio (pooled) for point prevalence abstinence at short, medium and long term for studies exploring the addition of bupropion (Fig. 2), varenicline (Fig. 3) and a specialist smoking intervention for people with SMI ( Fig. 4) were calculated. Funnel plots are not included in this review because we identified less than 10 studies eligible for inclusion in the meta-analyses.

Bupropion versus placebo
Eight trials that tested the addition of bupropion to a range of interventions in the control arm reported abstinence data. These studies were pooled to judge whether the addition of bupropion offered any additional benefit (Fig. 2). Pooling this data using a fixed-effects meta-analysis showed that the addition of bupropion improved quit rates significantly in the medium term and long term but not in the short term (short term RR = 6.42 95% CI 0.82-50.07; medium term RR = 2.93 95% CI 1.61-5.34; long term RR = 3.04 95% CI 1. 10-8.42). The median duration of the short term comparison was four weeks, 3.5 months for the medium term comparison, and 11.75 months for the long term comparison. There was no evidence of between study heterogeneity (I 2 = 0%).

Varenicline versus placebo
Five of these studies were pooled to evaluate whether the addition of varenicline offered any additional benefit (Fig. 3). Pooling this data using a fixed-effects meta-analysis showed that the addition of varenicline improved quit rates significantly in the medium term (RR = 4.13 95% CI 1.36-12.53), median time-point six months. None of these five studies gave long term quit data. There was no evidence of between study heterogeneity (I 2 = 0%). Participants in these studies received varenicline for between eight and 12 weeks. Removing the monotherapy study [27] from the meta-analysis did not substantially change the results and there was no overall change in heterogeneity (RR = 3.62 95% CI 0.68-38.69).

Specialist smoking cessation programme
The results from the studies exploring smoking cessation interventions were mixed in terms of results when compared to those exploring the effectiveness of smoking cessation medication. Whilst some studies reported positive findings others reported negative findings. This may be due to differences in the smoking cessation intervention being tested. It may be that some interventions or components of interventions are more effective than other smoking cessation interventions, however this cannot be certain. The setting, delivery mode and who delivers the intervention may also have some influence of the effectiveness of the intervention.
Four studies gave abstinence data, three of which gave medium term data and long term data and one gave long terms data only. These studies were pooled to assess whether a specialist programme offered any additional benefit (Fig.4). Pooling this data using a fixed-effects meta-analysis showed that there was no evidence of benefit for the specialist smoking cessation programme in the medium term (RR = 1.32 95% CI 0.85-2.06) or in the long term (RR = 1.33 95% CI 0.85-2.08). Median duration of comparison was six months in the medium term and 12 months in the long term. None of these five studies gave short term quit data. There was no evidence of between study heterogeneity (I 2 = 0%).

Secondary outcomes Change in psychiatric symptoms
Of the included studies, 22 used one or more validated symptom scales to ascertain whether psychiatric symptoms had altered during the course of the trial ( Table 3). None of the studies that tested outcomes for significance found any significant worsening of psychiatric symptoms in the intervention group and only one study found a significant worsening of cognitive score in the intervention group compared to placebo [17]. Therefore it does not appear that smoking cessation interventions worsened psychiatric symptoms however due to heterogeneity between the symptom scales and time points used no meta-analysis was conducted.
Only one study that included participants with bipolar disorder reported on the significance of any change in psychiatric symptoms (not significant). The rest of the studies that reported secondary outcome included participants with schizophrenia and schizoaffective disorder.

Change in BMI
Change in BMI was not routinely measured in the included studies and only two studies listed BMI as one of their outcomes [31,33]. Of these only one study reported change in BMI therefore no meta-analysis was conducted.

Adverse events
Of the included studies 14 included some reporting of adverse events (Table 3), although in four of these studies this was not fully reported. No standardised method for reporting adverse events was used and some studies differentiated between serious adverse events and adverse events whereas some did not.

Cost effectiveness
Only one study [33] set out to explore the costeffectiveness of the intervention. This study demonstrated that it was feasible to carry out a cost-effectiveness analysis of a bespoke smoking cessation intervention compared to usual care however as it was a pilot study it was not sufficiently powered for any firm conclusions could be drawn.

Discussion
Since our previous review there has been an increase in the evidence base of smoking cessation interventions for people with SMI. Previously we identified seven studies  meeting the inclusion criteria, in this review we have included 26 studies, 19 more than our previous review, indicating that this is a rapidly developing field. Despite the increase in the number of studies exploring the effectiveness of smoking cessation interventions for people with SMI, the studies are still generally of a small size and underpowered to detect a difference between the intervention and control. Overall studies were at high or unclear risk of bias with only two of the most recent studies being at low risk of bias [34,35].
In line with the results of our previous review, this updated review indicates that people with SMI can quit smoking and the same interventions that work for people in the general population work for people with SMI e.g. the use of varenicline, bupropion or NRT to support a quit attempt. The addition of bupropion gives a similar risk ratio at both medium and long term to that of our previous review [7]. In our previous review we calculated an RR = 2.76 (95% CI 1.48-5.16) CI 1.10-8.42) compared to 3.04 (95% CI 1.10-8.42) for long term point prevalence. For varenicline our review showed a slight increase in RR compared to a recent Cochrane review [39] where the RR = 2.27 (95% CI 2.02-2.55) whilst our meta-analysis gave a medium term RR of 2.93 (95% CI 1.61-5.34). A recent review of the effectiveness of varenicline in people with SMI which had slightly different inclusion criteria to our review also concluded that varenicline was clinically superior to placebo in helping people with SMI [40]. Due to the unclear or high risk of bias of 24 of the 26 included studies in our review our results need to be interpreted with some caution.
Point prevalence absolute quit rates at the final timepoint for intervention groups ranged from 1.1 to 75.0%, and for control groups ranged from 0.0 to 22.9%. In addition quitting smoking did not appear to worsen participants' mental state. In terms of varenicline and bupropion our review indicates that both medications appear to be effective in the medium terms as an aid to smoking cessation. A recent large trial comparing outcomes of people with psychiatric disorder has also found varenicline and bupropion to be effective with no increase in neuropsychiatric events [41], however this study was not eligible for inclusion in our review as the psychiatric cohort was not limited to people with SMI. The effectiveness of behavioural interventions in helping people with SMI to quit smoking is currently unclear and is the subject of on-going study [42].
We identified two studies [29,35] that included patients in an inpatient setting, however the majority of the studies were conducted in a psychiatrically stable population and it is therefore unclear as in our previous review how far these findings are generalisable to an acutely unwell population. It is important that further studies are conducted into what works in an acutely unwell population.
The use of e-cigarettes has been increasing in recent years [43] and a Cochrane review was conducted in 2016 exploring their effectiveness as a smoking cessation aid [44]. E-cigarettes have been shown to have a similar effect on quit rate as NRT [45]. However we did not identify any RCTs that explored the use of e-cigarettes as a smoking cessation aid for people with SMI. A subgroup analysis of people who took part in the ASCEND trial was conducted analysing the results for people with mental disorders however this was not limited to SMI [46]. This subgroup analysis indicated that e-cigarettes appear to be as effective in people with mental disorders as those without mental disorders. This topic deserves further research and there is a need for future trials of electronic cigarettes as an aid to smoking cessation amongst people who use mental health services.
Only one study investigated the cost effectiveness of a smoking cessation intervention and this was a pilot study so no clear conclusions could be drawn [33]. More trials are needed with a prospective cost effectiveness analysis. In addition how an intervention may fit into existing service structures needs to be explored.    Table 3 Outcomes Fatemi 2013 [30] Not reported  Only one study reported change in body weight and this was reported as mean change in BMI [31]. Given that weight gain is associated with the prescription of antipsychotic medication [47] and the health implications of obesity it is important that weight change is recorded and reporting in clinical trials. A recent systematic review demonstrated that whilst the mean increase in body mass 12 months after stopping smoking is four to five kilograms there was a wide variation in body mass change [48] (16% of participants had a reduced mass and 13% gained more than 10 kg).
The reporting of adverse events was not standardised. In 12 of the studies included in this review no details of adverse or serious adverse events were reported. It is important that adverse events are clearly reported as per the CONSORT guidelines [49] to allow a judgment to be made as to whether or not a pharmaceutical smoking cessation aid is suitable for people with SMI.

Strengths and limitations
A limitation of this review is that it only included articles that were written in English and this could have resulted in the exclusion of potentially important studies. The fact that all the titles and abstracts were not double screened is a possible limitation however the fact that both authors who screened the initial 10% of titles and abstracts were in agreement over which studies should go forward to full text review reduces the possibility that potentially suitable studies were missed. In addition reference lists of previous reviews of smoking cessation strategies were searched. There is currently a paucity of e-cigarette research. This is a technology that is rapidly evolving and where there has been uptake in the use of e-cigarettes in advance of randomised trials being conducted. However, a strength of this review compared to our previous review is that it includes the use of e-cigarettes as a smoking cessation aid.
Due to the heterogeneity of the scales used to assess psychiatric symptoms it was not possible to conduct a detailed analysis of the results or a meta-analysis. We have therefore summarised whether or not studies found a significant change in psychiatric symptoms and concluded that no significant worsening was found on giving up smoking.
It is possible that the results of this review are at risk of publication bias. To minimise the possibility of publication bias we checked trial registries to determine whether there were any trials registered that had not been published. Funnel plots are not included in this review because we identified less than 10 studies eligible for inclusion in the meta-anayses.

Recommendations for future research
It is currently unclear what proportion of people with SMI will engage with a smoking cessation intervention and trials are needed that will explore the use of very brief advice to encourage people with SMI to seek help with smoking. It is also recommended that the use of ecigarettes as a smoking cessation aid for people with SMI be explored in future high quality RCTs.

Conclusions
Despite evidence for the effectiveness of smoking cessation interventions for people with SMI the percentage of people with SMI who smoke in the UK still remains higher than the percentage of people without SMI who smoke.
In addition to our previous findings regarding the effectiveness of bupropion in helping people with SMI to quit smoking there is now trial based evidence to demonstrate that varenicline appears to be effective in helping people with SMI to quit smoking.