A bespoke smoking cessation service compared with treatment as usual for people with severe mental ill health: the SCIMITAR+ RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 11/136/52. The contractual start date was in March 2015. The draft report began editorial review in October 2018 and was accepted for publication in May 2019. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Catherine Hewitt is a member of the Health Technology Assessment Commissioning Board. Suzy Ker reports personal fees from the NHS outside the submitted work.

Copyright statement

© Queen’s Printer and Controller of HMSO 2019. This work was produced by Peckham et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2019 Queen’s Printer and Controller of HMSO

Smoking and severe mental ill health

People with severe mental ill health (SMI), such as schizophrenia and bipolar disorder, are more likely to smoke and to smoke more heavily than those without SMI. 2,3 Estimates of the percentage of people with SMI who smoke vary depending on the setting, with up to 70% of inpatients smoking. 4 The presence of mental ill health is associated with an elevated risk of smoking by a factor of odds ratio (OR) 2.2 [95% confidence interval (CI) 1.7 to 2.8]. 2 Smokers with SMI are more nicotine dependent, more likely to become medically ill with smoking-related diseases and less likely to access help in quitting than the general population. 5 People with SMI tend to:

• begin smoking at a higher rate before diagnosis or treatment for SMI than smokers without SMI6,7

• smoke each cigarette more intensely, extracting more nicotine per cigarette. 8

People with SMI are also more likely to be unemployed than the general population. Smoking is known to be part of the ‘culture’ of mental health services, among both staff and patients. In addition, people with SMI often lack self-esteem and see the future as ‘bleak’; as a consequence, they may not be motivated to look after their physical health. 9 Many people with SMI are also misinformed about the risks and benefits of smoking versus nicotine dependence treatment. 10 They often fear and overestimate the medical risks of nicotine replacement therapy (NRT). 11 Many believe that smoking relieves depression and anxiety12 (nicotine increases anxiety13).

Smoking contributes to the poor physical health of those with SMI. In the UK, the standardised mortality rate for all causes of death for people with schizophrenia was 289 (95% CI 247 to 337), amounting to a threefold increase in mortality compared with the population of England and Wales. 14 Although people with SMI are more likely to smoke than the general population, there is evidence that general practitioners (GPs) are less likely to intervene with smokers who have a mental disorder than those who do not have a mental disorder. 5 Although the number of people in the general population who smoke has declined over the last 20 years, the number of people with SMI who smoke has not seen a similar decline. 3 The smoking rate among people with SMI in England is 40.5%,15 which is more than double that of the general population. 16 Smoking rates in people with SMI vary across settings, with up to 70% of people in psychiatric units smoking. 4

It is within this context that a number of policy initiatives have emerged that emphasise improving the physical care of those with SMI, including taking initiatives to facilitate smoking cessation and the promotion of smoke-free environments in secondary care services. 2,17,18

Existing knowledge

Smokers most commonly cite ‘stress relief and enjoyment’ as their main ‘reason’ for smoking,19 although the major cause is nicotine dependence. Nicotine acts in the midbrain, creating impulses to smoke in the face of stimuli associated with smoking20 and producing what may be thought of as a kind of ‘nicotine hunger’ (a feeling of needing to smoke) when blood nicotine concentrations are depleted. Smokers also experience nicotine withdrawal symptoms, such as unpleasant mood swings and physical symptoms, that occur on abstinence and are relieved by smoking. 21 Nicotine dependence is the main reason that most unassisted quit attempts fail within a week. 22 Cochrane systematic reviews23–30 and guidance from the National Institute for Health and Care Excellence (NICE)31 highlight the following smoking cessation interventions (including medications used as smoking cessation aids) that help smokers reduce their tobacco intake and quit smoking.

Rationale for the SCIMITAR+ trial

Despite the higher prevalence of smoking, a substantial proportion of people with SMI express a desire to quit, but they expect to find it harder than those in the general population. 41 The introduction in 2004 of a new General Medical Services (GMS) contract42 created a policy impetus to improve the quality of primary care in priority areas. In terms of mental health, the new GMS contract specified that primary care is responsible for the provision of physical health care. Importantly, for smoking cessation initiatives, it ‘incentivised’ GPs to (1) produce a register of people with severe long-term mental health [Quality and Outcomes Framework (QOF) indicator MH8 – the SMI register] and (2) ensure that at least 90% of SMI patients have had a review that includes smoking status recorded within the previous 15 months (QOF indicator MH9 – SMI health check). This check included patients seen in primary care, secondary care and under shared care arrangements. Though this GMS target has now been withdrawn, this was an incentive at the time of conception of the SCIMITAR+ (smoking cessation intervention for severe mental ill health) trial.

In addition, for those who are admitted to hospital (including inpatient psychiatric units), the introduction of smoke-free policies provides an opportunity to address smoking (Public Health Guidance 4817). The admission of an individual to hospital, while being stressful and occurring at a time of personal crisis, also provides a unique opportunity to provide general health advice and to engage individuals in interventions targeted at smoking reduction and cessation.

Recent guidance issued by NICE17 offers clear statements of purpose to make secondary care services (including mental health services) entirely smoke free and to promote a smoke-free culture among staff and users of the services. Mental health services are highlighted as an area of priority and an unmet need in relation to smoking cessation, and there is clear guidance that services should be developed and implemented as a matter of some priority.

In 2009, the SCIMITAR pilot trial was devised to test whether or not a bespoke smoking cessation (BSC) intervention tailored to the needs of people with SMI would be acceptable to people with SMI. 43 In addition, it tested whether or not it was feasible to recruit and retain participants in a RCT to test this intervention. Recruitment to the pilot trial took place between May 2011 and May 2012. Results from the SCIMITAR pilot trial indicated that the intervention was acceptable to participants and that it was possible to recruit and retain participants in such a trial. 43 Although the trial was not powered to detect a difference between the bespoke intervention and usual care, the results indicated that there was a potential but non-statistically significant benefit of the bespoke intervention over usual care in terms of the proportion of self-reported quitters [12/33 (36%) vs. 8/35 (23%) adjusted OR 2.9, 95% CI 0.8 to 10.5]. Following this, the National Institute for Health Research Health Technology Assessment programme commissioned a fully powered RCT (SCIMITAR+) to explore the clinical effectiveness and cost-effectiveness of the SCIMITAR BSC intervention.

Research objectives44

• To establish the clinical effectiveness of a BSC intervention compared with usual care for people with SMI.

• To establish the cost-effectiveness of a BSC intervention for people with SMI.

Study design

This study was a pragmatic, two-arm, parallel-group RCT. The setting was in primary care and specialist mental health services within secondary care. Given that patients with SMI are a ‘hard-to-reach population’, a range of complementary strategies were used to try to identify and recruit eligible participants. A two-stage recruitment process was employed to check for eligibility, to check understanding of the study and to obtain consent. Participants were individually randomised to receive usual care or usual care plus a BSC service. Participants were followed up over the course of 12 months, with data collected at 6 and 12 months post randomisation. 44

Approvals obtained

Ethics approval was sought and granted on 15 March 2015 by Leeds East Research Ethic Committee (reference number 15/YH/0051). Approval was also obtained from the relevant research and development departments (see Appendix 1).

Trial sites

The study was conducted in 22 sites in England. Sites recruited throughout the duration of the study.

Participant eligibility

Identifying participants

We used five methods to recruit participants.

Screening for eligibility

After receiving an information pack and once a person gave consent to be contacted (either verbally to the recruiting clinician or in writing to the SCIMTAR+ researchers), a SCIMITAR+ researcher approached them by telephone, face to face or by e-mail, depending on the person’s preference. After briefly explaining the trial, the researcher checked the person’s eligibility by asking about their smoking habits, specifically (1) whether or not they smoke; (2) if so, how much they smoke (they needed to smoke at least five cigarettes per day to be eligible); (3) if they would consider quitting or cutting down, with a view to quitting within the next 6 months; and (4) whether or not they were currently receiving advice from a smoking cessation advisor. These questions ensured that the person currently smoked but was contemplating quitting. The researcher also asked screening questions about pregnancy and breastfeeding, and drug and alcohol usage, which led to exclusion if present. The researcher then arranged a meeting at a mutually convenient time and venue.

Consenting participants

When a potential participant met with the SCIMITAR+ researcher, they were given the opportunity to clarify any points they did not understand and to ask any questions. A full explanation of the trial was given by the SCIMITAR+ researcher. It was emphasised that they may withdraw their consent to participate at any time without loss of treatment options that would otherwise be available to them. They were also informed that, by consenting, they agreed to their GP being informed of their participation in the trial. Written informed consent was then obtained, with both the participant and the researcher signing and dating the consent forms prior to collecting baseline data (see Appendix 2 for participant information sheets and consent forms).

Baseline assessment

Once a participant had consented to take part in the trial, they completed the baseline questionnaires. Height and weight measurements were completed to calculate the participant’s body mass index (BMI) and an exhaled breath carbon monoxide (CO) reading was taken. These made up the baseline data set. The participant was randomised on completion of this data set.

Randomisation

Eligible, consenting individuals were randomised 1 : 1 to either the BSC service or usual care. The researcher contacted a secure telephone randomisation service run by the York Trials Unit. Simple randomisation was used, following a computer-generated number sequence devised by an independent data manager otherwise not involved in the conduct of the trial. Once given the details of the participant’s allocation, the researcher immediately informed the participant of their allocation and what would happen next. If allocated to the BSC arm, researchers at the University of York then contacted the MH-SCP in the participant’s area and organised for them to contact the participant to arrange a mutually convenient time to meet. A letter was sent to the GP and mental health specialist for the participant’s records and to advise them on subsequent smoking cessation management. Owing to the nature of the intervention, it was not possible to blind participants, GPs, researchers or the MH-SCPs to the treatment allocation.

Ineligible and non-consenting individuals

All ineligible and non-consenting persons were referred back to their GP/mental health service.

Sample size

Results from the SCIMITAR pilot trial were used to inform the sample size calculation for this trial. In the pilot trial, a quit rate of 23% was observed at 12 months in the usual-care arm and a quit rate of 36% was observed in the BSC arm. 43 The SCIMITAR+ trial was powered at 80% to detect a relative risk increase in quitting of 1.7, assuming a control quit rate of 20% (an increase to a quit rate of 34% in the intervention group), equal randomisation and a two-sided alpha of 0.05. Allowing for a 20% loss to follow-up at 12 months required a total sample size of at least 393 to be recruited and randomised. We therefore proposed to recruit 400 participants in total to ensure sufficient power according to the preceding statistical assumptions.

Description of interventions

Follow-up

Participants were followed up 6 months and 12 months after randomisation.

All assessments were carried out face to face where possible and, where not possible, a systematic approach was used to explore other avenues to collect self-report data; the participant was offered the option of completing the questionnaires over the telephone or having the questionnaires sent by post for them to complete and return. If it was not possible to make contact with the participant, a family member or friend, designated by the participant at their baseline interview, was contacted to verify the participant’s current smoking status.

Outcomes

Participant concordance

Mental health smoking cessation practitioners were asked to complete a treatment log that recorded all participants’ contacts (face-to-face meetings, telephone calls and joint appointments with MH-SCPs and GPs) to judge the degree to which the intervention, as designed on a session-by-session basis, was actually delivered in practice by MH-SCPs.

At each meeting, MH-SCPs would take a CO measurement from the participant. Measurements of ≥ 10 p.p.m. indicated that the participant had not ceased smoking. If the participant claimed to have stopped but their CO readings were ≥ 10 p.p.m., they were probed about when they last smoked and whether or not they had had any minor relapses during their quit attempt.

Statistical analysis

All analyses were conducted following the principles of intention to treat, including all randomised participants in the groups to which they were randomised, where data were available. Analyses were conducted in Stata^® version 15 (StataCorp LP, College Station, TX, USA). All statistical tests were two-sided at the 5% significance level.

Health economic analysis

The health economic analysis for the SCIMITAR+ trial assessed the cost-effectiveness of supplementing usual care with BSC, compared with usual care alone, within the trial. Although the impact of smoking cessation is likely to be observed in the long term, the within-trial evaluation could provide a relatively realistic estimate of intervention costs, necessary parameter inputs for model projection and potentially some insights into the difference in change, in the case of lacking in capacity of a long-term follow-up. The economic evaluation took the form of an incremental cost-effectiveness analysis from a NHS and Personal Social Services perspective, as recommended by NICE guidance. 57

The costs of BSC and usual care were identified, measured and valued. Costs and outcomes were collected at participant level and based on the trial population. Health outcomes were measured using quality-adjusted life-years (QALYs), as per NICE guidance. 57 QALYs were constructed from the EQ-5D-5L,54 which was collected at baseline and during follow-up.

Health-related quality of life

The EQ-5D-5L was used to measure health-related quality of life in both the BSC group and the usual-care group. The instrument was used at baseline and at 6- and 12-month follow-up. An index score was calculated using the mapping function developed by van Hout et al. 64 The index score ranges from –0.594 to 1, with higher scores indicating better health-related quality of life. QALYs were then derived from EQ-5D-5L index score at three time points by calculating the area under the curve. 65 The accompanied visual analogue scale (VAS) was also reported, ranging from 0 to 100, with higher numbers indicating better health-related quality of life.

After the original plan was finalised, a EQ-5D-5L value set for England was released by the Office of Health Economics. 66 Following that development, NICE released a position statement in August 2017 stating that, for the data gathered using EQ-5D-5L, the mapping function approach is still the preferred method for reference case analysis and the five-level version valuation set is not recommended for use at this time. 67 The updated statement in November 2018 retained this position. 68 Therefore, we maintained the original plan and disregarded the available five-level version valuation set for England.

Missing data

For the baseline covariates, little to no missing data were expected. These missing values were imputed by the mean of the parameter in the whole sample because these were information collected before the intervention began and the randomisation functioned to balance the parameters between two arms.

Methods for dealing with missing data on the primary outcome (CO-verified smoking cessation) followed the approach used in the previous statistical analysis. The missing data on costs and quality-of-life elements at follow-ups were imputed using multiple imputation. A chained equation model was developed, and predictive mean matching was used as the main imputation method for continuous variables, using 10 nearest neighbours to the prediction as a set from which to draw. All missing data were imputed separately by trial group. The imputation was performed on the aggregated level estimate, that is, EQ-5D-5L utility value converted from data provided by participants. As a rule of thumb, the number of imputations was set to the highest percentage of all missing values. 69

Analysis

Adverse events

A trial-specific procedure for detecting and reporting adverse events was implemented.

An adverse event was defined as any unexpected effect or untoward clinical event affecting the participant. Standard criteria for serious adverse events (SAEs) were used.

The participant’s MH-SCP, GP or mental health specialist was requested to inform the research team of any SAEs or non-SAEs. In addition, when participants indicated hospital attendance or use of emergency services either during a visit or in questionnaire responses, this was followed up by the research team as required.

All adverse events and SAEs were independently reviewed by a clinician and were routinely reported to the Data Monitoring and Ethics Committee and the Independent Trial Steering Committee (TSC).

Adverse event data are summarised descriptively by treatment group.

Suicide and self-harm risk protocol

A protocol for identifying and reporting potential suicide risk was implemented (see Appendix 3 for protocol). Item 9 on the PHQ-9, which asks if the participant has had ‘thoughts that you would be better off dead or of hurting yourself in some way’, was used to identify any suicide risk.

If the participant indicated a response of 3 (i.e. nearly every day) for this item at baseline or at any of the follow-up interviews, then the suicide protocol was implemented and the participant was asked if they had talked to their GP, psychiatrist or care co-ordinator/community psychiatric nurse (CPN) about these feelings. If they had not, consent was sought to contact the participant’s GP to inform them of the situation. If the participant refused, the relevant designated psychiatrist/health professional was contacted. If the health professional agreed, the participant’s GP or psychiatrist was contacted immediately. A Suicidal Intent Form was completed and, when applicable, a Suicidal Intent Form: Psychiatrist/Health Professional was also completed. These forms were stored with the participant’s trial records.

Patient and public involvement in research

The SCIMITAR+ trial benefited from the involvement of users of mental health services and carers of people with SMI throughout the research period. Our TSC included representation from carers and service users. Our protocol and study materials were scrutinised and supported by users and carers in the north-west of England.

We invited service users from a patient and public involvement (PPI) group based in the north-west to be involved in the study. One of the collaborators works at the University of Manchester, the institution that organised the PPI group. From this group, service users provided input on the recruitment materials. Service users from this group were also invited to sit on the TSC; we had two service users sit on the TSC and provide input into the running of the trial. We also had the benefit of a carer who sat on the TSC in the pilot trial and continued to sit on the TSC in the main trial.

A service user and carer were invited to attend one of our yearly SCIMITAR+ meetings, which we held for all the organisations taking part in the study. They gave a presentation on smoking and giving up smoking and provided useful input into the study. They continue to remain involved and provide advice on our work. We are seeking advice from service users regarding the dissemination of the SCIMITAR+ trial and have recently formed a PPI group at the University of York for service users with SMI. We will be asking this group to provide input into our dissemination plan.

Recruitment from service user groups

After the trial had been recruiting for 3 months, recruitment from service user groups was added as an additional recruitment method. This involved providing service user groups with information about the SCIMITAR+ trial, along with copies of the flyer. Interested service users could then contact the mental health professional named on the flyer or their care co-ordinator, or, if the service user preferred, the staff member at the service user group could contact the service users’ care co-ordinator on their behalf.

This recruitment method was added to broaden the range of recruitment methods and therefore broaden the range of participants invited to take part in the SCIMITAR+ trial. This method also had the potential to offer service users who may not routinely be attending a clinic, and therefore not approached by a clinician, the opportunity to take part in the study.

Recruitment from the Lifestyle Health and Wellbeing Survey

Using the Lifestyle Health and Wellbeing Survey to recruit was added as a recruitment option after the trial had been recruiting for 8 months. The survey is a questionnaire-based prospective study involving adults aged ≥ 18 years with SMI. Participants in the survey complete a series of questions about their health and well-being, including questions about their smoking. Participants in the survey who consented to being contacted by the research team and who smoked and were potentially interested in cutting down on or quitting smoking were invited to take part in the SCIMITAR+ trial.

The use of the Lifestyle Health and Wellbeing Survey was included to broaden the range of recruitment options and to offer people who may not have been invited by their mental health team or GP to take part in the SCIMITAR+ trial the opportunity to do so. During the course of the SCIMITAR+ trial, we noticed that some clinicians were gatekeeping: instead of checking whether or not a person wanted to do something about their smoking, they were stating that the person did not want to stop smoking. In the survey, each person was asked whether or not they wanted to do something about their smoking, thus removing the possibility of an incorrect assumption being made.

Recruitment

Recruitment started in October 2015 and ended in December 2016. Over the course of the trial, 40 GP surgeries mailed out recruitment packs, and 21 mental health trusts were enlisted to recruit participants.

A total of 526 participants were recruited to the trial. The target and actual rates of recruitment are shown in Figure 1, and monthly recruitment figures are shown in Figure 2. Recruitment took place at 21 mental health trusts and 16 GP surgery sites (one GP site was a standalone site and the other GP sites were acting as participant identification centres for the mental health trusts), although eligible and randomised participants were primarily identified from mental health trusts. Table 3 shows the recruitment broken down by method. Mental health trusts recruited between four and 52 participants per trust, and GP surgeries recruited between one and five participants. Participant flow through the trial is shown in CONSORT flow diagrams in Figures 3–5. Of the 526 participants, 265 were randomised to the BSC group and 261 were randomised to the usual-care group.

FIGURE 1.

The SCIMITAR+ target and actual recruitment rates.

FIGURE 2.

Monthly recruitment into the SCIMITAR+ trial.

FIGURE 3.

Primary care CONSORT flow diagram up to randomisation.

FIGURE 4.

Secondary care CONSORT flow diagram up to randomisation.

FIGURE 5.

The CONSORT flow diagram showing participant flow through the trial, from randomisation.

Baseline data

The baseline characteristics of participants are summarised in the following tables, and the treatment groups appear well balanced across baseline data. Table 4 summarises the sociodemographic and employment data for the participants. Overall, 58.7% of the participants were male, 41.1% were female and one participant identified as transgender (0.2%). The mean age of participants was 46 years, with a range from 19 to 72 years. The vast majority of participants were white (89.7%). Over two-thirds of the participants were unemployed but not seeking work owing to ill health (67.7%).

Table 5 presents data on the general health of the participants at baseline. Most participants reported at least moderate health in the last year (65.4%), despite most participants reporting that they experienced at least one of the listed medical conditions (85.6%). The majority of participants (83.5%) felt that smoking had negatively affected their health, and 70.9% reported that they had been advised to stop smoking by their GP. The mean BMI of participants was 29.9 kg/m² (range 16.6–59.8 kg/m²), which falls in the overweight range.

Table 6 summarises baseline mental health status. The most common severe mental health problems were schizophrenia or other psychotic illness (n = 342, 65.0%), bipolar disorder (n = 115, 21.9%) and schizoaffective disorder (n = 66, 12.5%). Nearly two-thirds of the participants had a care programme approach (CPA) co-ordinator (64.4%) and/or a CPN (64.6%), and 83.3% were under the care of a CMHT. On average, participants were aged 26 years when they were first diagnosed with their mental health problem and had required psychiatric treatment in hospital an average of 2.8 times (range 0–100 times) in the last 10 years. Most (64.3%) participants described their condition as ‘stable’, but 13.7% described their condition as ‘unstable’ (although each participant had been judged to be stable from the point of view of their condition by either their GP or a responsible mental health professional), and 22.0% either were unsure or did not respond to this question.

Participants reported that they started smoking, on average, at the age of 16 years (range 5–50 years) and had been smoking for a mean of 29.9 (SD 12.9) years (Table 7). The most common form of tobacco used was factory-made cigarettes, reportedly used by 63.3% of participants, followed by hand-rolled cigarettes (58.7%). The mean number of cigarettes smoked per day was 24.9 (range 3–100 cigarettes) (although all participants reported smoking at least five cigarettes per day when the eligibility check was carried out, one participant stated that they smoked three cigarettes per day in the self-complete smoking questionnaire). The median number of attempts to quit ever was 3, with a range of 0–200. The median duration of longest quit attempt reported was 40 days, with a range of 0 days to 188 months. The most common self-reported previous strategies used to stop smoking were ‘cold turkey’ (55.3%) followed by NRTs (20.3%).

The reasons for smoking and their importance are summarised in Table 8. The reasons most commonly reported as very important for smoking were helping to cope with stress (61.2%) and to give them something to do (39.7%).

Table 9 summarises the reasons for wanting to try to give up smoking; the reasons most commonly reported as very important were that it is bad for health (87.5%) and that it is expensive (71.9%).

Participant use of e-cigarettes is summarised in Table 10. Over two-thirds (70.3%) of the participants reported having ever tried an e-cigarette, of whom 77 (22.0%) had used e-cigarettes at least once per week in the past month. Of the participants who reported ever having tried an e-cigarette, over half stated that their use of tobacco had not changed since they started using e-cigarettes (53.5%), with only 34.1% saying that they now smoke less tobacco. The most commonly cited reason for using e-cigarettes was to try to quit smoking tobacco (51.1%).

Participant questionnaire return rates

Response rates to the participant questionnaires at 6 and 12 months are presented in Table 11. Over 95% of the forms given to participants at 6 months were returned (466/489, 95.3%; 88.6% of 526 randomised participants). The median number of days between the form being due and being returned was 6 (replacing those that were completed early with 0 days), and the majority were completed face to face (95.9%). At the primary time point of 12 months, 450 (93.8%) of the 480 forms sent out were returned (85.6% of 526 randomised participants). The median time to return was 4 days, and only eight (1.8%) were not completed face to face (six over the telephone and two on paper via post).

Primary analysis

The CO breath measurement readings are summarised by treatment group at baseline, 6 and 12 months in Table 12. The mean reading at baseline was 24.6 (SD 15.2) p.p.m. and was similar between the two groups. Seventy (13.3%) participants had a reading of < 10 p.p.m. at baseline. In both groups, the mean reading decreased from baseline to 6 months and then increased slightly at 12 months.

Self-reported smoking status and quit attempt data reported on the participant questionnaires at 6 and 12 months are presented in Table 13. At 6 months, 463 participants (88.0%) provided a valid response to the self-reported smoking status question, and 51 of these (11.0%) reported that they had not smoked even a puff in the last week [34 (14.5%) in the BSC group and 17 (7.4%) in the usual-care group]. At the primary time point of 12 months, 449 participants (85.4%) provided a valid response to the self-reported smoking status question, and 57 of these (12.7%) reported that they had not smoked even a puff in the last week [35 (15.5%) in the BSC group and 22 (9.9%) in the usual-care group). For those who said that they did still smoke, the most common hour of the day to have the first puff of a cigarette is between 08.00 and 09.00 (20% at both time points). At 12 months, of the participants with valid data for this question, 16.1% of BSC participants and 10.6% of usual-care participants reported that they had stopped smoking completely, and participants reported a median of one quit attempt in the previous 6 months, with the most recent quit attempt lasting an average of 3.5 weeks before the participant returned to smoking. Responses to the items in Table 13 relating to smoking status and smoking in the last week did not always agree, as might be expected. For example, replying ‘not a puff’ in relation to the amount smoked in the last week did not always imply that participants reported that they had ‘stopped smoking completely’, and vice versa. For instance, one participant who reported that they had not smoked a puff in the past week at 12 months in the usual-care group reported that they ‘smoke cigarettes (including hand-rolled) but not every day’, and this could be because they still smoke occasionally but just had not smoked in the previous week.

Smoking cessation at 6 and 12 months was defined as a CO measure of < 10 p.p.m. (indicating no smoking in the last 12 hours) and self-reported cessation (indicating no smoking in the last week). At 6 months, 446 participants (84.8%) provided a CO reading, 463 (88.0%) had valid data for self-reported smoking status and 443 (84.2%) had both (Table 14); 32 out of 226 (14.2%; 10.6% of 265 randomised participants) fulfilled the definition for smoking cessation in the BSC group and 14 out of 217 (6.5%; 5.4% of 261 randomised participants) fulfilled the definition in the usual-care group. At 12 months, 443 (84.2%) provided a CO reading, 449 (85.4%) had valid data for self-reported smoking status and 442 (84.0%) had both; 34 out of 223 (15.2%; 12.8% of 265 randomised participants) fulfilled the definition for smoking cessation in the BSC group and 22 out of 219 (10.0%; 8.4% of 261 randomised participants) fulfilled the definition in the usual-care group.

The OR for the difference between the two groups from the model adjusting for baseline number of cigarettes smoked, and site as a random effect, at 12 months was 1.6 (95% CI 0.9 to 2.8; p = 0.12). The OR at 6 months was 2.4 (95% CI 1.2 to 4.7; p = 0.01).

Sensitivity analysis

Secondary analysis

Fagerström Test for Nicotine Dependence

The FTND score is summarised by randomised group and time point in Table 15. Adjusted FTND means and group differences are presented in Table 21 and displayed in Figure 6. The adjusted mean difference at 6 months was –0.18 (95% CI –0.53 to 0.17) and at 12 months it was –0.01 (95% CI –0.39 to 0.38), both marginally favouring the BSC group but not statistically significantly (p = 0.32 and 0.97, respectively).

FIGURE 6.

Adjusted means for the FTND by treatment group and time point.

These results were extracted from a mixed-effects model in which a participant was treated as a random effect and observations over time (6 and 12 months) were nested within participant. The effect of randomised treatment group was assessed while adjusting for time, group-by-time interaction, baseline FTND score and baseline smoking severity (fixed effects), and site as a random effect. Different covariance structures were applied to the model. An unstructured pattern that models all variances and covariances separately resulted in the lowest AIC and so was used in the final model. Diagnostics of model fit revealed that the standardised residuals were normally distributed and were uniform against fitted values; therefore, data transformation was not required.

Motivation to Quit questionnaire

The MTQ score is summarised by randomised group and time point in Table 16. Adjusted MTQ means and group differences are presented in Table 21 and displayed in Figure 7. The adjusted mean difference at 6 months was 0.58 (95% CI –0.01 to 1.17) and at 12 months was 0.64 (95% CI 0.04 to 1.24), both marginally favouring the BSC group. The difference is almost statistically significant at the 5% level at 6 months (p = 0.06) and is below this level at 12 months (p = 0.04).

TABLE 16 - The MTQ scores by treatment group and time point

Max., maximum; min., minimum.

Time point	MTQ	BSC (N = 265)	Usual care (N = 261)	Overall (N = 526)
Baseline	Mean (SD)	13.9 (2.7)	13.7 (2.6)	13.8 (2.7)
	Median (min., max.)	14.0 (6.0, 19.0)	14.0 (5.0, 19.0)	14.0 (5.0, 19.0)
	Missing, n (%)	5 (1.9)	2 (0.8)	7 (1.3)
Month 6	Mean (SD)	13.2 (3.4)	12.4 (3.1)	12.8 (3.3)
	Median (min., max.)	14.0 (4.0, 19.0)	12.0 (4.0, 19.0)	13.0 (4.0, 19.0)
	Missing, n (%)	48 (18.1)	60 (23.0)	108 (20.5)
Month 12	Mean (SD)	13.0 (3.3)	12.3 (3.4)	12.6 (3.4)
	Median (min., max.)	13.0 (5.0, 19.0)	13.0 (4.0, 19.0)	13.0 (4.0, 19.0)
	Missing, n (%)	65 (24.5)	61 (23.4)	126 (24.0)

FIGURE 7.

Adjusted means for the MTQ by treatment group and time point.

These results were extracted from a mixed-effects model in which participant was treated as a random effect and observations over time (6 and 12 months) were nested within participant. The effect of randomised treatment group was assessed while adjusting for time, group-by-time interaction, baseline MTQ score and baseline smoking severity (fixed effects), and site as a random effect. Different covariance structures were applied to the model. A compound symmetric (or exchangeable) pattern that assumes equal correlation of errors between each pair of time points within participants resulted in the lowest AIC and so was used in the final model. Diagnostics of model fit revealed that the standardised residuals were normally distributed, and were uniform against fitted values; therefore, data transformation was not required.

Patient Health Questionnaire-9 items

The PHQ-9 score is summarised by randomised group and time point in Table 17. Adjusted PHQ-9 means and group differences are presented in Table 21 and displayed in Figure 8. The adjusted mean difference at 6 months was 0.20 (95% CI –0.85 to 1.24), marginally favouring usual care, and at 12 months was –0.12 (95% CI –1.18 to 0.94), marginally favouring BSC, but neither statistically significantly (p = 0.72 and 0.82, respectively).

TABLE 17 - The PHQ-9 scores by treatment group and time point

Max., maximum; min., minimum.

If you checked off problems, how difficult have these problems made it for you to do your work, take care of things at home, or get along with other people?

Time point	PHQ-9	BSC (N = 265)	Usual care (N = 261)	Overall (N = 526)
Baseline
Score
	Mean (SD)	10.3 (6.7)	10.8 (6.6)	10.5 (6.6)
	Median (min., max.)	10.0 (0.0, 27.0)	10.0 (0.0, 27.0)	10.0 (0.0, 27.0)
	Missing, n (%)	1 (0.4)	1 (0.4)	2 (0.4)
How difficult,a n (%)		n = 255	n = 254	n = 509
	Not at all	65 (25.5)	64 (25.2)	129 (25.3)
	Somewhat	106 (41.6)	111 (43.7)	217 (42.6)
	Very	42 (16.5)	45 (17.7)	87 (17.1)
	Extremely	42 (16.5)	34 (13.4)	76 (14.9)
Month 6
Score
	Mean (SD)	9.3 (6.7)	9.4 (6.4)	9.3 (6.6)
	Median (min., max.)	8.0 (0.0, 27.0)	8.0 (0.0, 27.0)	8.0 (0.0, 27.0)
	Missing, n (%)	42 (15.8)	47 (18.0)	89 (16.9)
How difficult,a n (%)		n = 216	n = 210	n = 426
	Not at all	67 (31)	56 (26.7)	123 (28.9)
	Somewhat	89 (41.2)	98 (46.7)	187 (43.9)
	Very	33 (15.3)	36 (17.1)	69 (16.2)
	Extremely	27 (12.5)	20 (9.5)	47 (11)
Month 12
Score
	Mean (SD)	9.0 (6.7)	9.7 (6.7)	9.3 (6.7)
	Median (min., max.)	8.0 (0.0, 27.0)	9.0 (0.0, 27.0)	8.0 (0.0, 27.0)
	Missing, n (%)	52 (19.6)	50 (19.2)	102 (19.4)
How difficult,a n (%)		n = 206	n = 206	n = 412
	Not at all	70 (34)	42 (20.4)	112 (27.2)
	Somewhat	70 (34)	93 (45.1)	163 (39.6)
	Very	47 (22.8)	47 (22.8)	94 (22.8)
	Extremely	19 (9.2)	24 (11.7)	43 (10.4)

FIGURE 8.

Adjusted means for the PHQ-9 by treatment group and time point.

These results were extracted from a mixed-effects model in which participant was treated as a random effect and observations over time (6 and 12 months) were nested within participant. The effect of randomised treatment group was assessed while adjusting for time, group-by-time interaction, baseline PHQ-9 score and baseline smoking severity as fixed effects, and site as a random effect. Different covariance structures were applied to the model. A compound symmetric (or exchangeable) pattern that assumes equal correlation of errors between each pair of time points within participants resulted in the lowest AIC and so was used in the final model. Diagnostics of model fit revealed that the standardised residuals were normally distributed, and were uniform against fitted values; therefore, data transformation was not required.

Generalised Anxiety Disorder Assessment-7 items

The GAD-7 score is summarised by randomised group and time point in Table 18. Adjusted GAD-7 means and group differences are presented in Table 21, and displayed in Figure 9. The adjusted mean difference at 6 months was –0.32 (95% CI –1.26 to 0.62) and at 12 months was –0.10 (95% CI –1.05 to 0.86), both marginally favouring BSC, but neither statistically significantly (p = 0.50 and 0.84, respectively).

TABLE 18 - The GAD-7 scores by treatment group and time point

Max., maximum; min., minimum.

If you checked off problems, how difficult have these problems made it for you to do your work, take care of things at home, or get along with other people?

Time point	GAD-7 questionnaire	BSC (N = 265)	Usual care (N = 261)	Overall (N = 526)
Baseline
Score
	Mean (SD)	8.4 (6.2)	8.4 (6.1)	8.4 (6.1)
	Median (min., max.)	7.0 (0.0, 21.0)	8.0 (0.0, 21.0)	8.0 (0.0, 21.0)
	Missing, n (%)	1 (0.4)	1 (0.4)	2 (0.4)
How difficult,a n (%)		n = 249	n = 249	n = 498
	Not at all	72 (28.9)	63 (25.3)	135 (27.1)
	Somewhat	93 (37.3)	113 (45.4)	206 (41.4)
	Very	53 (21.3)	41 (16.5)	94 (18.9)
	Extremely	31 (12.4)	32 (12.9)	63 (12.7)
Month 6
Score
	Mean (SD)	7.0 (5.9)	7.3 (5.8)	7.1 (5.9)
	Median (min., max.)	6.0 (0.0, 21.0)	7.0 (0.0, 21.0)	6.0 (0.0, 21.0)
	Missing, n (%)	41 (15.5)	44 (16.9)	85 (16.2)
How difficult,a n (%)		n = 202	n = 197	n = 399
	Not at all	70 (34.7)	52 (26.4)	122 (30.6)
	Somewhat	73 (36.1)	86 (43.7)	159 (39.8)
	Very	35 (17.3)	45 (22.8)	80 (20.1)
	Extremely	24 (11.9)	14 (7.1)	38 (9.5)
Month 12
Score
	Mean (SD)	7.0 (6.3)	7.4 (6.0)	7.2 (6.2)
	Median (min., max.)	5.0 (0.0, 21.0)	7.0 (0.0, 21.0)	6.0 (0.0, 21.0)
	Missing, n (%)	51 (19.2)	49 (18.8)	100 (19.0)
How difficult,a n (%)		n = 196	n = 202	n = 398
	Not at all	62 (31.6)	51 (25.2)	113 (28.4)
	Somewhat	74 (37.8)	88 (43.6)	162 (40.7)
	Very	38 (19.4)	41 (20.3)	79 (19.8)
	Extremely	22 (11.2)	22 (10.9)	44 (11.1)

FIGURE 9.

Adjusted means for the GAD-7 by treatment group and time point.

These results were extracted from a mixed-effects model in which participant was treated as a random effect and observations over time (6 and 12 months) were nested within participant. The effect of randomised treatment group was assessed while adjusting for time, group-by-time interaction, baseline GAD-7 score and baseline smoking severity as fixed effects, and site as a random effect. Different covariance structures were applied to the model. A compound symmetric (or exchangeable) pattern that assumes equal correlation of errors between each pair of time points within participants resulted in the lowest AIC and so was used in the final model. Diagnostics of model fit revealed that the standardised residuals were normally distributed, and were uniform against fitted values; therefore, data transformation was not required.

Short Form questionnaire-12 items

The SF-12 mental and physical health component subscale scores are summarised by randomised group and time point in Table 19. Adjusted SF-12 means and group differences are presented in Table 21 and displayed in Figure 10. The adjusted mean difference for the mental component score at 6 months was –0.73 (95% CI –2.82 to 1.36) and at 12 months was –0.41 (95% CI –2.35 to 1.53), both marginally favouring usual care, but neither statistically significantly (p = 0.49 and 0.68, respectively). The adjusted mean difference for the physical component score at 6 months was 1.75 (95% CI 0.21 to 3.28), indicating a statistically significant benefit of BSC (p = 0.03); however, this difference reduced at 12 months to 0.59 (95% CI –1.07 to 2.26) and was not statistically significant (p = 0.48).

TABLE 19 - The SF-12 mental and physical health component scores by treatment group and time point

Max., maximum; min., minimum.

Time point	SF-12	BSC (N = 265)	Usual care (N = 261)	Overall (N = 526)
Mental component score
Baseline	Mean (SD)	38.6 (12.6)	37.9 (11.7)	38.2 (12.2)
	Median (min., max.)	37.2 (14.8, 63.5)	37.4 (14.4, 62.6)	37.4 (14.4, 63.5)
	Missing, n (%)	8 (3.0)	5 (1.9)	13 (2.5)
Month 6	Mean (SD)	38.4 (13.1)	38.9 (12.2)	38.6 (12.6)
	Median (min., max.)	37.1 (9.3, 64.4)	38.3 (14.3, 65.1)	37.9 (9.3, 65.1)
	Missing, n (%)	51 (19.2)	53 (20.3)	104 (19.8)
Month 12	Mean (SD)	39.3 (11.9)	38.9 (11.9)	39.1 (11.9)
	Median (min., max.)	38.7 (12.2, 65.6)	38.1 (12.4, 63.9)	38.3 (12.2, 65.6)
	Missing, n (%)	53 (20.0)	54 (20.7)	107 (20.3)
Physical component score
Baseline	Mean (SD)	43.7 (10.4)	42.2 (11.0)	43.0 (10.7)
	Median (min., max.)	45.0 (17.0, 64.1)	42.0 (19.4, 62.1)	43.9 (17.0, 64.1)
	Missing, n (%)	8 (3.0)	5 (1.9)	13 (2.5)
Month 6	Mean (SD)	45.6 (9.8)	42.9 (11.0)	44.3 (10.5)
	Median (min., max.)	47.0 (22.7, 61.7)	43.4 (17.4, 63.8)	45.5 (17.4, 63.8)
	Missing, n (%)	51 (19.2)	53 (20.3)	104 (19.8)
Month 12	Mean (SD)	44.3 (10.1)	42.4 (11.4)	43.3 (10.8)
	Median (min., max.)	45.6 (20.2, 63.7)	44.1 (16.3, 61.3)	45.3 (16.3, 63.7)
	Missing, n (%)	53 (20.0)	54 (20.7)	107 (20.3)

FIGURE 10.

Adjusted means for the SF-12 mental and physical health components by treatment group and time point. (a) Mental; and (b) physical. MCS, mental component score; PCS, physical component score.

These results were extracted from mixed-effects models in which participant was treated as a random effect and observations over time (6 and 12 months) were nested within participant. The effect of randomised treatment group was assessed while adjusting for time, group-by-time interaction, baseline SF-12 mental/physical component score and baseline smoking severity as fixed effects, and site as a random effect. Different covariance structures were applied to the model. An unstructured pattern that models all variances and covariances separately resulted in the lowest AIC for both models and so was used. Diagnostics of model fit revealed that the standardised residuals were normally distributed, and were uniform against fitted values; therefore, data transformation was not required.

Body mass index

Body mass index is summarised by randomised group and time point in Table 20. Adjusted BMI means and group differences are presented in Table 21 and displayed in Figure 11. The adjusted mean difference at 6 months was 0.16 (95% CI –0.54 to 0.85) and at 12 months was 0.25 (95% CI –0.62 to 1.13), both marginally favouring usual care, but neither statistically significantly (p = 0.65 and 0.57, respectively).

TABLE 20 - Body mass index by treatment group and time point

Max., maximum; min., minimum.

Time point	BMI	BSC (N = 265)	Usual care (N = 261)	Overall (N = 526)
Baseline	Mean (SD)	30.2 (7.1)	29.7 (6.3)	29.9 (6.7)
	Median (min., max.)	29.0 (16.6, 59.8)	29.4 (16.9, 48.4)	29.3 (16.6, 59.8)
	Missing, n (%)	2 (0.8)	3 (1.1)	5 (1.0)
Month 6	Mean (SD)	30.5 (7.0)	29.9 (6.0)	30.2 (6.5)
	Median (min., max.)	30.0 (14.8, 60.7)	29.1 (16.6, 48.5)	29.4 (14.8, 60.7)
	Missing, n (%)	49 (18.5)	56 (21.5)	105 (20.0)
Month 12	Mean (SD)	30.4 (7.2)	29.7 (6.7)	30.0 (6.9)
	Median (min., max.)	29.7 (0.0, 61.9)	29.3 (0.0, 45.8)	29.4 (0.0, 61.9)
	Missing, n (%)	57 (21.5)	60 (23.0)	117 (22.2)

TABLE 21 - Adjusted means and group differences for continuous secondary outcomes

MCS, mental component score; PCS, physical component score; SE, standard error.

Time point	BSC, mean (95% CI)	Usual care, mean (95% CI)	Difference (95% CI), SE	p-value
FTND
Month 6	5.25 (5.00 to 5.50)	5.43 (5.19 to 5.67)	–0.18 (–0.53 to 0.17), 0.18	0.32
Month 12	5.42 (5.14 to 5.70)	5.43 (5.16 to 5.69)	–0.01 (–0.39 to 0.38), 0.20	0.97
MTQ
Month 6	13.1 (12.6 to 13.5)	12.5 (12.0 to 12.9)	0.58 (–0.01 to 1.17), 0.30	0.06
Month 12	12.9 (12.4 to 13.4)	12.3 (11.8 to 12.7)	0.64 (0.04 to 1.24), 0.31	0.04
PHQ-9
Month 6	9.6 (8.7 to 10.4)	9.4 (8.5 to 10.2)	0.20 (–0.85 to 1.24), 0.53	0.72
Month 12	9.3 (8.4 to 10.1)	9.4 (8.5 to 10.2)	–0.12 (–1.18 to 0.94), 0.54	0.82
GAD-7
Month 6	7.0 (6.3 to 7.7)	7.4 (6.7 to 8.1)	–0.32 (–1.26 to 0.62), 0.48	0.50
Month 12	7.1 (6.4 to 7.8)	7.2 (6.5 to 7.9)	–0.10 (–1.05 to 0.86), 0.49	0.84
SF-12 MCS
Month 6	37.9 (36.2 to 39.5)	38.6 (36.9 to 40.3)	–0.73 (–2.82 to 1.36), 1.07	0.49
Month 12	38.6 (37.0 to 40.1)	39.0 (37.4 to 40.5)	–0.41 (–2.35 to 1.53), 0.99	0.68
SF-12 PCS
Month 6	45.2 (44.1 to 46.3)	43.5 (42.4 to 44.6)	1.75 (0.21 to 3.28), 0.78	0.03
Month 12	43.6 (42.4 to 44.8)	43.0 (41.8 to 44.2)	0.59 (–1.07 to 2.26), 0.85	0.48
BMI
Month 6	30.3 (29.8 to 30.8)	30.1 (29.6 to 30.6)	0.16 (–0.54 to 0.85), 0.35	0.65
Month 12	30.2 (29.5 to 30.8)	29.9 (29.3 to 30.5)	0.25 (–0.62 to 1.13), 0.45	0.57

FIGURE 11.

Adjusted means for BMI by treatment group and time point.

These results were extracted from a mixed-effects model in which the participant was treated as a random effect and observations over time (6 and 12 months) were nested within participant. The effect of randomised treatment group was assessed while adjusting for time, group-by-time interaction, baseline BMI and baseline smoking severity as fixed effects, and site as a random effect. Different covariance structures were applied to the model. An unstructured pattern that models all variances and covariances separately resulted in the lowest AIC and so was used in the final model. Diagnostics of model fit revealed a minor violation: the ‘body’ (centre) of the distribution of the standardised residuals was normal, but the distribution was short-tailed; however, the residuals were uniform against fitted values. Data transformations of the outcome (log, square root and square) did not substantially improve the model fit and so analyses were conducted on untransformed data.

Compliance with intervention

Withdrawals

Some of the information reported in this section is given in Peckham et al. 1 This contains information licensed under the Non-Commercial Government Licence v2.0.

There were four categories of participant withdrawal:

1. Full withdrawal – participant withdrawn from the trial with regard to completion of both postal questionnaires and collection of GP data.

2. Withdrawal from follow-up – participant withdrawn from the completion of postal questionnaires but agreed to the continuing collection of GP data.

3. Withdrawal from treatment – participant withdrawn from trial intervention treatment, but agreed to continuing completion of postal questionnaires and collection of GP data.

4. Too unwell to continue – participant was deemed too unwell by medical staff to complete any questionnaires. This generally occurred only when a participant has been hospitalised.

Overall, there were 93 (17.7%) participant withdrawals (all types) during the trial: 23 participants (8.8%) from the usual-care group and 70 participants (26.4%) from the BSC group.

A total of 47 participants (8.9%) withdrew fully from the trial, two participants (0.4%) withdrew from follow-up, two (0.4%) were too unwell to continue and 42 (15.8%) withdrew from the intervention (Table 25).

Adverse events

There were 365 adverse events among 174 participants during the course of the trial. Of these, 111 were classed as serious. More participants in the BSC group (42.1%) experienced one or more adverse events than those in the usual-care group (24.5%). All adverse events (n = 58) that were deemed to be definitely or probably related to the intervention or trial participation were all non-serious adverse events (Table 26). This includes three events, related to NRT side effects, which were experienced by usual-care participants as part of a routine care quit attempt.

Of the 111 serious adverse events reported, the majority were related to unplanned hospital admission due to a deterioration in the participant’s mental health condition (n = 98). The remaining serious adverse events were associated with unplanned hospital admissions associated with comorbidities (n = 6) or death (n = 7). The one event deemed to be possibly related was an inpatient hospitalisation due to infective chronic obstructive pulmonary disease.

Of the 254 non-serious adverse events reported, the majority of events were associated with participant comorbidities (n = 124). Sixty-six events were associated with NRT side effects and 57 with the participant’s mental ill health. Of the remaining events, five were associated with accident and emergency or emergency ambulance use, but no information was available with regard to the reasons for this and two were unable to be classified (smoking teabags, n = 1; media coverage of family member, n = 1). Of the 56 events deemed to be probably related, the majority were associated with NRT side effects (n = 50) and of those possibly related, the majority were associated with participant comorbidities (n = 14). The two events deemed to be definitely related were both associated with NRT side effects (nausea following Champix use, n = 1; itching and redness caused by NRT patch, n = 1).

Summary of findings

Attempts were made to contact a total of 1606 people for eligibility and consent to partake in the trial, of which 526 were randomised (a rate of 33%). A primary reason for ineligibility, besides being unable to contact the person and non-consent, was that the person did not smoke or did not smoke at least five tobacco cigarettes a day. Over two-thirds of the participants were recruited from CMHT direct referral (72%). Between October 2015 and December 2016, 265 participants were randomised to BSC and 261 participants were randomised to usual care. This was over our original sample size target of 400 participants. Almost 60% of the participants were male and the average age was 46 years (range 19–72 years). The most common mental health diagnosis was schizophrenia or other psychotic illness (65%), followed by bipolar disorder (22%) and schizoaffective disorder (12%). The mean number of cigarettes smoked per day at baseline was 25, and the most commonly reported reasons for wanting to quit smoking were that it is bad for your health (88%) and is expensive (72%).

Retention rates in the trial were high. At the primary time point of 12 months, 450 (94%) of the 480 participant questionnaires intended to be completed were returned (86% of 526 randomised participants), with most (98%) being completed face to face with the MH-SCP. In total, 442 participants were included in the primary analysis (i.e. provided a CO breath measurement and completed the self-reported smoking status question on the 12-month participant questionnaire). A total of 34 out of 223 participants (15%; 13% of 265 randomised participants) fulfilled the definition for CO-verified smoking cessation in the BSC group, and 22 out of 219 participants (10%; 8% of 261 randomised participants) in the usual-care group. Therefore, participants in the BSC group were more likely to be a CO-verified 1-week quitter at 12 months post randomisation than participants in the usual-care group, but this difference was not statistically significant (OR 1.6, 95% CI 0.9 to 2.8; p = 0.12). A larger difference was, however, observed at the 6-month time point and this was statistically significant (OR 2.4, 95% CI 1.2 to 4.7; p = 0.01), perhaps suggesting an early benefit of the intervention, which waned over time.

The trial was originally powered at 80% to detect a relative increase of 1.7 in quitting from 20% in the usual-care group to 34% in the BSC group, allowing for a 20% loss to follow-up at 12 months. Ultimately, 526 participants were randomised and 442 (84%) were included in the primary analysis, so the loss to follow-up rate was lower than 20%. However, the quit rate was half that expected in the usual-care group (10% as opposed to 20%); therefore, the power needed to detect a relative increase in quit rate of 1.7 (to 17% in the BSC group) was ultimately 58%. Although this was not statistically significant at 12 months, the difference (10% vs. 15%) may still be clinically meaningful.

The primary result was robust to sensitivity checks, both the pre-planned and post hoc analyses, and similar results were seen when considering only self-reported smoking cessation. There was no evidence of a difference between the two groups in secondary outcomes at any time point except for the MTQ, which statistically significantly favoured the BSC group at 12 months, with a borderline p-value of 0.06 at 6 months. The physical component of the SF-12 also statistically significantly favoured the BSC group at 6 months (but not at 12 months).

The proportion of participants reporting that they had ever used e-cigarettes remained reasonably stable throughout the trial, as did the proportion reporting use in the previous month, suggesting that e-cigarettes were not frequently used as a NRT for those attempting to quit in the trial.

The uptake of the BSC intervention was reasonable, with 88% of participants randomised to the BSC group attending at least one session (median 6, range 1–14 sessions). The majority of sessions took place in the participants’ homes.

A total of 111 serious adverse events were reported in the trial and more participants in the BSC group (42%) reported one or more adverse events than in the usual-care group (25%); however, none of the SAEs was deemed to be related to trial participation.

Costs

Wider societal costs

Other sources of smoking cessation advice

In addition to the smoking cessation services provided by the NHS and Personal Social Services, the participants reported their use of other sources of help. These included a smoking cessation helpline run by non-NHS organisations, the internet and books. These were reported by only a small group of participants. The most often used source of advice was the internet. In the 6 months post randomisation, 18 participants reported, in total, 82 times of internet use for advice on smoking cessation in the BSC group, and 22 participants reported a total of 61 times of use in the usual-care group. From 6 months to 12 months post randomisation, 21 participants in the usual-care group reported a total of 86 times of internet use and 16 participants in the BSC group reported 64 times of use. Among them, the highest number of uses appeared in the usual-care group where one participant reported 28 times of use in the 7–12 months post randomisation. It was followed by two participants who reported 25 times of use and 26 times of use in the BSC group in the 6 months post randomisation. The costs of internet use were not estimated because of the lack of details. In the 1–6 months post randomisation, books were used by 18 participants in the BSC group and 10 participants in the usual-care group. In the 7–12 months post randomisation, 10 participants in the BSC group and nine participants in the usual-care group used books. Only two participants in the BSC group and five in the usual-care group called another smoking cessation helpline during the 12-month period.

Participants’ out-of-pocket expenses in relation to smoking cessation

Participants were asked about their expenses for the purchase of NRT products, the purchase of e-cigarettes and travel to GP/SSS at the follow-ups.

There were 157 participants (81% of 195 responders) in the BSC group who reported not purchasing any NRT products and 152 participants (77% of 197 responders) in the usual-care group in the 12 months post randomisation. Among the 38 participants who reported any purchase in the BSC group, 27 purchased a single form of NRT product, eight purchased two forms of NRT products and three purchased three forms of NRT products. Among the 45 participants who reported any purchase in the usual-care group, 27 purchased a single form of NRT product, 13 purchased two forms of NRT products and three purchased three forms of NRT products.

Among all the NRT products asked about the most purchased one was the nicotine patch. There were 27 participants who reported purchasing patches in the usual-care group, compared with 14 in the BSC group (Figure 12). The next most purchased were the lozenge and mouth spray. There were 12 participants who reported purchase of a lozenge in either group and the purchase of mouth spray was almost the same (11 in the usual-care group and 12 in the BSC group). These were followed by nicotine gum, which was reported by nine participants in the usual-care group and 11 participants in the BSC group. There were five participants who reported purchasing an inhalator in the usual-care group and two in the BSC group. Only one participant reported purchasing nasal spray (in the BSC group). None of the participants reported purchasing microtabs.

FIGURE 12.

Number of participants who acquire NRT products on prescription, purchase or both, by group.

Comparing the sources of NRT products in those who had both self-report response and prescription records (132 in the usual-care group and 129 in the BSC group), the majority of the NRT products in the BSC group were acquired on prescription, whereas the usual-care group acquired their NRT products, if any, over the counter. A small number of participants in both groups acquired their products via both prescription and purchase.

Given that the majority of the responders reported no purchase at all during the 12 months post randomisation, the mean amount of purchase of individual forms of NRT products was low and the SD was comparatively large, indicating a skewed distribution of purchase (Table 32).

TABLE 32 - The mean amount of purchase of NRT products (SD) in months 1–6 and months 7–12, by group

NRT products	BSC, mean (SD)	Usual care, mean (SD)
Months 1–6	n = 206	n = 209
Patches (pack)	0.02 (0.14)	0.20 (0.90)
Gum (pack)	0.20 (1.79)	0.09 (0.91)
Lozenges (pack)	0.64 (5.97)	0.11 (0.80)
Microtabs (pack)	–	–
Inhaler (bottle)	–	0.08 (0.75)
Nasal spray (bottle)	0.01 (0.21)	–
Mouth spray (bottle)	0.07 (0.57)	0.05 (0.33)
Months 7–12	n = 214	n = 209
Patches (pack)	0.11 (0.58)	0.19 (1.01)
Gum (pack)	0.02 (0.17)	0.03 (0.22)
Lozenges (pack)	0.62 (5.28)	0.13 (1.16)
Microtabs (pack)	–	–
Inhaler (bottle)	0.01 (0.15)	0.00 (0.07)
Nasal spray (bottle)	–	–
Mouth spray (bottle)	0.35 (3.40)	0.03 (0.21)

Applying Table 2 to the quantities in Table 32, the mean expense of NRT products was £8 (SD £30) among 209 responders in the usual-care group and £13 (SD £83) among 206 responders in the BSC group in months 1–6 post randomisation. In months 7–12, the mean expense for NRT products was £5 (SD £22) among 209 responders in the usual-care group and £17 (SD £97) among the 214 responders in the BSC group.

There were 136 participants who reported no purchase of e-cigarettes in the usual-care group and 152 participants in the BSC group in months 1–6 post randomisation. In months 7–12, there were 143 participants in the usual-care group and 148 participants in the BSC group who reported no purchase of e-cigarettes. The mean expense of e-cigarettes was £21 (SD £47) among 208 responders in months 1–6 and £21 (SD £59) among 205 responders in months 7–12 in the usual-care group. In the BSC group, the mean expense on e-cigarettes was £17 (SD £64) among 219 responders in months 1–6 and £19 (SD £68) among 210 responders in months 7–12.

The majority of the responders reported walking as their way of travelling to see their GP or SSS, followed by driving, taking a bus or taking a taxi. Fewer than five responders reported taking trains to see their GP or SSS. In the usual-care group, 173 responders in months 1–6 and 171 in months 7–12 reported no expense on travel. In the BSC group, 187 in months 1–6 and 176 in months 7–12 reported no expense on travel. The mean expense on travel to GP/SSS was £6 (SD £30) among 203 responders in months 1–6 and £3 (SD £11) among 199 responders in months 7–12, in the usual-care group. In the BSC group, it was £3 (SD £12) among 213 responders in months 1–6 and £2 (SD £9) among 205 responders in months 7–12. However, this expense might have been underestimated because some of the participants did not take into account when they used cars and others did.

Health-related quality of life

The EQ-5D-5L was returned by 260 participants in the usual-care group at baseline and 265 participants in the BSC group. The mean VAS score was 54.2 (SD 20.5) in the usual-care group and 54.8 (SD 22.9) in the BSC group. Two participants in the usual-care group had missing items on the descriptive system of EQ-5D-5L and, therefore, utility could not be calculated. Among the rest of the participants, the mean utility was 0.640 (SD 0.304) in the usual-care group and 0.638 (SD 0.280) in the BSC group. At 6 months, 217 participants in the usual-care group and 224 participants in the BSC group returned the questionnaire. The mean VAS score was 55.6 (SD 21.2) in the usual-care group and 58.3 (SD 21.8) in the BSC group. The mean utility was 0.659 (SD 0.307) in the usual-care group and 0.682 (SD 0.269) in the BSC group. At 12 months, 213 participants in the usual-care group and 216 participants in the BSC group returned information. The mean VAS score was 55.9 (SD 20.9) in the usual-care group and 55.8 (SD 23.9) in the BSC group. Three participants in the usual-care group and two in the BSC group had missing items on the descriptive system of EQ-5D-5L. Among the remaining participants, the mean utility was 0.662 (SD 0.306) in the usual-care group and 0.688 (SD 0.275) in the BSC group.

Missing data

The percentage of missing data varies from variable to variable (Table 33). As expected, the lowest missing percentage occurs among the variables at baseline. Missing data level ranges between 20% and 30% on participants’ self-reported items at follow-ups, which is consistent with the overall follow-up rate of the trial. The highest percentage of missing values is that of the costs of pharmacotherapy prescriptions in the trial period, which was collected from practices directly. This percentage of missing values is 44%; the number of multiple imputations was therefore set to 45. The missing values in the continuous variables were imputed using the predictive mean matching method, with 10 nearest neighbours to draw from. Binary variables were imputed using the logit method.

There were five participants who died before the 6-month follow-up and another two died before the 12-month follow-up. All of them were unable to be followed up at both follow-up points. For those who died before the 6-month follow-up, the costs and expenses in the first 6 months of the trial were to be imputed as others, while the costs and expenses in months 7–12 were set to zero. For the two who died after the 6-month follow-up but before the 12-month follow-up, their costs and expenses were to be imputed as others. The utility and VAS from EQ-5D-5L were set to zero at 6 months and 12 months for those died before the 6-month follow-up, and those who died after 6 months were to have 6-month values imputed but 12-month values set to zero. Other measures including cigarettes per day, alcohol drinking and FTND score were to be imputed conditioned on the death status at that time point.

Analysis

The BSC intervention cost, consisted of a fixed training cost and delivery cost, was estimated at £418 [standard error (SE) £10] per participant. The cost of usual GP care during the trial period was £65 (SE £6) per participant in the usual-care group and £52 (SE £6) per participant in the BSC group. The BSC group received more pharmacotherapies for smoking cessation on prescription than the usual-care group. The mean cost of pharmacotherapies on prescription was £91 (SE £13) in the BSC group, three times the mean cost in the usual-care group [£29 (SE £6)]. The total smoking cessation costs during the trial period were therefore £93 (SE £9) per participant in the usual-care group and £561 (SE £19) per participant in the BSC group.

In the 6 months before randomisation, both groups had similar levels of cost of smoking cessation from usual GP care [£38 (SE £4) vs. £37 (SE £4)]. The mean costs of emergency and hospital services were higher in the BSC group [£2030 (SE £494), compared with £1507 (SE £406) per participant in the usual-care group]. The costs of primary and community services were the first cost drive in both groups, with £2614 (SE £260) per participant in the usual-care group and £2746 (SE £223) per participant in the BSC group. The total health resource use costs were estimated at £4160 (£512) per participant in the usual-care group and £4813 (£548) per participant in the BSC group in the 6 months before randomisation.

Throughout the trial period, primary and community services remained the first cost driver in both groups, with the mean costs never dropping below £2000 per participant, and the mean costs of emergency and hospital services in the same period never reaching this threshold. The mean costs of emergency and hospital services were £1513 (SE £492) per participant and costs of primary and community services were £2391 (SE £204) in the usual-care group in the first 6 months of the trial. In the same period, the mean costs of emergency and hospital services were £982 (SE £285) per participant and costs of primary and community services were £2597 (SE £218) in the BSC group. In the second 6-month period of the trial, the mean costs of emergency and hospital services were £1404 (SE £361) per participant in the usual-care group and costs of primary and community services were £2320 (SE £232). In the BSC group, the mean costs of the mean costs of emergency and hospital services were £1004 (SE £211) per participant and costs of primary and community services were £2504 (SE £245) in the second 6-month period of the trial. The mean cost of antipsychotics on prescription in the trial period was estimated at £768 (SE £81) per participant in the usual-care group and £799 (SE £84) in the BSC group. The total health resource use costs were therefore £8396 (£774) per participant in the usual-care group and £7886 (SE £594) in the BSC group.

In the usual-care group, the mean VAS score was 54.2 (SE 1.3) at baseline, 54.9 (SE 1.9) at 6 months and 55.3 (SE 1.4) at 12 months. In the BSC group, the mean VAS score was 54.8 (SE 1.4) at baseline, 57.3 (SE 1.5) at 6 months and 55.1 (SE 1.6) at 12 months. In the usual-care group, the mean utility value was 0.640 (SE 0.019) at baseline, 0.647 (SE 0.020) at 6 months and 0.654 (SE 0.020) at 12 months. In the BSC group, the mean utility value was 0.638 (SE 0.017) at baseline, 0.670 (SE 0.018) at 6 months and 0.678 (SE 0.019) at 12 months. The mean QALYs were estimated at 0.647 (SE 0.017) in the usual-care group and 0.664 (SE 0.015) in the BSC group.

Sensitivity analysis

Mapping incremental costs and incremental QALYs estimated from 5000 bootstrapped replicates onto the cost-effectiveness plane, Figure 13a shows the overall majority (92%) of dots that represented ICERs of the replicates, scattered on the right side of the y-axis (zero effect), indicating that the BSC is very likely to achieve higher QALYs. The distribution of these dots in relation to incremental costs is less clear, where 62% are scattered below zero and 38% above. This suggests that the BSC is likely to be cost-saving, but the uncertainty is higher. Taking into account the combination of both, 57% fall in the south-east quadrant, where the BSC group dominates the usual-care group (less costly but more effective). Figure 13b is the cost-effectiveness acceptability curve with WTP thresholds at £20,000/QALY and £30,000/QALY as reference line. For the probability of BSC being cost-effective, compared with usual care, although it starts high (> 60%) and rises with the increase of WTP, the increase in the probability shows a trend of becoming smaller. This leads to a gradual upwards curve and almost flat line close to 90%.

FIGURE 13.

Cost-effectiveness plane and cost-effectiveness acceptability curve of primary analysis results. (a) Cost-effectiveness plane; and (b) cost-effectiveness acceptability curve.

Complete-case analysis

Taking into account the outcome measures and baseline variables for adjustment, there were 80 participants in the usual-care group and 88 participants in the BSC group included in the complete-case analysis. The intervention cost was similar in the usual-care group between primary analysis (£93, SE £9) and complete-case analysis (£92, SD £128) while in the BSC group it was slightly higher in complete-case analysis (£593, SD £338) than in primary analysis (£561, SE £19). In the usual-care group, the mean costs of health services were £4822 (SD £7881) in the 6 months before randomisation, £3440 (SD £3745) in months 1–6 of the trial and £4293 (SD £8490) in months 7–12 of the trial, in complete-case analysis. In the BSC group, the mean costs of health services were £4594 (SD £7449) in the 6 months before randomisation, £3599 (SD £4173) in months 1–6 of the trial and £3555 (SD £5078) in months 7–12 of the trial, in complete-case analysis. The difference between primary analysis and complete-case analysis was more prominent in the usual-care group than in the BSC group.

For the BSC group, the mean costs in complete-case analysis did not vary from those in primary analysis by much. For the usual-care group, however, the mean costs became higher than those in primary analysis in the 6 months before randomisation and months 7–12 of the trial, and were lower than in primary analysis in months 1–6 of the trial. This change led to a reverse of the relative position when comparing between groups and the reduction of the difference in the 6 months before randomisation and months 1–6 of the trial. In the second 6-month period of the trial, although the mean costs of health services were still higher in the usual-care group, the difference between the mean costs in the two groups was widened.

For EQ-5D-5L utility value, the pattern of change was more consistent. In complete-case analysis, the mean utility value was 0.595 (SD 0.319) at baseline, 0.612 (SD 0.336) at 6 months and 0.641 (SD 0.303) at 12 months in the usual-care group, while it was 0.661 (SD 0.257) at baseline, 0.686 (SD 0.254) at 6 months and 0.684 (SD 0.262) at 12 months in the BSC group. Although in the usual-care group the mean utility values were always lower in complete-case analysis than in primary analysis, the BSC group was the opposite.

The mean total cost was £911 (95% CI –£2769 to £2631) higher in the BSC group than in the usual-care group, after adjusting for costs of health resource use in the 6 months before randomisation, age, gender, pre-existing medical conditions and duration since first diagnosis of SMI (Table 35). The mean QALYs were 0.008 (95% CI –0.030 to 0.074) higher in the BSC group than in the usual-care group, after adjusting for utility value at baseline, age, gender, pre-existing medical conditions and duration since first diagnosis of SMI. The ICER ratio was therefore £113,875 (uncertainty: Figure 14) per QALY gained.

TABLE 35 - Results of complete-case analysis

Adjusted for health resource use in the 6 months before randomisation, age, gender, pre-existing medical conditions, duration since first diagnosis of SMI, with study centre as random effect.

Adjusted for EQ-5D-5L utility value at baseline, age, gender, pre-existing medical conditions, duration since first diagnosis of SMI, with study centre as random effect.

Results	BSC (n = 88)	Usual care (n = 80)
Costs
Smoking cessation, mean (SD)	£593 (£338)	£92 (£128)
Health resource use, mean (SD)	£7741 (£8649)	£8437 (£11,386)
Total, mean (SD)	£8434 (£8642)	£8530 (£11,405)
Adjusted difference,a mean (95% CI)	£911 (–£2768 to £2631)
Quality of life
QALYs, mean (SD)	0.679 (0.219)	0.615 (0.283)
Adjusted difference,b mean (95% CI)	0.008 (–0.030 to 0.074)
Incremental cost-effectiveness ratio
ICER, mean	£113,875 (uncertainty: Figure 14)

FIGURE 14.

Cost-effectiveness plane and cost-effectiveness acceptability curve of complete-case analysis results. (a) Cost-effectiveness plane; and (b) cost-effectiveness acceptability curve.

Compared with the primary analysis, Figure 14a shows the dots scattered more widely on the cost-effectiveness plane, indicating a higher level of uncertainty. The BSC is still likely to achieve higher QALYs as 79% of the dots fall to the right side of y-axis. The difference in costs, however, almost disappears as about half (49%) fall below zero. Only 39% of the dots fall in the south-east quadrant, suggesting a less costly but more effective intervention. The cost-effectiveness acceptability curve demonstrates that the probability of the intervention being cost-effective was 61–65% between £20,000 and £30,000 per QALY (see Figure 14b). Similar to the primary analysis, the increase of the probability becomes smaller the higher the WTP gets, and flattens close to 80%.

Discussion

This study collected extensive data to provide a comprehensive picture of costs to the NHS and Personal Social Services in this population. The mean intervention cost per quitter in the BSC group was around four times the mean intervention cost per quitter in the usual-care group (£3828 vs. £906). Although the cost per quitter of SSS in the general population was £690 per quitter for CO-validated 4-week quit in 2017/18 in England where data are available,72 this could not be compared directly. Our study offered a more flexible timeline to suit the needs of people with SMI and, therefore, did not measure 4-week smoking status. The participants in the usual-care group showed a more active engagement with usual smoking cessation services, especially in months 1–6 of the trial. This is understandable as the participants in the BSC group would have mostly been engaging with the intervention during that period while the help-seeking behaviour in the usual-care group might have been triggered by participation in the trial. Both groups became less engaged with usual-care services in months 7–12.

The use of various health-care and social services was highly intensive. The costs of primary and community services were the largest component of the overall costs. These costs were relatively stable from before the trial to the end of the trial and did not change significantly during the trial, indicating this population’s constant demand for community support. The lower total costs of the BSC group were mainly the results of the consistently lower costs of emergency and hospital services during the trial and higher costs in the 6 months before the trial, while the costs of emergency and hospital services remained on the same level in the usual-care group. This difference in costs of emergency and hospital services was probably related to the higher frequency and longer stay of hospital admission in the usual-care group than in the BSC group. However, it is unclear if this difference is related to the intervention or smoking status, or if it happened by chance. As the use of services was collected through self-report, the possibility of recall bias should be kept in mind when interpreting the results.

The data on prescriptions of antipsychotics and pharmacotherapies for smoking cessation were extracted from participants’ medical records. The information collected through this approach was more reliable and accurate than participants’ self-report, especially when there were multiple medicines involved. However, the information on prescriptions was completely missing for some participants because of the closure or lack of capacity of their general practices. Although attempts were made to ensure a uniform extraction of data, there were still discrepancies among data returns. As a result, clinic supervision and therefore staff cost necessitated by some antipsychotics [e.g. clozapine (Clozaril^®, Sandoz Pharmaceuticals)] were not included in the analysis as individual visits to the clinic could not be identified for all participants. For the same reason, the prescription charge was also not included in the participants’ expenses as individual prescription items were not always identifiable. The cost of pharmacotherapies for smoking cessation might be overestimated. As we extracted data from GPs’ prescription records, it should be noted that not all people redeem their prescriptions.

With the development and increased accessibility of the internet, people become rapidly used to seeking information online. The level and costs of smoking cessation help available online vary. The publicly available information provided by the NHS or charities costs only individuals’ internet charge, whereas the private organisations might offer more intensive online treatment for a fee. It is not clear if this particular population has a preference for internet use or not and to what extent they rely on the internet for smoking cessation help. A brief browse of free information might be negligible in terms of costs but more intensive help could add to participants’ financial burden.

Both groups had an upwards trend of EQ-5D-5L scores, although the BSC group’s increase was more drastic. This resulted in the positive QALY gains of the BSC intervention. This might be related to the lower frequency and shorter stay of hospital admission in the BSC group. Both appeared to reflect a better general health of the participants in this group. However, within such a short period, participants’ smoking status, successfully quit or not, is unlikely to have contributed to their better general health.

A disadvantage of collecting comprehensive service use data is that data are more likely to be missing because of the length of the questionnaire. However, owing to a high follow-up rate, the missing data level was not unacceptably high. When comparing primary analysis (multiple imputation of missing data) and complete-case analysis (remove all incomplete observations), it appears that missing data on costs have a larger impact on costs in the usual-care group than in the BSC group. The impact of missing data on EQ-5D-5L is less straightforward. In the usual-care group, it appears that the participants who managed better dropped out, whereas in the BSC group the participants who did not do well were the source of missing data.

The biggest impact of smoking on the individual’s private costs came from purchase of cigarettes itself. On average, the participants in the in the trial were estimated to spend > £3000 per year on cigarettes and there did not appear to be any difference between groups. Because the amount of money spent on cigarettes was estimated based on the data collected at three time points, the change in smoking patterns in between was not captured. The assumption of unchanging number of cigarettes per day throughout the estimated period might not hold. Another thing to keep in mind is that we estimated the expenses on smoking by converting all tobacco to cigarettes and costing it based on market price of manufactured cigarettes. This might lead to an overestimate as there are cheaper options available for purchasing tobacco. Although there were some participants purchasing e-cigarettes during the trial, the expenses did not appear significant in comparison to those of cigarettes. Previous evidence has recommended that people with SMI could reduce their antipsychotics intake if they stop smoking. Although the results showed that the participants who smoked throughout the trial had a significant increase in their costs of antipsychotics, the current study did not find evidence to support the reverse situation. One of the reasons might be that there were insufficient numbers of quitters in the trial to draw meaningful conclusions. Moreover, the quit status was measured by asking the question ‘have you smoked in the last week’, which is not a measure of abstinence. Clinical guidelines suggest that if a person makes a quit attempt their antipsychotic medication should be closely monitored as smoking affects the way these drugs are metabolised and the person may need to reduce their antipsychotic medication and if they subsequently relapse back to smoking their dose of antipsychotic medication may need to be increased. Finally, this comparison was undertaken based on the participants whose costs of antipsychotics and smoking status were both available. There is a possibility of selection bias.

The primary analysis showed a higher than 50% probability of the BSC intervention being cost-effective, compared with usual care, at a WTP threshold of £0 per QALY gained. This was largely attributable to the lower level of health-care services costs offsetting the intervention costs in the BSC group and a QALY gain with relatively high level of certainty. Although the likely cost-saving effect in the primary analysis disappeared in the complete-case analysis, the offset effect remained. This suggested that in the complete-case analysis, the mean health-care services costs in the BSC group remained lower than those in the usual-care group. Given that the primary outcome of CO-validated quit at 12 months post randomisation did not show a statistical significance, it could raise concerns over the more favourable results of the economic evaluation. As a perfect association between smoking status and health services and quality of life is not expected in 1 year, it is possible that the analysis shows a better outcome in terms of costs and QALYs. However, unless there were some unknown factors in participants’ life affected by the intervention that could be identified later, the result of this study might be hard to generalise.

Conclusion

The BSC intervention for people with SMI is likely (57%) to dominate usual GP care, from a NHS and Personal Social Services perspective. Although neither the difference in costs nor in QALYs was statistically significant, the probability of cost-effectiveness could reach 80% at a threshold of £30,000 based on the primary analysis and 65% based on the complete-case analysis. However, it remains uncertain whether or not this was the result of smoking cessation, as the smoking status at 12 months did not show a significant difference between groups. Furthermore, the impact of smoking cessation on health and wider health services use is unlikely to be observed in a short period of 12 months. A longer-term follow-up would be ideal to observe the sustainability of quit and impact on health.

Main findings

The main finding of the SCIMITAR+ trial is that smoking cessation can be achieved among people with SMI, and that, in people who express a desire to quit smoking, the use of a BSC intervention increased the chances of sustained quitting as estimated by a gold-standard biologically verified outcome measure (exhaled CO). 73 The observed odds of successful quitting at 12 months was higher among those who received BSC but this was not statistically significant. However, the observed odds of successful quitting at 6 months were higher among those who received BSC and this was statistically significant. The SCIMITAR+ trial used quitting as measured at 12 months post randomisation as its primary end point, and although there was a trend in favour of the BSC intervention, the difference in quit rates was no longer statistically significant at 1 year. This finding is in line with research in the general population, which shows that long-term quit rates are difficult to achieve, and this remains a challenge for the wider evidential basis of treatment for nicotine dependence in any population. 74 The results of the effect of the BSC intervention were seen in secondary outcomes, and we found an improvement in short-term physical health (as measured by the SF-12) and some evidence of reduced numbers of cigarettes smoked per day and increased motivation to quit. We also found that there were no between-group differences on measures of mental health, including depression and anxiety, supporting the notion that offering a smoking cessation intervention is not detrimental to mental health.

To our knowledge, this is the first large-scale UK RCT of a bespoke intervention designed for people with SMI. Trials to date have been small scale and with short periods of follow-up, and have focused on pharmacological treatments, with limited consideration of behavioural approaches. 40 The SCIMITAR+ trial adapted and enhanced an evidence-supported smoking cessation strategy that has been developed and forms the mainstay of successful UK SSSs. This structured intervention was delivered by a mental health professional and a ‘cut down to quit’ approach was also offered. The results of the SCIMITAR+ trial, alongside trials35 and systematic review evidence34,39,40 of the safety and effectiveness of pharmacological treatments for nicotine dependence in people with SMI, represents accumulating evidence of the effectiveness of smoking cessation interventions for people with severe mental ill health. Across both arms of the trial, people smoked a higher number of cigarettes per day and had a higher mean number of years of smoking history than those in the general population (25.5 per day compared with 11.3 per day in the general population). 75 This adds to the justification for the need to intervene and develop new and novel interventions for this population.

Limitations of the SCIMITAR+ trial

There were limitations to the SCIMITAR+ trial. First, we lost 16% of participants to follow-up, but note that this was better than we achieved in our pilot trial43 and that there was non-differential loss to follow-up. Second, a lower 12-month cessation rate was observed in the usual-care group than was hypothesised in the sample size calculation (10% vs. 20%), and the actual percentage increase was lower (5 vs. 14 percentage points). Therefore, although the trial recruited more participants than originally planned and the loss to follow-up rate was lower (16% vs. 20% hypothesised), the trial was ultimately underpowered to detect a difference in quit rates from 10% to 15%. Third, we noted difficulties in ensuring that some participants received pharmacological treatment, as are detailed previously. Given that smoking is the single most modifiable risk factor for early mortality, we recommend that local services have in place robust mechanisms to remove barriers to the provision of medication both when implementing the results of the SCIMITAR+ trial and to support smoking cessation more generally. Last, we were not able to ensure that all follow-up was blind to treatment allocation. However, we also put in place a biochemically verified end point that is less susceptible to observer biases.

Given the trial could not conclude that the intervention is effective in terms of smoking cessation at 12 months, it is uncertain if the result of the economic evaluation was associated with cessation itself. It is possible that the intervention triggered some changes in aspects other than smoking behaviour, which had a positive impact on health, or the differences in health-care costs occurred by chance. In addition, we lacked the capacity to conduct a longer follow-up and a well-correlated relationship between smoking status and health is not expected in the short term.

Conclusions

In the face of significant health inequalities for people with SMI, smoking is the most important modifiable risk factor for poor health and reduced life expectancy. 78 At 12 months we did not find a statistically significant difference between the intervention and usual care; however, we did find that there was an increased quit rate in the intervention at 6 months. Quitting smoking did not appear to have had any impact on mental health and there were no statistically significant differences between PHQ-9 scores in the intervention and control group at 6 or 12 months. Those in the intervention group had a slightly higher number of quit attempts than those in the usual-care arm.

In this trial, we have shown that people with SMI more readily engage with a bespoke intervention that results in increased 6-month quit rates; however, this difference was not sustained at 12 months. Further research is needed to establish how long-term quitting can be supported.

Trial Steering Committee members

Professor Ann McNeil (Independent Chairperson), Professor of Tobacco Addiction, King’s College London, London, UK.

Dr Andrew Elmslie, Consultant Psychiatrist, Tees, Esk and Wear Valleys NHS Foundation Trust, York, UK.

Dr David Shiers, Retired GP, North Staffordshire, UK.

Mr David Waldram (Service User Representative).

Data Monitoring and Ethics Committee members

Professor Mike Crawford (Chairperson), Professor of Mental Health Research, Imperial College London, London, UK.

Dr Elena Ratschen, Lecturer in Epidemiology/Tobacco Control, Faculty of Medicine and Health Sciences, University of Nottingham, Nottingham, UK.

Professor Richard Emsley, Professor of Medical Statistics and Trials Methodology, King’s College London, London, UK.

Hazel Cheeseman, Director of Policy, Action on Smoking and Health, London, UK.

Contributions of authors

Emily Peckham and Simon Gilbody wrote the original protocol.

Catherine Arundel, Della Bailey, Suzanne Crosland, Caroline Fairhurst, Paul Heron, Catherine Hewitt, Jinshuo Li, Steve Parrott, Tim Bradshaw, Michelle Horspool, Elizabeth Hughes, Tom Hughes, Suzy Ker, Moira Leahy, David Osborn, Joseph Reilly, Diane Brennan and Simon Gilbody refined the protocol.

Catherine Arundel, Suzanne Crosland, Tayla McCloud, Paul Heron and Thomas Steare were trial co-ordinators for the study.

Tayla McCloud, Thomas Steare, Emma Ballantyne, Polly Bidwell, Susan Bonner, Diane Brennan, Tracy Callen, Alex Carey, Charlotte Colbeck, Debbie Coton, Emma Donaldson, Kimberley Evans, Hannah Herlihy, Wajid Khan, Lizwi Nyathi, Elizabeth Nyamadzawo, Helen Oldknow, Jamie Rea, Crystal-Bella Romain-Hooper, Kaye Smith, Alison Stribling and Carinna Vickers recruited participants to the study, provided feedback on the recruitment methods and helped refine the study procedures.

Peter Phiri and Shanaya Rathod helped refine the study procedures.

Emily Peckham, Catherine Hewitt, Steve Parrott, Tim Bradshaw, Elizabeth Hughes, Tom Hughes, Suzy Ker, Moira Leahy, David Osborn, Joseph Reilly and Simon Gilbody were co-applicants on the Health Technology Assessment application.

Emily Peckham was the trial manger.

Della Bailey supervised the delivery of the intervention.

Caroline Fairhurst conducted the clinical analysis.

Catherine Hewitt oversaw the conduct of the analysis.

Jinshuo Li designed and undertook the economic analysis in conjunction with Steve Parrott.

Simon Gilbody was the chief investigator and oversaw the study.

Emily Peckham, Catherine Arundel, Della Bailey, Suzanne Crosland, Caroline Fairhurst, Paul Heron, Jinshuo Li and Simon Gilbody were the writing team.

All authors read and reviewed the transcript.

Publication

Gilbody S, Peckham E, Bailey D, Arundel C, Heron P, Crosland S, et al. Smoking cessation for people with severe mental illness (SCIMITAR+): a pragmatic randomised controlled trial. Lancet Psychiatry 2019;6:379–90.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to available anonymised data may be granted following this.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.