A pilot randomised trial to assess the methods and procedures for evaluating the clinical effectiveness and cost-effectiveness of Exercise Assisted Reduction then Stop (EARS) among disadvantaged smokers

Cognitive variables

Participants were asked at baseline, 8 weeks and 16 weeks how confident they were in their ability to (1) do at least 30 minutes of moderate-intensity PA on most days of the week over the next 6 months and (2) walk continuously for 15 minutes at a brisk pace, using a 7-point response scale from ‘not at all confident’ (1) to ‘extremely’ (7). Stage of readiness to use PA to control smoking was assessed by participants ticking one from five options as follows: (1) I do not use physical activity as a way of controlling my cigarette smoking and I do not intend to start; (2) I do not use physical activity as a way of controlling my cigarettes smoking but I’m thinking of starting; (3) I use physical activity once in a while as a way of controlling my cigarette smoking, but not regularly; (4) I use physical activity regularly as a way of controlling my cigarette smoking, but only started within the past 6 months; (5) I use physical activity regularly as a way of controlling my cigarette smoking and have been doing so for longer than 6 months. Responses reflected being in the pre-contemplation, contemplation, planning, action, or maintenance stage, respectively. All the above questions have been previously used. 67

To assess beliefs about quitting smoking the following questions were asked using a 7-point response scale from ‘not at all’ (1) to ‘extremely’ (7): How important is it for you to stop smoking permanently and completely in the next six months?; how confident are you that you can stop smoking permanently and completely in the next six months?; an important person in your life thinks you should quit smoking; how confident are you that over the next week (except at baseline when we referred to ‘over the next 4 weeks’) you will smoke only half the number of cigarettes you smoked at the time of entry into the study?

Behavioural measures

Physical activity data were collected using a tri-axial GT3X accelerometer (Actigraph, Pensacola, FL, USA). Data were recorded using a 1-second epoch, over a 7-day period. Self-reported PA was collected using the 7-day PA recall questionnaire. 79 Minutes of moderate to vigorous PA (MVPA) and average nightly sleep were recorded, and minutes of light activity were derived. Minutes of MVPA and energy expenditure were calculated. Smoking status was recorded by the question (1) How many cigarettes, cigars, or pipes, have you usually smoked each day over the past week (recorded in numbers smoked or grams/ounces of loose tobacco)? Those who reported no smoking in the past week (with expired air CO confirmation of < 10 p.p.m.) at 16 weeks were defined as not smoking at that point and we used this to report 16-week point prevalence abstinence. Alcohol consumption was assessed to monitor changes in another lifestyle behaviour often associated with smoking, using two questions taken from the World Health Organization’s (WHO) validated AUDIT alcohol screening tool:80 (1) How often do you have a drink containing alcohol? (0, ≤ 1 per month; 2–4 times per month; 2–3 times per week; ≥ 4 times per week); (2) How many drinks containing alcohol do you have on a typical day when you are drinking? (1–2; 3–4; 5–6; 7–9; ≥ 10); and the addition of the third question: how many drinks containing alcohol have you had in the past week? (1–2; 3–4; 5–6; 7–9; ≥ 10).

Emotional/affective measures

Strength of urge to smoke and time spent with those urges over the past week were assessed using two items (1–6 point response scale)81 and withdrawal symptoms assessed using an adapted 9-item Mood and Physical Symptoms Scale (MPSS). 82 Quality of life is assessed using the European Quality of Life-5 Dimensions (EQ-5D). 83 Smoking dependence was assessed using the Fagerström Test for Nicotine Dependence (FTND). 84,85 Subjective stress was assessed using the 4-item Perceived Stress Scale (PSS),86 to monitor any changes associated with smoking reduction or increasing PA. Satisfaction from smoking was assessed using the 10-item modified Cigarette Evaluation Questionnaire (mCEQ),87 as we expected that the pleasure derived from smoking may reduce with increased PA participation.

All questionnaire data were scanned and manually checked using Teleform Scanning software (Cardiff Software, Digital Vision, Highland Park, IL, USA). A second researcher double-checked 10% of the entered data against the raw data and if more than a 1% error was found, all questionnaire data would have been manually rechecked. The data were stored on a secure institutional server.

Primary outcome

The original intention was to encourage all participants who wished to quit smoking to agree to be referred to the local NHS SSS. Support for those in the intervention arm from the HT was also planned. We allowed participants to begin any quit attempt up until week 12 to ensure that we had a further period of 4 weeks (up until week 16) to have an expired air CO verified 4-week period of abstinence. Participants in the control group, at baseline and subsequent assessments, were also reminded that if they did plan to quit they should notify the HT/researcher and that the SSS was available to support them. These planned procedures were to ensure some control over capturing quit attempts and post-quit monitoring. In fact, we identified only two participants who contacted and used the support of the NHS SSS. Also, some participants wished to quit without further support from the HT. This created a challenge for collecting data on our flexible primary outcome, namely expired air CO continuous confirmed abstinence (captured on a weekly basis).

Thus, data we report for our primary outcome (i.e. number achieving 4 weeks’ expired air CO-confirmed continuous abstinence) are not as rigorous as we had planned. We had expected all participants to attend the SSS and be monitored weekly. However, those reported to have quit fitted the following criteria: (1) we were aware of their planned quit date and could therefore prospectively follow their progress; and (2) we recorded an expired air CO value of < 10 p.p.m. between 4 and 8 weeks after the quit date and participants self-reported that they had not smoked more than five cigarettes since quitting.

Point prevalence data at week 16 confirmed the data reported for 4-week continuous abstinence, with the exception of two intervention participants who had lapsed in week 15.

The difficulty of capturing data for our primary outcome was also due to the nature of the sample: scheduled assessments were routinely missed and rearranged (for a wide range of reasons). We make recommendations in the final chapter to address this issue, but this pilot study has provided valuable information on how to work with disadvantaged smokers and capture a flexible outcome.

Although participants were excluded from the study if they wanted to quit in the coming month at the time of entry into the study, one participant decided to quit only 1 week after starting the intervention. On average, it was 7 weeks before those participants who decided to quit began their attempt and it is not clear if the participants’ awareness that the intervention contact was due to finish after 8 weeks influenced this. Some participants needed longer than 8 weeks but, overall, a target to reduce (and, if desired, to quit) over 8 weeks was appropriate. The client-centred approach of the intervention certainly aimed to allow participants to decide if and when they wished to quit.

Other smoking-related measures

We aimed to assess the proportion in each arm of the trial who reduced the number of cigarettes smoked by at least 50% at 8 and 16 weeks. In our sample, approximately 60% smoked only, or predominantly, self-rolled cigarettes, with no one smoking pipes, and approximately 35% smoking only, or predominantly, factory-made cigarettes. The remaining 5% smoked a roughly equal combination of self-rolled and factory-made cigarettes. Given the intention to assess this outcome we decided to nevertheless create a common variable, namely number of cigarettes smoked at each assessment. We decided to base our conversion from grams of tobacco to number of cigarettes based on Laugeson89 who reported that 0.45 g was equivalent to one factory-made cigarette, following a rigorous laboratory assessment of number of puffs and nicotine inhalation. Also, for those reporting the number of factory-made cigarettes smoked, there was inevitably some random measurement error. Participants were encouraged to use a range of recall prompts to indicate how many cigarettes were typically or had recently been smoked (see detailed description of measures in Chapter 2), including number of packs consumed each day, number of cigarettes smoked in the morning, afternoon and evening, or in specific situations. The precise number of cigarettes smoked in the present study therefore has measurement error but data collection and conversion procedures were consistent across time and treatment arms.

Physical activity measures

We used the self-reported 7-day recall of PA to capture the type, intensity, frequency and volume, which has been widely used in trials and has been shown to be sensitive to changes as a result of behaviour change interventions. This measure tended to show over-reporting of MVPA when compared with accelerometer data. This was particularly true among the most active, perhaps because the most active tended to be those in manual labour who reported doing up to 10 hours of MVPA per day, whereas in reality this probably included some periods of inactivity. The value of using the self-report measure is that we were able to capture data from 60 participants at weeks 8 and 16. In contrast, just 44% and 39% participants provided a minimum of 3 days of accelerometer data during weeks 8 and 16, respectively. These accelerometer compliance figures are comparable with those from previous research with adult populations.

After recruiting the first 60 participants, we realised that a number of accelerometers had not been returned, at a cost of c. £250 each. In an attempt to avoid this, and to maximise compliance with wearing it, we initially provided a payment of £10 in exchange for returning the device after each complete week of wearing it: a total of £30 for wearing it for a period of 7 days at baseline, week 8 and week 16. This was then increased to £30 after each week of wearing. Overall, 13 were not returned, and all of the missing accelerometers were administered before the increase in payment for its return. It seems reasonable to assume that the increase in payment did impact on the successful returning of the accelerometers. The logistics of reimbursing participants was also complex and challenging. Some participants did not have bank accounts or preferred cash payments, and the research team had to set up a cash float. Eventually we worked out systems with the university for processing claims (in bulk) and reimbursing expenses from the float.

In any future study the use of 24-hour wrist-worn GENEActiv (Activinsights, Cambridgeshire, UK) accelerometers (c. £120 each) is recommended to increase compliance and minimise financial loss.

Both self-report and accelerometer data suggested that both groups were less active at week 16 than at baseline. Reactivity in trials is common, with participants tending to do more PA, possibly due to a heightened awareness of researchers’ interest in this behaviour. These findings must be interpreted with caution given that alternative explanations such as withdrawal of more active participants may have influenced the results.

Design

The intended intervention processes for the EARS intervention are described in Chapter 1. These were used as a basis for generating items for a checklist to assess intervention delivery and fidelity. Following a brief scoring-standardisation procedure, the checklist was applied to a purposive selection of consultation recordings and descriptive analyses were used to summarise the data.

A more detailed qualitative analysis, linking data from all of the above data sources (individual participant feedback with feedback from the HTs and the consultation recordings), will be conducted to examine the relationships between intervention delivery and smoking reduction and to refine the process model. However, this work is outwith the remit of the Health Technology Assessment (HTA)-funded research and so is not reported here.

Sampling frame

All consultation sessions were audio recorded subject to informed consent. Consent for this was taken on the main study consent form and this was checked verbally prior to starting the first consultation. A sample of four participants for each of the three HTs (12 participants in total) was selected to provide examples from early, late and in the middle of the study period (to smooth out any HT practice effects). For each client, three (out of a possible eight) consultations were selected for coding to provide examples of intervention techniques from early-stage motivation through to later-stage progress reviewing/relapse prevention.

Measures and procedure

To assess intervention fidelity (and at the same time quantify delivery in terms of predefined manualised elements), we used the Dreyfus system for assessing skill acquisition (Figure 5)95 to score recorded consultations with respect to a HT’s skill in delivering each of the 12 intervention processes (see Table 31). A scoring checklist and instructions were developed and these are provided in Appendix 5c. The checklist was applied initially by three researchers with expertise in behaviour change (AT, TT, CGVS) to a sample of six consultations from two participants. Scores were compared and reasons for any discrepancies were discussed to produce a consensus about how to apply the scoring system.

FIGURE 5.

The Dreyfus model of skill acquisition. 95

Two researchers (TT, CGVS) then each scored consultation data from two participants for each of the three HTs (using three consultations per participant) to produce an overall intervention fidelity rating for each item and for each HT (see Table 31). This was done by listening to the set of (three) recorded consultations for each participant, reading the transcripts of the same consultations and then rating the fidelity for each item on the checklist. Because of limitations in time and resources, we did not conduct formal inter-rater reliability checks; this would have required both researchers to rate fidelity for around 20–30 participants each. However, we did split the coding for each HT between the two researchers, so that each researcher coded two participants for each HT. The average score for the HT is therefore the average of the scores given by two coders.

Analysis and interpretation

Descriptive data were extracted and reported to highlight areas of good or bad practice in delivering the intended intervention processes. Owing to the clear descriptions associated with each score (see checklist scoring instructions in Appendix 5c) and the steps taken to establish a consensus between coders on the approach to scoring, interpretation of scores is relatively straightforward. Scores of 0 or 1 represented poor delivery (or no delivery) of the intended process. A score of 3 or more was considered to represent reasonable quality of intervention delivery. Scores of 5 or 6 represented very high (expert-level) quality, which we were not expecting to see very often with our trainers delivering this novel intervention for the first time. It was accepted that for item 9 (seeking to identify and reinforce shifts in identity), the opportunities to deliver this process would be scarce and so a lower score (1.5 or more) was considered acceptable for this item. Item 12 (referral to smoking-cessation services) was scored as either 0 or 1 (yes or no) and so is reported separately.

Examples of good intervention practice were highlighted as the coders went through the consultation transcripts and these will be collated and used as examples in future training for the EARS intervention (either in their current form as audio recordings and transcripts, or by conversion into video example using actors to play the roles of practitioner and participant).

Recruitment and sampling

All trial participants had consented at baseline to being approached by an independent qualitative researcher to capture their experiences associated with the study and with the intervention, for those in that arm of the study.

During the delivery of the intervention, the research team regularly discussed the progress of individual participant progress and the nature of engagement with the intervention. We were particularly keen to identify intervention participants who appeared to have benefitted from the intervention and be examples of good practice, possibly for future use in training. As many participants as possible who engaged with the intervention were interviewed by TT.

Control participants were selected at random. Participants sampled were contacted by telephone and a convenient time to conduct the interview by telephone was arranged. No participants who were contacted declined to be interviewed. Verbal consent was obtained for the interviewed to be digitally recorded. All interviews were digitally recorded and transcribed verbatim for further analysis.

Interviews

Participants were interviewed within 16 weeks of completing the study. The interviews followed the guide shown in Figure 6 for control and intervention participants, respectively.

FIGURE 6.

Generic and trial arm-specific interview guide.

Data analysis

Interview transcripts were analysed used the qualitative software package NVivo (version 9.2, QSR International, Southport, UK). The data were organised using a basic thematic analysis to provide a simple descriptive-level overview of the participants’ views and experiences. In-depth qualitative analysis procedures were not used here. However, these data will be analysed in more depth alongside the participant interviews and consultation recordings, using framework analysis6 to generate an integrated analysis of processes of behaviour change in the EARS intervention. This in-depth analysis will be reported as part of the PhD of one of the researchers (TT). The participant characteristics of those interviewed from both arms of the trial are shown in Table 32.

Clarity of invitation to take part in the trial

Targeting a population with expected low levels of literacy meant keeping the invitation as simple as possible without compromising the individuals’ understanding of what was being offered to them. Overall, the responses revealed a clear and concise understanding of what was being offered. For example:

So how clear was the information that you had?

Very. Very clear.

So how much did the information that you had read, how much did it match what you actually received in terms of the support and things like that?

Oh erm exactly . . . .

Female, aged 60–65 years, unemployed, moderate smoker, intervention

There were a couple of examples of where a participant reported misunderstanding, or elements of the study that surprised them. This included explanations of the requirements to wear an accelerometer, and confusion over the randomisation procedure:

The leaflet was kind of vague, it told you what it was actually based on but it didn’t tell you about you were going to get split into two groups. I don’t think it mentioned anything about wearing the accelerometer – was it called . . .?

No I didn’t know that, I thought ‘cause [the HT] rang up somebody else and they said, ‘You are in this section [trial arm],’ so I don’t know if they’d done any background research and said, ‘Well, she’ll be good in this group,’ or, ‘she’ll be good in that group’.

OK, right. So it doesn’t sound like that was explained to you properly then.

No, no not at all.

Female, aged 25–30 years, employed part-time, moderate smoker, intervention

Another example of confusion arose surrounding the payment for wearing and returning the accelerometer where the individual thought that rather than them getting paid for wearing and returning it, they had to pay the study. [Note: during the course of the study, a number of accelerometers were not being returned so we increased the reward for returning them from £10 to £30 in exchange for wearing it for 1 week and returning it].

Yeah, well, because [the HT] hadn’t mentioned it at the beginning, so I didn’t know anything about it, and then when it had been changed [the HT] said, ‘Oh, just to let you know that it’s gone from £10 to £30’. And I thought that I had to pay her and [the HT] said, ‘Why?!’

So that wasn’t made very clear to you.

No!

Female, aged 30–35 years, unemployed, moderate smoker, control

While this may demonstrate a high level of motivation for taking part in the study, it suggests that explanation of the reimbursement structure and randomisation may need a clearer explanation. This was the first time that most participants had engaged in a research trial and more practical explanations to participants of the need for and process of randomisation are suggested by these findings.

Appeal of the invitation and support for reduction

The novel approach of actively promoting support for reduction in smoking was shown to be well received by the majority of the sample. Many framed the appeal of reduction against the alternative of stopping abruptly. The appeal of reducing appeared to stem from an underlying desire to change behaviour but, owing to a lack of confidence in stopping abruptly, reduction seemed like a much more manageable approach to tackling their smoking behaviour:

Yeah, I think yes, I think that was probably it, the reduction thinking. Well you know, rather than sort of go cold turkey and completely stop I thought, ‘Oh you know, you could help me reduce it,’ which you did, so you know that obviously it worked.

Female, aged 60–65 years, unemployed, moderate smoker, intervention

What was it about that then that appealed to you?

Erm [coughs] it’s easier than cutting it out altogether, I guess it should be easier actually. But . . . well I needed to try something as opposed to nothing, that’s what appealed to me.

Male, aged 60–65 years, unemployed, moderate smoker, intervention

Well it . . . I didn’t feel as if I was undermined in any way you know, until it’s to stop altogether you know. I felt that it was, well, maybe an easier solution.

Female, aged 60–65 years, employed full-time, moderate smoker, intervention

One individual explicitly expressed this notion as reason for coming into the study, revealing that cutting down would be a good idea at a time when they feel they are not ready to stop:

Cutting down, is that something that perhaps appealed to you?

Yeah it did as well, because I thought, ‘I don’t really want to, I am not ready to stop yet,’ and I thought cutting down is quite good. I tend to smoke if I am a bit stressed. I use it as an excuse to smoke so [laughs] I thought, ‘You know, that’s quite a good way to do it really is just cut down bit by bit’.

Female, aged 35–40 years, employed part-time, moderate smoker, control

For some, past experiences of failed quitting heightened the appeal of support for reduction as a novel approach to tackling their smoking behaviour:

Well, for three or four years I’ve been trying to give up smoking [and] last month [I] done ten months, and then I had a smoke. The year before six months, the year before that was four months, and it’s always a quit, you know, stop, go on to nicotine gum and lozenges and patches and whatsoever. This one appealed to me because you cut down, you know, every week you cut down two cigarettes a week and you just cut down and cut down and I eventually got down to none.

Male, aged 60–65 years, unemployed, moderate smoker, intervention, successful quitter

The message of support to cut down, in the trial invitation, did not appear to threaten people’s sense of control over their own behaviour compared with a message around abrupt quitting. For some, it was clear that a pervading message of the need to ‘stop smoking’ would have completely alienated them from engaging in the study:

Well, [the HT] said to me, ‘We are not here to make you stop smoking, make you stop smoking. We are here to at least try and cut you back, if we can help you just to cut down, cut down by whatever, just to help you cut down, we are happy with just helping you cut down.’ And I thought, ‘Well, this is someone talking to me that’s not saying, ‘We can stop you smoking. We can do this,’ ’ but [the HT] was saying, ‘We can help you to cut down. We may in time be able to stop you smoking,’ and I said, ‘Well, that’s a very sensible attitude to take,’ because someone telling me, ‘I’m going to stop you smoking,’ I’d tell them to . . . go away! So that’s what made me do it initially, because they weren’t threatening me that they could stop me smoking. But even at this time, there is no one that can tell me, ‘I can stop you smoking,’ you know what I mean?

Male, aged 55–60 years, unemployed, heavy smoker, intervention

There was also evidence to suggest that smokers can feel saturated and alienated by the ‘hard sell’ of the abrupt quit message. The invitation was designed, as was the intervention, to be supportive and client centred and a step away from traditional services, and the supportive and pressure-free nature of the invitation was well received:

You know, I say, when [the HT] gave me the leaflet I thought, ‘Yeah, all right, I’ve heard all this before,’ and I thought, ‘Well, here we go with the hard sell’. But [they were] totally different. [They were] so relaxed, so friendly, and that’s what pushed me towards it. If [they] had tried to come across with the hard sell I would most probably have just ignored [them] and said cheerio. But I think just approaching people in a friendly manner . . . I mean, sometimes it helps.

Male, aged 55–60 years, employed full-time, heavy smoker, intervention

Oh, I expected the trial to be like, ‘Oh, you’ve got to do this, you’ve got to do that,’ you know! And it wasn’t nothing like that because I said to my chap, I said, ‘If I go down there and they say, ‘You’ve got to give up smoking now,’ I’m saying no, I can’t do that!’ But [the HT] wasn’t nothing like it, [the HT] was really nice. I said to her, ‘I can do what I can do.’ I said, ‘You can set me out some tasks and I will try and do them.’ So that’s what [the HT] done for me.

Female, aged 50–55 years, employed full-time, moderate smoker, intervention, failed quitter

The intention of providing a supportive and pressure-free message seems to have been achieved and effective in engaging with the targeted population.

For one individual, they recognised both the reduction and the PA support side of the intervention as being something they would like to take advantage of:

I was actually thinking about doing something to help myself so when it came up and then it said about the smoking reduction I thought it was like a bonus that I can get healthy and reduce my smoking at the same time.

Female, aged 25–30 years, employed part-time, moderate smoker, intervention

This was a relatively isolated case in responding to the appeal of the invitation, but this was to be expected due to the intentional ‘down-playing’ of the PA side of the intervention in an attempt not to alienate the less active, or to recruit only those most interested in PA.

Effect of invitation

It became apparent that the invitation itself had a motivating effect on people’s desire for change. As was expected, most smokers were already contemplating, and had been for some time, the idea of changing their smoking behaviour. The invitation acted as a prompt to change their smoking habits:

Oh exactly, exactly, it actually gave me the push that I needed [laughs] just getting the letter and then, you know, yeah it just gave me the kick that I needed. I’d been thinking about it for a long time but never doing anything about it. As I say, the first getting the letter was the kick that I needed. I mean I could really have stopped in the next couple of years, I don’t know about that one but I hope I would have done, but it was just that initial for the letter coming through the door and I thought, ‘Yeah, I would do that’.

Female, aged 60–65 years, unemployed, moderate smoker, intervention

Um, well, because I recently tried to give up smoking a few months before I started and obviously it didn’t happen, but I was on the smoking replacement stuff, and I just thought, ‘You know, I’ll give anything a try,’ if you know what I mean. So when I got the letter I just thought, ‘Oh yeah, I’ll go for that’. You know, it was just something that I immediately thought, ‘Yeah, I want to do it’.

Female, aged 35–40 years, employed part-time, moderate smoker, intervention

Motivations for taking part

Not surprisingly, a common motivation was health related, either through personal health or through ill health of family or friends. Although there is little evidence to suggest that cutting down smoking has significant health benefits, it was perceived by people to offer health benefits with quitting still an underlying long-term goal:

Because I have had a few friends who have died of cancer and that recently, see? So it’s getting more of a thing to, like, try [cutting down].

Male, aged 45–50 years, employed full-time, moderate smoker, control

Because I keep on having a permanent cough and it won’t go away. And the doctor has given me this, that and the other but he said, ‘Your cough ain’t going unless you give up smoking’. So it really is . . . it irritates my throat and that’s why I cough all the time.

Female, aged 50–55 years, employed full-time, moderate smoker, intervention, failed quitter

Oh, I’d been thinking for about a year I must give up smoking, it’s expensive and it’s not healthy, it’s antisocial and you know, all these things, but never quite doing anything about it.

Female, aged 60–65 years, unemployed, moderate smoker, intervention

The last person also mentioned the financial cost of smoking as a motivating factor, a motive that was borne out in other dialogues:

So you weren’t necessarily interested in quitting altogether, but just the idea of reducing was quite a good idea for you, was it?

Yeah, I was averaging about 20 a day and being on a pension I went over it and I thought, ‘This is getting too expensive. I am going to have to knock it on the head’.

Male, aged 60–65 years, employed full-time, heavy smoker, intervention, successful quitter

Unsuccessful and negative previous experiences of using NRT and other medicinal therapies emerged as a strong theme linked positively to motivation for taking part. It was envisaged that the intervention would offer a novel alternative to the use of NRT, and people’s description of past experiences seemed to confirm this particular aspect of the intervention.

Umm [sighs] well I suppose the fact that I’d been on the 3-month course [varenicline] . . . I started smoking again [on holiday] and when I got back and this cropped up I thought, ‘Well you know, anything I can do to help,’ you know, ‘stop people smoking including myself,’ I thought, ‘Well I’ll take part in it,’ you know and whether it was going to be of any use to me personally or not at the end of the day I thought I’d just wait and find out.

Male, aged 55–60 years, employed full-time, heavy smoker, control

Oh I’ve tried a couple of times to cut out smoking totally ‘cause I’ve tried the smoking aids and all the things you know, the puffer and the patches and that hasn’t worked, so I thought, ‘Oh well, I’ll give this a try then try and cut down,’ yeah.

Male, aged 40–45 years, employed part-time, very heavy smoker, intervention

There were a couple of reports of people taking part through altruistic motivation and the interest of being part of a research study:

I just thought I’d see. I don’t know, it just sort of appealed really, and also I suppose out of interest of a scientific enquiry.

Male, aged 45–50 years, employed full-time, moderate smoker, intervention

Well, I’m up at the uni anyway, and helping out with studies is a good idea.

Female, aged 30–35 years, unemployed, moderate smoker, control

There was an isolated example of somebody coming forward to enter the study because of the incentives offered to the intervention group in terms of subsidised leisure facility access:

Well obviously I thought I might be one of the lucky ones that would get the support, obviously they had gym membership and obviously I was still wanting to do all of that as well so, and obviously where I am it costs too much a month to actually go to the gym, so . . .

Female, aged 35–40 years, unemployed, moderate smoker, control

It appeared that some participants were only taking part for the financial incentive offered for returning the accelerometer:

Well, I’m not sure. I mean, I’ve spoken to a few people who’s been on it and when it was first mentioned, somebody said, ‘Oh you get £30 at the end of the 3 months.’ I said, ‘So you did it for the £30?’ He said, ‘Well, yeah.’ I said, ‘Well, how sad is that? That’s really sad. So are you going to pack up smoking?’ ‘Oh, no.’ You know, which I think is rather sad. I would just cut out the incentive of any money.

Male, aged 55–60 years, employed full-time, heavy smoker, intervention

We are also aware of one participant who withdrew because the financial incentive was ‘not enough’.

Randomisation: understanding the process and acceptability

Participating in a RCT was a new experience for the majority of the participants. With the exception of the one example discussed earlier, the procedure was well understood by those interviewed, albeit explained and understood in a variety of terminology:

[The HT] said to me that there’s going to be two groups and after we’d had the discussion about the whole thing, they [phoned] somebody in Exeter and they would allocate me to a particular group.

Female, aged 50–55 years, employed full-time, moderate smoker, control

Yeah, yeah. No, the person who was mentoring me at the time, he said, ‘You may be lucky, you may not. It depends’.

Male, aged 60–65 years, employed full-time, heavy smoker, intervention, successful quitter

The effect of the randomisation procedure on motivation to remain in the study (reluctance to remain in the study if not receiving any support) showed no particular influence either way. Some people showed no disappointment in being allocated to the control arm, while some did:

Right, and how was that for you? Was it a disappointment or was it . . .?

No, no, not at all, no, it was fine. I was just interested to see what would happen actually!

Female, aged 50–55 years, employed full-time, moderate smoker, control

Yeah, I was a bit gutted by it but like you say, they did explain to me beforehand that that may be the case.

Male, aged 25–30 years, unemployed, heavy smoker, control

The HTs also revealed that participants were on the whole ‘disappointed’ to be allocated to the control condition, with one participant immediately withdrawing from the study as a result.

Locality of assessments

Assessments took place in a centralised and well-known health service building and, where appropriate, at the participant’s GP surgery. From those interviewed there were no complaints about the location of sessions.

One of the successful quitters in the study praised the location of the sessions as the SSS was based in the same building, and so when he did decide to make a quit attempt the process was relatively seamless. Only two of the quitters did make use of the SSS, so this was not relevant for many participants.

Yeah. I would just like to mention that at the [centre] where it was held, to actually stop smoking and then getting the nicotine aids things, it’s made a damn sight easier because you are in the hospital and they have got a non-smoking unit there. So I saw this [stop smoking advisor], prescription, straight away, for gum, nicotine gum, patches and all the rest of it, you know. Because the location is brilliant because you did not have to go away and find somewhere, go to my surgery.

Male, aged 60–65 years, unemployed, moderate smoker, intervention, successful quitter

Intensity of support

Behavioural support for helping people who do not want to quit in the immediate future to cut down is a new area of research and little is known about the amount of support required. A major consideration was the frequency, duration and length of time over which the support should take place, and the mode of delivery. The pre-pilot development work resulted in a very flexible style of intervention delivery, with a mix of face-to-face sessions and telephone support sessions.

There was support for weekly sessions, which some participants found to be acceptable:

Oh no, no that was fine. I think once a week was brilliant and that’s what’s helped me.

Male, aged 60–65 years, unemployed, moderate smoker, intervention, successful quitter

I think it was probably about right, I don’t think, you know, it didn’t seem onerous particularly being part of the study.

Male, aged 45–50 years, employed full-time, moderate smoker, intervention

Well I think I, all the ideas were there but you know I thought it was probably enough.

Male, aged 60–65 years, unemployed, moderate smoker, intervention

No, it was just right, I saw [the HT] on a Tuesday morning at 10 and I was more or less out of there by 11, just before. Yeah, no, I was happy with that, that was fine.

Female, aged 35–40 years, employed part-time, moderate smoker, intervention

The lengths of the sessions were generally acceptable as well:

No, I think it was OK. It was mostly an hour a week, sometimes it went a bit longer, but no it was fine.

Male, aged 40–45 years, employed part-time, very heavy smoker, intervention

The mixture of face-to-face contact and telephone support was generally well received by those interviewed, allowing for increased flexibility and engagement with the support around continually changing circumstances for some:

[The HT] always used to ring me. Because some days I couldn’t make it and I just said to [them], ‘Can I speak to you on the phone instead?’ And then we’d do the paperwork over the phone because I was working quite a bit then.

Female, aged 50–55 years, employed full-time, moderate smoker, intervention, failed quitter

Having the telephone contact appeared to play an important part of the perceived support participants experienced, whether or not they used it, having been provided with the HT’s telephone number and being told they could always telephone if they had any problems.

I always had the mobile number anyway, not that I needed it, but they said you know whenever I did need it for anything.

Male, aged 40–45 years, employed part-time, very heavy smoker, intervention

I think it worked really well because I always had [the HT] on the other end of the phone, if I needed her, do you know what I mean? You know, [the HT] was always there for me and so they really helped me a lot.

Female, aged 50–55 years, employed full-time, moderate smoker, intervention, failed quitter

No, no it was fine it was fine. I mean when, well, they were always there at the end of the phone whoever you were dealing with, that was quite nice, if you had to change an appointment or what have you they were always there.

Female, aged 60–65 years, employed full-time, moderate smoker, intervention

The overall feeling regarding the intensity of the intervention was generally mixed. There was no report from those interviewed that the support was too intensive or overbearing, but there was a recurring theme that the support could have been more intensive than once a week:

Maybe erm a little more often to see somebody. I mean as I say I was fine, but maybe some people would like that little bit more . . . even a phone call maybe just to see how you’re getting on [in between sessions].

Female, aged 60–65 years, unemployed, moderate smoker, intervention

We were doing regular meetings but I found if there had have been a little bit more contact like maybe just a phone call like halfway through the week . . . I think as well the weekly meetings maybe like a midweek telephone call to give you that little bit of encouragement and everything.

Female, aged 25–30 years, employed part-time, moderate smoker, intervention

Some participants reported that not only was the support not intensive enough, but it could have gone on for longer to support further behaviour change:

I would say the only thing is, I mean like now I’ve done it and everything, it was too short and it wasn’t intense enough . . . I mean, [the HT] was excellent and I am personally saying my idea is that if it was more intense it would be more helpful. It would be better.

Male, aged 55–60 years, unemployed, heavy smoker, intervention

Erm I think . . . if you got people to give up smoking I think it would be, you know, a longer period of time would be better.

Male, aged 60–65 years, unemployed, moderate smoker, intervention

There was some evidence to suggest that participants were not receiving the weekly level of support that was intended.

Well, when I say that, without being rude, I mean . . . my counsellor was very good but I saw [them] once every few weeks, of a couple of weeks maybe or maybe I saw [them] like next week, then I wouldn’t see [them] for two weeks, or three weeks maybe, or two weeks, you know what I mean?

Male, aged 55–60 years, unemployed, heavy smoker, intervention

There was a period when it was sort of four, eight and sixteen weeks was it I can’t remember what the time was but erm, yeah, maybe say every two weeks would have been helpful to some people.

Female, aged 60–65 years, unemployed, moderate smoker, intervention

Some instances of limited contact were due to staff illness, the three HTs all working part-time and on different days, and HTs being fairly rigidly assigned to individual participants (to maintain continuity). Field and meeting notes did reveal a large variability in participants’ abilities to stick to regular scheduled meetings depending on unpredictable and changeable personal circumstances. One participant withdrew from the study before receiving any intervention support sessions due to a delay between the baseline assessment and the first intervention session. Fortunately, these difficulties were few and participants generally appreciated the intensity and flexibility of support received.

Style of delivery

The EARS client-centred intervention was designed to maximise adherence among a traditionally service resistant group. The HTs were trained to establish a strong rapport and engage participants in taking control of any behaviour change. We were keen to understand how this worked and was perceived by participants.

There was good support for the HTs and the strong rapport they developed with the participants interviewed, with discussions around the targeted behaviours moving forwards in an individualised way.

Yeah, fine. [The HT’s] all right. Yeah, [they’re] lovely. I used to go down and see them or they used to ring me if I was too busy at work or something. Yeah, they was fine. I got on really well with them. They was a good help.

Female, 50–55 years, employed full-time, moderate smoker, intervention, failed quitter)

But [the HT] was very good at what they done, you know what I mean? [The HT] was, I mean, like first, like most people with a doctor or something like that, ‘How are you feeling? How’s your problem?’ Well with [the HT], with their consulting ways, it was like, ‘How’s your week been? Have you had a good time? Did you do anything?’ then you had a 5-minute chat, only about basics like, you know, blah blah blah, and then slowly they brought in the smoking, you know, and what I was doing and the conversation. They was very good at what they done.

Male, aged 55–60 years, unemployed, heavy smoker, intervention

[The HT] was a brilliant help, [they were] fantastic to be honest.

Male, aged 45–50 years, unemployed, heavy smoker, intervention

The skill of the HTs in developing and maintaining good relationships with the participants was widely acknowledged among those interviewed. Advanced interpersonal skills were a high priority for the appointment of the HTs within the trial and they appeared to be effective:

Well I mean, there was never any question, you know, it was always like, ‘You know it’s down to you,’ you know, so no, I found it, you know . . . really quite good actually you know, I thought they were there was no question but lots of support I guess, and interest.

Male, aged 60–65 years, unemployed, moderate smoker, intervention

[The HT] said it was entirely up to me, what I was doing, and yeah, they were a good help, that’s all I can say, really. [The HT] was a good help.

Female, aged 50–55 years, employed full-time, moderate smoker, intervention, failed quitter

Oh totally, like my decisions, I wasn’t pressurised into it at all. I was encouraged and I was supported but I wasn’t pressurised.

Did you feel like you were in control of your decisions then about how much to cut down?

All the time, yeah, all the time.

Female, aged 60–65 years, unemployed, moderate smoker, intervention

Oh yeah, I was in control. I mean, I told [The HT] how many I was going to cut down a week or a day and they said, ‘OK, well that’s the goal you are going for.’

Male, aged 60–65 years, unemployed, moderate smoker, intervention, successful quitter

Oh no, not at all, [the HT] gave me some useful, like, suggestions and how to plan what I was going to do. But no, [the HT] was never like suggesting, ‘You are going to do this, you’re going to do that,’ no, never like that, no.

Female, aged 25–30 years, employed part-time, moderate smoker, intervention

Participants experienced a very client-centred style of delivery, one which was not threatening and placed them firmly in control of the decision-making process. There was also repeated evidence of people responding well to the strongly supportive and non-directive nature of the intervention delivery:

Well, no one likes people telling them what to do really, do they, unless it was discussed, so it was a study about me and my smoking habits then obviously it’s got to come from me and not someone else, ‘cause everyone’s different ain’t they with smoking and everything.

Female, aged 25–30 years, employed part-time, moderate smoker, intervention

It’s all down to the person but as I say, I’m not weak-minded and people can’t tell me to stop smoking but [the HT], the way they put things across to me, explained things, and, you know? They brought me down to earth in a nice way.

Male, aged 55–60 years, unemployed, heavy smoker, intervention

[The HT] was just, you know, not advising me but . . . [the HT] wasn’t even telling me, [the HT] was just saying, ‘You do what you think best,’ and ‘Do what you can do,’ they said. But they did say to me to get out a bit more because I never did go out very much. No, it was from me. They didn’t force it on to me, they just advised.

Female, aged 50–55 years, employed full-time, moderate smoker, intervention, failed quitter

Yeah, I mean, it was advice, it was good advice. [The HT] used to say, ‘We are not pushing you to stop smoking, but if you want to cut down, this is the ways you can do it.’ I had the forms each week down there each time like, and it was always good advice, it was always pleasant. There was no hard pressure, no hard sell.

Male, aged 55–60 years, employed full-time, heavy smoker, intervention

I wonder how well would the intervention have worked if it was a bit more directive?

Yeah I, yeah I probably wouldn’t have reacted you know in the same way, yeah. Because I think it’s something, it’s nice to talk to somebody one to one but at the end of the day it’s yourself that must be in control and be able to say, you know, ‘Well yeah, this is okayOK, I can do this,’ or, ‘I can do that,’ but what do you know sort of to be told that you should be do this, you should do that, because it’s up to each and every one of us to decide.

Female, aged 60–65 years, employed full-time, moderate smoker, intervention

Key to promoting ownership and supporting people in taking control of their own behaviour change is eliciting the individuals’ own solutions to problems. The HTs mostly prompted participants to reflect on behaviours and consequences, and identify solutions:

Yes I did, yes, and as I said you know you didn’t, they weren’t intrusive, they weren’t saying, ‘You must do this,’ or . . . they were helpful and just made suggestions, but it’s probably things that you would have thought about yourself but their suggestions were what you were thinking really, you know, the ones that reacted, but it was nice to know that they weren’t sort of saying, ‘Well you should do this, you should do that,’ you know, it was a one-to-one thing, it was very good.

Female, aged 60–65 years, employed full-time, moderate smoker, intervention

Support was voiced for the non-judgemental nature of the intervention, which was designed to avoid failure and always reframe any failures as learning experiences or opportunities for reflection:

The next fortnight that I saw [the HT], they said, ‘How did you get on?’ and I showed them the sheet and they said, ‘Well, you kept to your goal.’ And that was it. They didn’t tell me off a lot or belabour me or . . . I did try and keep to my goal [but] a couple of days I missed out, you know.

Male, aged 60–65 years, unemployed, moderate smoker, intervention, successful quitter

Yeah I think, you know, there wasn’t, I didn’t have any sort of great pressure or anything, you know, I mean there was no, ‘You are a bad person, you must quit,’ sort of thing or attitude if you know what I mean.

Male, aged 45–50 years, employed full-time, moderate smoker, intervention

One participant favourably compared the different approach with other support they have received in the past:

I wasn’t told, you know, ‘You should be doing, this you should be doing that,’ like when I have, when I’ve gone to other places to try and give up smoking [laughs] yeah, ordered to do things, you know what I mean, we just talked about things, talked about ideas.

Male, aged 40–45 years, employed part-time, very heavy smoker, intervention

Multiple behaviour change

Fundamental to the intervention structure was its aim of addressing two behaviours simultaneously, an approach that is not widely advocated in behaviour change literature, and in most cases avoided, for those wishing to quit. The complementary nature of the two behaviours (one a ‘stopping’ and one a ‘starting’ behaviour) meant that it was postulated that addressing the two behaviours alongside each other could be acceptable.

Resulting from the pre-pilot work and work with early intervention participants, the behaviours were introduced sequentially with smoking behaviour taking precedence over PA behaviours. How acceptable this was to participants and how much they engaged with the idea was of particular interest, especially how they viewed the utility of using PA to influence their smoking behaviour.

While it was difficult to explicitly elicit from participants how difficult or challenging they found it to try changing two behaviours simultaneously, there was qualitative evidence that participants actively engaged in the process of using PA as a way of managing and influencing their smoking habits:

I did make some changes, yeah, and when I do go out for a walk with the dog I don’t take my cigarettes with me and I could be out for hours, so I don’t bother taking my fags with me.

Female, aged 50–55 years, employed full-time, moderate smoker, intervention, failed quitter

Well, there was the exercise thing, getting out and being more active and stuff, which I am quite an active person anyway. I walk everywhere and you know, I am not a sitty-home person. I might want to do it once a week but I’m sort of like an active person anyway. Yeah, things like that helped me. You know, the gym helped.

Female, aged 35–40 years, employed part-time, moderate smoker, intervention

Physical activity was important in highlighting the negative health consequences of smoking and this appeared to change the way some people thought about their smoking habits:

As I say, doing exercises and that, you feel healthier and then obviously if you are doing exercises and silly things like that, you feel healthier and then when you go to put a cigarette in your mouth, it’s obviously psychological, you’re going, ‘Oh no, I am wasting myself going in the gym for that hour,’ if you know what I mean, ‘to go and have a cigarette now’.

Male, aged 55–60 years, unemployed, heavy smoker, intervention

There was also support for an ‘identity shift’ from a smoker to an exerciser, and the value attached to being active:

Yeah, well as I say, it did help and not only that, you see, getting confident was . . . in this gym, it was state-of-the-art, you know, I tell you if I was still smoking there is no way I could use some of those machines that I’ve been using, the rowing machine and the boxing machine thing. No way I could use those if I was smoking! I’d just be coughing my lungs up.

Male, aged 60–65 years, unemployed, moderate smoker, intervention, successful quitter

And I found out it’s a drug in a way ‘cause when you go to a gym you come back out you feel great for it and then smoking, you don’t even think about smoking ‘cause you’ve gone in there and you’ve done like an hour and a half workout, the last thing you want to do is come out and have a cigarette.

Female, aged 25–30 years, employed part-time, moderate smoker, intervention

Physical activity was an acceptable way for some of distracting themselves from smoking and coping with the cravings:

Erm, well it’s still difficult to cut down the cigarettes I’ve got to agree, but if I walk more then I don’t need to, you know, I don’t want to smoke, put it that way.

Female, aged 60–65 years, employed full-time, moderate smoker, intervention

Well, I get about a bit more now because I never used to get out very much but yeah, I get out a bit more now with the dog and that, and [the HT] said, you know, [the HT] said, ‘Go out walking and try and keep your mind off cigarettes,’ which I did do . . . Yeah, I take the dog out and I do things, you know? Definitely. Yeah, it works, most definitely works, yeah. It’s something there to stop me . . . yeah, to stop me from thinking about it, distraction, yeah.

Female, aged 50–55 years, employed full-time, moderate smoker, intervention, failed quitter

I’m doing my exercises three afternoons a week. And that takes my mind off the smoking. It takes your mind off of smoking actually, the exercise. It really does.

Male, aged 60–65 years, unemployed, moderate smoker, intervention, successful quitter

When [the HT] was trying to put across all the activity side, I found [that when] I was engrossed in doing something, which may have been whatever, I didn’t smoke so much [ . . .]. But then, if I wasn’t doing anything I was smoking.

Male, aged 60–65 years, employed full-time, heavy smoker, intervention, successful quitter

It did become clear that PA was not necessarily the primary way in which people would control their smoking behaviour. It was continually aligned with other ‘distraction’ activities which participants would engage in to manage their smoking behaviour. PA was, for many, just another way to keep busy:

Well I like walking so you know, I’d go out for a walk you know or something like that or, you know, read a book; anything to take your mind off it really, break the routine, ‘cause that’s what smoking is really.

Erm well, keeping myself busy basically, because if I am doing other things I am not smoking so much. I go swimming too you know, but first thing I do when I get up in the morning is have a cigarette you know, so if you can cut that one out you know a good part of the, it is a different sort of day so instead of, you know, having a cup of tea and having a cigarette I would do something for an hour or so.

Male, aged 60–65 years, unemployed, moderate smoker, intervention

Some participants did see more benefit in PA as a way of distracting themselves from smoking than other activities:

It’s more effective because if you’re sat down reading or anything you’ve still got that opportunity to think about a cigarette, whereas when I’m like huffing and puffing along to the aerobics the last thing I want to do is stick a cigarette in my mouth.

Female, aged 25–30 years, employed part-time, moderate smoker, intervention

In one case, the participant had made explicit links between PA and the potential for gaining weight while cutting down their smoking:

I made myself cut down but I haven’t put on any weight, you know, and I think that’s how I explain it [PA], you know, ‘cause some of my friends who have cut down smoking the first thing you hear about is putting on weight, putting on weight, putting on weight, and I haven’t, so I’ve cut down on my cigarettes – well, I have and I haven’t put on any weight whatsoever you know, nothing like that has changed at all.

Male, aged 40–45 years, employed part-time, very heavy smoker, intervention

On the whole, those interviewed found the idea of addressing behaviours simultaneously acceptable, to varying degrees. There were cases where the promotion of PA was not completely appropriate, and people failed to see how the two behaviours linked together. This was particularly the case if the participant already viewed himself or herself as being highly active.

No I didn’t you see, you know [the HT] was telling me about exercises and I mean I do dance and I do swim and I do walk you know this sort of thing, and I am a pretty sort of active person anyway to do an exercise to, ‘cause I think that was one of the sort of objectives of it to up your exercise to lessen your smoking, but with me that didn’t or wouldn’t have worked because if I’d really wanted a cigarette I’d have done my exercise and then had a cigarette, so [coughs] it didn’t stop me from smoking if you know what I mean, to do more exercise.

Female, aged 60–65 years, unemployed, moderate smoker, intervention

For one individual, PA may have reduced their desire for a cigarette while actually doing it, but heightened their desire for a cigarette afterwards:

I said to [the HT] you know, the truth of the matter is that whenever I actually have a cigarette, whenever I did physical exercise at the end of it I would sort of in some case smoke more [laughs] if that makes sense. It was sort of a reward you know.

OK, that’s interesting. So for you, you don’t find that sort of doing something active tends to lessen your desire for a cigarette then.

Well when I’m doing it, but not afterwards, no.

Male, aged 45–50 years, employed full-time, moderate smoker, intervention

Design

Qualitative research was conducted using a basic thematic analysis of individual face-to-face interviews to elicit and describe the HTs’ experiences and views about delivering the intervention.

Sampling frame

An opportunity sample consisting of all three part-time HTs (with a dual role of also collecting data) employed in the study was used to maximise the diversity of opinions in the data. Interviews were conducted both early in the study (1–2 months after starting to deliver the intervention in the pilot trial, to capture experiences of the training course while they were still fresh in the HTs’ memories), and in the last 1–2 months of the 16-month pilot trial (to capture any changes in practice or opinion following extended experience of delivering the intervention).

Measures and procedure

Semi-structured, individual, face-to-face interviews were conducted within a few months of completing the training and at the end of the intervention period using topic guides developed by AT, CGVS and TT, designed to elicit data relating to the aims above. The first interview started with general questions about the HTs’ experiences of delivering the intervention and then asked specific questions about the training course, recruitment processes, intervention delivery (what was working well or badly) and the HTs’ understanding of the different intervention processes. The second interview (at the end of the intervention phase) asked about their ongoing experiences in delivering the intervention and how these might have changed since the initial interview. The interviews were digitally recorded and transcribed verbatim.

Analysis

The data were organised using a basic thematic analysis to provide a simple descriptive-level overview of the HTs’ views. In-depth qualitative analysis procedures were not used here. However, these data will be analysed in more depth alongside the participant interviews and consultation recordings, using framework analysis96 to generate an integrated analysis of processes of behaviour change in the EARS intervention. This in-depth analysis will be reported as part of the PhD of one of the researchers (TT).

The training course (see Appendix 5)

Overall, the training was well received and valued in building the necessary skills for delivering both the research and the intervention. The training was an exploratory process to help formulate the intervention, drawing on the individual skills and experiences that the HTs brought to the training; they were all involved in some form of health promotion work in their other part-time employment. The developmental nature of the training gave the HTs an impression that the training process was somewhat disjointed, but equally one HT valued the opportunity to be able to contribute to the intervention development.

When asked ‘what worked well?’ the HTs identified practice sessions (with volunteer clients) and reflecting on recorded consultations, having an intervention manual, and getting formative feedback from a health psychologist as important. There were differences of opinion about how many practice sessions were useful, apparently based on the differences in individual learning styles.

One aspect that the HTs found particularly difficult was the initial assessment of their ‘instincts’ for the intervention (a simulated client interview in session 2 of the training). The intention was to allow the training team to adapt the training to the existing skill levels of the trainees. However, this was perceived to be challenging and disconcerting by the HTs but useful, nonetheless, for the research team.

The several months between the end of training and the start of intervention delivery, due to the logistics of taking occupancy of an office in an NHS facility and delays caused by co-ordinating the recruitment process, was a frustration. However, it was clear that the HTs’ confidence in using the techniques increased greatly with practice and their use of these became more ‘automatic’ as time went on.

Ideas for improvement in the training course included:

• having more content for what to do beyond the initial motivation stage

• having a longer session on motivational interviewing or client-centred approaches and techniques

• practising the protocol for collecting the research data/baseline assessments

• having a short self-reflection period built in at the end of each session

• having the fidelity scale available from the outset to provide a steer on what was expected

• having the trainers model the intended delivery style

• providing a much clearer structure to the training to ‘put the cogs right in the correct order’

• guidance on working with smokers with other substance misuse issues

• including training on how to how to rein people in if they go off track

• reinforcing the idea that a client-centred approach still allows the HT the opportunity, with the participant’s consent, to offer (or exchange) information

• guidance on how to deal with people who are already active

• including more supervision and formative feedback, particularly in the early stages of delivery.

These suggested changes will be very useful in designing the training course for future implementations of the EARS intervention.

Recruitment (see Appendix 5)

The HTs recognised the relative efficiency of recruiting participants through GP and SSS invitation letters after identifying smokers on surgery databases. There were minor initial teething problems in working with surgeries. In contrast, recruiting without an invitation letter from the community was much harder, with many of the strategies used producing few participants. There was a belief that approaching existing group leaders as advocates (rather than directly approaching individuals) could be more promising but that it may take more time to build relationships within the community. The study timeline may therefore have placed constraints on what could be achieved.

When asked ‘What attracted participants to the study?’, the HTs identified a desire to reduce smoking as the primary factor, along with an interest in the research itself and the idea of getting one-to-one support. All the HTs felt that potential participants were ‘not hearing the physical activity side of it’. Reasons reported for non-participation included having other priorities, a lack of time, illness or a lack of interest in reducing smoking.

Making contact with potential participants by telephone was reported to be acceptable, but there were mixed views from different HTs: one felt she was intruding into people’s lives ‘like a salesman’. All HTs reported that this process was time-consuming, as they often needed to make several calls to get hold of the participant.

The most difficult part of the recruitment process was the disappointment expressed by participants who were allocated to the control group. It was noted that increasing the study-completion payment later in the study might have helped with this. No major problems were reported regarding access, non-attendance or workload for participants after they were enrolled.

Suggestions for improvement included:

• offering the intervention to controls at the end of the study to counteract the sense of disappointment (possibly in a group format or as a condensed version)

• developing a strategy for what to say if someone else answers the telephone

• not focusing on PA during the recruitment stage as it ‘sort of confuses people’

• having a more prolonged engagement with workplaces, perhaps via occupational health professionals or existing public health team contacts, which could help recruitment.

Delivering the intervention (see Appendix 5)

The HTs reported no problems in using most of the intended intervention techniques, including exploring a typical day; encouraging self-monitoring (which was seen as particularly useful); problem solving; empathy-building/person-centred counselling; exploring importance (including pros and cons and using 1 to 10 scales); exploring confidence; using motivational interviewing techniques (including affirmation and reflective listening); reviewing progress; assessing existing smoking; offering alternative strategies for smoking reduction; setting realistic/SMART (Specific, Measurable, Attainable, Realistic, Time-referenced) goals (usually verbally, or with the HT writing them down) and making coping plans.

When asked ‘what worked well?’, the HTs identified regular contact, encouraging self-monitoring, MI (client-centred) techniques, pedometers and offering a choice of clear behavioural strategies for smoking reduction.

In terms of delivering the PA aspect of the intervention, the HTs felt that that most people were willing to try to do some PA. However, making the link between PA and smoking reduction was easier for some participants than others. The idea that PA could help to reduce cravings was not generally well understood or accepted. However, the idea that PA could provide a distraction was felt to be more useful/more easily accepted by participants. The HTs also tried to encourage people to do experiments to test the link, with mixed success. Other strategies were sometimes useful (e.g. focusing on the general health benefits, addressing misconceptions about what PA entails). The need for individually tailored strategies was highlighted by all three HTs:

Everybody’s very, very different aren’t they, sort of, receptive to different things.

Tailoring the intervention applied to assessing motivation, identifying and addressing barriers, trying to make the link between PA and smoking reduction and deciding which behavioural strategy (or mix of strategies) to use. One HT reported that making the link to PA was often easier for these participants, although it was not clear why this might be.

For the ‘harder-to-reach’ participants (e.g. those with higher levels of mental health problems or low literacy or analytic skills), flexible tailoring of the intervention seemed to be particularly important although this did not seem to diminish the chances of a successful intervention.

There were mixed views about delivering the process measures alongside intervention sessions (NB: this was also noted as being a potentially difficult process when reviewing the consultation recordings). Some HTs felt that this was not a problem, but others identified a ‘tension between the HT and the researcher role’.

All the HTs felt that the short timescale of the intervention (8 weeks) was a limitation and would have liked to have more flexibility to maintain contact with participants who were starting to make progress.

Encouraging engagement of social support (which was identified by the intervention fidelity analysis as largely lacking) was identified as being potentially problematic/provocative as not everyone had good sources of (positive) support. This was also identified as not being high on the agenda, perhaps reflecting a lack of emphasis on this during the training. Similarly, the issue of identity change was not considered to be a major element of the intervention process.

Overall, the HTs reported feeling very positive about their experiences in being part of the research study and delivering the intervention. They felt that this was a job they would enjoy doing and would apply for it again if it the opportunity arose.

Ideas for improvement included:

• Consider using a solution-focused approach rather than a problem-solving approach.

• Possibly encourage participants to take a longer-term personal appraisal of the benefits of PA, rather than any acute effects.

• Include training on how to avoid or minimise dependence/reliance on the HT.

• Use text reminders as a useful way to remind people about appointments.

• Include training on how to deal with passive resistance (participants who avoid engagement by agreeing/going along with the HT, but then do not make the changes discussed).

• There was a further suggestion that some type of debriefing supervision might be useful, where the HTs could discuss difficult cases or their own feelings about certain difficult clients.

• Finally, there was also a suggestion from one HT that at the baseline interviews, there was a tension between the need to build empathy at this stage and the need not to engage therapeutically. This might result in some contamination of the control group (albeit quite low level).

Early engagement with the intervention

Coming in to the first intervention session, John had completed a week of self-monitoring his smoking behaviour, recording when and how many he smoked. He found this activity to be particularly thought-provoking, highlighting patterns in his smoking behaviour he had not previously considered:

And with further exploration of perceived challenges for smoking reduction:

The HT gave John a chance to reflect on his smoking behaviour:

After discussion about techniques and approaches to cut down, John adapted his own way to approach cutting down, which did not exactly fit any of the four strategies the HT had identified. He broke his day down into morning, afternoon and evening and began thinking about which cigarettes would be easiest to cut down (hierarchical reduction) and times of the day he would not smoke (smoke-free periods), and talked about extending the time between when he smoked (scheduled reduction).

John lacked confidence to cut down the first three cigarettes in the morning, and recognised the strong association between his morning coffee and cigarettes. The morning cigarettes would be the last ones John would tackle, and the ones he was least confident about cutting out.

Early discussions about PA revolved around John’s walking habits. He did not smoke when he walked, but he had not made this connection, and there was little engagement in any idea to increase PA at this stage. The HT provided him with a pedometer and he agreed to monitor the step counts. Surprisingly, John raised the possibility of using a gym.

John left the first session with a goal, after the first three cigarettes in the morning, to extend the period between cigarettes to at least an hour and a half and cut out smoking in the evening. He also planned to obtain information about opportunities at the local sports centre.

Health trainer time input to Exercise Assisted Reduction then Stop

Health trainers are the primary delivery point for the EARS intervention. For the current analyses, and based on experience within development research, and within the pilot RCT, HTs on delivery of EARS are expected to be employed on Agenda for Change Band 497 (salary mid point assumed for base-case analyses). It is assumed here that the EARS intervention, delivered by HTs, would be implemented in a way that placed EARS within a broader provider environment, where EARS was one of the interventions available, for example integrated with other NHS SSSs or health promotion services.

Health trainer time input is for specific EARS-related contact time, participant-level contacts (e.g. face-to-face meetings, telephone contacts), participant-related non-contact time/activities (e.g. planning, organising and preparation for participant contact), supervision time, and other more general activities related to recruitment of caseload and service development. Within the pilot trial, electronic records were kept by HTs to record the duration of each contact as well as the frequency of administrative contacts specific to participants (e.g. email, text messaging, written letter). HTs were also asked to record those who did not attend, and the uptake of financial subsidies. To assess the level of participant-related HT time for non-contact activities, a work-sampling approach can be used. 98 In the pilot trial a work-sampling form was developed, in consultation and with design input from the trial co-ordinator and the three HTs employed on the trial, in order to improve ease of use and acceptability. The work-sampling form was piloted and tested over a 2-week period in the trial (completed by two of the three HTs); HTs recorded the number of minutes they spent on each of the specified areas of work activity over the days worked in the data collection period (see Appendix 9 for the form used).

In order to assess the level of HT time unrelated to intervention delivery (e.g. recruitment activity, service development), activity logs for trial recruitment activities and training activity were reported/recorded by the trial co-ordinator.

Other resource use in Exercise Assisted Reduction then Stop intervention

Data on supervisory time input, assumed to be from more senior staff (Agenda for Change Band 6), exercise aids, and resources required for training of HTs in the EARS approach, were recorded by the trial co-ordinator (trial team).

Sensitivity analysis (Exercise Assisted Reduction then Stop intervention cost)

Where allowance is made for difference in use of the NHS SSS, and pharmacotherapies, assuming that 50% of those making a quit attempt use these interventions (see Appendix 9 for detail/assumptions) the estimate of the EARS intervention cost is £242.62 per participant (i.e. an increase of £50.45 per participant). However, there is no strong evidence in the pilot trial of a difference (significant difference) in use of services, with two EARS participants and no controls reporting use of services/interventions. A future full RCT will further inform this point.

Summary of reviews

Woolacott et al. 100 present a narrative review of fifteen economic evaluations (published 1986–2001) of smoking cessation. Included studies considered pharmacological aids delivered as part of face-to-face interventions, that is to say in addition to advice or counselling. The review of methods used in the included studies (n = 15) offers guidance in terms of potential data requirements (and sources); however, recommendations by the authors do not extend to modelling methods. The review highlights some of the salient issues associated with undertaking economic evaluation in this field, particularly challenges and limitations in the data available to inform CEAs studies.

Bolin99 documents a formal systematic review that identifies 30 studies (identified from 1995) using a mathematical modelling framework, applying simulation techniques, to assess smoking-cessation therapies. Most of the identified studies used a Markov-type modelling approach, and most were cohort-based models rather than individual-level simulation models. The review charts the development of the included mathematical models chronologically. Particular emphasis is given to a widely adapted model, the Benefits of Smoking Cessation on Outcomes (BENESCO) model,102 which was used in 10 of the 30 modelling studies identified. The BENESCO model is used by Bolin et al. 99 as a point of reference to frame the review of the studies identified, with commentary presented/organised around central considerations of ‘model structure’, ‘data’ used to drive and to populate the model, and on the approach taken towards assessing ‘uncertainty/consistency’. The critique of models offered by Bolin et al. 99 concludes that, while the modelling methods published to date are reasonable, there is scope for improvement. The suggestion is that a more complex modelling framework (than that currently reported) may be helpful, and that models could consider heterogeneity in more detail.

A further review from Ronckers et al. 103 was identified as potentially useful. This review considered the results from 14 model-based economic evaluations, reanalysing cost-effectiveness ratios, by applying standardised methods for estimating costs and effects. Ronckers et al. 103 do not offer commentary on modelling methods (in terms of model type and structure) and so this study is not considered further here.

Summary of model-based cost-effectiveness analysis studies in smoking cessation

Table 34 presents summary detail on 11 studies using modelling methods in CEAs for smoking-cessation interventions. Here we provide an outline on nine of these studies. Further detail is provided in Appendix 9d. The models by Feenstra et al. 104 and Hurley and Matthews105 are not discussed further here, given the particular perspective in these studies, that is to say from a Dutch and Australian perspective, respectively, and their focus on national evaluations. The BENESCO model, presented by Howard et al. 102 has been used in multiple settings and populations and is presented, albeit in slightly different forms, in other papers. Given that these papers document the same underlying model, the focus in the current review is on the original paper.

Coleman et al.101

The HTA report presented by Coleman et al. 101 uses a Markov modelling approach, with cohort-level analyses, to examine the cost-effectiveness of relapse prevention initiatives in smoking cessation. The perspective of the model was lifetime, arranged in 6-month model cycles. The risk of smoking-related disease is calculated for current, never and former smokers using the observational data from Doll et al. 111 The epidemiologically driven model derives ratios for mortality rates (by smoking status) and uses prevalence data/rates to attribute/decompose standard population mortality rates by smoking (Figure 7).

FIGURE 7.

Example of cohort-level simulation (simple schematic) used by Coleman et al. 101 Status by age and sex category. CHD, chronic heart disease; COPD, chronic obstructive pulmonary disease; LC, lung cancer; MI, myocardial infarction.

The structure of the model is based on a simple Markov process, but with some complexities. For instance, the likelihood of an individual being a smoker or former smoker and of developing comorbidities varies with their age, based on the prevalence of smoking status from the Health Survey for England. 119 RRs of diseases by smoking status are used to decompose the population-based prevalence rates of five major smoking related diseases. Further detail on prediction of mortality by smoking status is given in the next section [see Modelling mortality: mortality rate by smoking status, and projection of life expectancy (life-years) over time].

Overview

The models identified in the literature are either decision-tree or Markov-type models, or a combination of the two frameworks. Decision-tree models have been used to model the first 12 months of the model time frame, typically an intervention related time period (and/or post intervention duration), in order to establish the probability of 12-month sustained abstinence. The complexity of the modelling of short-term abstinence was in many cases dependent upon the nature of the interventions being compared and the trial evidence available. The scope of the models (in terms of treatment strategies and referral pathways considered) tended to be more comprehensive if the model was produced to inform specific policy contexts.

A significant number of the cohort-level Markov models identified101,102,104,105,107 were population-based smoking-cessation models developed to guide budget-holders and policy makers in their decision-making. In general, these models replicated national populations, explicitly modelled individual smoking-related diseases using data from epidemiological registries and used simulation methods (typically cohort-level Markov models with multiple states and time dependency) to determine the benefits of quitting smoking.

The evidence review has been used to inform planning for model development related to EARS, and here we outline considerations on structure and data, specifically around the outcomes used, the modelling of abstinence and relapse (data inputs) and the use of available data to inform on mortality data/rates in order to model life-years and related QALYs. We also consider, in brief, issues related to data inputs on resource use and associated cost, and the way models had presented outcomes and considered issues related to uncertainty (structural and parameter uncertainty in modelling methods) and consistency (model validation).

Quality-adjusted life-year weights, and estimation of quality-adjusted life-years over time

Modelling methods described here have adjusted mortality data over time using health-state values (QALY weights) in order to estimate QALYs by treatment strategy, and to compare incremental QALY gains by treatment (Table 36). Within the literature review undertaken here, we identify a recent study presenting health-state values (QALY weights) by smoking status. 126 Vogl et al. 126 present health-state values obtained by use of EQ-5D within the Health Survey for England119 by age (year of life as opposed to decile of life), sex and smoking status for the English population.

Relevance of resource use and costs

All studies discussed here considered costs directly relevant to the intervention, and the majority used a unit costing approach and reported units of resource separately from cost. The majority of included models considered the costs of treating the main smoking-related diseases, either in the base case or in the sensitivity analysis (notable exceptions being studies by Wang et al. 15 and Woolacott et al. 100). In terms of the smoking-related diseases, lung cancer, myocardial infarction, stroke, and COPD were typically included, and diseases such as CHD and asthma more seldom. The epidemiological methods for determining smoking-attributable disease differed, as discussed by Bolin et al. 99 in their review. There was a tendency for studies to report cost-effectiveness both with and without the future cost of lifetime smoking-related diseases, with discounting of costs only usually applied in the latter. No study considered the possibility of higher future lifetime costs (‘survivor’ costs) of treating former smokers relative to continuing smokers, despite this being an area of contention in the literature. 127,128

Uncertainty

All models discussed above reported some attempt to consider uncertainty in input parameters. This typically included a series of univariate analyses, but in addition many studies apply best-/worst-case scenarios to examine the potential cost-effectiveness of smoking-cessation interventions. Probabilistic sensitivity analysis (PSA) was not widely reported in the literature. Discounting of future outcomes and costs was a common consideration in sensitivity analysis, as one of the challenges facing public health interventions, where the payoff from current investment may be many years into the future, is the fact that discounting reduces the estimated magnitude of health gains. For example, Coleman et al. 101 highlight that 1 QALY today is worth approximately 0.25 QALYs in 40 years’ time, where a discount rate of 3.5% per year is applied.

Conclusions/reflections

The literature overview presented above outlines how the current literature has been used to inform the structure and development of a CEA model, and to inform the identification of data to populate a model, for application in estimating the cost-effectiveness of the EARS intervention.

The review here has not identified modelling studies that have put emphasis on interventions with a specific focus on PA as a key component of the intervention, as is the case with EARS. This may be due to the nature of the literature review and its consideration of specific smoking-cessation interventions, rather than searching more broadly on PA. However, as the main outcome of interest for EARS is the quit rate, the model set out for EARS (in the next section), draws on the previously published models for smoking cessation. It may be that in a future, more comprehensive systematic search of the literature, and in further development of the modelling framework set out in this report, the use of PA data could be integrated into a more complex model structure.

A specific consideration for EARS is that the target population is a disadvantaged target treatment population. The models identified to inform on the smoking-cessation analyses were not in disadvantaged populations, nor is there a specific focus in the evidence base around abstinence, relapse or mortality, on a hard-to-reach population. In only one of the models described above is there a focus/reference to a specific target population, with Bauld et al. 110 considering NHS smoking services in Glasgow, highlighting higher levels of social deprivation in its description of baseline characteristics of clients entering one of two community-based interventions. Therefore, the modelling development in this report draws on the current evidence that relates to a broader smoking-cessation target group.

A further feature of the EARS intervention is the targeting of smokers who do not wish to make an abrupt quit attempt (as trial inclusion criteria). In our review of the literature we identified only one study15 that explicitly examined the costs and consequences of ‘cutting down’. In that study, Wang et al. 15 did not consider the long-term benefits of cutting down, in terms of additional benefits to life-year gains and QALY gains.

In the literature considered here on modelling methods for smoking cessation, there is a reliance/emphasis on use of data on mortality by smoking status from Doll111,113 and on data from Fiscella and Franks106 to inform QALY values. The recently published study by Vogl et al. 126 provides a useful contribution/addition to the literature on health-state (QALY) values by smoking status. Vogl et al. 126 publish health-state values by age and sex and smoking status based on data obtained in the Health Survey for England119 through use of the EQ-5D (generic preference-based health status measure).

In the following section we document the development of a smoking-cessation model, a simple model to inform exploratory analyses and to set the foundations for future cost-effectiveness analyses for EARS, alongside a future full RCT. Using the findings from the EARS pilot trial (as presented in Chapter 3 and Chapter 4) we also present illustrative CEAs.

Statement of the problem/objective of research

A modelling framework is needed to predict longer-term outcomes from end points in smoking-cessation trials, in this instance to predict longer-term outcomes from effectiveness data in the trial comparing the EARS intervention with brief advice. The model set out here is developed to estimate the impact of smoking status over time on mortality and QALYs, in order to inform on estimates of the cost-effectiveness of the EARS intervention compared with brief advice. The model developed and presented here is a simple model, aligned to the current exploratory research and pilot RCT of the EARS intervention. The model development research set out here seeks to inform on a framework for CEA that will be appropriate for a future CEA alongside a full RCT of the EARS intervention.

Perspective/viewpoint

The model developed is aligned to a UK policy context, and the primary perspective of the analysis is that of the third-party payer, that is to say UK NHS and Personal and Social Services, consistent with the commonly applied analytical ‘reference case’ recommended by NICE in the UK. 129 The framework developed here can be applied to a broader analytical perspective, where additional detail, complexity and data inputs are added to the simple model presented here, for exploratory purposes.

Model type and rationale for structure

A Markov-type model has been developed to estimate the cost-effectiveness of smoking-cessation interventions. The model structure has been informed by a review of the literature describing modelling methods used in assessment of the cost-effectiveness of interventions in smoking cessation (discussed/presented above). The model developed here is a simple Markov model, a cohort simulation model, with an annual cycle length and a lifetime time horizon for analysis (up to maximum age 85 years); both of these model characteristics are informed by review of the literature and are consistent with methods used for assessment of cost-effectiveness in smoking-cessation interventions. The model includes two health states for smoking status: either ‘smoker’ or ‘former smoker’, and death/dead (Figure 8). It is assumed at the start of the model that people will be in one of the two states for smoking status, with former smokers being recent quitters (i.e. at least 4 weeks post quit and 3–4 months post EARS intervention/comparator).

FIGURE 8.

Model structure/schematic for the EARS model.

The model considers a starting cohort of people, by age group (decile age bands) and sex, with a distribution across smoking status, either smoker or former smoker, and predicts future status over time (in each cycle) in these states, or death. For the CEAs, cohorts with differing start point distributions (all other characteristics being equal) are modelled over time, with cost and consequences estimated and compared between competing interventions.

The model is structured to allow transitions from former smoker (successful quitters) to smoker, using a risk of relapse in each of the first eight 1-year cycles of the model, and people can continue to be in states for ‘smoker’ and ‘former smoker’ (continued abstinence) in each cycle of the model. The current model does not allow for transition from ‘smoker’ to ‘former smoker’ (spontaneous quit) in each cycle of the model (as evidenced in Figure 8 – this could be included, e.g. 2% quit rate in literature, Coleman et al. 101). Each person faces a risk of death in each model cycle. The main driver in the model is differential mortality rates by smoking status, with comparison of life years in each health state, and difference in life expectancy and QALYs over time, being the focus of comparisons by differing starting distribution (smoking status) for subsequent CEAs.

The primary outcomes, and economic end points, of interest in this development modelling are number of long-term quitters (12 months and beyond), that is to say cost per quitter, the cost per life-year gained, and cost per QALY gained. The model therefore aims to estimate, in the first instance, smoking status over time, life-years and QALYs in the cohorts considered for comparison in CEAs.

The model does not model smoking-related morbidity directly; however, people are given a health-state value, according to smoking status, age and sex, using published data (see discussion of Vogl et al. 126 in Health-related quality of life/quality-adjusted life-years, below), and in such a way differences in health-related quality of life/QALYs are estimated over time, by cohort distribution (smoking status). The model does not estimate costs associated with smoking-related morbidity, although there is an opportunity to develop the model in this direction for future analyses. In the present model, sensitivity analyses consider differential costs by comparison interventions through the use of a cost associated with death, smoking-related death, where estimated costs are assumed to capture additional health care expenditures associated with treating smoking-related diseases.

Modelling smoking status over time

The framework used here models smoking status over time. At the start of the model, the cohort entering the model is defined by proportions (distribution) in states for ‘smoker’ and ‘former smoker’. The model is developed here to support analyses for the EARS smoking-cessation intervention, as described in this report. In applying the modelling framework to EARS and comparators, in a decision-analytic context, we have a distribution (comparison of differing starting distributions) of people entering the model where a proportion will be recent quitters, that is to say former smokers. In the current analyses data from the end point of the EARS pilot RCT is used to inform the starting distribution of smoking status for cohorts modelled over time. In the first 12 months of the model, former smokers face a relatively high risk of relapse (to smoker); thereafter, in years 2 to 8 of the model, these former smokers face a continued, but relatively lower, risk of relapse. Data to inform transition from former smoker to smoker have been taken from the current literature,101,117,121 informed by our review of the literature on modelling methods.

Relapse/continuous quit rate at 12 months

In order to estimate the continuous quit rate at 12 months, and the related relapse rate, we use data from the published literature, applying data from Coleman et al. 101 in the base-case analyses. In practice, we do not know how many successful quitters within the EARS trial reported at 4 weeks post quit will relapse at 12 months (a future long-term trial may inform on this). Coleman et al. 101 report a high-quality systematic review, presenting estimates of abstinence rates and relapse patterns in smokers who had made pharmacologically aided quit attempts with NRT, bupropion and varenicline (see previous discussion). In the current model we use a relapse rate of 28% for the first 12-month cycle of the model. This input/assumption is informed by data from Coleman et al. 101 on relapse (continued abstinence) in RCTs reporting on NRT. Data on difference in abstinence between 3 months (post intervention) and 12 months is used to inform the current analyses (see Coleman et al. ; Table 35; Figure 7), with a relapse rate of 28% applied, taken from data on five RCTs for NRT. Coleman et al. 101 report a relapse rate of 40% (years 2–8) in combined analyses, using 16 RCTs; however, in some of these RCTs the intervention was over a duration of up to 52 weeks; this latter estimate is used in sensitivity analyses.

Relapse/long-term quitters beyond 12 months

Data to inform on longer-term smoking-cessation status, and relapse, beyond 12 months are sparse. A high-quality literature review has been reported by Etter and Stapleton. 121 From the literature review, we identified long-term continuous cessation data (8 years post quit) from a study by Yudkin et al. 117 who report a longitudinal study of 1686 trial participants (NRT) based in Oxfordshire, UK. Yudkin et al. 117 report a relapse rate of 46% for former smokers to smokers, over the period after a 1-year trial, up to 8 years (the rate was 44% for placebo and 47% for those in the NRT arm of the trial). The median time to follow-up was 8.3 (SD 0.35) years. Of the 153 participants who had stopped smoking for 1 year in the trial, 83 were still not smoking at follow-up, giving a relapse rate of 46%. The study assumed that all those lost to follow-up were still smoking. We use the data from Yudkin et al. 117 (46% relapse) in the base case of the model. In sensitivity analyses we apply data from the meta-analysis reported by Etter and Stapleton,121 which includes the study by Yudkin et al. ,117 in which a relapse rate of 30% is reported, over an average follow-up of 4.3 years (after year 1).

In order to apply these data, on rate of continued abstinence and on rate of relapse over years 2 to 8, in the framework of the decision model developed, using a 1-year model cycle, we derived a profile for continued abstinence and relapse rate by year over years 2 to 8. Data were derived using an exponential survival function for remaining smoke free (time to event analyses). It would be possible to use other functional forms here but the rationale for using the exponential function is that it is consistent with evidence suggesting that the proportion of quitters follows a decreasing trend.

We calculate the hazard or (instantaneous event rate) within the 7-year period of follow-up as h = –[ln(S_m/S₀)]/(t_m – t₀)] (Box 1), representing a constant hazard rate over years 2 to 8. The hazard rate for relapse is applied to the number of people in the health state of ‘former smoker’ from 12 months. Applying these methods in the modelling framework provides outputs over time (number who relapse and revert to smoking) that are consistent with the implicit relapse rate used from Yudkin et al. 117

The model framework assumes that smokers who are abstinent at 8 years post quit would remain abstinent. The model does not allow non-smokers [or cut-downers (at trial follow-up)] to make a subsequent quit attempt if they failed to do so within the prior trial (prior to entering the model). Both assumptions are potential limitations with this model and studies of this type. Any future modelling, alongside a full RCT and economic evaluation, will explore the impact of these assumptions. The current literature commonly refers to the spontaneous quit rate at around 2% per year (see Coleman et al. 101). However, it is expected that in the EARS target population, of hard-to-reach smokers who are not planning a quit attempt, spontaneous quit rates will be relatively low, and the differences between comparisons (EARS vs. brief advice in current context) will be very small/negligible.

Modelling mortality over time by smoking status

The model estimates life expectancy over time, survival in each model cycle, using methods presented by Coleman et al. 101 for estimating mortality by age, sex and smoking status (see the earlier evidence review for discussion/detail of the approach used by Coleman et al.). Methods presented by Coleman et al. 101 have been adapted for the current model, in order to use data from Doll et al. 113 (Coleman et al. 101 used data reported by Doll et al. 111). As discussed earlier, Doll et al. 111 report mortality in British doctors over time by smoking status. We adapt methods in order to model mortality by smoking status, and to model mortality rates dependent on time since quit, in order to model the relation between sustained abstinence (in deciles age bands) and smoking-related mortality.

The estimation of mortality data by smoking status uses (1) current mortality rate data, by age group and sex and (2) data on the prevalence attached to smoking status; the data from Doll et al. 113 on mortality by age and smoking status is used to represent the ratio of mortality rate by smoking status, the latter analogous to a RR of mortality by smoking status and age group (and here referred to as RR). Mortality is modelled from age 35 upwards, using age groups reflecting 10-year age groups (e.g. 35–44 years), aligned to the data reported by Doll et al. 113 (Doll et al. report that mortality by smoking status below this age is subject to wide random variation. Evidence from actuary life tables suggest that total mortality prior to the age of 30 is low, in the region of < 1 per 1000, and that while mortality rates between smokers and non-smokers begin to diverge at this age the divergence in mortality rates between former and current smokers starts at age 35) (Table 37).

As the approach described by Coleman et al. ,101 RRs are derived from data presented by Doll et al. 113 to reflect relative mortality rates by smoking status (see Table 37). The RRs derived in the current analyses, for former smokers, distinguish between the age of the quit attempt in terms of deciles since quitting. These data are combined with mortality data reported in current life tables, sorted by age decile, in the general population (see Table 7) (www.ons.gov.uk/ons/rel/lifetables/decennial-life-tables/no-16-–2000–2002-/index.html) and data on the prevalence of smoking for each age and sex category, using data presented in the Health Survey for England,119 (see Table 40). Data were not available, to our knowledge, on prevalence of former quitters (in terms of time in deciles since quitting in general population statistics), and therefore we apply a simplifying assumption, that prevalence is uniformly distributed between quit ages 35–44, 45–54 and 55–64 years.

The above data for former smokers can be presented using years of smoking abstinence as the independent variable. This cross-tabulation facilitates the calculation of time dependent state transitions based on years since quitting.

Applying the methods used by Coleman et al. ,101 we estimate/derive the mortality rates for use in the model by smoking status using the above data. We model mortality data for each potential combination of age, sex and smoking status category, using the formula below. This is an adaptation of the formula, from Coleman et al. ,101 described earlier.

For each age and sex population mortality rate, Q, we use the Coleman et al. 101 formula to derive mortality rate by smoking status:

As applied in Coleman et al. ,101 in order to solve for mortality risks associated with current (E_c), non (E_n), and former smokers (E_f) we define ratios for mortality rates between smoking status categories as E_n/E_c = RR_n and E_f/E_c = RR_f.

We adapt this formula to estimate mortality rates for former smokers according to time since quit, using the data reported from Doll et al. 113 (see Table 37). In the younger age group(s), those aged 45–54 years (and below), former smokers will have quit smoking too recently (1 year < quit time > 10 years) to derive statistically significant variations in mortality for the current ages, while for older age groups, such as those aged 55–64 years, former smokers will include those who have quit smoking for less than 10 years (1 year < quit time < 10 years) and for between 10 and 20 years. (Note: the benefits of early quitting are realised in terms of higher mortality rates as the cohort ages within the model. See the following section for a description of how these rates are implemented in the model.)

To illustrate the approach taken, consider the example of a female who is 55–64 years old. The mortality rate for former smokers in this formula, E_fP_f, can be represented by the following expression:

where time since quit is represented by t₁ < 10 years and 10 < t₂ < 20 years.

Substituting (2) into (1) gives

Information on Q is taken from (given by) mortality life tables; in addition, data on the expected ration of mortality rates by smoking status (RR) are available from Doll et al. ,113 and therefore by adapting the methods presented by Coleman et al. ,101 it is possible to solve the equation for E (by smoking status) by replacing RRs by smoking status relations, for E_n and E_f, solving for E_c.

Substituting E_n = E_cRR_n and E_ft1 = E_cRR_ft1 and E_ft2 = E_cRR_ft2:

and rearrange to

Using date from the above tables (female aged 55–64 years), when P_c = 0.19, RR_n = 1, P_n = 0.36, RR_ft1 = 0.7664, P_ft1 = 0.22, RR_ft2 = 0.4206, P_ft2 = 0.22, P_ft and Q = 0.0026, then E_c = 0.004389; and substituting back into equation (13) solves for E_n = 0.3925, E_ft1 = 0.003363 and E_ft2 = 0.001846. Using these data, the mortality rate for a former smoker in this category is E_f = E_ft1P_ft1/P_f + E_ft2P_ft2/P_f.

Using the above methods, we derive estimates of mortality by smoking status for age and sex age bands, stratified by time since quit for former smokers (Table 41).

Health-related quality of life/quality-adjusted life-years

As described above, the model developed and used here predicts mortality over the (cycles) time horizon, up to maximum age of 85 years. In addition, in order to consider health-related quality of life over time, using the QALY approach, for each year in the model a health-state value is attached by smoking status, and according to age and sex characteristics. The health-state values are used as QALY weights to reflect health-related quality of life in each smoking category. Data used on health-state values are from a recent study published by Vogl et al. 126 Vogl et al. 126 also present date by smoking intensity, and the base-case inputs used in the current analyses are for ‘heavy smokers’. Vogl et al. 126 estimate health-state values through use of the EQ-5D, the generic preference based measure recommended by NICE for their ‘reference case’ analyses. 129 Table 42 presents data on health-state valuations from Vogl et al. 126

The above rates are used to inform risk of death in each 1-year time cycle of the model, and to therefore estimate mortality over time (in a cohort analysis).

Smoking attributable (related) mortality

In sensitivity analyses, where the model is used to inform CEAs, we use an estimate of ‘smoking-related’ mortality to capture potential costs associated with smoking-related morbidity. In order to determine mortality attributable to smoking and not to other causes, the excess risk of death for a smoker over and above a non-smoker is calculated by subtracting E_n (risk of death non-smoker) from E_f (risk of death former smoker) and E_c (risk of death current smoker), for each category. This allows the model to estimate smoking attributable mortality. This outcome also allows calculation of the cost of smoking-related disease leading to death (see ‘Costs’).

Intervention(s)

The interventions compared in the CEA are the EARS intervention, versus brief advice. Interventions have been described in detail earlier in this report (see Chapter 1).

Effectiveness: comparing alternative intervention strategies

The model as simply set out produces estimates of outcomes (and costs) over time for a given cohort of people described by age, sex and smoking status. The distribution by smoking status will differ where intervention strategies differ, assuming that people are subject to an intervention prior to the modelling of longer-term outcomes, and associated costs. The model is configured in this way for the development stages of research; however, it is possible to integrate intervention strategies within the first cycle of the model where appropriate. In this simple modelling framework, outcomes over time are compared in different cohort by intervention scenarios, yet it is possible, in developments of this model, to introduce other mechanisms for comparing effectiveness (e.g. applying adjustment, such as a relative risk, to transit probabilities in a base/control cohort).

Base-case cohort characteristics

The modelling framework is developed to predict outcomes, and related costs, associated with a given starting distribution for smoking status for a cohort (defined by the decision-making context) of people over time. The modelling framework estimates average (expected) outcomes against specific baseline characteristics; in the current framework it is by age and sex groups. In the current exploratory analyses we run the model for the distribution of smokers and ex-smokers reported at the final follow-up of the EARS pilot RCT, and also for a cohort with the distribution by smoking status reported in the control participants having brief advice (the EARS trial reports the number of participants achieving expired air CO-confirmed abstinence at 4–8 weeks post quit date). Table 43 presents these starting distributions for smoking status. We undertake analyses by age and sex stratification to represent these population data available to inform the model and to reflect policy-relevant analyses. However, we apply the trial end point distributions uniformly across the age-/sex-specific analyses. In future analyses it is possible to use age-/sex-specific results, and to model a weighted average result, based on stated demographics in the overall population cohort (e.g. using population weights from national data).

Costs

The additional cost associated with delivery of the EARS intervention has been estimated at a mean cost of £192 (described in Table 33). This is the only additional cost included in this exploratory CEA. The intervention cost reflects the additional resource use and cost associated with delivery of the intervention. We introduce uncertainty around this cost estimate in the probabilistic CEA, reported in the cost-effectiveness plane and acceptability curves, through the use of a potential range of costs £154 to £230, reflected via a standard error input of £19.61.

The model framework here does not include a description of, or modelling over time of, smoking-related morbidity. These costs can be considered in future development of the model, for example using the methods applied and described by Coleman et al. 101 In the current analyses, a simple and crude estimation of costs associated with smoking-related morbidity is introduced in a sensitivity analysis where death attributable to smoking is associated with an increased cost. The cost of smoking-related death used in sensitivity analysis is a mean cost of £27,120 as reported by Bauld et al. 110 This was applied to the smoking-attributable deaths in both the comparator and intervention scenarios.

Discounting

Future outcomes and future costs (beyond the first year) are discounted at 3.5% in the base-case analyses, consistent with common practice in the UK, and with the reference case for analyses recommended by NICE. 109 Sensitivity analyses also report estimates where future costs and outcomes are not discounted. Note that intervention costs were not discounted as they were accrued in the base year.

Cost-effectiveness analyses: methods/presentation

In order to estimate the relative effectiveness and cost-effectiveness of EARS versus brief advice, we compare the outcomes from the model, quit rate over time, life-years and QALYs over time (up to a maximum age of 85 years) and we combine incremental effects with incremental costs to present the ICER associated with the comparison of interventions, for the base-case scenario and for scenarios considered in sensitivity analyses. The formula for the ICER is:

Results are presented in a tabular form for comparative analyses, showing data in a disaggregated form (by intervention), and as incremental costs and effects.

Analyses are undertaken using deterministic and probabilistic methods. 130 Probabilistic analyses are used to propagate uncertainty associated with parameter inputs simultaneously, in order to reflect joint uncertainty in model inputs. This is undertaken in an exploratory way in the current analyses/framework, applying data on distribution (standard error) by parameter input, and using assumptions on the nature of the distribution (e.g. normal, beta or gamma distributions). Appendix 9e reports these data inputs with distributions used for probabilistic analyses.

Estimates from probabilistic analyses allow presentation of CEA results via the cost-effectiveness plan, and via cost-effectiveness acceptability curves (CEACs). In these presentations of these data we use the net-benefit statistic,130 alongside assumptions on decision-maker willingness to pay per QALY gained, to represent the proportion of analyses (in simulations undertaken) where the net-benefit statistic shows the intervention to be cost-effective, that is to say where the incremental cost per QALY is equal to or lower than the decision-maker’s willingness to pay. These methods are consistent with good practice in economic evaluation, as presented by Drummond et al. 131

Sensitivity analyses/assumptions

Sensitivity analyses are reported to assess uncertainty in model parameter inputs. Probabilistic sensitivity analyses capture uncertainty in input parameters, although many input parameters have small standard errors (see Appendix 9). To address uncertainty over the choice of input parameters we set out a range of sensitivity analyses, one-way and multiway scenarios, where alternative data inputs and/or assumptions are used. The key areas for sensitivity analyses (to date, can do others) are:

1. start point distributions by smoking status (base case informed by EARS pilot RCT)

2. relapse rate in first 12 months post quit (base case at 28%)

3. relapse rate over years 2–8 (base case derived from rate of 46% over 2–8-years)

4. use of discount rate of 3.5% in base-case analyses

5. estimate of EARS intervention cost (base case at £192)

6. assumption (simplifying) that there are no smoking-related morbidity costs

7. alternative QALY weight data used.

Other key simplifying assumptions, not subject to sensitivity analyses in current exploratory modelling/results, include:

1. no spontaneous quit rate included in the model (focus on EARS effectiveness only)

2. time horizon up to maximum of age 85 years

3. no specific smoking-related morbidity events/costs included

4. simple method used for estimating smoking-attributable mortality

5. health-state (QALY) values from Vogl et al. ;126 no sensitivity analyses undertaken.

Sensitivity analyses

In order to consider the sensitivity of cost-effectiveness results to inputs and assumptions, sensitivity analyses have been undertaken to assess alternative data inputs/assumptions. Tables 47 and 48 present a range of sensitivity analyses, with CEA results (i.e. a cost-effective profile) being robust to most changes/variations in model inputs and assumptions.

Table 47 presents results in which no discounting of future health outcomes has been included, presenting a greater mean health gain over time, especially relevant in analyses such as these where expected gains are accrued over the longer-term time horizon.

Table 48 presents sensitivity analysis in which the mean participant cost is assumed to be much higher, at £500, compared with a base case of £192, with EARS versus control continuing to show a cost-per-QALY range of between £9400 and £20,000 across age and sex groups.

Tables 49 and 50 present sensitivity results where the effectiveness of EARS versus control has been varied by ± 25% through adjustment of expected starting in which by smoking status, compared with control with 4% former smokers. These analyses indicate that results are robust to these variations. In later scenario analyses, exploratory threshold analyses examine this area further.

Tables 51 and 52 present results from sensitivity analyses in which different data sources have been used to inform the relapse rate, in the first 12 months (see Table 51) and over years 2–8 (see Table 52). The adjustment to the 12-month relapse rate, from 28% to 40%, using data from Coleman et al. 101 is a more pessimistic assumption, while the adjustment to the longer-term relapse rate, from 46% to 30%, using data from Etter and Stapleton,121 reflects a more optimistic assumption. In univariate analyses, neither of these assumptions shows a big impact on the overall estimates of cost-effectiveness. These rates are applied uniformly across comparison interventions, to a relatively small proportion of the overall cohort, and so mean estimates of difference have not shown sensitivity.