Relapse prevention in UK Stop Smoking Services: current practice, systematic reviews of effectiveness and cost-effectiveness analysis

Smoking as a drug dependence

There is strong evidence of pharmacological, as well as psychological, dependence on cigarettes. 1 Nicotine obtained from cigarettes meets all the standard criteria used to define a drug of addiction or dependence and most smokers continue to smoke because they are addicted to nicotine. Smoking is therefore a difficult behaviour to change, often taking several attempts over a period of years before permanent cessation is successfully achieved and as such, can be seen as a chronically relapsing behaviour. This is similar to other drug taking behaviour that persists in the face of serious negative health consequences. 7 Hence, although the majority of smokers report wanting to stop, each year only around a third make a quit attempt and only 2% succeed. 1

The process of relapse

Attempts to stop smoking are characterised by a period of initial abstinence followed by an extended period when the abstainer is at a high risk of relapse, but this risk reduces over time. 8 The risk of relapse is greatest in the first few weeks after quitting when withdrawal effects peak, hence relapse to smoking occurs quickly, with many smokers not even managing to stay abstinent for 1 day. 9 Lapses (defined as isolated events or slips followed by a renewal of abstinence) predict subsequent relapse (defined as fully going back to smoking), and those who maintain abstinence for the first 2 weeks are more likely to be abstinent 6 months later. 10 After the first few weeks of abstinence, withdrawal symptoms reduce and confidence in remaining abstinent increases. 9 However, abstainers continue to relapse for months, even years, after the quit attempt, although a recent study of natural relapse patterns in different populations indicated that relapse dropped to around 5% after more than 2 years. 11

The process of relapse is not yet fully understood, and theoretical models focus on specific mechanisms of drug motivational processes (e.g. Herd),12 which increasingly recognise relapse to smoking as a dynamic process difficult to predict and prevent, and likely to involve a complex interaction of factors. 13 Individual factors, such as more dependent smoking, craving, self-efficacy, perceived benefits of smoking, and cues, such as the presence of smokers, situations or behaviours associated with previous smoking, as well as negative effects (e.g. stress), have been shown to trigger lapses and relapse. 14–16

Little is known about how to prevent relapse. Early theorists12 believed that abstainers should try to recognise the high-risk situations in which they are likely to relapse and then learn to use cognitive and behavioural strategies to cope with these situations. Over time, the use of effective medications has also been added to these strategies. Shiffman13 recently summarised current thinking suggesting that treatment needed to ‘imbue the person with the ability to respond more effectively to a wide range of situational challenges’. However, very few interventions have been shown to be effective for relapse prevention, an issue we discuss further below. 17

The need for support with quitting and the NHS SSS

In contrast to the evidence base for relapse prevention treatments, evidence for the use of acute cessation treatments has grown rapidly over recent decades and a variety of effective treatments now exist which can increase the success of stopping smoking up to fourfold18 compared with no support. In the UK, NHS SSS were set up to offer evidence-based treatment to smokers wanting to stop, in recognition of the fact that many smokers needed such support and although these services are cost-effective, relapse is still the most common outcome for those using them. 6

The English Department of Health (DoH) publishes guidance for the NHS SSS to ensure they target their efforts at those most in need, in particular routine and manual smokers, and deliver the most effective treatment,19 referring to guidance on efficacy. 20 The NHS SSS aim to reach as many smokers as possible through health professionals’ brief interventions, their onward referral to the NHS SSS of those who need more support and also with smokers’ self-referrals. 19 In general, the NHS SSS offer flexible treatment, which usually involves the most effective combination of individual or group behavioural therapy, backed-up by the offer of smoking cessation medications, such as nicotine replacement therapy (NRT), bupropion and, more recently, varenicline. 19

Historically, the success of NHS SSS has been measured by monitoring the numbers of smokers who access services and set dates for quitting smoking, and, of these, the numbers who are not smoking 4 weeks after their quit dates; challenging primary care trust (PCT) targets have been set for delivery against these measures. 21 Recent guidance indicates an aspiration that, in time, at least 5% of smokers from areas served by NHS SSS should set quit dates with services’ help annually but, currently, monitoring of ‘4-week quitters’ is still required. Although services are encouraged to offer, for at least 4 weeks after smokers’ quit dates, the most effective cessation treatments (including behavioural support), there is little contemporary information on the treatments actually offered. Also concerns have been expressed that the focus on 4-week quit rates militates against the provision of support beyond this initial period, despite the possibility that treatments like relapse prevention interventions might sustain successful quit attempts in the longer term. 3 In addition, although the services are encouraged to monitor longer-term success at 1 year,19 their funding is not dependant on this and many services find such monitoring difficult in practice. For these reasons the focus on NHS SSS delivering short-term quitters is likely to deter services from offering relapse prevention treatments.

Given the addictive nature of smoking, this focus on short-term therapy is at odds with the extended duration of therapy and follow-up recognised for other drugs of dependence. 22 Understanding relapse, methods to reduce relapse and the feasibility and effectiveness of introducing these in NHS SSS are therefore important issues to explore.

Qualitative research

Qualitative research

Of the 23 individuals who had indicated their willingness to participate in the study, four could not be contacted and three potential participants indicated that they no longer wished to participate, leaving a total of 16 individuals with whom interviews were conducted (12 with females). Fifteen interviewees were smoking cessation professionals who were also responsible for overseeing the day-to-day activities of their respective services or actively involved in managing one aspect of the service in which they worked and one was a regional tobacco control lead.

Three broad themes emerged from interview data: beliefs, knowledge and understanding of relapse prevention; RPIs for abstinent/lapsed and relapsed smokers, and barriers and challenges (Box 2).

The following sections describe these themes in more detail. Quotes are reported in Boxes 3–6 and are attributable to individual interviewees by code numbers.

Summary of findings

Strengths and limitations

To our knowledge, these are the first studies to explore smoking cessation professionals’ experiences of, beliefs about, and current provision of RPIs within the NHS SSS or indeed in any other health system and the findings have highlighted relevant issues that could hinder or facilitate the introduction of RPIs into the NHS. Our qualitative research sample was small and it is possible that by interviewing a larger, more diverse group, we would have found other important factors relating to the use of RPIs, however, similarities inherent in the accounts we obtained, suggest that we have identified most of the major issues. Also, as our sample comprised volunteers from those attending a smoking cessation conference, one might expect their views on RPIs to be better informed than those of others working in the field. Our survey results are based on smoking cessation managers’ self reports and may have overestimated the rates of RPI provision; for example, managers may have been unwilling to reveal that they do not provide some form of RPI in their services. However, we believe that biases in the reporting among this sample are an unlikely explanation for our findings, given the lack of current evidence on the effectiveness of RPIs at the time the survey was carried out. The survey is also limited by the response rate as just over half of managers completed the online questionnaire. A greater response rate may have produced more varied results; non-responders to the survey may have different experiences of providing RPIs in their services, although a comparison of the two waves of responses to the survey showed that there were no differences in variables relevant to the focus of this report.

Important emergent issues

The varied descriptions of RPIs in the qualitative research indicated that health professionals working in the NHS SSS did not share a common definition of what these interventions should involve. While some believed relapse prevention should be aimed at lapsed smokers, more commonly it was perceived appropriate for preventing relapse among smokers who were still abstinent and attempting to quit, in concordance with the definition we use in this report. However, often no distinction was made between RPIs and routine cessation treatment. It is, perhaps, unsurprising to find differences of opinion about RPIs among health professionals, because the scientific literature on the subject at the time the research was carried out described studies which conceptualised relapse prevention in different ways too23 (as discussed in the next chapter).

When provided with a definition of RPIs, over half the managers in the survey reporting providing them, despite the then weak evidence base,23 and two-thirds reported being likely to provide them in the future. This enthusiasm may have reflected a desire to improve support provided to smokers, who have a high relapse rate from acute cessation treatment, or their anecdotal experience of the effectiveness of RPI, or a perception that smokers appreciate being provided with RPIs. While this wasn’t explored in our research, there was little support for the latter explanation as participants in the qualitative research expressed concern about the low uptake of RPIs, which had often led to these being withdrawn, and was perceived as a major potential barrier to their successful introduction. They also thought that relapsed smokers were embarrassed about their ‘failure’ and hence did not return for treatment. This view is consistent with the literature; lapsed smokers often feel guilty12 and experience negative affect and decreased confidence in their ability to quit. 29

Managers favoured behavioural intervention over drug treatments and, at the time of the survey, there was insufficient evidence to say whether either was, indeed, effective. However, our subsequent systematic review (see Chapter 3), found drug treatments and not behavioural ones effective for smokers who achieve abstinence after receiving support, as those using NHS SSS do. This suggests that overall findings from this report could help the NHS SSS clinical practice to become more evidence based. Survey respondents preferred telephone follow-up of clients for behavioural RPIs, though there is no evidence that this is effective and, conversely, the qualitative interviewees found little or no enthusiasm for this, citing that few abstinent smokers welcomed this type of contact that was also logistically challenging to undertake.

We examined whether the preference for the provision of behavioural treatments reflected provision of acute cessation treatment. While this could explain the use of telephone support (81% of services provided telephone support as acute cessation treatment), this explanation does not clarify the disparity between the high percentage of services offering bupropion and varenicline for acute cessation treatment and the low percentage offering these pharmacotherapies for relapse prevention, despite at the time some, albeit weak, evidence for their effectiveness. 23 In the absence of any barriers, individual and group counselling were also favoured for relapse prevention, while the pharmacotherapies were the least favoured. However, when provision of RPIs was explored in the qualitative research, interviewees believed that pharmacotherapy used as extended treatment for relapse prevention could be easily integrated into current routine practice, as long as costs were addressed.

In addition to cost, the other substantive perceived barrier to any form of RPI provision were the current performance management targets for the NHS SSS. For example in October 2002, the English DoH stated an aim that 800,000 smokers would successfully quit following help from smoking cessation services by 200621 and this was subsequently translated into an aspiration that services should treat at least 5% of their local population of smokers in the course of a year, with an expected validated success range of 35% to 70% at 4 weeks. 1 For these targets, success involves smokers achieving a 4-week period of abstinence, so there is considerable pressure on services to achieve a high throughput of smokers who manage to stop smoking for at least 1 month. Interviewees’ comments reflected this perceived pressure and indicated that the provision of RPIs, which, if effective, could be expected to improve longer term but not short-term quit rates, could not be a priority for services without a change in the targets against which they are measured.

The survey data indicated that current provision of acute cessation treatment largely reflects UK guidance,30,31 and our survey suggests that similar evidence-based guidance is needed to support managers wishing to provide RPIs in order to ensure that, in the future, the most promising and effective RPIs can be introduced into routine care, assuming that any perceived barriers are removed.

Search strategy

The search strategies undertaken were identical to those used in the 2005 Cochrane review23 and intended to update this and so were conducted between 2004 (when searches for this review concluded) and July 2008. In addition to searching the Cochrane Tobacco Addiction Group register of trials, we also searched MEDLINE, the Cochrane Central Register of Controlled Trials (CENTRAL), EMBASE, PsycINFO, the Science Citation Index and Social Science Citation Index. To identify grey literature we also searched the abstracts of the annual meeting of the Society for Research on Nicotine and Tobacco, contacting authors for further information about their research, where relevant. The bibliographies of retrieved references were also scanned for further relevant publications and studies in all languages/from all countries were considered. Full details of search strategies used are documented in Appendix 3.

The titles and abstracts identified from searches were assessed independently by two authors (TC and AMcN). Studies were excluded if it was clear that the study did not refer to a randomised controlled trial (RCT) of an intervention used for relapse prevention. Two authors then independently assessed each study to determine whether it met the pre-specified selection criteria, with any difference being resolved through discussion with remaining authors.

Inclusion and exclusion criteria

Main outcome measure

The primary outcome for this review was abstinence from smoking, ascertained preferably as continuous abstinence, with point prevalence abstinence used if this was not recorded.

Data extraction strategy

All authors participated in data extraction with two authors working independently on each paper, using a specially designed data extraction form and differences being resolved through discussion. The data extracted included:

• Description of study (design, duration of trial, unit of randomisation, unit of analysis).

• Description of treatments (type, dose, duration).

• Participants (inclusion and exclusion criteria, demographic data, number lost to follow-up).

• Outcome measure (primary and secondary).

• Results (abstinence at 1–3 months, 6–9 months and 12–18 months).

Attempts to obtain missing data from the authors of the manuscripts were made, where possible. Data were checked and entered into review manager, version 5, by one author (SA).

Quality assessment

The methodological quality of the included studies was assessed using the Cochrane Collaboration’s recommended tool for assessing the risk of bias which evaluated aspects of trial design: methods of randomisation sequence generation and allocation concealment; blinding; adequate assessment of incomplete data; comparability of groups at baseline; and whether treatments were adequately described.

Data analysis

For studies with similar types of intervention, a meta-analysis was performed to calculate a weighted intervention effect across trials using a random effect (DerSimonian and Laird) model and results are expressed as pooled ORs with 95% CIs. Numbers needed to treat (NNT) for additional beneficial outcomes were calculated using pooled baseline event rates, where significant comparisons were seen. Statistical heterogeneity was assessed using I² and where heterogeneity levels were detected (I² > 80%), studies were summarised individually. 32 Follow-up time points were defined as follows: (1) short term, 1 month (permitted range 1–3 months), (2) medium term, 6 months (range 6–9 months) and (3) long term, 12 months (range 12–18 months), with the long-term follow-up point considered as the primary end point time. In trials of pregnant/postpartum women, abstinence was also reported at delivery or the last follow-up prior to this. Where there were multiple intervention groups within a trial, pair-wise comparisons were made for each active intervention versus control, with the results in the non-control being divided by the number of pair-wise comparisons, so participants in the control group were not double counted. Funnel plots were used to assess publication bias.

Number of studies identified

The search strategy generated 1598 potentially relevant references and after title/abstract assessment 144 papers were retrieved and examined; results of this process are summarised in Figure 1.

FIGURE 1.

Flow chart of study retrieval process.

Included studies

Fifteen studies met our inclusion criteria, and these were added to the 39 studies from the earlier Cochrane review,23 giving 54 studies in total that are summarised in Tables 4–10; 18 randomised smokers and 36 abstinent smokers.

Excluded studies

Details of excluded studies are summarised in Table 3; these were retrieved and examined, but failed to meet one or more of the inclusion criteria in terms of study participants, type of interventions and type of outcome measures. The 2009 Cochrane review17 excluded one study which we have included;39 this randomised abstinent smokers to receive 5-mg or 20-mg rimonabant or placebo for 42 weeks after 10 weeks of open-label treatment with rimonabant. We excluded two studies which are included in the Cochrane review:17 one randomised abstinent smokers who had lapsed, rather than smokers or abstinent smokers61 and the other included many participants who were not smokers and analysed their outcomes with smokers. 62

Comparison categories

To enable interpretation of findings and avoid heterogenic comparisons, we have divided trials into nine categories according to their participants (i.e. smokers or abstinent smokers) and the types of intervention delivered (i.e. pharmacological or behavioural and subcategories within each). Within each category, details of included studies, characteristics of interventions and outcome assessments are provided and the effectiveness of interventions is investigated, where appropriate, by meta-analysis. The 10 comparison categories are as follows:

Behavioural interventions in pregnant and postpartum women

Analyses

Interventions delivered during pregnancy

Pooled analyses of behavioural interventions delivered during pregnancy at delivery/the longest follow-up prior to delivery failed to detect a significant effect (pooled OR 1.18, 95% CI 0.85 to 1.62, I² = 0%, five studies; Figure 2). Three of these studies40,63,75 do not provide details of the exact follow-up point, and abstinence was measured at delivery. One study reported abstinence during the 34th week of pregnancy,75 while the last study in this category measured abstinence at 36 weeks. 72 Only one study in this category assessed abstinence at short-term follow-up68 and the intervention did not appear to have an effect on abstinence (OR 0.83, 95% CI 0.43 to 1.58). Pooled analysis of three studies at the medium-term follow-up was not performed due to significant levels of heterogeneity between the studies (I² = 81%). In all three studies individual behavioural counselling did not significantly appear effective at reducing relapse (OR 0.88, 95% CI 0.49 to 1.58),63 (OR 0.35, 95% CI 0.18 to 0.70)68 and (OR 3.45, 95% CI 0.95 to 12.62)70 respectively. Only one study reported data at long-term follow-up72 that found no significant effect with individual counselling (OR 0.76, 95% CI 0.36 to 1.57).

FIGURE 2.

Behavioural relapse prevention interventions in pregnant and postpartum women. (a) Individual counselling delivered during pregnancy at longest follow-up prior to delivery; (b) individual and telephone counselling delivered postpartum at long-term follow-up; (c) individual and telephone counselling delivered during pregnancy and postpartum at longest follow-up prior to delivery; and (d) individual and telephone counselling delivered during pregnancy and postpartum at medium-term follow-up.

Interventions delivered postpartum

Individual counselling

Individual behavioural counselling did not appear effective at the medium-term follow-up in one study (OR 0.83, 95% CI 0.51 to 1.34)74 but appeared effective in the long term in one large trial (OR 1.38, 95% CI 1.05 to 1.82). 73 No data were available for short-term follow-up in either study.

Individual and telephone counselling combinations

Two trials64,69 randomised participants to receive individual counselling plus telephone booster sessions. Pooled analysis of both trials failed to detect a significant effect at medium-term follow-up (pooled OR 1.33, 95% CI 0.94 to 1.90, I² = 4%), and at long-term follow-up (pooled OR 0.91, 95% CI 0.63 to 1.32, I² = 0%, Figure 2). No data were available for short-term follow-up.

Interventions delivered during pregnancy and continued postpartum

Self-help interventions and telephone counselling

A combination of self-help booklets (written materials intended to prevent relapse) and pre- and postpartum telephone counselling exerted a moderately significant effect on abstinence rates at short-term follow-up in one study (OR 1.60, 95% CI 1.15 to 2.23) and in the medium term (OR 1.60, 95% CI 1.16 to 2.21). 65 This effect was not present at long-term follow-up (OR 1.04, 95% CI 0.76 to 1.44).

Individual counselling

One study71 delivered individual behavioural counselling to participants during pregnancy and postpartum, but failed to detect evidence for an effect at the longest follow-up prior to delivery (OR 0.91, 95% CI 0.46 to 1.79) and at long-term follow-up (OR 1.02, 95% CI 0.53 to 1.96). No data were available for short- and medium-term follow-ups.

Individual and telephone counselling combined

Two trials66,67 examined the effectiveness of individual counselling followed by pre- and postpartum telephone counselling calls. The interventions did not appear effective at delivery (pooled OR 1.07, 95% CI 0.53 to 2.14, I² = 12%) or at medium-term follow-up (pooled OR 1.23, 95% CI 0.80 to 1.88, I² = 0%; Figure 2). Only one study reported data at short- and long-term follow-ups66 that found no significant effect at either follow-up.

Behavioural interventions in other population groups

There were 14 studies in this category41,76–88 and Table 5 summarises these.

Details of included studies

Two studies involved military recruits41,82 and eight studies community volunteers,76–79,84,85,87,88 two trials were conducted in hospital in-patients80,81 and two were conducted in the workplace;83,86 two studies were conducted in women only. 41,79 We have divided studies in this section according to three types of participant: unaided abstinent smokers (five studies);76–78,80,81 aided abstinent smokers (seven studies in which participants had achieved abstinence with support from a formal smoking cessation programme);79,83–88 and individuals undergoing enforced abstinence (two studies). 41,82 The length of abstinence prior to randomisation in trials of unaided abstinent smokers ranged from 1 week to 1 month.

Unaided abstinent smokers

Characteristics of interventions

These five studies examined the effectiveness of self-help booklets and letters, telephone counselling and individual counselling. 76–78,80,81

Self-help booklets and letters

Three trials assessed the effectiveness of repeated mailings of relapse prevention-orientated booklets and letters:76–78 one investigated the impact of booklets’ content and mailing frequency in a 2 × 2 factorial design;78 a second compared mailing of booklets with a ‘non-mailing’ control group;77 and a third compared computer-generated (tailored) letters and standard self-help materials. 76 In the former two studies above, participants had achieved at least 1 weeks’ abstinence prior to enrolment and in the latter study participants required at least 24 hours’ abstinence to participate.

Telephone counselling

One study examined the effectiveness of postdischarge telephone counselling in hospital in-patients who had stopped smoking within 31 days of admission into hospital. 81

Individual counselling

One study examined the effect of a combination of individual behavioural interventions in individuals admitted to hospital after myocardial infarction (MI) or for cardiac bypass surgery, and who had not smoked since admission. Interventions included verbal advice, a self-help booklet, a written quiz on the contents of the booklet, carbon monoxide readings, signed declaration of commitment, contact with other people giving up and a sticker in hospital notes. 80

Outcome assessment

Two studies present point prevalence abstinence only;77,81 three validated smoking status biochemically;77,80,81 and two reported abstinence at one follow-up point only. 77,78

Effectiveness of interventions

Two trials detected no effect79,80 and three studies reported that the interventions appeared effective at preventing relapse to smoking. 76–78

Analyses

Pooled analysis of data from the three studies of self-help booklets and letters at long-term follow-up indicated a significant effect on abstinence rates76–78 (pooled OR 1.52, 95% CI 1.15 to 2.01, I² = 0%, NNT = 14; Figure 3); and one study found significant increases in abstinence rates at medium-term follow-up (OR 1.72; 95% CI 1.01 to 2.92). 76 No data were available for short-term follow-up.

FIGURE 3.

Behavioural relapse prevention interventions in unaided abstinent smokers. Self-help booklets and letters at long-term follow-up.

Telephone counselling did not have an effect on abstinence rates at any of the three follow-ups respectively (short term, OR 1.52, 95% CI 0.59 to 3.92; medium term, OR 1.06, 95% CI 0.47 to 2.38; long term, OR 1.23, 95% CI 0.57 to 2.64).

Individual counselling did not appear to have an effect on abstinence rates in the short and long terms respectively (OR 1.04, 95% CI 0.73 to 1.47; OR 0.86, 95% CI 0.60 to 1.23). No data were available for medium-term follow-up.

Aided abstinent smokers

Characteristics of interventions

Of these seven studies,79,83–88 four investigated behavioural counselling delivered to groups,85–88 one84 studied the effect of telephone call content, one83 compared proactive telephone counselling to workplace group counselling and another79 examined the effect of individual behavioural counselling.

Group counselling

These four trials compared group behavioural counselling to usual care or no intervention. 85–88 One study randomised participants who had achieved 1 weeks’ abstinence with counselling and NRT,87 two after 1 week88 and 5 days85 of group counselling and, in one trial, 3 months’ abstinence, achieved with group behavioural support and NRT,86 were required before randomisation.

Telephone counselling (content)

One study assessed the effect of the content of telephone calls, randomising participants into a ‘basic’ content group or ‘enhanced’ content group;84 they received group counselling and pharmacotherapy in a 7-week abstinence period prior to randomisation.

Telephone counselling (proactive)

One study83 compared workplace group counselling to proactive telephone counselling; participants were abstinent for 3 months, using a smoking cessation programme prior to randomisation.

Individual counselling

One study50 randomised abstinent smokers who had achieved abstinence with 2 weeks of group counselling to receive either group-based or individually-tailored relapse prevention treatment; the authors hypothesised that smoking abstinence rates would be higher in individuals assigned to the individually-tailored intervention.

Outcome assessment

One study reported point prevalence and continuous abstinence rates,79 four trials reported point prevalence only83–85,87 and two studies presented continuous abstinence only;86,88 all trials biochemically validated outcomes.

Effectiveness of interventions

Interventions in four trials were reportedly effective at reducing relapse rates,79,84,85,88 but the remaining trials had negative outcomes. 83,86,87

Analyses

Pooled analysis of data from two studies of group behavioural counselling at short-term follow-up appeared to have a significant effect (pooled OR 2.55, 95% CI 1.58 to 4.11, I² = 0%),85,88 but a pooled analysis of data from four studies at the medium-term follow-up did not detect a significant effect (pooled OR 1.20; 95% CI 0.72 to 2.00, I² = 68%),85–88 and a pooled analysis of data from three of the studies at the long-term follow-up point also did not detect a significant effect (pooled OR 0.98, 95% CI 0.55 to 1.76, I² = 63%; Figure 4). 85,87,88

FIGURE 4.

Behavioural relapse prevention interventions in aided abstinent smokers. (a) Group counselling at short-term follow-up; (b) group counselling at medium-term follow-up; and (c) group counselling at long-term follow-up.

Enforced abstinent smokers

Characteristics of interventions

Two studies41,82 assessed the effect of behavioural interventions in individuals that were undergoing enforced abstinence.

Group counselling

One study82 randomised military personnel after 6 weeks of a smoking ban and presented abstinence at long-term follow-up only.

Telephone counselling

One study41 randomised military personnel into either a telephone counselling group, a self-help group and a control group after 8 weeks of a smoking ban respectively.

Outcome assessment

Both trials presented point prevalence abstinence, and neither validated outcomes.

Effectiveness of interventions/analyses

The group counselling intervention did not appear effective at long-term follow-up (OR 0.93, 95% CI 0.85 to 1.02). Telephone counselling was effective in the short term (OR 1.26, 95% CI 1.06 to 1.49); however, the effect was lost in the medium term (OR 0.89, 95% CI 0.76 to 1.05) and in the long term (OR 0.96, 95% CI 0.82 to 1.13).

Pharmacotherapy interventions – bupropion

Analyses

Pooled analysis of two heterogeneous trials of bupropion did not detect any effect in the short term (pooled OR 1.38, 95% CI 0.64 to 2.96; I² = 76%; Figure 5),36,37 or with the inclusion of other studies, in the medium term (pooled OR 1.56, 95% CI 0.95 to 2.56, I² = 56%). However, the estimated effect for relapse prevention treatment with bupropion reached statistical significance when assessed at long-term follow-up (pooled OR 1.49, 95% CI 1.10 to 2.01, I² = 0%; NNT = 11, four studies). Pooled analysis of two trials that combined bupropion and NRT did not show any effect in either the medium term (pooled OR 1.94, 95% C1 0.50 to 7.51, I² = 64%; Figure 6) or long term (pooled OR 1.31, 95% CI 0.46 to 3.79, I² = 30%; Figure 6). No data were available for short-term follow-up analysis.

FIGURE 5.

Bupropion for relapse prevention. Participants were randomised after variable treatment periods with either open-label bupropion and/or nicotine replacement patch. (a) Short-term follow-up; (b) medium-term follow-up; and (c) long-term follow-up.

FIGURE 6.

Bupropion and NRT for relapse prevention. Participants were randomised after variable treatment periods with bupropion and NRT. (a) Medium-term follow-up; and (b) long-term follow-up.

Pharmacotherapy interventions – nicotine replacement therapy

Analyses

Pooled analysis of the two trials of NRT which assessed effectiveness at short-term follow-up was not performed due to high levels of heterogeneity (I² = 85%), but results from these individual studies both showed statistically significant increases in the odds of abstinence (Figure 7). 35,38 More conclusive evidence of an effect for NRT was seen in a pooled analysis of these studies and two more smaller trials with low levels of heterogeneity, at the medium- (pooled OR 1.56, 95% CI 1.16 to 2.11, I² = 37%, NNT = 14, four trials) and long-term follow-ups (pooled OR 1.33, 95% CI 1.08 to 1.63, I² = 0%, NNT = 20, four trials).

FIGURE 7.

Nicotine replacement therapy for relapse prevention. Participants randomised after unassisted abstinence or variable treatment periods with nicotine inhaler or bupropion plus nicotine replacement patch. (a) Short-term follow-up; (b) medium-term follow-up; and (c) long-term follow-up.

Pharmacotherapy interventions – varenicline

Pharmacotherapy interventions – rimonabant

Behavioural interventions matched for programme length

Analyses

Pooled analysis of nine trials42–46,48–51 in the short term did not detect evidence for the effectiveness of behavioural interventions among smokers; OR 0.67 (CI 0.44 to 1.01, I² = 14%). No evidence for the effectiveness of these interventions was detected in the medium- and long-term analyses; OR 0.99 (CI 0.71 to 1.39, I² = 0%, seven trials) and OR 1.09 (CI 0.72 to 1.67, I² = 14%, seven trials) respectively (Figure 8).

FIGURE 8.

Studies randomising smokers with interventions matched for programme length. Interventions were administered for equal durations in both control and intervention groups. (a) Short-term follow-up; (b) medium-term follow-up; and (c) long-term follow-up.

Studies investigating behavioural interventions of differing intensities

Six studies (Table 8) randomised participants into intervention and control groups of varying intensities. 52–71 These studies investigated the impact of behavioural RPIs delivered to smokers as additional components of cessation programmes which resulted in longer treatment programmes; trials in this category do not have intervention and control groups matched for length, rather intervention group programmes are administered for longer periods because of additional relapse prevention content, or deliver content with additional components not included in control group treatments.

Analyses

Pooled analysis revealed that the interventions were not effective in the short, medium and long term respectively with pooled OR 1.11 (95% CI 0.49 to 2.49, I² = 69%, six studies, short term), 1.01 (95% CI 0.57 to 1.80, I² = 47%, six studies, medium term) and 0.86 (95% CI 0.56 to 1.31, I² = 0%, five studies, long term; Figure 9).

FIGURE 9.

Studies randomising smokers with different intensity programmes. Interventions varied in intensity and timing. (a) Short-term follow-up; (b) medium-term follow-up; and (c) long-term follow-up.

Relapse prevention adjuncts to cessation programmes

Two studies (Table 9) provided adjunctive interventions, aimed at preventing relapse to smoking, in addition to smoking cessation treatments. One provided telephone support at specific intervals after a multisession smoking cessation clinic and the other provided access to a specially designed computer program in addition to the control condition components.

Pharmacotherapy interventions – bupropion

Table 10 gives details of the single study that randomised smokers and then tested a pharamacotherapy (bupropion) intervention plus 30 minutes of behavioural counselling at clinic visits after open-label bupropion treatment. 60

Overview of methods

A cohort simulation model was designed to estimate the costs and QALYs associated with interventions to reduce relapse following initial smoking cessation and to determine and compare different interventions’ relative cost-effectiveness. A hypothetical cohort of 1000 smokers who had recently initiated quit attempts (‘recent quitters’) was assembled; this was intended for use in a simulated ‘population cohort’ approach with modelling in 6-monthly cycles over cohort smokers’ lifetimes. In each cycle, ‘recent quitters’ could:

• relapse (i.e. become a ‘smoker’)

• remain a non-smoker (i.e. a ‘former smoker’)

• die.

Figure 10 demonstrates the relationships between smoking status and smokers’ comorbidities permitted by the model which are explained further below.

FIGURE 10.

Movement between health states (note that a smoker can have more than one comorbidity).

In each cycle, smokers and former smokers have a chance of experiencing one or more of five potential comorbidities:

• lung cancer (LC)

• coronary heart disease (CHD)

• chronic obstructive pulmonary disease (COPD)

• MI

• stroke.

First, estimates for the prevalences of each comorbidity within regular and former smokers of different ages and genders were calculated; below, further details of methods used are given. To calculate the number of people with comorbidities, in each cycle, the numbers of smokers and former smokers were multiplied by the estimated prevalences (e.g. to calculate the number of smokers with LC, the number of smokers in each cycle was multiplied by the prevalence of LC among smokers). One qualification is that, as there were insufficient available data on the relative risks (RRs) of former smokers experiencing comorbidities, the model could not take into account the impact of the length of abstinence from smoking amongst former smokers; therefore, the same smoking-related costs are attributed to recent or long-term ex-smokers.

The likelihood of a cohort individual being a smoker or former smoker and also of developing one or more comorbidities in each cycle varies with their age. Each comorbidity has an associated cost and utility [quality of life (QoL)]; to enable the total costs and utilities of the interventions to be compared with ‘no intervention’, the number of people with each comorbidity was, within each cycle, multiplied by the associated cost/utility of that comorbidity, giving an estimated cost/utility for each comorbidity, and these were summed together to calculate an overall estimate for total cost/utility. The cost-effectiveness of individual relapse prevention treatments were determined by inputting the costs of these treatments into the simulation cohort and modelling their impact on cohort utilities, using estimates for effectiveness derived from Chapter 3.

Study population

The cohort is flexible and the cost and QALY outcomes for each combination of age and gender were estimated (e.g. for a 16-year-old man, 16-year-old woman, 17-year-old man, 17-year-old woman, etc.). Population weights derived from population estimates provided by the Office for National Statistics101 (see Appendix 4) were then applied to each cohort group, to ensure that the cohort was representative of the England and Wales population. The costs and QALY outcomes for each age–gender group were also multiplied by these weights to ensure overall QALY outcomes were similarly representative. We did not weight cohort simulations to reflect the sociodemographic characteristics; theoretically this would only have been possible using data on variations in model parameters (e.g. rates of complications, comorbidities, smoking status, etc.) with these characteristics, but such data were unavailable.

Data

This section describes the data sources from which estimates for parameters used in the cohort simulation were derived.

Economic evaluation

Cost-effectiveness models are used to assess the relative benefits of a given treatment using patient outcomes and the costs incurred in achieving those outcomes. The calculation of the additional cost per additional unit gain of benefit (i.e. QALYs) is known as the incremental analysis and results are presented as ICERs. After incremental costs and QALYs were estimated, the ICERs were calculated using the following formula:

The incremental cost per QALY is calculated for all the interventions modelled, allowing the user to compare any two interventions.

Discounting

Costs and outcomes were discounted at 3.5% per year. 122

Sensitivity analyses

Sensitivity analyses were carried out for each intervention to examine the impact of changing model values for:

• background quit rate

• cost and effectiveness

• persistence of intervention benefits.

Model background quit rates of 2% (base case), 1.2% and 2.8% were investigated, as was varying intervention costs between – 50% and + 50% of the base-case value. For effectiveness data, the proportion of additional abstainers at 6 and 12 months was assessed using a range between 10% and 150% of the original value for additional abstainers. For example if, in the control group, the number of abstainers in the control group is 40% and the number of abstainers in the treatment group is 50%, then this equates to an additional 10% of abstainers. Therefore, the sensitivity analysis would assess the impact of changing this value between 1% (i.e. 41% abstaining) and 15% (i.e. 55% abstaining). The persistence of short-term benefits from interventions was investigated as described for the three scenarios outlined above (see Analysis of intervention effectiveness).

Base-case results

Table 20 provides the base-case costs and QALYs, per person, associated with each intervention, using a 2% background cessation rate.

All interventions result in increased QALYs compared with ‘no intervention’ and net costs, including all medical costs incorporated in the model and intervention costs, generally increased, with the exception of bupropion.

Figure 12 illustrates the total cost and total QALYs for all the interventions and ‘no intervention’.

FIGURE 12.

Total cost and total QALY results.

Incremental analysis – comparison against ‘no intervention’ using trial data

Incremental analysis was carried out to compare each intervention to ‘no intervention’ in terms of the total costs and QALYs. All treatments show a very low cost per QALY (maximum = £2106), therefore, they can be considered cost-effective against a willingness-to-pay threshold of £20,000. However, bupropion is the cheapest and the most effective intervention and, therefore, assuming that the interventions are mutually exclusive it dominates all the other interventions (i.e. is more effective and less costly).

Should any decision maker/health service commissioner wish to choose between the different interventions included in the analysis, the incremental costs and benefits should be assessed. For example, if varenicline was compared against, say, bupropion, we can see that it produces an additional 0.03 QALYs at an additional cost of around £45 (Figure 13). The ICER for varenicline versus bupropion is, therefore, £1500 per QALY and varenicline would be considered cost-effective compared to bupropion. However, as mentioned below, caution should be taken when comparing results from different trials; robust data comparing interventions is best obtained from head-to-head comparisons are made within studies and none were available for bupropion and varenicline.

FIGURE 13.

Hypothetical example of long-term benefits (as assumed in ‘base-case’ model). Assuming abstinence rates observed at 1 year continue for a further 10 years.

Table 21 shows the results of comparing each intervention to ‘no intervention’. Note that the effectiveness of ‘no intervention’ varies, as each treatment was modelled against the relevant ‘no intervention’ abstinence rate derived from control groups in trials of that treatment. However, as baseline differences may exist between different trial participants, comparisons between drugs are, indirect and need to be treated with caution. Incremental costs and QALYs are shown, as well as the ICER.

When we excluded those NRT trials, which recruited participants who had been abstinent for very short periods and compared remaining trials with ‘no intervention’, there was very little change in findings for NRT; the incremental QALYs gained were 0.08 (12.66 compared to 12.58) and NRT was also associated with an overall reduction in total costs of £47 (of which, the £100 cost of NRT was offset by £147 savings on other costs). As such, NRT (when used for relapse prevention with smokers who have been abstinent for longer periods) dominated ‘no intervention’ in terms of cost-effectiveness.

Background quit rate variation

Intervention costs

For each of the three interventions, cost and effectiveness parameters were varied to assess their impact upon the cost-effectiveness results in the model. Tables 24–26 detail the impact of changing the attributable costs of the three interventions. Bupropion remains dominant, and hence cost-effective at all projected costs (i.e. ± 50% of those included in the base case) while NRT is more sensitive to cost changes, becoming dominant at an intervention cost of £70, which is below that in the base case. At the highest projected intervention cost, varenicline appears less cost-effective; however, all interventions remain cost-effective as judged by the National Institute for Health and Clinical Excellence (NICE) cost-effectiveness benchmark of £20,000 per QALY.

Effectiveness of interventions

Tables 27–29 demonstrate the impact of changing the level of effectiveness of each intervention. As described above, the proportion of additional abstainers at 6 and 12 months was assessed using a range between 10% and 150% of the original value for additional abstainers. For example if, in the control group, the number of abstainers in the control group was 40% and the number of abstainers in the treatment group was 50%, then this equates to an additional 10% of abstainers. Therefore, the sensitivity analysis would assess the impact of changing this value between 1% (i.e. 41% abstaining) and 15% (i.e. 55% abstaining). Changes in the postulated effectiveness of interventions affects RPIs’ cost-effectiveness more than alterations in cost. Bupropion dominates unless hypothesised effectiveness falls below 50% of used in the base case, and cost-effectiveness falls with decreasing effectiveness as shown by an incremental cost-effectiveness ratio of £8831 per QALY at 10% of base-case effectiveness. NRT only dominates when effectiveness is set at a higher level than estimated from clinical trials and used in the base case, but the ICER for NRT would exceed the NICE threshold if this were to fall below 10% of the base-case level. Varenicline is never dominant and although the NICE benchmark is only exceeded when effectiveness falls below 20% of observed, it appears less cost-effective than either NRT or bupropion with consistently higher ICER values throughout the range of effectiveness explored.

Additional analysis: varying persistence of intervention benefits

It is recognised that the benefits of smokers achieving permanent abstinence are greater than those arising from temporary abstinence periods, but there were no available data with which to predict the long-term smoking status of those smokers who achieved temporary abstinence but subsequently relapsed to smoking. Consequently, we assessed three different potential scenarios for the persistence of benefits arising from intervention delivery (described above) and findings are related below for the base case (i.e. a constant ‘background’ relapse rate occurring after all interventions), assuming abstinence rates observed at 1 year continue for a further 10 years and assuming that differences in abstinence rates attributable to interventions persist for only 1 year.

For this analysis, it was assumed that abstinence rates observed at 1 year would continue for a further period of 10 years, after which the ‘background’ rate would be observed for each treatment arm (Figure 14). As expected, increasing the level of effectiveness of the treatments to 10 years increases the cost-effectiveness of each intervention (Table 30). However, it should be recognised that abstinence rates usually decline with time,123 so this scenario is likely to overestimate the cost-effectiveness of treatments.

FIGURE 14.

Hypothetical example of benefits remaining for 10 years.

Assuming abstinence rates at 1 year do not persist beyond this and differences between groups disappear

In this scenario, we assumed that the benefits of an intervention would last only 1 year, after which there would be no differences between intervention and ‘no intervention’ groups (Figure 15). Table 31 shows that, in this instance, interventions do not appear to be cost-effective as judged by the usual NICE benchmark. This makes intuitive sense, since the benefits are only accrued for 1 year, while the cost of the intervention has remained at its full amount. However, it is very unlikely that the benefits of averting relapse would not be apparent after the first year and, as such, these results represent a highly unlikely worst-case scenario. Differences in abstinence rates between arms of smoking cessation trials have been observed up to 8 years after randomisation. 123

FIGURE 15.

Hypothetical example of benefits remaining for 1 year.

Limitations

There are a number of limitations inherent within the model. A lack of data on how comorbidities varied with smoking status made it impossible to categorise former smokers as achieving either ‘recent’ or ‘long-term’ abstinence and the impact of this on our findings is unclear. If at some point after permanently stopping smoking, the probability of developing some or all of the model comorbidities returns to that of non-smokers, the model will have overestimated the numbers of people with comorbidities and, hence, comorbidity costs, resulting in an underestimation of interventions’ cost-effectiveness.

The model assumes that smokers use only one type of cessation intervention in any one quit attempt but, in ‘real life’, some smokers try stopping smoking repeatedly and some use many different cessation methods. However, the incorporation of a background quit rate into the model addresses this limitation, and sensitivity analysis demonstrated that RPIs appeared effective across a wide range of different background rates.

Model estimates for the effectiveness of interventions were taken from RCTs but interventions often show greater efficacy in trials than in routine clinical care and this could have contributed to an overestimation of cost-effectiveness. Also most pharmacotherapy relapse prevention trials required between 2 and 3 months abstinence from smoking before relapse prevention treatments were started and abstinence was generally achieved with the help of evidence-based support, so the model relates primarily to interventions used in this context. NRT trials provide some insight into the potential influences of abstinence period length and the use of support to achieve abstinence on cost-effectiveness. When trials with very short abstinence periods (< 48 hours), achieved without cessation support, were excluded from NRT analyses, NRT appeared to be even more cost-effective. Consequently, the effectiveness and cost-effectiveness of NRT, and perhaps of other RPIs, is likely to be dependent upon abstinent smokers’ characteristics, such as how they achieved abstinence and also upon the timing of relapse prevention intervention delivery in relation to the start of any quit attempt. Unfortunately, the model cannot investigate these issues in the absence of sufficient empirical trial data.

Again, due to a lack of available data, the model assumed that, when a person had multiple comorbidities, their QoL was equivalent to that experienced with the most severe of these. From the perspective of assessing the impact of the interventions, this is a conservative assumption as each intervention, by encouraging abstinence from smoking, is likely to reduce the prevalence of combinations of comorbidities (e.g. LC and CHD would both become less likely in the event of smoking cessation). Improvements in the QoL experienced by some of those people with more than one comorbidity who remain abstinent would, therefore, be greater than the model predicts; QALY gains from eliminating such comorbidity combinations would be greater than predicted within the model and interventions would appear even more cost-effective.

One interesting aspect of this model is that, as recommended by the UK Treasury, all future costs and health outcomes have been discounted at 3.5% per year. However, it should be noted that the costs of RPIs are borne in the immediate future (i.e. undiscounted), while the benefits are likely to be accrued in the long-term future and hence, discounted. With discounting at 3.5% rate, one QALY today is equivalent to around 0.25 QALYs in 40 years’ time and health gains experienced at this future time are, therefore, reduced fourfold. Recommended discount rates vary between different countries and even within countries over time; NICE’s recommended rate for health outcomes was 1.5% until 2003. Because the costs of the interventions are accrued in the short-term and the benefits (i.e. reduced comorbidities) occur in the future, the results of this analysis will underestimate the undiscounted outcomes.

Summary

Our model shows that bupropion, NRT and varenicline (based on one trial) used by recently abstinent quitters to prevent relapse to smoking are cost-effective when, compared with ‘no intervention’ and judged against a ‘willingness-to-pay’ threshold of £20,000 per QALY. A direct comparison of the incremental cost-effectiveness of these treatments was not possible because data underpinning analyses were generally derived from mutually-exclusive trials, with insufficient data comparing interventions within trials to permit direct comparisons and any comparisons, therefore, must be indirect. All three of these pharmacological interventions for relapse prevention appear to have a similar magnitude of cost-effectiveness as smoking cessation-orientated interventions.

Rationale for studies’ inclusion

We anticipated that the most robust data on the frequency of relapse by smokers in quit attempts would be collected during clinical trials and used the following considerations to decide which clinical trials were appropriate for consideration.

Process for identifying studies

The Cochrane Database of Systematic Reviews was searched in December 2009 to identify all reviews that potentially assessed the impact of smoking cessation treatments using the term ‘smoking cessation’. Forty-one systematic reviews were identified, of which 12 focused on pharmacotherapies for smoking cessation. 138–140,144–152 The 12 systematic reviews were screened to identify studies that assessed specific pharmacotherapies with evidence of effectiveness. The full texts of the papers identified from the included Cochrane reviews were then screened independently by two authors (JL-B and SA or AMcN and TC) using a specially designed data extraction form to assess whether they fulfilled the pre-specified inclusion criteria (see Box 7, Figure 16).

FIGURE 16.

Flow chart for included studies.

Assessment of methodological quality

Two authors (JL-B and SA) independently assessed the methodological quality of the eligible primary trials. The quality assessment for each trial included an evaluation of the method of generation of the randomisation sequence, the method of allocation concealment, whether blinding was used, percentage of participants lost to follow-up in the treatment group, and whether the participants were analysed using the intention to treat principle.

Data extraction

Two authors (JL-B and SA) performed data extraction independently and any differences were resolved through discussion. To derive curves for relapse to smoking by smokers using optimal, evidence-based smoking cessation treatments (as defined above), the following data were extracted from the arms of all included trials in which participants used such support:

1. (a) Where survival curves were reported, these were converted to abstinence rates at the pre-specified time points using visual assessment.

2. (b) Proportion of trial participants reporting continuous abstinence from smoking at all pre-specified time points between quit date and final follow-up.

3. (c) Proportion of trial participants reporting point abstinence from smoking at pre-specified time points between quit date and final follow-up.

4. (d) Number enrolled into the pharmacotherapy treatment group at randomisation.

The standard error for the prevalence of abstinence from smoking was estimated using the following formula:

If possible, missing data were obtained by contacting the authors of the primary papers. One author (JL-B) entered the data, and double-checking was performed by another author (SA).

Statistical analysis

Meta-analyses were performed to calculate a weighted prevalence of abstinence in the cessation treatment groups across trials using a random effect model (DerSimonian and Laird model) at each pre-specified time point to allow for heterogeneity being anticipated between the studies. The results are presented as prevalence of abstinence estimates with 95% CIs. Statistical heterogeneity between the trials was quantified using I². 153 Separate pooled analyses were conducted based on the method used to report abstinence from smoking (point prevalence or continuous abstinence) were performed. Sensitivity analyses were conducted to assess the effect of excluding trials which yielded heterogeneous findings. The pooled estimates were then plotted to derive relapse curves for prolonged/continuous abstinence, and abstinence curves for point prevalence abstinence. Meta-analyses were conducted using review manager 5.

Definition of abstinence

The majority of the included trials used point prevalence estimates to ascertain abstinence from smoking; 3-day point prevalence was used by one trial,131 7-day point prevalence by six,130,154,156,158,159,161 and three did not report a minimum abstinence period required for point prevalence to be considered achieved. 163–165 Four trials used continuous abstinence from quit date,132,133,157,160 one trial used continuous abstinence for the first month of follow-up followed by 7-day point prevalence,162 and the remaining trial used 7-day point prevalence for all time points, except at 12 months when prolonged abstinence was reported. 155 None of the four trials reporting continuous abstinence reported point prevalence abstinence. 132,133,157,160 All of the included studies used biochemical verification to confirm non-smoking status.

Behavioural support

In addition to drug treatments, all included trials provided behavioural support (see definition above) outside of routine clinical care; four used group-based sessions and remaining studies used individual consultations to deliver this support. Further details are given in Table 33.

Risk of bias in included studies

Of the 16 included trials, eight reported methods used to generate their randomisation sequence, with five using computer-generated lists,130,132,133,159,160 one stratified randomisation162 and two block randomisation. 131,154 Only seven studies reported using an adequate method for allocation concealment, using centrally performed systems. 130,132,133,154,155,159,160 Thirteen studies used double-blinding,130–133,154–157,159,160,162,163,165 all studies reported the proportion of loss to follow-up, and 11 stated using an intention-to-treat analysis. 130,131,154–159,161,162,164

Pooled estimates for prevalences of abstinence by treatment

Nine estimates of prevalence from eight studies contributed to a pooled analysis which assessed the prevalence of abstinence from smoking after using NRT as a smoking cessation treatment; five assessed continuous/prolonged abstinence and the remaining four assessed point prevalence abstinence. In studies that reported continuous/prolonged estimates, abstinence decreased over time from 86% (95% CI 79% to 92%) at the first time point in the first month to 20% (95% CI 16% to 24%) at 12 months’ follow-up (Table 34, Figure 18). Very high levels of heterogeneity were seen in the analyses conducted at the first three time points within the first month (I² = 93%, 94% and 93%, respectively). In studies that reported point prevalence estimates, smaller reductions in abstinence were seen over time, which decreased from 51% (95% CI 42% to 61%) at the first time point in the first month to 30% (95% CI 17% to 42%; Figure 19). Substantially lower levels of heterogeneity were seen between the studies. Studies that reported continuous abstinence for all timings showed higher abstinence rates than studies which reported point prevalence abstinence rates. This is slightly unusual because point prevalence estimates for smoking cessation are usually higher than prolonged/continuous ones, however, these abstinence estimates were obtained from different trials, so comparisons are indirect and must be made cautiously because outcome differences may reflect the different characteristics of smokers within and settings of trials.

FIGURE 18.

Prevalence of point abstinence after using NRT for smoking cessation in addition to behavioural support using a meta-analysis of RCTs. Data are presented as prevalence of abstinence and are pooled using random effect models. Squares represent the point estimate and the horizontal lines denote 95% CIs. Size of the data markers correspond to the weight of the study in the meta-analysis.

FIGURE 19.

Prevalence of point abstinence after using bupropion SR for smoking cessation in addition to behavioural support using a meta-analysis of RCTs. Data are presented as prevalence of abstinence and are pooled using random effect models. Squares represent the point estimate and the horizontal lines denote 95% CIs. Size of the data markers correspond to the weight of the study in the meta-analysis.

Six trials provided eight point prevalence of abstinence estimates that contributed to the pooled analysis assessing relapse after using bupropion; abstinence decreased over time from 35% (95% CI 21% to 49%) at the first time point within the first month to 22% (95% CI 17% to 27%) at 12 months’ follow-up (Figure 20). Very high levels of heterogeneity were seen in the analyses conducted for the three follow-up points within the first month, at 3 months’ and 12 months’ follow-up (I² = 98%, 95%, 95%, 85%, 88%, respectively). A subgroup analysis based on studies that reported continuous abstinence for all timings could not be performed as all of the included studies reported point prevalence of abstinence only. However, when a sensitivity analysis based on excluding outlier results131,155,164 was performed, the estimates for pooled prevalence were generally less heterogeneous and abstinence rates were lower (first month: first time point = 30%, second time point = 30%, third time point = 29%; 3 months = 31%; 6 months = 26%; 12 months = 23%).

FIGURE 20.

Prevalence of point abstinence after using varenicline for smoking cessation in addition to behavioural support using a meta-analysis of RCTs. Data are presented as prevalence of abstinence and are pooled using random effect models. Squares represent the point estimate and the horizontal lines denote 95% CIs. Size of the data markers correspond to the weight of the study in the meta-analysis. BID, twice per day; NT, non-titrated; T, titrated.

Five trials provided 10-point prevalence of abstinence estimates that contributed to the analysis-assessing relapse after varenicline use. Abstinence did not decrease appreciably over time; it was 36% (95% CI 32% to 40%) at the first time point within the first month and 31% (95% CI 28% to 35%) at 12-month follow-up (Figure 21) with moderate levels of heterogeneity in analyses. A sensitivity analysis based on methods of ascertainment of smoking cessation outcome was not possible because all studies reported only point prevalence abstinence, and a sensitivity analysis excluding ‘outliers’ was not performed as no studies were deemed to have outlier results.

FIGURE 21.

Abstinence curves (shown as line graph) in participants from the included trials of evidence-based smoking cessation treatments using pooled estimates of point prevalence of abstinence from meta-analyses of RCTs. Crosses represent the point estimates for NRT, squares represent the point estimates for bupropion, triangles represent the point estimates for varenicline, and stars represent the point estimate for the combined medications derived from the meta-analyses. Horizontal lines denote the 95% CIs for the combined medications estimate at each of the pre-specified points of follow-up.

Comparison of relapse patterns, measured by point prevalence, for different cessation treatments

The smoking abstinence patterns for NRT, bupropion and varenicline, as measured by point prevalence estimation, differed markedly over the first 3 months after quitting, became similar after this and showed a similar pattern of steady decline from 6 months onwards (Figure 21). One caveat was that, when drawing smoking abstinence curves from point prevalence estimates, imputation of data between the discrete time points at which prevalence’s were obtained is necessary. However, Figure 21 summarises the best currently available data for considering smoking abstinence patterns after supported cessation attempts.

Treatment with NRT resulted in most abstinence at the first follow-up point after quitting; approximately 51% of those receiving NRT were abstinent at around 2 weeks into quit attempts, compared with approximately 35% of those receiving bupropion and varenicline. The abstinence rate with NRT then remained fairly constant, and much higher than with other treatments, for the rest of the first month after quitting, followed by a smooth decline for the remainder of the year. In contrast, the proportion abstinent with bupropion changed little over the initial 3 months and subsequently followed a near-identical pattern of decline to that with NRT. Abstinence patterns with varenicline were different again; varenicline treatment resulted in a sharp increase in the proportion abstinent within the first month of treatment followed by a slower rate of increase in abstinence until around 3 months, when over half were abstinent (55%). After 3 months, point prevalence abstinence rates after varenicline began declining and did so at a similar rate to those after treatment with on bupropion and NRT, but with higher overall abstinence rates recorded. This is demonstrated by the varenicline curve having a very similar shape to others after the 3-month point, but being higher in relation to the y-axis.

Limitations

Comparisons across treatments require caution because most studies used only point prevalence measures to ascertain smoking status and few trials reported prolonged abstinence measures, such as continuous cessation between quit dates and follow-up points. Point prevalence measures can be more volatile than continuous or prolonged measures of abstinence and the latter are more appropriate for deriving relapse curves, however, in the absence of continuous outcome data being reported at all time points we have been restricted to using point prevalence to derive our abstinent curve. We have only been able to contrast curves derived from prolonged and point prevalence abstinence measures for NRT trials and curves presented are derived from different trials as, within individual studies, continuous and point prevalence outcome measures were often not reported for simultaneous follow-up points. However, apart from within the first month where an increase in abstinence was seen in the abstinence curves, the curves derived using these two different outcome measures had broadly similar shapes. When comparing the shapes of the abstinent curves derived using prevalence data, very different relapse patterns are seen within the first 3 months of quit attempts assisted by bupropion or varenicline as compared to NRT. This difference may be real and due to differences in the responses of patients to these treatments or perhaps in the characteristics of smokers using them. Another caveat is that 14 of the 16 included studies assumed participants with missing outcome data or who were lost to follow-up had returned to smoking; this could have had an impact on our findings by artificially lowering the abstinence rates at the later follow-up time points since the percentage of participants not completing the studies ranged from 0% to 50%, although a third of studies had less than 10% non-completers.

A further caveat is that no trials reported the proportion of participants who did not attempt to stop smoking on their quit date and none explicitly reported that patients must attempt cessation on their intended quit dates (though this is implied in many cessation studies). Consequently, by including in our analyses those who fail to stop smoking at all, our relapse curves may overestimate initial relapse rates at the outset of cessation attempts; future trials should consider reporting such data. Findings should be seen as preliminary due to observed heterogeneity between some prevalence estimates. By only including trials that satisfied Cochrane review standards, heterogeneity attributable to interventions employed is likely to have been minimised, but different doses of drug treatments and/or intensities of behavioural support may still have contributed towards this, as there were no striking differences in participants’ demographic characteristics or nicotine addiction levels to help explain it. We believe that most heterogeneity has arisen from trials’ use of slightly different follow-up time points within the first month after quitting which explains the more acceptable heterogeneity levels observed at the later, more uniform follow-up time points. We attempted to account for heterogeneity by using random effect models in analyses, though these methods may still result in confidence intervals which are too narrow and hence insufficiently conservative. 166 However, it is unlikely that bias was introduced into the meta-analysis since we identified eligible studies from published Cochrane reviews and included only relevant RCTs which collect data prospectively and are the most robust form of research evidence. Finally, the generalisibility of our findings to NHS SSS might be questioned; while all three pharmacotherapies studied are used frequently by NHS SSS, treatment protocols for the delivery of behavioural support are likely to differ and interventions used in clinical trials often demonstrate less effectiveness when translated into routine care.

A key finding of this review is that, irrespective of what happens in the very early stages of quit attempts, for all three treatments considered, substantial relapse occurs after 3 months and most cessation treatments will have been discontinued either prior to or at this time point. Eliminating relapse after 3 months could potentially increase 12-month quit rates by 13%, 14% and 19% for NRT, bupropion and varenicline respectively (estimates taken from Figures 2, 4 and 5 but are preliminary due to heterogeneity levels). This illustrates the potential impact of extending cessation treatment and complements the findings from Chapter 3 in which we demonstrated the efficacy, for smoking cessation, of extending the provision of smoking cessation medications beyond the usual acute cessation treatment period. Clearly, for the three treatments considered here and in Chapter 3, an extension of therapy periods, as previously suggested for NRT134 might reduce relapse to smoking, facilitating higher rates of permanent smoking cessation. Additionally, as the rate of relapse early in quit attempts involving NRT is so rapid, any relapse prevention treatments that have an adjunctive impact at this time could have the substantial effects.

Research

1. Further research investigating the use of NRT, bupropion and varenicline (the three pharmacotherapies used in the NHS SSS) for relapse prevention is required, including the following:

   1. Placebo RCTs to investigate the (cost) effectiveness of these RPIs as an extension to current NHS SSS cessation support – most review trials were conducted in countries without organised cessation services and, hence relapse prevention interventions may have different outcomes in the UK.

   2. Studies of the acceptability of extended use of pharmacotherapies for relapse prevention in NHS SSS users, and particularly of bupropion, which is the least frequently used cessation therapy in England; the acceptability of these pharmacotherapies for relapse prevention will influence their uptake.

   3. Whether or not the addition of behavioural relapse prevention interventions, delivered in the early stages of quit attempts using NRT can have an adjunctive, positive impact on cessation rates.

   4. Confirmation of whether the different trajectories of relapse that we observed for NRT, bupropion and varenicline are valid (i.e. a more rapid relapse rate for users of NRT in the first month compared with the other two drug treatments) and occur when these treatments are used in routine NHS SSS clinical practice.

2. The following research into behavioural relapse prevention interventions is required:

   1. RCTs to confirm or refute the finding that self-help interventions, delivered to smokers who have achieved abstinence unsupported, have long-term effectiveness for preventing relapse to smoking.

   2. RCTs to investigate whether or not self-help interventions delivered to smokers who have achieved abstinence with NHS SSS support are effective.

   3. Further research to refine interventions that showed potential in our effectiveness review, such as individual counselling for pregnant women and the use of telephone support after cessation treatment and test whether or not these might have long-term effectiveness.

3. Methodological standardisation: among relapse prevention trials identified for this report, there was huge variation in the definition of RPIs, the characteristics of smokers these were delivered to, follow-up periodicity, and outcome measurement after randomisation. Also among cessation trials used to derive relapse curves, reporting of outcomes seriously restricted the data available. In order to permit coherent synthesis of future research findings in this important field, we recommend that practitioners and researchers investigating this field agree common standards for:

   1. The definition of RPI: in particular, consensus is needed as to whether behavioural RPIs, delivered alongside smoking cessation interventions to smokers either prior to or soon after quit attempts have started can or should be categorised as different to smoking cessation interventions. If there is consensus about such interventions being different, clear definitions for both are required.

   2. Methodological standards for the conduct and reporting of behavioural and pharmacological relapse prevention trials.

   3. Cessation trials should report the percentage of participants who make no attempt to stop smoking on target quit dates and should report continuous and point prevalence smoking cessation measures simultaneously at all follow-ups.

Implications for health care

Some NHS SSS are providing RPIs, but where this occurs, those with the weakest evidence base are generally used, illustrating a requirement for the emerging evidence base, and guidance, to be made available as soon as possible. Should the NHS decide to encourage and fund the use of RPIs for smokers who have become abstinent with NHS SSS support, new incentives are likely to be required before NHS SSS will substantially adopt their use. Currently, NHS SSS are performance-managed on their ability to achieve targets set for short-term (i.e. 4-week) periods of cessation; managers perceived these targets were a clear disincentive to spending on interventions such as RPIs, which might enhance longer term abstinence but not their clients’ initial, monitored cessation rates. Any integration of RPIs into NHS SSS care should include sufficient monitoring such that an assessment of their cost-effectiveness in routine use can be made.

MEDLINE (Ovid Web); 2004–7/August week 1; 8 August 2007

579 records were retrieved.

1. Smoking Cessation/

2. “Tobacco Use Disorder”/

3. Tobacco/

4. Nicotine/

5. Tobacco, Smokeless/

6. Smoking/pc, th

7. ((quit$or stop$or ceas$or giv$) adj smok$).ti,ab.

8. Tobacco Smoke Pollution/

9. or/1-8

10. Randomised controlled trial.pt.

11. controlled clinical trial.pt.

12. Randomised Controlled Trials/

13. Random Allocation/

14. Double-Blind Method/

15. Single-Blind Method/

16. clinical trial.pt.

17. exp Clinical Trials/

18. (clin$adj trial$).ti,ab.

19. Placebos/

20. placebo$.ti,ab.

21. random$.ti,ab.

22. Research Design/

23. ((singl$or doubl$or trebl$or tripl$) adj (blind$or mask$)).ti,ab.

24. (volunteer$or prospectiv$).ti,ab.

25. exp Evaluation Studies/

26. exp Cross-Sectional Studies/

27. Prospective Studies/

28. Retrospective Studies/

29. Follow-Up Studies/

30. exp Health Education/

31. exp Health Behaviour/

32. exp Community Health Services/

33. Health Promotion/

34. exp Behaviour Therapy/

35. or/10-34

36. Recurrence/

37. (relaps$or recurr$).ti,ab.

38. (resum$or restart$or re start$or lapse$).ti,ab.

39. maintenance.ti,ab.

40. (maintain$or sustain$).ti,ab.

41. or/36-40

42. 9 and 35 and 41

43. Animals/

44. Humans/

45. 43 not (43 and 44)

46. 42 not 45

47. (2004$or 2005$or 2006$or 2007$).ed.

48. 46 and 47

EMBASE (Ovid Web); 2004–7/week 31; 8 August 2007

406 records were retrieved.

1. (random$or factorial$or crossover$or cross over$or cross-over$or placebo$or (doubl$adj blind$) or (singl$adj blind$) or assign$or allocat$or volunteer$).ti,ab.

2. crossover procedure/or double-blind procedure/or Randomised controlled trial/or single-blind procedure/

3. 1 or 2

4. smoking cessation.mp.

5. exp Smoking Cessation/

6. exp SMOKING/

7. ((quit$or stop$or ceas$or giv$or prevent$) adj smok$).mp.

8. exp Passive Smoking/or exp Smoking Habit/or exp Cigarette Smoking/

9. or/4-8

10. Relapse/

11. (relaps$or recurr$).ti,ab.

12. (resum$or restart$or re start$or lapse$).ti,ab.

13. maintenance therapy/

14. maintenance.ti,ab.

15. (maintain$or sustain$).ti,ab.

16. or/10-15

17. 3 and 9 and 16

18. exp animal/

19. Nonhuman/

20. (rat or rats or mouse or mice or hamster or hamsters or animal or animals or dogs or dog or cats or bovine or sheep or ovine or pig or pigs or porcine).ti,ab, sh.

21. or/18-20

22. exp human/or exp human experiment/

23. 21 not (21 and 22)

24. 17 not 23

25. (2004$or 2005$or 2006$or 2007$).em.

26. 24 and 25

CENTRAL (Cochrane Library/Wiley); 2007:3; 8 August 2007

228 records were retrieved.

1. #1 MeSH descriptor Tobacco Use Disorder explode all trees

2. #2 MeSH descriptor Tobacco, Smokeless explode all trees

3. #3 MeSH descriptor Tobacco Smoke Pollution explode all trees

4. #4 MeSH descriptor Tobacco Use Cessation explode all trees

5. #5 MeSH descriptor Nicotine explode all trees

6. #6 (smoking AND cessation)

7. #7 (antismok*)

8. #8 (quit*):ti

9. #9 (smok*):ti

10. #10 (cigar*):ti

11. #11 (tobacco):ti

12. #12 (nicotine):ti

13. #13 MeSH descriptor Smoking explode all trees

14. #14 (#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12 OR #13)

15. #15 MeSH descriptor Recurrence explode all trees

16. #16 (relaps* or recurr*)

17. #17 (resum* or restart* or (re start*) or lapse*)

18. #18 (maintenance)

19. #19 (maintain* or sustain*)

20. #20 (#15 OR #16 OR #17 or #18 or #19)

21. #21 (#14 AND #20), from 2004 to 2007

PsycINFO (Ovid Web); 2004–7/July week 5; 8 August 2007

555 records were retrieved.

1. smoking cessation.mp. or exp Smoking Cessation/

2. (antismoking or anti-smoking).mp.

3. (quit$or cessat$).mp.

4. (abstin$or abstain$).mp.

5. (control$adj smok$).mp. [mp = title, abstract, heading word, table of contents, key concepts]

6. exp Behaviour Modification/

7. 2 or 3 or 4 or 5 or 6

8. exp Tobacco Smoking/

9. (smok$or cigar$or tobacco$).mp.

10. exp Prevention/

11. 8 or 9

12. 7 and 11

13. 10 and 11

14. 1 or 12 or 13

15. relapse prevention/

16. exp Maintenance Therapy/

17. (relaps$or recurr$).ti,ab.

18. (resum$or restart$or re start$or lapse$).ti,ab.

19. maintenance.ti,ab.

20. (maintain$or sustain$).ti,ab.

21. or/15-20

22. 14 and 21

23. (2004$or 2005$or 2006$or 2007$).up.

24. 22 and 23

Science Citation and Social Science Citation Index (ISI Web of Science); 2004–7/August 8; 8 August 2007

646 records were retrieved.

(((smoking cessation) OR (smok* SAME (quit or stop or prevent*))) AND (random* or trial or control* or (study same smok*)) AND (relaps* or recur* or resum* or restart* or (re start*) or lapse* or maintenance or maintain or sustain)) NOT (rat or rats or mouse or mice or hamster or hamsters or animal or animals or dogs or dog or cats or bovine or sheep or ovine or pig or pigs or porcine)

Time span = 2004–7

Productivity losses and absenteeism

Annual costs of lung cancer and stroke in the UK

Utilities: myocardial infarction; chronic obstructive pulmonary disease; lung cancer; coronary heart disease; and stroke

Association between smoking and COPD/stroke: separated into current, former and never smokers