Effectiveness of preloading
Summary of main findings
In this pragmatic open-label trial testing the effect of nicotine preloading for 4 weeks prior to a quit day, there was no evidence that preloading increased the proportion of people achieving smoking abstinence. Adherence to the nicotine patch used for preloading was good, with the large majority using treatment as prescribed. Preloading was tested in a clinical setting where smokers could opt to use either NRT or non-nicotine pharmacotherapies for continued cessation treatment after preloading had ended. Preloading encouraged the use of NRT for cessation treatment or deterred the use of varenicline as post-quit day medication. Varenicline is more effective than other medications, so in a planned analysis we adjusted for its use after quit day. This led to clear evidence that preloading itself did increase the likelihood of achieving abstinence. There was no evidence that use of varenicline undermined the effect of preloading. Most people continued with the preloading treatment, implying that it is well tolerated. Around 1 in 20 people suffered AEs of moderate severity and above as a consequence of the treatment, principally nausea. There was no evidence of an excess of SAEs.
Public involvement
The study was run under the auspices of the UK Centre for Tobacco and Alcohol Studies (UKCTAS).89 UKCTAS has a standing panel of public members who guide the research programme, helping select topics that matter to people who smoke and shaping the way that research is carried out. Before the study started, we discussed the study and the methods with the panel. One member of the panel joined the trial steering committee that provided oversight to the trial.
In the early days of the study, we encountered a much slower rate of recruitment than we had anticipated. We used the method of recruitment that we had employed in three of the four centres in previous trials.17,32,36 In the past, around 5–10% of people who received an invitation letter from their GP had enrolled in our studies. As in this trial, in the previous trials the invitation letter invited people wanting to stop smoking to seek support, rather than emphasising the novel treatment. We know from discussions with previous participants that it is the opportunity to receive support to quit that is the main draw for potential participants. In the Preloading trial, only around 1% of people who were invited enrolled in the trial. We asked GPs to write to batches of participants, 100 at a time, to ensure that we had a reasonable number of initial appointments to fill the clinic but did not generate a waiting list. However, initial recruitment was much slower than anticipated and, therefore, accrual into the study was much slower than planned.
As a consequence, we consulted with eight members of the UKCTAS panel to review our recruitment material and two other members of the public from the Clinical Research Network. They commented on the documents and answered questions about our material. Our recruitment material in the Preloading trial, as in previous trials, was very brief, which is probably important in attracting people who smoke. Consequently, only a few minor amendments were suggested, which we implemented after approval from the ethics committee. However, the response to the recruitment letter remained as poor as it had been. In fact, the Preloading trial coincided with a downturn in interest in NHS Stop Smoking Services and a rise in the use of e-cigarettes; it is thought that these two factors may be linked. We enhanced our recruitment rate by sending more letters per practice in larger batches and using Docmail® (CFH Docmail Ltd, Radstock, UK) to send the invitations, easing the burden on general practices. We also added other means of recruitment including text messages from practices and advertising.
At the time of writing, the results of this trial have only just become available. We plan to return to the UKCTAS panel with the results and discuss them with the members. To help guide implementation into policy and practice, it will be important to obtain the public’s views on what the results mean and what would make a person who wants to stop smoking take up preloading.
Strengths and limitations
This trial has several strengths. The trial had power to detect the modest benefits that may have been expected, given the previous meta-analyses.12 It was conducted with rigour, with a prespecified protocol and analysis plan; it followed standards set out in CONSORT (Consolidated Standards of Reporting Trials) guidelines90 and achieved high follow-up rates compared with many trials of smoking cessation interventions, increasing confidence in the findings. It is the first trial of preloading to assess AEs to standards set out in Good Clinical Practice. In an open-label trial such as this, it is non-intuitive, both to participants and to trial staff, to report and enquire about AEs occurring in people on no medication. We trained our staff to enquire in the same way regardless of trial group, but it remains possible that the open-label design may have led to over-reporting of AEs in the preloading group. By also actively soliciting reports of likely AEs by questionnaire in all participants, we may have removed some of this bias. However, both methods revealed a somewhat similar profile, with nausea being the most common adverse consequence of preloading specifically. Patients also reported other well-established AEs of using a nicotine patch relating to its stimulant (poor sleep and dreaming) and adhesive (skin irritation) properties, but these were uncommon. However, taken together, both methods showed that severe AEs are rare and that the medication is well tolerated. A further strength of the study is that we collected data on potential mechanisms of action of preloading; these are addressed in Chapter 4.
The open-label design could be viewed as a limitation. Most notably, the open-label design led to 29.5% of participants in the control group choosing to use varenicline for post-quit day medication compared with 22.1% in the preloading group. This is a modest absolute difference, but, in this study, use of varenicline more than doubled the odds of achieving the primary outcome compared with the use of an alternative pharmacotherapy (chiefly NRT) and, in our pre-stated analysis, adjustment for this difference led to a statistically significant effect of preloading on the primary and secondary outcomes. There were also more people in the preloading group who used nicotine patches only, possibly because they received 5 weeks’ supply of the medication. The two study arms also had different expectations, which may have affected their smoking behaviour and their responses to various subjective ratings meant to assess the effects of preloading on smoking reward, enjoyment and urges to smoke. A blinded design would probably have shown more clearly an effect of preloading on smoking abstinence and would allow an unbiased analysis of mediators.
The open-label design could be seen as a strength because it produced an outcome that would not have been detected in a blinded design. If patch preloading reduces the use of varenicline, this would suggest that, within the NHS Stop Smoking Services, preloading is an efficacious strategy, but not an effective one, because preloading would be used in an open-label manner in that context. It is unclear, however, whether the difference in varenicline use arose because patients who were happy on NRT opted to continue it or because smoking cessation therapists or prescribers (mainly GPs) were unhappy to prescribe varenicline to someone already using NRT. We sought to counter the latter effect in discussions and with a letter of referral from the research team to the NHS Stop Smoking Services, which took on the pre- and post-quit support and prescribing. However, we know of several instances where prescribers could not be persuaded that co-prescribing the medications was safe, as NICE guidance was viewed as ‘trumping’ this. Revised NICE guidance could easily remove the apparent prohibition on concurrent use. If the difference was due to patients feeling content with patches and disinclined to add another medication, informing patients that adding varenicline improves outcomes would be another potential remedy.
Thus, the open-label design may be viewed as a weakness because it clouded the answer to the scientific question of efficacy and mechanisms of action of preloading, and as a strength because it revealed the effect of preloading as it would play out in the current UK NHS or similar health systems. That said, it is likely that we would be able to overcome the effect on use of varenicline subsequently through changes to guidelines and discussion with patients.
Comparison with other literature
Our previous meta-analysis yielded a RR of 1.16 (95% CI 0.97 to 1.38) for prolonged abstinence, but a RR of 1.26 (95% CI 1.03 to 1.55) for nicotine patches.12 In this trial, the RR for the primary outcome (also prolonged abstinence) was very similar, at 1.27 (95% CI 1.03 to 1.57), in the analysis adjusted for the use of varenicline. The previous meta-analysis of nicotine patches alone was from an exploratory subgroup analysis and the main analysis showed no evidence of effectiveness, with a RR of 1.16 (95% CI 0.97 to 1.38). Thus, compared with the current literature, the trial has strengthened the evidence for the efficacy of preloading. However, our trial is the only trial to be carried out in a context in which the therapists who prescribed the preloading (our trial team) were different from those who prescribed the post-cessation support (the NHS Stop Smoking Service). Arguably, the context of this trial is more likely to reflect that of most health systems, where there is less behavioural support and dedicated smoking cessation expertise available to people who want to stop smoking than has been the case in the previous efficacy trials, thus revealing an effect that could negatively affect the impact of preloading.
One other factor also deserves attention. In this trial, we asked centres to record the proportion of people who did not make a quit attempt. We were concerned that deferring a quit attempt may mean that a person loses motivation to make one, thereby undermining the effect of preloading. This variable was incompletely recorded by centres, and it was clear that practice with respect to recording also differed in a way that makes relying on this variable unwise. From cross-tabulating other relevant data recorded, it appears that 112 (6.3%) people did not make a quit attempt: 60 (6.7%) of these were in the preloading group and 52 (5.8%) were in the control group. As the proportion was similar in both arms, this is not an effect of preloading per se, but it could reflect the fact that participants waited 4 weeks to quit smoking and we know that plans to quit smoking change quickly.91 Not many smoking-related trials record the proportion of participants who do not make attempts to quit smoking. However, some of our previous trials have recruited in largely similar ways to this trial and, in these trials, 3% and 12% made no attempt to quit smoking, with a much shorter delay (1–2 weeks and 1 day) between trial enrolment and the target quit day.33,36 A more recent trial reported that 17% of people who had undertaken 2 weeks of preloading failed to make a quit attempt, but this was defined as not achieving 24 hours of abstinence after quit day.17 Evidence suggests that around half of all quit attempts do not achieve this milestone, meaning that perhaps only 8–9% failed to make a quit attempt.92 Taken with these other trials, it appears that preloading and the necessary delay in making a quit attempt is not deterring people from following through with that attempt.
Interpretation
This trial has clarified the role of preloading in the treatment of tobacco dependence. Preloading is an efficacious strategy that could lead to around 3% of people who are seeking help to quit smoking achieving 12-month prolonged abstinence, who might not otherwise have done so. This effect may seem small, but the current 12-month abstinence rate in the UK specialist Stop Smoking Services is 8%93 and so an additional 3% would represent a worthwhile improvement. A failed quit attempt is likely to cost a person about 3–5 years of life expectancy.34 Thus, around 12 people need to use preloading to gain around 1 year of life. However, within the specific setting of the NHS Stop Smoking Services, without changing NICE guidelines and informing advisors, the efficacy of preloading could be undermined by the effect it has on deterring use of the most effective cessation medication (varenicline) and, possibly, by delaying quit attempts in those motivated to quit immediately. Thus, it seems sensible to confine it as a strategy for use in highly controlled circumstances. If used in NHS Stop Smoking Services, we would recommend that NHS advisors discuss the choice of post-cessation treatment at the start of any period of preloading so that experience of preloading does not deter the use of varenicline.
Our study is not the only study that has produced observational evidence of the superiority of varenicline in routine NHS practice. The recent HTA-funded ELONS (Evaluating Long-term Outcomes of NHS Stop Smoking Services) study showed that the OR for use of varenicline compared with non-use of varenicline was 1.7 (95% CI 1.3 to 2.3),93 very similar to the OR of 2.2 in this study. A Cochrane network meta-analysis of trials gave an OR of 1.57 (95% credible interval 1.29 to 1.91) for varenicline compared with single-form NRT and an OR of 1.59 (95% credible interval 1.29 to 1.96) for varenicline compared with bupropion.6 A subsequent very large trial produced ORs of 1.68 (95% CI 1.46 to 1.93) for varenicline compared with the NRT patch and 1.75 (95% CI 1.52 to 2.01) for varenicline compared with bupropion.94 The randomised trials suggest that the superiority of varenicline in these observational data is probably because varenicline is more effective both within and outside the context of a clinical trial. This suggests that a way to improve the outcome of the NHS Stop Smoking Services is to increase the proportion of patients who use varenicline. The ORs of 1.7 and 2.2 for use of varenicline compared with non-use of varenicline compare favourably with those for preloading (1.25 for the primary analysis and 1.34 for the varenicline-adjusted analysis); the use of varenicline seems simpler, and requires less service change, than implementing a new strategy of preloading.
It is important to note that these conclusions relate to preloading as we implemented it in this trial. It remains possible that preloading performed in other ways or in contexts other than the UK NHS Stop Smoking Services, which provide a range of medications and behaviour support free of charge, may improve outcomes to a greater extent. We used a 4-week period of preloading, which, theoretically, ought to be more effective than the 2 weeks used in several previous studies. However, there is no clear evidence that this is so; therefore, limiting preloading to a shorter period could overcome the possible disadvantages that significantly delaying a quit attempt may entail. Theoretically, it is also possible that higher doses of nicotine could lead to greater effectiveness.95 The additional 21 mg of nicotine per 24 hours preceding quit day was generally well tolerated and other research shows that 63 mg of nicotine per day appears to be well tolerated too.91 In addition, the higher-dose NRT reduced consumption to a greater extent than the 21-mg patch that was used, suggesting that higher doses may be effective.
It is also possible that varenicline preloading may be effective. One small trial showed that varenicline preloading more than doubled the number achieving 12-week abstinence.21 A further trial produced somewhat less positive results, but the trial was too small to offer a precise estimate of effect.96 Consequently, there is no evidence that varenicline preloading improves long-term abstinence, but, given that the RR of achieving abstinence through use of a smoking cessation medication appears to be constant, whether measured early or late in a quit attempt,97 we might expect that it does. Thus, future trials of preloading might want to examine whether or not this promising effect of varenicline holds in a larger trial and manifests in an increased proportion of participants achieving long-term abstinence.
Mediation
Summary of findings
There was no evidence that preloading was more effective for those showing a greater dependence on smoking. Participants allocated to preloading compared with the control condition showed a reduced reward from smoking, a reduced urge to smoke and reduced smoke inhalation prior to quitting and lowered their cigarette consumption. After attempting abstinence, participants using preloading experienced a reduced craving for cigarettes. Of these possible mediators, only a reduced urge to smoke before and after quit day and reduced consumption while using preloading were associated with abstinence. Together, these changes appeared to mediate most of the effect of preloading on abstinence. Although preloading increased feelings of aversion to smoking, these feelings were not associated with an increased likelihood of achieving abstinence and, thus, there is no evidence that this was an important mediating mechanism. There is also no evidence that the effect of preloading on abstinence was explained by increased confidence in quitting or increased adherence to post-cessation medication.
Strengths and limitations
This study has several strengths. Chief among these are that we planned a comprehensive evaluation of the mechanism of action, assessing all steps in the pathway, and we included analyses of competing hypotheses that were advanced to explain the preloading effect. Being the largest trial to date and the trial with the most comprehensive data on mediators means that the pathway of effect has been clarified. The trial also has some limitations. Our follow-up of participants to assess the effect of preloading on mediators while smoking was carried out only 1 week after commencing treatment. This follow-up was timed to maximise adherence to preloading and to allow the dose to be adjusted if needed. However, previous trials have suggested that the effect of preloading on smoking increases with duration;21,96 therefore, we may have seen larger associations between preloading and change in mediators and between change in mediators and abstinence if this assessment had taken place later. Another issue is that our analysis does not account for the multistep nature of mediation. Our hypothesis was that preloading reduces the strength of urges to smoke while in the pre-quit phase and these reduced urges reduce cigarette consumption. However, we modelled these mediators as separate ‘parallel’ mediators rather than in a structural equation model. Finally, the open-label trial leaves open the possibility that some of these effects are subject to bias, such as participant expectations. Participants in the intervention group were told about the proposed mechanism of preloading to motivate them to adhere to the medication. Thus, the effect of preloading on the proposed mediators could at least partly be caused by this effect rather than by the pharmacological action of the additional nicotine. However, we might have seen an increase in confidence in the preloading group if this was driven by expectations of a positive outcome and this was not apparent. It, therefore, seems less likely that participants’ expectations could explain the increase in abstinence and, therefore, this may be an unlikely explanation of the apparent mediation.
Comparison with other studies
Our analysis of whether or not the effect of preloading depends on baseline tobacco dependence is unique. We used two markers of dependence, FTCD and exhaled CO concentration. In our study, as in previous studies, for example Aveyard et al.,36 FTCD predicted the likelihood of achieving abstinence. The ORs were 0.85 (95% CI 0.62 to 1.17) for moderate dependence compared with low dependence and 0.61.(95% CI 0.42 to 0.90) for moderate dependence compared with high dependence. The p-value for trend through those categories is 0.013. Thus, in this population of people who seek help to stop smoking and who are more dependent than the average person who smokes, this marker functions as a predictor of abstinence and might, therefore, be expected to moderate the effect of preloading. However, exhaled CO concentration did not function as a useful predictor of the likelihood of abstinence; compared with a CO concentration of < 10 p.p.m. at baseline, for the categories in Figure 5b, the ORs were 0.67, 0.63, 0.79 and 0.74, with a p-value for trend of 0.84. This may, therefore, explain why exhaled CO concentration did not moderate the association.
Our mediation analysis is complemented by previous analyses in other trials of preloading that examined the effect of preloading on several potential mediators. Unlike previous analyses,12 we found that preloading did reduce both positive and negative reward from smoking, but our study is much larger than the previous studies that have examined these mediators, so these trials may have missed this association. As two previous studies have also discovered,12 we found a small reduction in pre-quit urges to smoke. Most previous trials examined changes in cigarette consumption caused by preloading and, as we found, noted reductions in the pre-quit period.12 Those trials that examined reduction in dependence reported the same kind of reductions that we observed in this trial. Previously, we concluded that preloading did not appear to reduce craving after quit day,12 but there was clear evidence in this trial that preloading did reduce craving. Although previous studies examined the effect of preloading on potential mediators, none undertook a full mediation analysis, as we have done here. Doing so revealed new insights. Although people’s perception of reward may have changed, the main apparent mechanism seems to be through reduction in the urge to smoke. It is possible that this effect is indeed mediated by reward, but that people’s perception of it may not be very sensitive. Alternatively, the drive to smoke may weaken in smokers saturated with nicotine from an alternative source without any effect on their subjective reactions when they do smoke. The main drive to smoke does not require conscious appreciation for it to be experienced.98 No previous studies have examined whether or not there is effect modification by strength of dependence on tobacco.
Interpretation
This study and previous literature together have produced clear evidence that preloading reduces the intensity of urges to smoke and that it reduces the intensity of smoking during the pre-quit period. The reduced drive to smoke appears to be the best candidate mechanism for the effect of preloading, with reduced smoke intake being one of its consequences.
This has important implications for clinical practice. If preloading were to be used routinely in NHS Stop Smoking Services, then advisors could monitor its effect using markers of urge strength, asking users whether or not they have experienced reduced urge intensity, and also by measuring exhaled CO concentration, which is routine in NHS clinics. This could allow clinicians to advise their patients to abandon preloading if it appears to be having little effect, confident that the evidence now supports this.
Many trials of treatment for tobacco dependence enrol people who smoke a minimum of 10 cigarettes per day, but around one-third of the population who smoke consume fewer cigarettes than this per day. We enrolled such participants if they were seeking help to quit and smoked at least one cigarette per day, providing they manifested other features of dependence, including difficulty attaining abstinence. This may have contributed to including a greater range of participants than in some trials of pharmacotherapy. Including a wide range of participants should make it easier to detect effect modification between markers of dependence. Although there was no clear evidence of effect modification, both point estimates suggested a greater relative benefit in people with a higher level of dependence. As we did not plan for the study to have power to detect this interaction, it may be that we simply failed to find evidence for an effect, rather than this representing clear evidence against an effect. Given the mechanisms of action of preloading and this evidence together, it would seem appropriate for smoking cessation advisors to consider this strategy in anyone seeking treatment, but perhaps particularly for people with a higher level of dependence.
Within-trial cost-effectiveness
Summary of main findings
The primary outcome of the cost-effectiveness analysis was the incremental cost per additional case of 6-month prolonged abstinence. The prolonged abstinence rate at 6 months post quit date according to the Russell Standard was 17.5% in the intervention group and 14.4% in the control group, a difference of 3.1% (95% CI –0.3% to 6.4%). Mean total costs, from randomisation to 6 months post quit date, were £978 (SE £105) in the intervention group and £1090 (SE £110) in the control group. The adjusted incremental costs were £4 (95% CI –£188 to £222). The incremental costs per additional 6-month prolonged abstinence were, therefore, £129 (95% CI –£16,527 to £17,782). The wide CI crossed the zero point, which amounts to no clear evidence that the intervention was more cost-effective than the control condition. The analysis of data at 12 months’ follow-up suggested a possible reduction in health service costs in the intervention group (–£233, 95% CI –£747 to £42). This pointed to a potential cost-saving scenario in the long run, given that the abstinence rate at 12 months was not significantly different between arms. Sensitivity analysis did not change the pattern of these results. There were indications that the costs might have been underestimated in the complete-case analysis.
Strengths and limitations
The strength of this analysis lies in the comprehensive analysis of all health and social care costs, thus providing a clear picture for the NHS when deciding on treatment. As in all such studies, however, we suffered from missing data. In particular, people who try to quit smoking with support and fail to do so are always somewhat reluctant to return for follow-up. For the sake of the primary outcome, we maximised the return of data on smoking status among participants who were reluctant to be followed up. This sometimes meant that we excused people from answering the health resource questions and, as a result, these data were more often missing than data on smoking status. As the data showed, people with complete data were more likely to be abstinent, but there is no reason to imagine that health service use differed between arms among people who were not abstinent and this is unlikely to bias the results.
With the complicated nature of health resource use data, it was inevitable that some assumptions had to be made before any analysis could be undertaken. The less than reliable approach of self-reporting only added to the uncertainties. However, short of accessing participants’ usage directly from the providers, it was the best way to collect this information. The assumption that no response equates to no use is generally seen to be compatible with the way that people complete health resource use questionnaires. The assumptions made about monthly prescriptions are less certain. The prescribing patterns varied between GPs and, without more detailed information, it was not possible to estimate the cost of medication more accurately. Collecting these data from participants directly in a trial of this kind would add a considerable burden and probably increase withdrawal rates. Even if this burden could be disregarded, participants’ reports of medication use are often less than satisfactory for the purpose of costing. In spite of these uncertainties, the assumptions were applied to both arms and, therefore, should not bias the results towards either group.
Participants with complete information on health resource use had much lower use than that suggested from the imputed data. This is not surprising: participants with greater health needs may find it more burdensome to complete questionnaires, be less likely to achieve abstinence and be more likely to drop out of follow-up. However, this effect is unlikely to differ between arms.
The data presented are a within-trial analysis of cost-effectiveness. The reason that people try to stop smoking, and the reason the NHS invests in smoking cessation services, is to support long-term abstinence and to avoid the costs of smoking-related illnesses in the future. It is not possible to collect such data in trials, but it is possible to model them and we present these data in Chapter 7. A second issue relates to the design of the trial itself and the constraints that this imposed on the cost-effectiveness analysis. To be able to assess AEs and the mechanism of action of preloading in an unbiased way, those in the control group delayed their quit attempt to the same extent as those in the preloading group. To ensure that people returned to give us these data, we gave behavioural support to those in the control group and set this group up to believe that this would be useful to them. There is no clear evidence that this is the case and, in any case, the support that we gave is commonly incorporated into standard NHS smoking cessation support, which both the intervention group and the control group received later. In this within-trial analysis, we included the cost of the behavioural support in the control group. The NHS decision problem, however, was to assess the difference between preloading and standard NHS smoking cessation support, with only a minimal period of preparation before quitting. Thus, the control group behavioural support sessions would not happen in any future NHS scenario. Assuming that the outcomes remain the same, the simple exclusion of two behavioural support sessions from the control group would result in an increase in the adjusted difference in costs between arms and an increase in the ICER in turn. For the complete-case analysis, the difference in costs between trial arms would increase from £138 to £155, resulting in a modest change in the ICER of around £700, whereas the change in the adjusted difference in costs between arms in the imputed data (from £4 to £17) would lead to an increase in the ICER of £550. However, the assumption that extra behavioural support in preparation to quit has no impact on outcomes cannot be easily made. These numbers are only estimates and further assessment is required to provide supporting or opposing evidence. In addition, the cost-effectiveness analysis appropriately focuses on the primary analysis of the primary outcome. However, this is a biased estimate of the efficacy of preloading because it is confounded by differences in the use of varenicline. Had we used the estimate for the difference in the effect of preloading adjusted for varenicline, then the cost-effectiveness might have more strongly favoured preloading over the control.
Interpretation
This is the first assessment of the cost-effectiveness of nicotine preloading. Indeed, relatively few trials have conducted within-trial analyses of the cost-effectiveness of smoking cessation medication. One comparator is the NHS Stop Smoking Service, which collects data on the cost per person who has quit smoking and remained abstinent for 4 weeks. Statistics on NHS Stop Smoking Services in England April 2014 to March 201565 showed, among local authorities, a cost per person who quit smoking ranging from £98 to £1185, with a mean of £456. The cost per person who quit smoking at 4 weeks in this trial was £501 in the preloading group and £417 in the control group, somewhat in the middle of the distribution. That said, the NHS counts self-reported abstinence as abstinence, which we excluded; if we had included self-reported abstinence, it would reduce the cost per quitter by nearly one-third.
Interpreting the data here is harder because there is no available WTP threshold for 6-month prolonged abstinence; therefore, it is not possible to draw clear conclusions with regard to whether or not the intervention can be seen to be cost-effective. An incremental cost of £129 per additional 6-month abstinence might be considered cost-effective if the policy-makers judge the abstinence to be beneficial, given the impact of smoking cessation on long-term quality of life. What may encourage policy-makers to consider preloading are the data showing that, at 12 months, preloading reduced overall health service costs. In the short term at least, smoking cessation should reduce the demand for health services by reducing the incidence of both minor and major ill-health consequences of smoking. Future modelling may provide further data to support this.
Modelling the long-term cost-effectiveness of preloading
Summary of findings
The smoking cessation costs within the intervention group over the course of the first year were higher than those in the control group, with intervention group costs of £295, compared with control group costs of £265. However, more smokers quit smoking in the intervention group (n = 139) than in the control group (n = 112) and consequently health-care costs associated with smoking were lower in the intervention group over the course of the lifetime. Overall, the total costs were £29 lower in the intervention group. The intervention was also more effective than the control condition, resulting in 14.0% of smokers in the intervention group quitting at 1 year compared with 11.3% of smokers in the control group, and consequently produced more QALYs and life-years. The intervention would be considered cost-effective as it resulted in greater efficacy at a lower cost than the control condition. The results were robust to changes in assumptions regarding discount rates and relapse after the first year of follow-up.
Strengths and limitations
The major strength of this approach is that it builds on the comprehensive within-trial cost-effectiveness assessment presented in Chapter 6. However, it takes the same perspective as people who are stopping smoking: that the main reason to stop smoking is to avoid future smoking-related disease.99 As such, it better represents the views of possible service users of NHS Stop Smoking Services and of the community at large. The major weakness is that, as with practically all long-term cost-effectiveness studies, the results are modelled on the future likelihood of disease. Confidence in this modelling is supported by the robust epidemiological evidence available on the association between smoking and incidence of and mortality from disease and the difference in risk arising as a result of stopping smoking at different ages. The model has been extensively discussed and described elsewhere (forthcoming supplement in the journal Addiction) and is used by NICE to examine the cost-effectiveness of smoking cessation medications.67 However, nobody has prospectively validated models of incidence, mortality and health-care expenditure of cohorts of people who continue or stop smoking, so ultimately the data should be regarded as somewhat conjectural. This emphasises the importance of sensitivity analyses and a strength of the findings here is that they appear largely unaltered by both deterministic and probabilistic sensitivity analyses; all answers converged to suggest that preloading is dominant over the comparator in the large majority of model runs.
Research recommendations
- Will changes to NICE guidance and interventions with clinicians and/or patients overcome the effect of preloading on the subsequent use of varenicline?
- Are higher doses of nicotine preloading more effective, especially in those who do not seem to suppress the urge to smoke in response to the standard dose?
- Are people who reduce their urge to smoke in response to preloading more likely to stop smoking than those who experience no reduction in the urge to smoke?
- Is reduced urge to smoke translated into reduced consumption of cigarettes? Would this work as a marker of the effectiveness of nicotine preloading?
- Is response to use of cessation pharmacotherapy while in the period prior to quit day an effective way of personalising treatment for people seeking help to stop smoking?
- Does varenicline preloading increase the proportion of people achieving long-term abstinence? Is varenicline preloading cost-effective in the long term?
Recommendations for practice
Using nicotine preloading requires a person to make a planned quit attempt. Our trial shows that, when there is dislocation between the person prescribing the NRT preloading and the subsequent smoking cessation service, the choice of subsequent pharmacotherapy undermines the benefit of preloading. The most likely scenario for implementation in the NHS is that a GP would refer a patient to NHS Stop Smoking Services, which replicates the scenario in our trial. In such a circumstance, preloading is unlikely to add greatly to the effectiveness of standard care and we would not recommend its use.
In other circumstances, it is possible that preloading would prove to be effective. A clinician looking to maximise the chances of successful cessation in a patient would prescribe varenicline as post-cessation medication. Both observational data from our trial, other observational data and trial data (discussed in Modelling the long-term cost-effectiveness of preloading) testify to the superiority of this medication. Suggesting this to a patient first and gaining commitment to its use before discussing the use of preloading may overcome the deterrent effect of preloading on the subsequent use of varenicline.
Conclusion
Use of nicotine patch preloading for 4 weeks prior to attempting to stop smoking can increase the proportion of people who stop successfully, but its benefit is undermined because it reduces the use of varenicline after preloading. If this latter effect could be overcome, then nicotine preloading appears to reduce health service costs in the long term. Using the primary outcome of 6-month prolonged abstinence, there is no clear evidence that preloading is more cost-effective than the control condition, but there is a suggestion that over a 12-month period it may be more effective and it is very likely to be cost-effective over a lifetime perspective.
Nicotine preloading appears to work because it reduces the strength of urges to smoke during preloading and after preloading has finished. This seems to be the central mechanism underlying its effect. None of the other theories proposed for its mechanism of action have any support and are likely to be part of the mechanism of action.