- Email: [email protected]
Comparison of the Effectiveness of Varenicline and Combination Nicotine Replacement Therapy for Smoking Cessation in Clinical Practice
ORIGINAL ARTICLE
Comparison of the Effectiveness of Varenicline and Combination Nicotine Replacement Therapy for Smoking Cessation in Clinical Practice Leonie S. Brose, PhD; Robert West, PhD; and John A. Stapleton, MSc Abstract Objective: To compare the effectiveness of varenicline and combination nicotine replacement therapy (NRT) for treating smokers attempting to stop, as it is unclear which of the 2 is more effective. Patients and Methods: Data from 167,487 treatment episodes in patients from 42 National Health Service Stop Smoking Services (April 1, 2009, through June 30, 2011) using varenicline or combination NRT were analyzed. The outcome was carbon monoxideevalidated 4-week smoking abstinence. Potential predictors were age, sex, occupational grade, exemption from prescription charges, intervention setting (specialist or other), support (group or other), and year of quit attempt. Results: Observed smoking abstinence rates were higher with varenicline (43.5% vs 36.9%). However, there was evidence of systematic variation in medication effect across clinical services and differences in predictors of outcome between medications. Allowing for these influences indicated a small mean advantage for varenicline (odds ratio, 1.080; 95% CI, 1.003-1.163; difference, 1.86%; 95% CI, 0.07%3.67%; P¼.04). The relative effectiveness of the 2 medications was not moderated by setting, type of support, or year. Conclusion: The relative effect of medication varies substantially according to clinical practice and population treated. Averaged across current English clinical practice and populations, varenicline is marginally more effective than combination NRT. Demographic and intervention characteristics associated with success predict varenicline use. ª 2013 Mayo Foundation for Medical Education and Research From the National Centre for Smoking Cessation and Training and Department of Clinical, Educational, and Health Psychology (L.S.B.) and Department of Epidemiology and Public Health (R.W., J.A.S.), University College London, London, England.
226
T
obacco is a leading cause of death, illness, and impoverishment.1 Smokers wanting to quit can be supported by medication. Evidence-based medication options for treating smokers attempting to stop include nicotine replacement therapy (NRT), the atypical antidepressant bupropion, and the nicotinic acetylcholine receptor partial agonist varenicline.2 Nicotine replacement therapy is available in a range of forms, including longer-acting nicotine patches and faster-acting gums, lozenges, and inhalators. Nicotine replacement therapy is used as a single product or a combination of products,2,3 usually patches with one of the faster-acting forms. Bupropion and varenicline are taken as a tablet once daily for the first week and twice daily for the rest of the course. Compared with no medication or a placebo, these medications have all been reported to increase the chance of quitting successfully in
n
Mayo Clin Proc. 2013;88(3):226-233
randomized controlled trials (RCTs).3-5 This leaves open the important question of which is most effective for smokers who have the option to use any of them. A single NRT product and bupropion have similar efficacy in RCTs,5 and there appears to be little difference in their relative effectiveness in clinical practice.6 However, varenicline has been reported to be more efficacious than bupropion in RCTs(relative risk [RR], 1.52; 95% CI, 1.221.884), and there is also evidence from RCTs (RR, 1.13; 95% CI, 0.94-1.354,7) and clinical practice (odds ratio [OR], 1.78; 95% CI, 1.572.026; and OR, 1.96; 95% CI, 1.15-3.338) that varenicline is more efficacious and effective than a single NRT product. Similarly, the combination of 2 or more forms of NRT has been reported to be more efficacious than a single NRT product in clinical trials (RR, 1.35; 95% CI, 1.11-1.633), and there is evidence that it is
Mayo Clin Proc. n March 2013;88(3):226-233 n http://dx.doi.org/10.1016/j.mayocp.2012.11.013 www.mayoclinicproceedings.org n ª 2013 Mayo Foundation for Medical Education and Research
VARENICLINE VERSUS COMBINATION NRT
more effective in clinical practice (OR, 1.42; 95% CI, 1.06-1.916,9). Although the evidence indicates that both varenicline and combination NRT are more effective than other medications, to date there has been no direct comparison of these 2 therapies to establish which is more effective. Placebo-controlled RCTs comparing varenicline and combination NRT would be impracticable and probably too costly to mount. In practice, many combinations of 2 or more NRT products are used, and it would be difficult to include all these in a trial. Restriction to just one combination would seriously limit generalizability. In addition, costs would be prohibitive because many thousands of participants would be required to detect a likely small difference. However, an opportunity for a comparison in a real-world setting is available using data provided by the Stop Smoking Services of England’s National Health Service (NHS). These services offer the full range of licensed pharmacologic treatments in combination with behavioral support. Each year, more than 750,000 quit attempts are made with the help of these services.10,11 At the start of a treatment episode the type of pharmacotherapy used is decided by the smoker after consultation with the support therapist and is subject to any contraindications and cautions that might apply. The aim of the present study was to compare the effectiveness of varenicline and combination NRT as practiced within the Stop Smoking Services of England. We also examined whether there were factors related to which of the 2 medications was used and whether any difference in medication effectiveness was dependent on (1) whether the smoker had been treated in a specialist or nonspecialist setting, (2) whether the behavioral support had been provided on a one-to-one basis with a stop smoking practitioner or in a group with other smokers trying to quit, and (3) the year of treatment. METHODS Sample Data were obtained from QuitManager (North 51), an online database system for recording information on patient demographic features, intervention characteristics, and outcomes in accordance with the Department of Health’s standard monitoring requirements.2 Of the Mayo Clin Proc. n March 2013;88(3):226-233 www.mayoclinicproceedings.org
n
152 services in England, 58 were using this database at the time of the present audit, and 49 agreed to share their anonymized data. All completed treatment episodes from April 1, 2009, through June 30, 2011, were included as individual records. As defined by the Department of Health for England and Wales,2 a treatment episode begins when a smoker sets a quit date with the service and ends 4 weeks after the quit date, when the short-term success of the quit attempt is determined. The data represent completed treatment episodes with quit dates between April 2009 and June 2011 for which either varenicline or combination NRT was used (189,795 of 313,193 episodes [60.6%]). To test for differences in medication effect according to the type of support given, the analysis was restricted to the 42 services (of 49) that provided both individual and group support options (20,965 episodes [11.0%] not included). Data were not recorded on key variables for 1434 (0.8%) episodes, and these episodes were excluded to leave a total of 167,487 treatment records for the current study. The clinical services we were able to include in this analysis appeared to be representative for all services in England in terms of patients seen and interventions provided (Table 1). They were also representative in terms of deprivation.12 The index of multiple deprivation is a weighted area-level aggregation of 7 specific domains of deprivation (each derived from a number of indicators): income deprivation; employment deprivation; health deprivation and disability; education, skills, and training deprivation; barriers to housing and services; crime; and living environment deprivation. The least deprived service is ranked as 1, the most deprived as 151; thus, the mean rank is 76. The 42 services included in the present sample ranked from 1 to 149, with a mean rank of 76.5, thus providing a good representation of all services in England. Outcome Measure The outcome measure was successful quitting as defined by the Department of Health criterion (ie, the attempt is successful if the smoker reports continuous abstinence from smoking between weeks 2 and 4 and records an expired-air carbon monoxide reading of <10 ppm at the 4-week follow-up).2 As is
http://dx.doi.org/10.1016/j.mayocp.2012.11.013
227
MAYO CLINIC PROCEEDINGS
TABLE 1. Patient and Intervention Characteristics in the 42 Services From Which the Present Sample Was Drawn and in All English Servicesa,b
Characteristic Female White ethnicityc Occupational group Full-time student Never worked/long-term unemployed Retired Home carer Sick/disabled and unable to work Managerial/professional Intermediate Routine and manual In prison Unable to code Exempt from prescription charge Medication used No medication/unknown Single NRT Combination NRT Bupropion (Zyban) Varenicline (Champix) Intervention setting Specialist clinic Primary care Pharmacy Other Intervention type One-on-one Drop-in Open group Closed group Telephone Other 4-Week quit rate Carbon monoxide validated Self-reported
Patients (%) (N¼299,670)
All patients in England10,11 (2009-2010 and 2010-2011 combined) (%) (N¼1,545,064)
52.4 87.2
52.1 87.7
4.5
4.5
13.0 11.9 5.6
12.4 10.7 5.3
6.6 13.8 7.2 22.4 1.3 13.7
5.8 12.0 7.4 22.1 1.3 18.5
56.5
52.2
10.4 27.3 35.0 1.0 26.3
9.0 64.2d . 1.0 24.5
33.8 40.9 21.1 4.1
30.7 43.9 18.5 6.9
80.7 10.7 3.6 2.1 1.4 1.0
79.7 10.7 3.8 2.0 1.3 2.4
34.7 49.0
34.0 49.0
Baseline Characteristics Available demographic and intervention characteristics were included in the modeling as potential mediators and confounders. Demographics were sex, age at the quit date, exemption from being required to pay the NHS prescription charge (as a proxy measure of economic deprivation), and occupational status. For these analyses, occupational status was dichotomized into employed, which included those studying and those retired from their employment (professional/managerial, intermediate, routine/manual, full-time student, retired), and not employed (never worked/longterm unemployed, home carer, sick/disabled and unable to work, in prison, other). Tobacco dependence as measured by the Heaviness of Smoking Index (HSI), with higher scores (range, 0-6) indicating higher dependence,16 is an optional clinical measure for Stop Smoking Services. In this sample, the HSI score was recorded in 40,145 treatment episodes (24.0%). Intervention characteristics included were the type of behavioral support (group support vs other support types), intervention setting (specialist clinic vs other setting), which have both been associated with cessation rates,6 and the year of the quit attempt.
NRT ¼ nicotine replacement therapy. Missing data in 42 services: sex: n¼60 (0.02%) missing; intervention setting: n¼225 (0.07%) missing; and intervention type: n¼1643 (0.55%) missing. No missing data reported for all services in England. c Ethnicity is self-reported using standard UK Census categories. d Single and combination NRT not reported separately. a
b
standard practice, those who did not report their smoking status or did not attend the session for carbon monoxide verification at 4-week follow-up were regarded as continuing smoking.2,13 Approximately 75% of carbon monoxideevalidated 4-week quitters will relapse 228
Mayo Clin Proc.
within the following year, most within the first 6 months.14,15 However, relative smoking abstinence rates have been found to be stable over short- and long-term follow-ups,15 and we can therefore have reasonable confidence that effects observed in the short term will translate to the long term.
n
Statistical Analyses Because the effects of interest are nested within 42 clinical services having potentially varying practices and populations, 2-level logistic regression models were used with random effects for clinical service (level 2). The odds of success for varenicline compared with combination NRT were estimated, adjusting for demographic and intervention characteristics. Single-level logistic regression models were used to obtain the adjusted OR for medication comparisons within each service. Analyses were conducted using SPSS statistical software, version 18 (SPSS Inc), and MLwiN software, version 2.14.17 The significance level was set at P<.05 for all analyses, and 95% CIs were calculated. Differences between the distributions of individual
March 2013;88(3):226-233
n
http://dx.doi.org/10.1016/j.mayocp.2012.11.013 www.mayoclinicproceedings.org
VARENICLINE VERSUS COMBINATION NRT
demographic and intervention characteristics for the 2 medication options were calculated using either the t test or c2 test, followed by a 2-level multivariable logistic regression model to examine predictors of medication use. The HSI score for dependence was not included in the modeling analyses because of the small proportion of cases for which it was recorded. Instead, the likely effect of dependence on outcome and differences between the 2 medications was assessed separately. Ethical approval was not required for this audit of anonymous NHS clinical data. RESULTS Patient Characteristics Combination NRT was used more frequently (57.4%, n¼96,174) than varenicline (42.6%, n¼71,313). The measured demographic and intervention characteristics were significantly different between those receiving varenicline and combination NRT (Table 2), which is to be expected in a sample of this size when using c2 statistics and t tests. Characteristics that have been found to be associated with successful quitting6 were more frequent in those who received varenicline compared with those treated with combination NRT. In the case of age and sex, the differences were of modest clinical significance. However, more patients receiving varenicline were employed, not exempt from paying
NHS prescription charges, and treated in a specialist clinic and in a group, all of which are associated with higher success rates. Dependence was the only characteristic with an opposing pattern. In the subsample of cases with HSI scores recorded, those receiving varenicline had a higher level of tobacco dependence, which is associated with less successful quit attempts. Predictors of Medication Selection Systematic differences in the characteristics of those receiving the 2 medication options were further examined in a multivariable model (Table 3). After adjusting for all other characteristics and allowing for variation across the clinical services, all characteristics remained significant predictors of whether varenicline or combination NRT had been used in the treatment episode, with the single exception of intervention setting. Varenicline treatment was associated with being male, being older, being employed, not being exempt from paying the NHS prescription charge, receiving group support, and being treated in 2009 (rather than 2010 or 2011). Outcome Overall, 39.7% of patients were carbon monoxideevalidated abstinent from smoking at 4-week follow-up. As anticipated, some characteristics associated with a greater likelihood of receiving
TABLE 2. Distribution of Demographic and Intervention Characteristics in Those Treated With Varenicline and Combination NRTa,b Characteristic
Varenicline (n¼71,313)
Combination NRT (n¼96,174)
Comparison statistic
Mean (SD) age (y) Male Employed NHS prescription charge Paying for prescription Exempt from payment Undisclosed Specialist clinic Group intervention Treatment year 2009 2010 2011 Mean (SD) HSI scorec
43.2 (13.4) 34,851 (48.9) 49,966 (70.1)
42.3 (15.3) 44,808 (46.6) 54,153 (56.3)
t¼14.1, P<.001 c21 ¼85.3, P<.001 c21 ¼3296, P<.001
30,731 33,156 7426 28,482 6587
(43.1) (46.5) (10.4) (39.9) (9.2)
24,958 64,778 6438 34,435 3988
(26.0) (67.4) (6.7) (35.8) (4.1)
14,142 31,602 25,569 3.51
(19.8) (44.3) (35.9) (1.35)
16,169 42,273 37,732 3.35
(16.8) (44.0) (39.2) (1.43)
c22 ¼7351, P<.001 c21 ¼298, P<.001 c21 ¼1794, P<.001
c22 ¼331, P<.001 t¼11.4, P<.001
HSI ¼ Heaviness of Smoking Index; NHS ¼ National Health Service; NRT ¼ nicotine replacement therapy. Data are presented as No. (percentage) of patients unless otherwise indicated. c The HSI scores were recorded in 40,145 cases (24.0%). a
b
Mayo Clin Proc. n March 2013;88(3):226-233 www.mayoclinicproceedings.org
n
http://dx.doi.org/10.1016/j.mayocp.2012.11.013
229
MAYO CLINIC PROCEEDINGS
TABLE 3. Predictors of Varenicline vs Combination Nicotine Replacement Therapy Usea Predictor
Odds ratio (95% CI)
P value
Male vs female Age per 10-y increase Employed vs not employed National Health Service prescription charge Exempt vs pays Undisclosed vs pays Specialist clinics vs other setting Group support vs other support Quit year 2010 vs 2009 2011 vs 2009
1.05 (1.03-1.08) 1.08 (1.07-1.09) 1.31 (1.28-1.34)
<.001 <.001 <.001
0.43 1.15 1.01 1.30
(0.42-0.44) (1.10-1.20) (0.99-1.04) (1.23-1.37)
<.001 <.001 .39 <.001
0.81 (0.79-0.84) 0.66 (0.64-0.69)
<.001 <.001
a
Two-level multivariable logistic regression. Level 1 is treatment episode, and level 2 is Stop Smoking Service.
varenicline treatment were also independently prognostic of success. Being male, older, and employed, paying for medication prescriptions, and receiving group support all increased the odds of success (Table 4). In the subsample with HSI data, a higher HSI score, which was associated with varenicline treatment, was predictive of failure (OR, 0.936; 95% CI, 0.922-0.949). The observed success rate was 36.9% with combination NRT and 43.5% with varenicline (absolute percentage difference, 6.6%) For combination NRT and varenicline, 23.3% and 20.6%, respectively, failed to attend the 4-week follow-up and were assumed to be smoking. However, despite within-service adjustment for all measured patient and treatment characteristics in the 2-level model, there remained
TABLE 4. Predictor Variable Model for Carbon MonoxideeValidated 4-Week Abstinencea Predictor
Odds ratio (95% CI)
P value
Male vs female Age per 10-y increase Employed vs not employed National Health Service prescription charge Exempt vs pays Undisclosed vs pays Specialist clinics vs other setting Group support vs other support Quit year 2010 vs 2009 2011 vs 2009
1.00 (0.98-1.02) 1.20 (1.19-1.21) 1.24 (1.22-1.27)
.92 <.001 <.001
0.81 0.80 1.01 1.39
<.001 <.001 .31 <.001
a
(0.79-0.87) (0.77-0.83) (0.99-1.04) (1.32-1.46)
0.97 (0.94-0.99) 1.03 (0.99-1.06)
.02 .14
Two-level multivariable logistic regression. Level 1 is treatment episode, and level 2 is Stop Smoking Service.
230
Mayo Clin Proc.
n
evidence of large systematic differences in the relative effect of medication across the services (c241¼385; P<.001). The Figure shows the adjusted ORs for treatment success in favor of varenicline in each of the 42 clinical services. In 20 services varenicline was more effective than combination NRT, in 16 services there was no significant difference, and in 7 services varenicline was less effective than combination NRT. The ORs ranged from 1.96 in favor of varenicline to 3.13 in favor of combination NRT. The effect averaged across services in the 2level model indicated a small benefit for varenicline (OR, 1.08; 95% CI, 1.003-1.163). Applying this OR to the adjusted treatment rates gives an adjusted RR of 1.047 (95% CI, 1.002-1.092) and an adjusted absolute percentage difference of 1.86% (95% CI, 0.07%-3.67%). In models that included terms representing interactions between medication and intervention characteristics, there was only weak evidence that intervention setting (P¼.08), type of support (P¼.30), or year of treatment (P¼.09) moderated the relative effect of medication. DISCUSSION To our knowledge, this is the first systematic comparison of varenicline and combination NRT in smoking cessation, using a very large sample from real-world clinical practice. Treatment using varenicline compared with combination NRT in the English NHS Stop Smoking Services was associated with a small mean increase in short-term smoking abstinence rates. The mean benefit (adjusted absolute percentage difference) was less than 2%, which would suggest an additional 0.5% permanently quitting on the assumption of equivalent relapse rates over time and equivalent rates of subsequent quit attempts between the 2 treatments in this setting.15,18 Although a small effect, it would translate to 500 additional quitters for every 100,000 treated with varenicline rather than combination NRT in services and patients similar to the present sample. This benefit was not significantly affected by whether the treatment episode had taken place in a specialist clinic or nonspecialist setting, by whether group or one-on-one behavioral support had been given, or by the year in which the quit attempt had taken place. However, the selection of medication was found to be associated with demographic and intervention characteristics that
March 2013;88(3):226-233
n
http://dx.doi.org/10.1016/j.mayocp.2012.11.013 www.mayoclinicproceedings.org
VARENICLINE VERSUS COMBINATION NRT
Adjusted OR in favor of varenicline
3.0
2.5
Adjusted OR and 95% Cl for each service Overall adjusted OR and 95% Cl
2.0
1.5
1.0
0.5
0.0
FIGURE. Adjusted odds ratios (ORs) for single-level models of carbon monoxideevalidated 4-week abstinence. Error bars indicate CIs.
were also predictive of success or failure of the attempt, and the relative medication effect varied considerably across services. On the basis of our findings of relatively similar effectiveness and the relatively similar cost of the 2 medications (approximately £160 or $250 for varenicline and £140-£180 or $220-$280 for combination NRT), a substantial difference in cost-effectiveness is unlikely. Studies in a number of countries have confirmed the cost-effectiveness of varenicline compared with NRT19; therefore, cost should not be a significant factor in future prescribing decisions. The present findings confirm earlier results that reported group interventions to be more successful than one-on-one interventions and that socioeconomic variables (eg, paying prescription charges and being employed) are positively associated with success.6 Previous findings revealed specialist clinics to be more successful than interventions set in primary care; this finding has not been replicated here. It is possible that the restriction of the sample to interventions using varenicline or combination NRT has reduced the effects of settings. An overarching feature of this study was the finding of systematic variation among the 42 clinical services in the relative effectiveness of the 2 medications. The size of the systematic variation in effect size between the services was not anticipated. Although in almost half of services there was a significant benefit for varenicline, in Mayo Clin Proc. n March 2013;88(3):226-233 www.mayoclinicproceedings.org
n
a substantial few (7/42 [17%]) there was a significant benefit for combination NRT. Such differences are unlikely to be the result of population differences. They were evident after controlling for the included individual characteristics predictive of both medication selection and outcome, and although it is possible that some other unrecorded measures were also predictive, it is difficult to envisage individual factors that would exert such a large influence as to explain the differences we observed. Instead, the differences are likely to be attributable to differences in clinical practice, with some practices tending to favor varenicline being more successful and others favoring combination NRT being more successful. We did not have data on the style of treatment delivery in the 42 services and are therefore unable to empirically address this issue. Although national guidance on smoking cessation treatment should be equally applicable to all services, it is clear that delivery is not consistent. We can only speculate that there are likely biases in whether or not a full course of medication is prescribed, how well the behavioral support is aligned to the treatment, how therapists and patients perceive medication efficacy, or how adverse drug responses are treated. Reducing such differences to engender optimal treatment is a major task of the National Centre for Smoking Cessation Training (www.ncsct.co.uk). Although not sufficient to explain the major part of the relative differences in
http://dx.doi.org/10.1016/j.mayocp.2012.11.013
231
MAYO CLINIC PROCEEDINGS
medication efficacy across the services, we identified characteristics that were associated with medication selection. Those taking varenicline were more likely to be older, to be employed and therefore not exempt from paying the NHS prescription charge (currently approximately £7.50 or $12), and to have received group support, all of which were found to be associated with successful outcome. Again, it is unclear as to why there were such differences in treatment delivery. In the subsample with HSI scores, those taking varenicline were also more dependent on tobacco. Dependence on tobacco was an uncontrolled factor in the main analyses, being only recorded in a few episodes. Had the HSI score been available in all cases and had the same association with medication selection and outcome been evident, then inclusion of the HSI score in the model would slightly have improved the relative benefit of varenicline. A major strength of these analyses is that they are based on a very large national sample in real-world clinical practice. The included services were representative for all English NHS Stop Smoking Services. Limitations were the use of short-term 4-week cessation data as the outcome measure and the availability of only a small set of background characteristics. Regarding the outcome measure, a longer follow-up was not feasible because few services attempt to follow up smokers beyond 4 weeks and those who are followed up longer are likely to be unrepresentative. However, as noted in the “Methods” section, short-term relative cessation rates have been found to be predictive of longterm relative rates.14,15 In this analysis we were not able to identify individuals, only treatment episodes. Consequently, we could not model the effect of some individuals appearing with more than one treatment episode during the years of study. However, anecdotal evidence from the services would suggest that this was the case in a few episodes. Further, any bias is likely to be fairly neutral due to some reattending individuals using different treatments, in which case our analysis would be less sensitive to differences than it should be, and some individuals using the same treatment, in which case our analysis would be more sensitive to differences than it should be. Future research is needed to identify factors that may explain the heterogeneity in relative 232
Mayo Clin Proc.
n
effectiveness in clinical practice. Further studies are also needed to investigate the effectiveness of combining NRT and varenicline. To date, the safety and efficacy of this combination have only been researched in one small trial,20 which reported no benefit over the usual care. Another small trial has been completed, but its results have not yet been published (ClinicalTrials. gov identifier: NCT01184664). CONCLUSION Overall, there appears to be a small mean advantage for varenicline over combination NRT in clinical practice, although variations in population characteristics or clinical practice appear to be capable of determining whether varenicline or combination NRT is more effective. Demographic and intervention characteristics associated with success also predict use of varenicline. Abbreviations and Acronyms: HSI = Heaviness of Smoking Index; NHS = National Health Service; NRT = nicotine replacement therapy; OR = odds ratio; RCT = randomized controlled trial; RR = relative risk Grant Support: This study was funded by Cancer Research UK (CR-UK C1417/A7972), the National Centre for Smoking Cessation, and Pfizer. Potential Competing Interests: Dr Brose is employed by the National Centre for Smoking Cessation. Before 2005, Mr Stapleton acted as adviser to the manufacturers of smoking cessation medications, for which he has received remunerations and hospitality, and conducted clinical trials with support from the manufacturers of smoking cessation medications. Dr West undertakes research and consultancy for companies that develop and manufacture smoking cessation medications (Pfizer, J&J, McNeil, GSK, Nabi, Novartis, and Sanofi-Aventis). He also has a share of a patent for a novel nicotine delivery device and is a trustee of QUIT, a charity that provides stop smoking support. Correspondence: Address to Leonie S. Brose, PhD, Department of Clinical, Educational, and Health Psychology, University College London, 1-19 Torrington Place, London WC1E 7HB, England ([email protected]).
REFERENCES 1. World Health Organization. Tobacco. http://www.who.int/ mediacentre/factsheets/fs339/en/. Accessed August 9, 2012. 2. Department of Health. Local Stop Smoking Services: Service Delivery and Monitoring Guidance 2011/12. London, England: Department of Health; 2011. 3. Stead LF, Perera R, Bullen C, Mant D, Lancaster T. Nicotine replacement therapy for smoking cessation. Cochrane Database Syst Rev. 2008;(11):CD000146. 4. Cahill K, Stead LF, Lancaster T. Nicotine receptor partial agonists for smoking cessation. Cochrane Database Syst Rev. 2012;(4):CD006103.
March 2013;88(3):226-233
n
http://dx.doi.org/10.1016/j.mayocp.2012.11.013 www.mayoclinicproceedings.org
VARENICLINE VERSUS COMBINATION NRT
5. Hughes JR, Stead LF, Lancaster T. Antidepressants for smoking cessation. Cochrane Database Syst Rev. 2010;(4):CD000031. 6. Brose LS, West R, McDermott MS, Fidler JA, Croghan E, McEwen A. What makes for an effective stop-smoking service? Thorax. 2011;66(10):924-926. 7. Mills EJ, Wu P, Lockhart I, Thorlund K, Puhan M, Ebbert JO. Comparisons of high-dose and combination nicotine replacement therapy, varenicline, and bupropion for smoking cessation: a systematic review and multiple treatment meta-analysis. Ann Med. 2012;44(6):588-597. 8. Stapleton JA, Watson L, Spirling LI, et al. Varenicline in the routine treatment of tobacco dependence: a pre-post comparison with nicotine replacement therapy and an evaluation in those with mental illness. Addiction. 2008;103(1):146-154. 9. Loh WY, Piper ME, Schlam TR, et al. Should all smokers use combination smoking cessation pharmacotherapy? using novel analytic methods to detect differential treatment effects over 8 weeks of pharmacotherapy. Nicotine Tob Res. 2012;14(2):131-141. 10. The NHS Information Centre. Statistics on NHS Stop Smoking Services: England, April 2009 e March 2010. Leeds, England: The Health and Social Care Information Centre; 2010. 11. The NHS Information Centre. Statistics on NHS Stop Smoking Services: England, April 2010 e March 2011. Leeds, England: The Health and Social Care Information Centre; 2011. 12. McLennan D, Barnes H, Noble M, Davies J, Garratt E. The English Indices of Deprivation 2010. London: Department for Communities and Local Government; 2011.
Mayo Clin Proc. n March 2013;88(3):226-233 www.mayoclinicproceedings.org
n
13. West R, Hajek P, Stead L, Stapleton J. Outcome criteria in smoking cessation trials: proposal for a common standard. Addiction. 2005;100(3):299-303. 14. Ferguson J, Bauld L, Chesterman J, Judge K. The English smoking treatment services: one-year outcomes. Addiction. 2005; 100(suppl 2):59-69. 15. Stapleton J. Cigarette smoking prevalence, cessation and relapse. Stat Methods Med Res. 1998;7(2):187-203. 16. Kozlowski LT, Porter CQ, Orleans CT, Pope MA, Heatherton T. Predicting smoking cessation with self-reported measures of nicotine dependence: FTQ, FTND, and HSI. Drug Alcohol Depend. 1994;34(3):211-216. 17. Rasbash J, Steele F, Browne WJ, Goldstein H. A User’s Guide to MLwiN, v2.10. Bristol, England: University of Bristol, Centre for Multilevel Modelling; 2009. 18. Stapleton JA, West R. A direct method and ICER tables for the estimation of the cost-effectiveness of smoking cessation interventions in general populations: application to a new cytisine trial and other examples. Nicotine Tob Res. 2012;14(4): 463-471. 19. Bolin K, Wilson K, Benhaddi H, et al. Cost-effectiveness of varenicline compared with nicotine patches for smoking cessation: results from four European countries. Eur J Public Health. 2009; 19(6):650-654. 20. Ebbert JO, Burke MV, Hays JT, Hurt RD. Combination treatment with varenicline and nicotine replacement therapy. Nicotine Tob Res. 2009;11(5):572-576.
http://dx.doi.org/10.1016/j.mayocp.2012.11.013
233
Comparison of the Effectiveness of Varenicline and Combination Nicotine Replacement Therapy for Smoking Cessation in Clinical Practice Leonie S. Brose, PhD; Robert West, PhD; and John A. Stapleton, MSc Abstract Objective: To compare the effectiveness of varenicline and combination nicotine replacement therapy (NRT) for treating smokers attempting to stop, as it is unclear which of the 2 is more effective. Patients and Methods: Data from 167,487 treatment episodes in patients from 42 National Health Service Stop Smoking Services (April 1, 2009, through June 30, 2011) using varenicline or combination NRT were analyzed. The outcome was carbon monoxideevalidated 4-week smoking abstinence. Potential predictors were age, sex, occupational grade, exemption from prescription charges, intervention setting (specialist or other), support (group or other), and year of quit attempt. Results: Observed smoking abstinence rates were higher with varenicline (43.5% vs 36.9%). However, there was evidence of systematic variation in medication effect across clinical services and differences in predictors of outcome between medications. Allowing for these influences indicated a small mean advantage for varenicline (odds ratio, 1.080; 95% CI, 1.003-1.163; difference, 1.86%; 95% CI, 0.07%3.67%; P¼.04). The relative effectiveness of the 2 medications was not moderated by setting, type of support, or year. Conclusion: The relative effect of medication varies substantially according to clinical practice and population treated. Averaged across current English clinical practice and populations, varenicline is marginally more effective than combination NRT. Demographic and intervention characteristics associated with success predict varenicline use. ª 2013 Mayo Foundation for Medical Education and Research From the National Centre for Smoking Cessation and Training and Department of Clinical, Educational, and Health Psychology (L.S.B.) and Department of Epidemiology and Public Health (R.W., J.A.S.), University College London, London, England.
226
T
obacco is a leading cause of death, illness, and impoverishment.1 Smokers wanting to quit can be supported by medication. Evidence-based medication options for treating smokers attempting to stop include nicotine replacement therapy (NRT), the atypical antidepressant bupropion, and the nicotinic acetylcholine receptor partial agonist varenicline.2 Nicotine replacement therapy is available in a range of forms, including longer-acting nicotine patches and faster-acting gums, lozenges, and inhalators. Nicotine replacement therapy is used as a single product or a combination of products,2,3 usually patches with one of the faster-acting forms. Bupropion and varenicline are taken as a tablet once daily for the first week and twice daily for the rest of the course. Compared with no medication or a placebo, these medications have all been reported to increase the chance of quitting successfully in
n
Mayo Clin Proc. 2013;88(3):226-233
randomized controlled trials (RCTs).3-5 This leaves open the important question of which is most effective for smokers who have the option to use any of them. A single NRT product and bupropion have similar efficacy in RCTs,5 and there appears to be little difference in their relative effectiveness in clinical practice.6 However, varenicline has been reported to be more efficacious than bupropion in RCTs(relative risk [RR], 1.52; 95% CI, 1.221.884), and there is also evidence from RCTs (RR, 1.13; 95% CI, 0.94-1.354,7) and clinical practice (odds ratio [OR], 1.78; 95% CI, 1.572.026; and OR, 1.96; 95% CI, 1.15-3.338) that varenicline is more efficacious and effective than a single NRT product. Similarly, the combination of 2 or more forms of NRT has been reported to be more efficacious than a single NRT product in clinical trials (RR, 1.35; 95% CI, 1.11-1.633), and there is evidence that it is
Mayo Clin Proc. n March 2013;88(3):226-233 n http://dx.doi.org/10.1016/j.mayocp.2012.11.013 www.mayoclinicproceedings.org n ª 2013 Mayo Foundation for Medical Education and Research
VARENICLINE VERSUS COMBINATION NRT
more effective in clinical practice (OR, 1.42; 95% CI, 1.06-1.916,9). Although the evidence indicates that both varenicline and combination NRT are more effective than other medications, to date there has been no direct comparison of these 2 therapies to establish which is more effective. Placebo-controlled RCTs comparing varenicline and combination NRT would be impracticable and probably too costly to mount. In practice, many combinations of 2 or more NRT products are used, and it would be difficult to include all these in a trial. Restriction to just one combination would seriously limit generalizability. In addition, costs would be prohibitive because many thousands of participants would be required to detect a likely small difference. However, an opportunity for a comparison in a real-world setting is available using data provided by the Stop Smoking Services of England’s National Health Service (NHS). These services offer the full range of licensed pharmacologic treatments in combination with behavioral support. Each year, more than 750,000 quit attempts are made with the help of these services.10,11 At the start of a treatment episode the type of pharmacotherapy used is decided by the smoker after consultation with the support therapist and is subject to any contraindications and cautions that might apply. The aim of the present study was to compare the effectiveness of varenicline and combination NRT as practiced within the Stop Smoking Services of England. We also examined whether there were factors related to which of the 2 medications was used and whether any difference in medication effectiveness was dependent on (1) whether the smoker had been treated in a specialist or nonspecialist setting, (2) whether the behavioral support had been provided on a one-to-one basis with a stop smoking practitioner or in a group with other smokers trying to quit, and (3) the year of treatment. METHODS Sample Data were obtained from QuitManager (North 51), an online database system for recording information on patient demographic features, intervention characteristics, and outcomes in accordance with the Department of Health’s standard monitoring requirements.2 Of the Mayo Clin Proc. n March 2013;88(3):226-233 www.mayoclinicproceedings.org
n
152 services in England, 58 were using this database at the time of the present audit, and 49 agreed to share their anonymized data. All completed treatment episodes from April 1, 2009, through June 30, 2011, were included as individual records. As defined by the Department of Health for England and Wales,2 a treatment episode begins when a smoker sets a quit date with the service and ends 4 weeks after the quit date, when the short-term success of the quit attempt is determined. The data represent completed treatment episodes with quit dates between April 2009 and June 2011 for which either varenicline or combination NRT was used (189,795 of 313,193 episodes [60.6%]). To test for differences in medication effect according to the type of support given, the analysis was restricted to the 42 services (of 49) that provided both individual and group support options (20,965 episodes [11.0%] not included). Data were not recorded on key variables for 1434 (0.8%) episodes, and these episodes were excluded to leave a total of 167,487 treatment records for the current study. The clinical services we were able to include in this analysis appeared to be representative for all services in England in terms of patients seen and interventions provided (Table 1). They were also representative in terms of deprivation.12 The index of multiple deprivation is a weighted area-level aggregation of 7 specific domains of deprivation (each derived from a number of indicators): income deprivation; employment deprivation; health deprivation and disability; education, skills, and training deprivation; barriers to housing and services; crime; and living environment deprivation. The least deprived service is ranked as 1, the most deprived as 151; thus, the mean rank is 76. The 42 services included in the present sample ranked from 1 to 149, with a mean rank of 76.5, thus providing a good representation of all services in England. Outcome Measure The outcome measure was successful quitting as defined by the Department of Health criterion (ie, the attempt is successful if the smoker reports continuous abstinence from smoking between weeks 2 and 4 and records an expired-air carbon monoxide reading of <10 ppm at the 4-week follow-up).2 As is
http://dx.doi.org/10.1016/j.mayocp.2012.11.013
227
MAYO CLINIC PROCEEDINGS
TABLE 1. Patient and Intervention Characteristics in the 42 Services From Which the Present Sample Was Drawn and in All English Servicesa,b
Characteristic Female White ethnicityc Occupational group Full-time student Never worked/long-term unemployed Retired Home carer Sick/disabled and unable to work Managerial/professional Intermediate Routine and manual In prison Unable to code Exempt from prescription charge Medication used No medication/unknown Single NRT Combination NRT Bupropion (Zyban) Varenicline (Champix) Intervention setting Specialist clinic Primary care Pharmacy Other Intervention type One-on-one Drop-in Open group Closed group Telephone Other 4-Week quit rate Carbon monoxide validated Self-reported
Patients (%) (N¼299,670)
All patients in England10,11 (2009-2010 and 2010-2011 combined) (%) (N¼1,545,064)
52.4 87.2
52.1 87.7
4.5
4.5
13.0 11.9 5.6
12.4 10.7 5.3
6.6 13.8 7.2 22.4 1.3 13.7
5.8 12.0 7.4 22.1 1.3 18.5
56.5
52.2
10.4 27.3 35.0 1.0 26.3
9.0 64.2d . 1.0 24.5
33.8 40.9 21.1 4.1
30.7 43.9 18.5 6.9
80.7 10.7 3.6 2.1 1.4 1.0
79.7 10.7 3.8 2.0 1.3 2.4
34.7 49.0
34.0 49.0
Baseline Characteristics Available demographic and intervention characteristics were included in the modeling as potential mediators and confounders. Demographics were sex, age at the quit date, exemption from being required to pay the NHS prescription charge (as a proxy measure of economic deprivation), and occupational status. For these analyses, occupational status was dichotomized into employed, which included those studying and those retired from their employment (professional/managerial, intermediate, routine/manual, full-time student, retired), and not employed (never worked/longterm unemployed, home carer, sick/disabled and unable to work, in prison, other). Tobacco dependence as measured by the Heaviness of Smoking Index (HSI), with higher scores (range, 0-6) indicating higher dependence,16 is an optional clinical measure for Stop Smoking Services. In this sample, the HSI score was recorded in 40,145 treatment episodes (24.0%). Intervention characteristics included were the type of behavioral support (group support vs other support types), intervention setting (specialist clinic vs other setting), which have both been associated with cessation rates,6 and the year of the quit attempt.
NRT ¼ nicotine replacement therapy. Missing data in 42 services: sex: n¼60 (0.02%) missing; intervention setting: n¼225 (0.07%) missing; and intervention type: n¼1643 (0.55%) missing. No missing data reported for all services in England. c Ethnicity is self-reported using standard UK Census categories. d Single and combination NRT not reported separately. a
b
standard practice, those who did not report their smoking status or did not attend the session for carbon monoxide verification at 4-week follow-up were regarded as continuing smoking.2,13 Approximately 75% of carbon monoxideevalidated 4-week quitters will relapse 228
Mayo Clin Proc.
within the following year, most within the first 6 months.14,15 However, relative smoking abstinence rates have been found to be stable over short- and long-term follow-ups,15 and we can therefore have reasonable confidence that effects observed in the short term will translate to the long term.
n
Statistical Analyses Because the effects of interest are nested within 42 clinical services having potentially varying practices and populations, 2-level logistic regression models were used with random effects for clinical service (level 2). The odds of success for varenicline compared with combination NRT were estimated, adjusting for demographic and intervention characteristics. Single-level logistic regression models were used to obtain the adjusted OR for medication comparisons within each service. Analyses were conducted using SPSS statistical software, version 18 (SPSS Inc), and MLwiN software, version 2.14.17 The significance level was set at P<.05 for all analyses, and 95% CIs were calculated. Differences between the distributions of individual
March 2013;88(3):226-233
n
http://dx.doi.org/10.1016/j.mayocp.2012.11.013 www.mayoclinicproceedings.org
VARENICLINE VERSUS COMBINATION NRT
demographic and intervention characteristics for the 2 medication options were calculated using either the t test or c2 test, followed by a 2-level multivariable logistic regression model to examine predictors of medication use. The HSI score for dependence was not included in the modeling analyses because of the small proportion of cases for which it was recorded. Instead, the likely effect of dependence on outcome and differences between the 2 medications was assessed separately. Ethical approval was not required for this audit of anonymous NHS clinical data. RESULTS Patient Characteristics Combination NRT was used more frequently (57.4%, n¼96,174) than varenicline (42.6%, n¼71,313). The measured demographic and intervention characteristics were significantly different between those receiving varenicline and combination NRT (Table 2), which is to be expected in a sample of this size when using c2 statistics and t tests. Characteristics that have been found to be associated with successful quitting6 were more frequent in those who received varenicline compared with those treated with combination NRT. In the case of age and sex, the differences were of modest clinical significance. However, more patients receiving varenicline were employed, not exempt from paying
NHS prescription charges, and treated in a specialist clinic and in a group, all of which are associated with higher success rates. Dependence was the only characteristic with an opposing pattern. In the subsample of cases with HSI scores recorded, those receiving varenicline had a higher level of tobacco dependence, which is associated with less successful quit attempts. Predictors of Medication Selection Systematic differences in the characteristics of those receiving the 2 medication options were further examined in a multivariable model (Table 3). After adjusting for all other characteristics and allowing for variation across the clinical services, all characteristics remained significant predictors of whether varenicline or combination NRT had been used in the treatment episode, with the single exception of intervention setting. Varenicline treatment was associated with being male, being older, being employed, not being exempt from paying the NHS prescription charge, receiving group support, and being treated in 2009 (rather than 2010 or 2011). Outcome Overall, 39.7% of patients were carbon monoxideevalidated abstinent from smoking at 4-week follow-up. As anticipated, some characteristics associated with a greater likelihood of receiving
TABLE 2. Distribution of Demographic and Intervention Characteristics in Those Treated With Varenicline and Combination NRTa,b Characteristic
Varenicline (n¼71,313)
Combination NRT (n¼96,174)
Comparison statistic
Mean (SD) age (y) Male Employed NHS prescription charge Paying for prescription Exempt from payment Undisclosed Specialist clinic Group intervention Treatment year 2009 2010 2011 Mean (SD) HSI scorec
43.2 (13.4) 34,851 (48.9) 49,966 (70.1)
42.3 (15.3) 44,808 (46.6) 54,153 (56.3)
t¼14.1, P<.001 c21 ¼85.3, P<.001 c21 ¼3296, P<.001
30,731 33,156 7426 28,482 6587
(43.1) (46.5) (10.4) (39.9) (9.2)
24,958 64,778 6438 34,435 3988
(26.0) (67.4) (6.7) (35.8) (4.1)
14,142 31,602 25,569 3.51
(19.8) (44.3) (35.9) (1.35)
16,169 42,273 37,732 3.35
(16.8) (44.0) (39.2) (1.43)
c22 ¼7351, P<.001 c21 ¼298, P<.001 c21 ¼1794, P<.001
c22 ¼331, P<.001 t¼11.4, P<.001
HSI ¼ Heaviness of Smoking Index; NHS ¼ National Health Service; NRT ¼ nicotine replacement therapy. Data are presented as No. (percentage) of patients unless otherwise indicated. c The HSI scores were recorded in 40,145 cases (24.0%). a
b
Mayo Clin Proc. n March 2013;88(3):226-233 www.mayoclinicproceedings.org
n
http://dx.doi.org/10.1016/j.mayocp.2012.11.013
229
MAYO CLINIC PROCEEDINGS
TABLE 3. Predictors of Varenicline vs Combination Nicotine Replacement Therapy Usea Predictor
Odds ratio (95% CI)
P value
Male vs female Age per 10-y increase Employed vs not employed National Health Service prescription charge Exempt vs pays Undisclosed vs pays Specialist clinics vs other setting Group support vs other support Quit year 2010 vs 2009 2011 vs 2009
1.05 (1.03-1.08) 1.08 (1.07-1.09) 1.31 (1.28-1.34)
<.001 <.001 <.001
0.43 1.15 1.01 1.30
(0.42-0.44) (1.10-1.20) (0.99-1.04) (1.23-1.37)
<.001 <.001 .39 <.001
0.81 (0.79-0.84) 0.66 (0.64-0.69)
<.001 <.001
a
Two-level multivariable logistic regression. Level 1 is treatment episode, and level 2 is Stop Smoking Service.
varenicline treatment were also independently prognostic of success. Being male, older, and employed, paying for medication prescriptions, and receiving group support all increased the odds of success (Table 4). In the subsample with HSI data, a higher HSI score, which was associated with varenicline treatment, was predictive of failure (OR, 0.936; 95% CI, 0.922-0.949). The observed success rate was 36.9% with combination NRT and 43.5% with varenicline (absolute percentage difference, 6.6%) For combination NRT and varenicline, 23.3% and 20.6%, respectively, failed to attend the 4-week follow-up and were assumed to be smoking. However, despite within-service adjustment for all measured patient and treatment characteristics in the 2-level model, there remained
TABLE 4. Predictor Variable Model for Carbon MonoxideeValidated 4-Week Abstinencea Predictor
Odds ratio (95% CI)
P value
Male vs female Age per 10-y increase Employed vs not employed National Health Service prescription charge Exempt vs pays Undisclosed vs pays Specialist clinics vs other setting Group support vs other support Quit year 2010 vs 2009 2011 vs 2009
1.00 (0.98-1.02) 1.20 (1.19-1.21) 1.24 (1.22-1.27)
.92 <.001 <.001
0.81 0.80 1.01 1.39
<.001 <.001 .31 <.001
a
(0.79-0.87) (0.77-0.83) (0.99-1.04) (1.32-1.46)
0.97 (0.94-0.99) 1.03 (0.99-1.06)
.02 .14
Two-level multivariable logistic regression. Level 1 is treatment episode, and level 2 is Stop Smoking Service.
230
Mayo Clin Proc.
n
evidence of large systematic differences in the relative effect of medication across the services (c241¼385; P<.001). The Figure shows the adjusted ORs for treatment success in favor of varenicline in each of the 42 clinical services. In 20 services varenicline was more effective than combination NRT, in 16 services there was no significant difference, and in 7 services varenicline was less effective than combination NRT. The ORs ranged from 1.96 in favor of varenicline to 3.13 in favor of combination NRT. The effect averaged across services in the 2level model indicated a small benefit for varenicline (OR, 1.08; 95% CI, 1.003-1.163). Applying this OR to the adjusted treatment rates gives an adjusted RR of 1.047 (95% CI, 1.002-1.092) and an adjusted absolute percentage difference of 1.86% (95% CI, 0.07%-3.67%). In models that included terms representing interactions between medication and intervention characteristics, there was only weak evidence that intervention setting (P¼.08), type of support (P¼.30), or year of treatment (P¼.09) moderated the relative effect of medication. DISCUSSION To our knowledge, this is the first systematic comparison of varenicline and combination NRT in smoking cessation, using a very large sample from real-world clinical practice. Treatment using varenicline compared with combination NRT in the English NHS Stop Smoking Services was associated with a small mean increase in short-term smoking abstinence rates. The mean benefit (adjusted absolute percentage difference) was less than 2%, which would suggest an additional 0.5% permanently quitting on the assumption of equivalent relapse rates over time and equivalent rates of subsequent quit attempts between the 2 treatments in this setting.15,18 Although a small effect, it would translate to 500 additional quitters for every 100,000 treated with varenicline rather than combination NRT in services and patients similar to the present sample. This benefit was not significantly affected by whether the treatment episode had taken place in a specialist clinic or nonspecialist setting, by whether group or one-on-one behavioral support had been given, or by the year in which the quit attempt had taken place. However, the selection of medication was found to be associated with demographic and intervention characteristics that
March 2013;88(3):226-233
n
http://dx.doi.org/10.1016/j.mayocp.2012.11.013 www.mayoclinicproceedings.org
VARENICLINE VERSUS COMBINATION NRT
Adjusted OR in favor of varenicline
3.0
2.5
Adjusted OR and 95% Cl for each service Overall adjusted OR and 95% Cl
2.0
1.5
1.0
0.5
0.0
FIGURE. Adjusted odds ratios (ORs) for single-level models of carbon monoxideevalidated 4-week abstinence. Error bars indicate CIs.
were also predictive of success or failure of the attempt, and the relative medication effect varied considerably across services. On the basis of our findings of relatively similar effectiveness and the relatively similar cost of the 2 medications (approximately £160 or $250 for varenicline and £140-£180 or $220-$280 for combination NRT), a substantial difference in cost-effectiveness is unlikely. Studies in a number of countries have confirmed the cost-effectiveness of varenicline compared with NRT19; therefore, cost should not be a significant factor in future prescribing decisions. The present findings confirm earlier results that reported group interventions to be more successful than one-on-one interventions and that socioeconomic variables (eg, paying prescription charges and being employed) are positively associated with success.6 Previous findings revealed specialist clinics to be more successful than interventions set in primary care; this finding has not been replicated here. It is possible that the restriction of the sample to interventions using varenicline or combination NRT has reduced the effects of settings. An overarching feature of this study was the finding of systematic variation among the 42 clinical services in the relative effectiveness of the 2 medications. The size of the systematic variation in effect size between the services was not anticipated. Although in almost half of services there was a significant benefit for varenicline, in Mayo Clin Proc. n March 2013;88(3):226-233 www.mayoclinicproceedings.org
n
a substantial few (7/42 [17%]) there was a significant benefit for combination NRT. Such differences are unlikely to be the result of population differences. They were evident after controlling for the included individual characteristics predictive of both medication selection and outcome, and although it is possible that some other unrecorded measures were also predictive, it is difficult to envisage individual factors that would exert such a large influence as to explain the differences we observed. Instead, the differences are likely to be attributable to differences in clinical practice, with some practices tending to favor varenicline being more successful and others favoring combination NRT being more successful. We did not have data on the style of treatment delivery in the 42 services and are therefore unable to empirically address this issue. Although national guidance on smoking cessation treatment should be equally applicable to all services, it is clear that delivery is not consistent. We can only speculate that there are likely biases in whether or not a full course of medication is prescribed, how well the behavioral support is aligned to the treatment, how therapists and patients perceive medication efficacy, or how adverse drug responses are treated. Reducing such differences to engender optimal treatment is a major task of the National Centre for Smoking Cessation Training (www.ncsct.co.uk). Although not sufficient to explain the major part of the relative differences in
http://dx.doi.org/10.1016/j.mayocp.2012.11.013
231
MAYO CLINIC PROCEEDINGS
medication efficacy across the services, we identified characteristics that were associated with medication selection. Those taking varenicline were more likely to be older, to be employed and therefore not exempt from paying the NHS prescription charge (currently approximately £7.50 or $12), and to have received group support, all of which were found to be associated with successful outcome. Again, it is unclear as to why there were such differences in treatment delivery. In the subsample with HSI scores, those taking varenicline were also more dependent on tobacco. Dependence on tobacco was an uncontrolled factor in the main analyses, being only recorded in a few episodes. Had the HSI score been available in all cases and had the same association with medication selection and outcome been evident, then inclusion of the HSI score in the model would slightly have improved the relative benefit of varenicline. A major strength of these analyses is that they are based on a very large national sample in real-world clinical practice. The included services were representative for all English NHS Stop Smoking Services. Limitations were the use of short-term 4-week cessation data as the outcome measure and the availability of only a small set of background characteristics. Regarding the outcome measure, a longer follow-up was not feasible because few services attempt to follow up smokers beyond 4 weeks and those who are followed up longer are likely to be unrepresentative. However, as noted in the “Methods” section, short-term relative cessation rates have been found to be predictive of longterm relative rates.14,15 In this analysis we were not able to identify individuals, only treatment episodes. Consequently, we could not model the effect of some individuals appearing with more than one treatment episode during the years of study. However, anecdotal evidence from the services would suggest that this was the case in a few episodes. Further, any bias is likely to be fairly neutral due to some reattending individuals using different treatments, in which case our analysis would be less sensitive to differences than it should be, and some individuals using the same treatment, in which case our analysis would be more sensitive to differences than it should be. Future research is needed to identify factors that may explain the heterogeneity in relative 232
Mayo Clin Proc.
n
effectiveness in clinical practice. Further studies are also needed to investigate the effectiveness of combining NRT and varenicline. To date, the safety and efficacy of this combination have only been researched in one small trial,20 which reported no benefit over the usual care. Another small trial has been completed, but its results have not yet been published (ClinicalTrials. gov identifier: NCT01184664). CONCLUSION Overall, there appears to be a small mean advantage for varenicline over combination NRT in clinical practice, although variations in population characteristics or clinical practice appear to be capable of determining whether varenicline or combination NRT is more effective. Demographic and intervention characteristics associated with success also predict use of varenicline. Abbreviations and Acronyms: HSI = Heaviness of Smoking Index; NHS = National Health Service; NRT = nicotine replacement therapy; OR = odds ratio; RCT = randomized controlled trial; RR = relative risk Grant Support: This study was funded by Cancer Research UK (CR-UK C1417/A7972), the National Centre for Smoking Cessation, and Pfizer. Potential Competing Interests: Dr Brose is employed by the National Centre for Smoking Cessation. Before 2005, Mr Stapleton acted as adviser to the manufacturers of smoking cessation medications, for which he has received remunerations and hospitality, and conducted clinical trials with support from the manufacturers of smoking cessation medications. Dr West undertakes research and consultancy for companies that develop and manufacture smoking cessation medications (Pfizer, J&J, McNeil, GSK, Nabi, Novartis, and Sanofi-Aventis). He also has a share of a patent for a novel nicotine delivery device and is a trustee of QUIT, a charity that provides stop smoking support. Correspondence: Address to Leonie S. Brose, PhD, Department of Clinical, Educational, and Health Psychology, University College London, 1-19 Torrington Place, London WC1E 7HB, England ([email protected]).
REFERENCES 1. World Health Organization. Tobacco. http://www.who.int/ mediacentre/factsheets/fs339/en/. Accessed August 9, 2012. 2. Department of Health. Local Stop Smoking Services: Service Delivery and Monitoring Guidance 2011/12. London, England: Department of Health; 2011. 3. Stead LF, Perera R, Bullen C, Mant D, Lancaster T. Nicotine replacement therapy for smoking cessation. Cochrane Database Syst Rev. 2008;(11):CD000146. 4. Cahill K, Stead LF, Lancaster T. Nicotine receptor partial agonists for smoking cessation. Cochrane Database Syst Rev. 2012;(4):CD006103.
March 2013;88(3):226-233
n
http://dx.doi.org/10.1016/j.mayocp.2012.11.013 www.mayoclinicproceedings.org
VARENICLINE VERSUS COMBINATION NRT
5. Hughes JR, Stead LF, Lancaster T. Antidepressants for smoking cessation. Cochrane Database Syst Rev. 2010;(4):CD000031. 6. Brose LS, West R, McDermott MS, Fidler JA, Croghan E, McEwen A. What makes for an effective stop-smoking service? Thorax. 2011;66(10):924-926. 7. Mills EJ, Wu P, Lockhart I, Thorlund K, Puhan M, Ebbert JO. Comparisons of high-dose and combination nicotine replacement therapy, varenicline, and bupropion for smoking cessation: a systematic review and multiple treatment meta-analysis. Ann Med. 2012;44(6):588-597. 8. Stapleton JA, Watson L, Spirling LI, et al. Varenicline in the routine treatment of tobacco dependence: a pre-post comparison with nicotine replacement therapy and an evaluation in those with mental illness. Addiction. 2008;103(1):146-154. 9. Loh WY, Piper ME, Schlam TR, et al. Should all smokers use combination smoking cessation pharmacotherapy? using novel analytic methods to detect differential treatment effects over 8 weeks of pharmacotherapy. Nicotine Tob Res. 2012;14(2):131-141. 10. The NHS Information Centre. Statistics on NHS Stop Smoking Services: England, April 2009 e March 2010. Leeds, England: The Health and Social Care Information Centre; 2010. 11. The NHS Information Centre. Statistics on NHS Stop Smoking Services: England, April 2010 e March 2011. Leeds, England: The Health and Social Care Information Centre; 2011. 12. McLennan D, Barnes H, Noble M, Davies J, Garratt E. The English Indices of Deprivation 2010. London: Department for Communities and Local Government; 2011.
Mayo Clin Proc. n March 2013;88(3):226-233 www.mayoclinicproceedings.org
n
13. West R, Hajek P, Stead L, Stapleton J. Outcome criteria in smoking cessation trials: proposal for a common standard. Addiction. 2005;100(3):299-303. 14. Ferguson J, Bauld L, Chesterman J, Judge K. The English smoking treatment services: one-year outcomes. Addiction. 2005; 100(suppl 2):59-69. 15. Stapleton J. Cigarette smoking prevalence, cessation and relapse. Stat Methods Med Res. 1998;7(2):187-203. 16. Kozlowski LT, Porter CQ, Orleans CT, Pope MA, Heatherton T. Predicting smoking cessation with self-reported measures of nicotine dependence: FTQ, FTND, and HSI. Drug Alcohol Depend. 1994;34(3):211-216. 17. Rasbash J, Steele F, Browne WJ, Goldstein H. A User’s Guide to MLwiN, v2.10. Bristol, England: University of Bristol, Centre for Multilevel Modelling; 2009. 18. Stapleton JA, West R. A direct method and ICER tables for the estimation of the cost-effectiveness of smoking cessation interventions in general populations: application to a new cytisine trial and other examples. Nicotine Tob Res. 2012;14(4): 463-471. 19. Bolin K, Wilson K, Benhaddi H, et al. Cost-effectiveness of varenicline compared with nicotine patches for smoking cessation: results from four European countries. Eur J Public Health. 2009; 19(6):650-654. 20. Ebbert JO, Burke MV, Hays JT, Hurt RD. Combination treatment with varenicline and nicotine replacement therapy. Nicotine Tob Res. 2009;11(5):572-576.
http://dx.doi.org/10.1016/j.mayocp.2012.11.013
233