Abstract
Background and aims:
The Phase-Based Model of smoking cessation treatment suggests that treatment needs may vary across phases (e.g., pre-cessation, cessation). This study tested the comparative effects of varenicline and combination nicotine replacement therapy (C-NRT) relative to nicotine patch monotherapy on pre-cessation and cessation phase candidate withdrawal, expectancy, and motivation mediators; relations between mediators and abstinence; and indirect effects of enhanced treatments on abstinence via candidate mediators.
Design:
Secondary mediation analysis of data from the open-label, randomized Wisconsin Smokers’ Health Study 2, a comparative effectiveness trial of varenicline or C-NRT, versus patch monotherapy, in adults who smoked, recruited via media and community outreach.
Setting:
Research clinics in Madison and Milwaukee, Wisconsin, USA.
Participants:
1,051 daily smokers motivated to quit smoking (52.5% female; Mage=48.1, SD 11.6).
Interventions:
12 weeks of varenicline (n=407) or 12 weeks of combination nicotine patch and nicotine lozenge therapy (n=421), both compared with 12 weeks of patch control condition (n=230), with individual smoking cessation counseling.
Measurements:
The primary abstinence outcome was biochemically verified 7-day point-prevalence abstinence 4 weeks post-target quit day (TQD). Candidate mediators (craving, positive smoking expectancies, withdrawal symptoms, and quitting motivation) were assessed via ecological momentary assessment from 1 week prior (pre-cessation phase) to 4 weeks after (cessation phase) the TQD.
Findings:
Pre-cessation and cessation mean levels and slopes of craving (AOR=0.34-0.79), smoking expectancies (AOR=0.46-0.79), and quitting motivation (AOR=1.35-7.21) significantly predicted 4-week post-TQD abstinence (p<.05). Significant varenicline mediation occurred via greater suppression in pre-cessation craving (ab=0.09, SE=0.03, 95% CI=0.04-0.14) and smoking expectancies (ab=0.06, SE=0.02, 95% CI=0.02-0.12). C-NRT mediation occurred via greater reduction in pre-post-TQD changes in craving (ab=0.04, SE=0.02, 95% CI=0.01-0.08) and expectancies (ab=0.03, SE=0.02, 95% CI=0.001-0.07), relative to patch monotherapy.
Conclusion:
Among adult smokers seeking to quit, varenicline seems to work through its effects on suppression of craving and smoking expectancies pre-cessation while combination nicotine replacement therapy mediation seems to work through cessation-related reduction in craving and smoking expectancies changes.
Keywords: Mediation, varenicline, combination nicotine replacement therapy, nicotine lozenge, nicotine patch, nicotine withdrawal, smoking cessation
Introduction
Two pharmacotherapies produce the highest levels of long-term abstinence from smoking: varenicline and combination nicotine replacement therapy (C-NRT, i.e., nicotine patches paired with a fast-acting nicotine medication, such as nicotine lozenges). While these therapies have been studied extensively [1-5], important questions about their effectiveness and mechanisms relative to simpler treatments remain. Greater understanding of when and how these therapies improve abstinence relative to patch alone may spur treatment refinement and help identify treatment non-response before a return to smoking, which may facilitate adaptive treatment. For instance, if pre-cessation varenicline treatment improves abstinence indirectly by reducing craving, this intermediate outcome may serve as a marker of treatment response. Those who do not show pre-cessation craving suppression may be at risk for treatment failure and benefit from adaptive treatment [6]. Comparative effectiveness results may provide insights into why some treatments work better than others and suggest how such treatments may be used more strategically.
The current study assessed differential indirect effects of varenicline and C-NRT (nicotine patch and lozenge), versus patch alone (an active, standard-care comparator), on short-term abstinence via putative mediators (e.g., cigarette craving, negative affect, and other withdrawal symptoms) using secondary data from an open-label, randomized, comparative effectiveness trial [7]. Although the primary analyses of the comparative efficacy of the 3 pharmacotherapy regimens yielded null effects on abstinence [7], this does not preclude differences in proximal effects on mediators or significant indirect effects [8]. Examining indirect effects can yield insights into differential treatment effects on intervention targets and outcome predictors even when treatments do not differ in abstinence endpoints [8].
Varenicline better suppresses craving [2, 3, 9, 10], smoking satisfaction [9, 11-12], anticipated reward from smoking [13, 14], and nicotine withdrawal symptoms [2], including negative affect [3, 9], after a target quit day (TQD) relative to placebo pills or nicotine patch. Similarly, C-NRT reduces post-TQD craving better than NRT monotherapy, and craving reduction mediates C-NRT benefits on long-term abstinence [1, 15]. Most studies have focused on the cessation-phase of treatment, even though varenicline treatment typically begins at least 7 days pre-TQD [cf. 16 which showed mediation of pre-quit varenicline effects on smoking via craving suppression]. In addition, studies of varenlicine and C-NRT mediators have been limited by a narrow scope of candidates (e.g., craving) omitting important constructs (e.g., positive smoking expectancies or reliance on retrospective or laboratory measures of mediators [2-3,13,16]) [1,15].
The current study addresses these gaps by examining the comparative effects of two enhanced therapies, varenicline and C-NRT, versus patch, on a broad range of candidate mediators (craving, positive smoking expectancies, positive affect, negative affect, difficulty concentrating, hunger, cessation fatigue, quitting motivation) in both the pre-cessation and cessation phases of treatment [17]. Candidate mediators were examined separately to identify specific targets sensitive to enhanced treatment, and those not markedly improved by enhanced treatment versus simpler, less expensive patch therapy. These exploratory analyses were not pre-registered.
Candidate mediators in this study were assessed via ecological momentary assessment (EMA) nightly to reduce recall biases and enhance ecological validity [18]. The mediators were selected based on research showing that they both predict abstinence (i.e., they are important targets in conceptual models of abstinence) and are affected by treatment (i.e., they are important outcomes in treatment action models [19]). Cigarette craving is a robust proximal predictor of smoking lapses and relapses [20] and is sensitive to both varenicline and C-NRT treatment [1-3]. Other nicotine withdrawal symptoms (e.g., negative affect, difficulty concentrating, hunger) are both affected by NRT and predictive of quitting success [21]. Individual withdrawal symptoms follow different time-courses and have differential relations with both treatments and abstinence [22, 23], so discrete symptoms (e.g., negative affect, hunger) were analyzed separately. Positive smoking expectancies were examined, as these are predictive of smoking [12], and may be particularly affected by varenicline [2, 3, 16]. Finally, positive affect and motivation to quit were examined, as these variables have been shown to both predict abstinence and mediate pharmacotherapy effects on abstinence [23, 24].
This study aimed to: 1) assess comparative effects of varenicline or C-NRT relative to patch monotherapy on 8 candidate mediators from 1 week pre-TQD (pre-cessation) to 4 weeks post-TQD (cessation), 2) examine relations of pre-cessation and cessation mean levels and slopes of candidate mediators with 4-week abstinence post-TQD, and 3) estimate indirect, mediated effects of varenicline or C-NRT on abstinence via candidate mediators. The two enhanced therapies (varenicline and C-NRT) were not compared head-to-head, as the purpose of the current analyses was to identify intermediate outcomes that may indicate enhanced treatment response and to test the action models of enhanced treatments, relative to standard care. Also, the current analyses allow for the detection of different mediators specifically responsive to the two enhanced treatments, and that the treatments may affect the candidate mediators across different phases of treatment.).
Methods
Design
Secondary analyses were conducted using data from the Wisconsin Smokers’ Health Study 2 [7], an open-label randomized clinical trial of 12-week nicotine patch therapy (n=230), varenicline (n=407), and combination nicotine patch and lozenge (n=421), all offered with 6 sessions of cessation counseling in a sample of 1051 adult smokers motivated to quit smoking who provided at least one EMA report on candidate mediators. Piecewise single-mediator multilevel models were used to estimate enhanced treatment vs. patch effects on candidate mediators and to derive individual estimates of mediator pre-cessation and cessation intercepts and slopes. These estimates were used as predictors of 4-week biochemically confirmed abstinence in logistic regression analyses and mediators of indirect effects of treatment on abstinence in SAS PROCESS macro [25] analyses.
Participants
Participants were recruited from Madison and Milwaukee, Wisconsin and surrounding areas via advertisements, earned media, and outreach to Wisconsin Smokers’ Health Study participants who were smoking at the final follow-up. Inclusion criteria were: at least 18 years old; smoked at least 5 cigarettes per day with an expired carbon monoxide (CO) level of at least 5 parts per million (ppm) at enrollment; able to read, write, and speak English; planned to remain in the area for at least 12 months; medically able and willing to use study medication; willing to refrain from using electronic cigarettes during the study; willing to respond to interactive voice response phone prompts up to 3 times per day for up to 5 weeks; and willing to prevent pregnancy during treatment. Exclusion criteria included: severe depression; current suicidal ideation or attempt in the last 5 years; diagnosis of or treatment for schizophrenia, other psychotic disorder, or bipolar disorder within the last 10 years; currently taking bupropion; history of stroke, heart attack, congestive heart failure or past-year diabetes; and pregnancy or breastfeeding. Additional details and the full CONSORT diagram are available in [7]. The study was approved by the University of Wisconsin Health Sciences Institutional Review Board.
Procedures
Participants were phone-screened to determine eligibility and attended an information session where written informed consent was obtained. Participants then completed a physical exam and baseline assessments prior to randomization and treatment initiation at a 1-hour visit one week pre-TQD. Randomization was stratified by site (Madison or Milwaukee), gender, and race in this open-label (unblinded) trial. Participants completed 4 more 45-minute office visits for assessment and treatment and a 10-minute treatment call over the first 12 weeks post-TQD. The primary abstinence for this study was collected 4 weeks post-TQD.
Treatment
Nicotine patch therapy comprised one patch per day for 12 weeks, starting on the TQD. Those who smoked more than 10 cigarettes per day were instructed to use 21-mg patches for 8 weeks, reduce to 14-mg patches for 2 weeks, and further reduce to 7-mg patches for 2 weeks. Those who smoked 5-10 cigarettes per day were instructed to use 14-mg patches for 10 weeks, and 7-mg patches for 2 weeks. Varenicline therapy started 1 week pre-TQD, with 0.5 mg once daily for 3 days, 0.5 mg twice daily for 4 days, and then 1 mg twice daily for 11 weeks post-TQD. C-NRT included the patch (as dosed and dispensed in the patch-only condition) and 12 weeks of nicotine lozenges for use beginning on the TQD. Those who reported smoking within 30 minutes of waking received 4-mg lozenges; those with longer smoking latencies received 2-mg lozenges. Participants were instructed to use at least 9 lozenges per day and to use 1 every 1-2 hours for 6 weeks, reduce to 1 every 2-4 hours for 3 weeks, and taper to 1 every 4-8 hours for the final 3 weeks.
All participants received 6 individual, 10-20-minute counseling sessions with bachelor’s-level health educators supervised by licensed psychologists between 1 week pre-TQD and 12 weeks post-TQD. Counseling content focused on problem-solving and support, consistent with the 2008 Public Health Service Clinical Practice Guideline [26].
Measures
At baseline, staff collected vital signs, CO, demographics (age, sex, race, ethnicity, and income), smoking history (including longest period of past abstinence coded as a 4-level categorical variable, see Table 1), and tobacco dependence measures including the Fagerström Test of Cigarette Dependence (FTCD [27]).
Table 1.
Demographic Characteristics and Baseline Smoking History (N= 1,051)
| Overall (N=1051, 96.8% of 1086 randomized) |
Nicotine Patch (n=230, 95.4% of 241 randomized) |
Varenicline (n=407, 96.0% of 424 randomized) |
C-NRT (n=414, 98.3% of 421 randomized) |
|
|---|---|---|---|---|
| Female, N (%) | 550 (52.5) | 121 (52.6) | 213 (52.6) | 216 (52.3) |
| Race, N (%) | ||||
| White | 707 (67.3) | 157 (68.3) | 269 (66.1) | 281 (67.9) |
| Minority | 344 (32.7) | 73 (31.7) | 138 (33.9) | 133 (32.1) |
| Annual Household Income, N (%) | ||||
| Under $10,000 | 200 (20.0) | 46 (21.1) | 78 (20.0) | 76 (19.3) |
| $10,000 - $24,999 | 210 (21.0) | 41 (19.9) | 81 (20.8) | 88 (22.4) |
| $25,000 - $49,999 | 272 (27.2) | 59 (27.2) | 111 (28.5) | 102 (26.0) |
| $50,000 or greater | 318 (31.8) | 71 (32.7) | 120 (30.8) | 127 (32.3) |
| Longest past abstinence, N (%) | ||||
| One week or less | 389 (37.2) | 85 (37.3) | 162 (40.1) | 142 (34.4) |
| 8 days to 1 month | 116 (11.1) | 28 (12.3) | 39 (9.7) | 49 (11.9) |
| 1-12 months | 355 (34.0) | 72 (31.6) | 144 (35.6) | 139 (33.7) |
| More than 1 year | 185 (17.7) | 43 (18.9) | 59 (14.6) | 83 (20.1) |
| Age, mean (SD) | 48.1 (11.6) | 49.2 (10.9) | 47.0 (11.6) | 48.5 (11.9) |
| Cigarettes per day, mean (SD) | 17.0 (8.1) | 16.7 (7.9) | 16.9 (7.6) | 17.2 (8.7) |
| Years smoked, mean (SD) | 28.6 (12.1) | 29.3 (11.4) | 27.7 (12.0) | 29.1 (12.6) |
| Number of past quit attempts, mean (SD) | 4.0 (6.1) | 4.1 (7.1) | 3.6 (3.6) | 4.3 (7.4) |
| FTCD score, mean (SD) | 4.8 (2.1) | 4.9 (2.2) | 4.8 (2.1) | 4.8 (2.0) |
C-NRT: Combination Nicotine Replacement Therapy (nicotine patch and nicotine lozenge); FTCD: Fagerström Test of Cigarette Dependence; SD: Standard Deviation. Note: Percentages do not always sum to 100 due to missing data.
Candidate mediators and recent smoking were assessed in real time using EMA [18] via automated interactive response (IVR) calls. Calls occurred 3 times a day (morning, evening, and one random prompt during the day) from 1 week pre-TQD to 2 weeks post-TQD, and then every other day for the next 2 weeks (to 4 weeks post-TQD). Candidate mediators were assessed using items adapted from the Positive and Negative Affect Schedule (PANAS [28]), the Wisconsin Smoking Withdrawal Scale [29], and previous EMA studies [1, 16, 22, 30]. Participants were asked to rate how they felt in the last 15 minutes using single items to assess: craving (“want to smoke”), positive affect (“excited or enthusiastic”), negative affect (“negative mood”), difficulty concentrating (“can’t concentrate or think clearly”), hunger (“thinking about food or hungry”), all rated on a scale from 1 (not at all) to 7 (extremely). Additional items assessing positive smoking expectancies (“I would enjoy having a cigarette right now”), cessation fatigue (“I am tired of trying to quit smoking”), and quitting motivation (“quitting smoking is important to me”) were rated on a scale from 1 (don’t agree) to 7 (agree). Previous research with brief EMA items such as these has shown that such measures are sensitive to cessation and predictive of later abstinence [23, 24, 31].
Intent-to-treat, 7-day point-prevalence abstinence confirmed by CO testing (CO≤5 ppm) at 4 weeks post-TQD was the abstinence outcome. We focused on 4-week abstinence rather than the primary outcome (26-week) for the parent trial [7] because a more proximal abstinence outcome may be more sensitive to pharmacotherapy effects [17, 19].
Data analyses
We conducted multilevel models (MLMs; in HLM 7.03, Scientific Software International, Lincolnwood, IL, USA) to assess enhanced treatment effects (vs. patch alone) on candidate mediators. Piecewise MLMs were tested, using maximum likelihood estimation, to examine pharmacotherapy effects on the mean level and the rate of change (i.e., slope) of the mediator separately in the 1 week pre-TQD and the 4-week post-TQD epochs. EMA reports were nested within individuals.
The general form of the level 1 equation for each candidate mediator was: Mit = β0 + β1 (Pre-TQD week) + β2 (Post-TQD indicator) + β3 (Post-TQD weeks) + error where each candidate mediator (M) was a function of time in fractional weeks from the TQD, centered around the mid-point of each epoch (week 0.5 pre-TQD and week 2 post-TQD), so the pre-TQD intercept (β0) can be interpreted as the pre-TQD mediator mean level and the post-TQD indicator coefficient (β2) can be interpreted as the change in the mean level of the mediator from pre-to-post-TQD (pre-post change). The slope variables reflect the linear change per week in the mediator pre-TQD (β1) and post-TQD (β3). Growth coefficients were set to random if they varied significantly (all ps<.001) across subjects and could be estimated with at least minimal reliability (>0.3). All models converged.
Separate dummy-coded indicators of varenicline and C-NRT (with patch only as the reference group) were included at the subject level to assess the effects of randomized pharmacotherapy regimen on candidate mediators pre- and post-TQD. Models also included the following time-invariant covariates: gender (0=female, 1=male), minority status (0=White, 1=member of a racial minority group), and grand mean-centered longest period of past abstinence, years of age, baseline FTCD score, and time-varying recent smoking status(0=no smoking, 1=any smoking in the last 2 hours or since the last EMA report).
Empirical Bayes’ estimates of pre-TQD individual mean and slope, and pre-post-TQD change mean and post-TQD slope for each candidate mediator were derived from unconditional MLM models. These are estimates of random coefficients for individual subjects that can be used as predictors of distal outcomes (with some shrinkage in variance) [32] and do not require experimenter-specified priors. Subject-level estimates were used as predictors of 7-day biochemically confirmed point-prevalence abstinence 4 weeks post-TQD in logistic regression analyses using SAS version 9.4 (SAS Institute, Cary, NC), controlling for treatment (entered as dummy-coded indicators of varenicline or C-NRT) and baseline covariates (gender, racial minority status, age, longest period of past abstinence, and FTCD score). The Benjamini-Hochberg approach was applied to final multilevel model and logistic regression models to control the false discovery rate [33], as shown in Table 3.
Table 3.
Estimated Varenicline Effects (vs. Patch) and Combination-Nicotine Replacement Therapy (C-NRT, Patch and Lozenge) Effects (vs. Patch) on Mediator Intercepts and Slopes in Conditional Multi-level Modelsa, and Relations between Mediator Empirical Bayes’ Coefficient Estimates and 4-Week Abstinence
| Candidate Mediators | Predictors | Varenicline | C-NRT | Abstinence 4 weeks post- target-quit-day |
|||
|---|---|---|---|---|---|---|---|
| Est | (Standard Error) |
Est | (Standard Error) |
Adjusted Odds Ratio |
(95% Confidence Interval) |
||
| Craving | Pre-TQD Mean | −0.290* | (0.115) | 0.066 | (0.119) | 0.698*** | (0.603-0.809) |
| Pre-TQD Slope | −0.244 | (0.174) | −0.080 | (0.175) | 0.786** | (0.675-0.916) | |
| Pre-Post Difference | −0.086 | (0.129) | −0.3290* | (0.135) | 0.663*** | (0.568-0.774) | |
| Post-TQD Slope | 0.015 | (0.033) | 0.061 | (0.033) | 0.344** | (0.191-0.621) | |
| Positive smoking expectancies | Pre-TQD Mean | −0.3230* | (0.124) | 0.083 | (0.125) | 0.790** | (0.689-0.905) |
| Pre-TQD Slope | −0.396± | (0.189) | 0.169 | (0.189) | 0.794** | (0.694-0.909) | |
| Pre-Post Difference | −0.024 | (0.137) | −0.441** | (0.142) | 0.773** | (0.665-0.899) | |
| Post-TQD Slope | −0.016 | (0.035) | 0.039 | (0.035) | 0.457* | (0.264-0.788) | |
| Positive affect | Pre-TQD Mean | −0.013 | (0.124) | 0.069 | (0.124) | 1.155* | (1.026-1.300) |
| Pre-TQD Slope | −0.200 | (0.176) | 0.080 | (0.177) | 1.054 | (0.927-1.199) | |
| Pre-Post Difference | 0.040 | (0.084) | −0.008 | (0.090) | 1.573* | (1.146-2.159) | |
| Post-TQD Slope | −0.006 | (0.029) | 0.016 | (0.029) | 1.013 | (0.508-2.022) | |
| Negative affect | Pre-TQD Mean | −0.076 | (0.075) | −0.021 | (0.075) | 0.690** | (0.556-0.855) |
| Pre-TQD Slope | 0.162 | (0.147) | 0.049 | (0.149) | 0.880 | (0.741-1.045) | |
| Pre-Post Difference | −0.131 | (0.078) | −0.164± | (0.077) | 0.645 | (0.409-1.018) | |
| Post-TQD Slope | 0.032 | (0.022) | 0.050* | (0.020) | 0.624 | (0.213-1.829) | |
| Difficulty concentrating | Pre-TQD Mean | −0.061 | (0.089) | −0.049 | (0.089) | 0.892 | (0.756-1.051) |
| Pre-TQD Slope | −0.038 | (0.142) | −0.124 | (0.149) | 0.987 | (0.841-1.159) | |
| Pre-Post Difference | −0.031 | (0.093) | 0.008 | (0.097) | 0.766 | (0.574-1.022) | |
| Post-TQD Slope | 0.026 | (0.022) | 0.015 | (0.021) | 0.622 | (0.231-1.672) | |
| Hunger | Pre-TQD Mean | −0.019 | (0.097) | −0.023 | (0.095) | 0.765** | (0.649-0.901) |
| Pre-TQD Slope | −0.238 | (0.150) | −0.176 | (0.150) | 1.158 | (0.963-1.393) | |
| Pre-Post Difference | 0.071 | (0.084) | 0.013 | (0.083) | 1.072 | (0.638-1.801) | |
| Post-TQD Slope | 0.019 | (0.028) | 0.022 | (0.028) | 0.577 | (0.231-1.441) | |
| Cessation fatigue | Pre-TQD Mean | −0.098 | (0.161) | 0.099 | (0.157) | 0.992 | (0.915-1.075) |
| Pre-TQD Slope | −0.068 | (0.189) | 0.156 | (0.191) | 0.950 | (0.862-1.047) | |
| Pre-Post Difference | 0.069 | (0.116) | 0.005 | (0.119) | 1.076 | (0.894-1.295) | |
| Post-TQD Slope | 0.038 | (0.034) | 0.023 | (0.032) | 0.529 | (0.290-0.064) | |
| Motivation | Pre-TQD Mean | 0.119 | (0.096) | 0.033 | (0.097) | 1.617*** | (1.295-2.020) |
| Pre-TQD Slope | −0.079 | (0.155) | −0.064 | (0.150) | 1.352*** | (1.160-1.577) | |
| Pre-Post Difference | −0.040 | (0.077) | −0.019 | (0.074) | 1.881** | (1.253-2.824) | |
| Post-TQD Slope | 0.014 | (0.018) | −0.001 | (0.018) | 7.214** | (1.936-26.885) | |
Significant at:
p<.001
p<.01
p<.05 after applying the Benjamini-Hochberg approach to control the false discovery rate.
p<.05, but effect no longer significant after applying the Benjamini-Hochberg approach.
Multilevel models controlled for age, gender, minority status, longest period of abstinence, baseline Fagerström Test of Cigarette Dependence (FTCD) score, and recent smoking. Age, FTCD score and longest period of abstinence were grand-mean centered.
SAS PROCESS macro v3.1 [25] was used to assess mediated effects for each candidate mediator, with a binary outcome and 1000 bootstrapped samples with resampling. To test the significance of each indirect (mediated) effect, the PROCESS macro generated non-parametric, percentile-based bootstrapped 95% confidence intervals (CI) using the bootstrapped sampling distribution (rather than a normal distribution) [34]. Significant mediation was established if the CI around the mediated effect did not include zero.
Missing data were handled using maximum likelihood estimated in MLM. In logistic regression models, missing abstinence data were imputed as smoking using the intent-to-treat principle. In PROCESS macro analyses of indirect effects, listwise deletion was used; the macro does not yet support multiple imputed datasets.
Results
Descriptive analyses
A total of 1051 smokers (96.8% of the 1086 randomized in the parent trial) with at least one EMA report were included in this analysis. Table 1 summarizes demographic and smoking history information in the retained sample. These participants provided 14553 of 22071 (65.9%) scheduled reports in the pre-TQD week and 43799 of 63060 (69.5%) scheduled reports in the 4 weeks post-TQD, for an average of 56 of 81 scheduled reports per person (69.1%) (SD=20.0, Median=59, Mode=77, Range=1 to 86). These response rates are not prorated for attrition. A total of 981 (93.3%) of the retained sample provided abstinence data at 4-week post-TQD.
Descriptive statistics for mediators are shown in Table 2. Bivariate correlations among mediators were small (rs<.28), with the exception of medium-sized correlations between negative affect and difficulty concentrating (r=.46) and between craving and positive smoking expectancies (r=.67).
Table 2.
Means, standard deviations and correlations among 8 candidate mediators collapsed across pre- and post-TQD
| Mean | SD | M1 | M2 | M3 | M4 | M5 | M6 | M7 | |
|---|---|---|---|---|---|---|---|---|---|
| M1. Craving | 2.98 | 2.06 | |||||||
| M2. Positive smoking expectancies | 3.31 | 2.25 | 0.67*** | ||||||
| M3. Positive affect | 3.37 | 2.03 | −0.06*** | −0.09*** | |||||
| M4. Negative affect | 1.82 | 1.50 | 0.27*** | 0.21*** | −0.05*** | ||||
| M5. Difficulty concentrating | 1.79 | 1.49 | 0.22*** | 0.16*** | 0.00 | 0.46*** | |||
| M6. Hunger | 2.53 | 1.87 | 0.23*** | 0.14*** | 0.10*** | 0.16*** | 0.20*** | ||
| M7. Cessation fatigue | 3.60 | 2.57 | 0.12*** | 0.14*** | 0.00 | 0.09*** | 0.10*** | 0.00 | |
| M8. Motivation | 6.45 | 1.36 | −0.04*** | −0.05*** | 0.09*** | −0.09*** | −0.07*** | −0.01** | 0.09*** |
p <.001
p<.01
SD: Standard Deviation.
Mean, standard deviation and correlation among candidate mediators were computed using a total sample (58,352 EMA reports from 1,051 participants), collapsed across epochs (pre-and post-TQD) and treatment conditions (varenicline, C-NRT, and nicotine patch monotherapy). Craving, positive affect, negative affect, difficulty concentrating and hunger were rated on a 7-point scale from 1 (not at all) and 7 (extremely). Positive smoking expectancies, cessation fatigue and motivation were rated on a 7-point scale from 1 (don’t agree) and 7 (agree).
Pharmacotherapy effects on abstinence
As reported in Baker et al. [7], intent-to-treat 7-day biochemically confirmed point-prevalence abstinence 4 weeks post-TQD was not significantly affected by randomized pharmacotherapy regimen in the sample retained for these analyses: nicotine patch 32.6% (75/230); varenicline 35.6% (145/407); C-NRT 35.0% (145/414); chi-square (N=1051)=0.62, p=0.73.
Pharmacotherapy effects on candidate mediators relative to patch monotherapy
Results of conditional MLMs assessing the effects of varenicline (vs. patch alone) and C-NRT (vs. patch alone) on candidate mediators controlling for recent smoking and baseline covariates are shown in Table 3. Varenicline significantly reduced mean pre-cessation levels of both craving and positive smoking expectancies better than patch alone (Figure 1). Mean pre-post cessation changes in candidate mediators did not differ significantly between varenicline and patch conditions, suggesting that the difference between symptom levels that emerged pre-cessation was not reversed by the start of patch treatment.
Figure 1.
Estimated mean ratings of craving (a), positive smoking expectancies (b) and negative affect (c) from 1 week pre-target quit day (TQD) (pre-cessation phase) to 4 weeks post-TQD (cessation phase) as a function of treatment condition
C-NRT suppressed pre-post cessation increases in mean craving, positive smoking expectancies, and negative affect levels better than patch alone (Table 3, Figure 1). C-NRT was also associated with a slower decline in negative affect over the cessation phase than was patch, however. No other significant treatment effects on mediators were detected after applying the Benjamini-Hochberg [33] approach.
Candidate mediator relations with abstinence
Pre-cessation mean levels of all symptoms except difficulty concentrating and cessation fatigue were predictive of lower log odds of abstinence (Table 3). In both the pre-cessation and cessation phases, craving and positive smoking expectancies that were both high and sustained, predicted lower odds of abstinence. In contrast, greater positive affect pre-cessation and increases in positive affect from pre- to post-TQD positively predicted abstinence. High and increasing motivation were predictive of greater log odds of abstinence in both epochs.
Mediated effects
Mediated effects were significant for varenicline-induced reductions (vs. patch) in mean pre-cessation craving (ab=0.09, SE=0.03, 95% CI=0.04-0.14) and positive smoking expectancies (ab=0.06, SE=0.02, 95% CI=0.02-0.12). Significant mediated effects of C-NRT (vs. patch) were smaller pre to post-TQD increases in mean levels of craving (ab=0.04, SE=0.02, 95% CI=0.01-0.08) and positive smoking expectancies (ab=0.03, SE=0.02, 95% CI=0.001-0.07). C-NRT did not have a significant mediated effect on abstinence via post-TQD negative affect slope (ab=0.02, SE=0.01, 95% CI=−0.002-0.04).
Discussion
Relative to patch monotherapy, varenicline suppressed mean levels of pre-cessation craving and positive smoking expectancies. These results are consistent with findings that varenicline reduces pre-cessation craving and smoking enjoyment better than patches alone or placebo [16, 36]. Varenicline effects on 4-week abstinence were mediated by its pre-cessation effects: i.e., suppressing craving and smoking expectancies, relative to patch alone.
C-NRT significantly reduced cessation-phase increases in craving, positive smoking expectancies, and increased post-TQD negative affect slope, relative to patch monotherapy. The effects of C-NRT (vs. patch) on 4-week abstinence were mediated by suppression of cessation-phase craving and positive smoking expectancy spikes. These findings are consistent with research showing that C-NRT and high-dose NRT effects on abstinence are mediated by craving reductions [1, 15]. Thus, although the enhanced therapies produced indirect effects in different phases in the cessation process, their effects occurred via the same mediators: craving and positive smoking expectancies, ratings tapping moderately correlated but conceptually distinct experiences (wanting vs. anticipated enjoyment).
The only other significant pharmacotherapy effects were that, relative to the patch, C-NRT slowed the decrease in negative affect post-TQD, but this did not have a significant indirect effect . Despite the theoretical and empirical evidence supporting the importance of negative affect in smoking [35], studies often fail to find benefits of pharmacotherapy on abstinence via negative affect [15, 23, 24].
The current findings have implications for treatment. First, this research adds to evidence that craving suppression is a valid index of pharmacotherapy benefit [23, 24, 36], and suggest that positive smoking expectancies may also be useful as a treatment response marker. According to the Phase-Based Model of smoking treatment [17], identifying interventions that are especially effective in each smoking treatment phase may lead to cumulative improvements in abstinence rates. While varenicline did not significantly improve 4-week abstinence rates relative to patch, the evidence of mechanistic benefit in the pre-cessation phase suggests that varenicline may set the stage for quitting by reducing craving and smoking expectancies pre-cessation. This highlights the potential value of research on extended pre-cessation varenicline treatment supported by promising initial findings [37,38]. Pre-cessation NRT may work via similar mechanisms, although its effects are inconsistent [39], which may heighten the value of pre-cessation varenicline [36-38]. Finally, the overlap in mediators for varenicline and C-NRT effects (vs. patch) raises questions about the additivity or redundancy of their effects (i.e., would NRT and varenicline be especially potent in reducing craving and positive smoking expectancies?) as suggested by some promising data [40].
Limitations of this research include the fact that data were drawn from a trial that yielded atypical results in which varenicline and C-NRT did not enhance abstinence rates relative to patch alone [26, 41], and did not include a placebo condition. In addition, mediators were assessed pragmatically via single items and response rates were below recommended levels [42] and data may not have been missing at random. An intensive longitudinal study paradigm like this offers tradeoffs between the benefits of enhanced ecological validity, increased temporal granularity, and reduced recall biases at the cost of more missing data [43-44]. Other EMA methods (e.g., passive monitoring, text-message or in-app responding rather than IVR) may yield higher response rates [45-46]. Other limitations include the lack of examination of potential interactions between pre- and post-TQD effects or moderation of treatment effects by study site. Finally, the study was not designed or powered for these secondary analyses. This concern is mitigated by the rigor of the randomized design and a sample size that was selected to confer power >.80 to detect moderate comparative effects (e.g., 9-10%) in a 6-month binary abstinence outcome, and the fact that power is often greater for indirect than direct effects [34].
Conclusions
Suppression of craving and anticipated enjoyment of smoking mediated the effects of both varenicline and C-NRT on short-term abstinence. These mediated effects were significant for both agents when they were first used: for varenicline during the pre-cessation phase of smoking treatment and for C-NRT during the cessation phase. Thus, cigarette craving and anticipated smoking enjoyment may sensitively index enhanced pharmacotherapy effects in the context of open-label use with psychosocial support and may be useful markers of treatment response for the purpose of treatment adjustment.
Supplementary Material
Acknowledgements
The authors wish to acknowledge Drs. Michael Fiore and James Stein, principal investigators of the parent study; researchers at the University of Wisconsin Center for Tobacco Research and Intervention; and participants in the study.
Funding: This research was supported by grant R01HL109031 from the National Heart, Lung, and Blood Institute at the National Institutes of Health and by the Wisconsin Partnership Program.
Footnotes
Conflict of Interest Declaration: Drs. Baker, McCarthy, and Piper are conducting a research study for which Pfizer Inc. donated varenicline and placebo medication.
Clinical trial registration: ClinicalTrials.gov NCT01553084
Compliance with Ethical Standards: The study was approved by the University of Wisconsin Health Sciences Institutional Review Board. All participants provided written informed consent for study participation.
References
- 1.Bolt DM, Piper ME, Theobald WE, Baker TB. Why two smoking cessation agents work better than one: role of craving suppression. J Consult Clin Psychol. 2012;80(1):54–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Gonzales D, Rennard SI, Nides M, Oncken C, Azoulay S, Billing CB, et al. Varenicline, an alpha4beta2 nicotinic acetylcholine receptor partial agonist, vs sustained-release bupropion and placebo for smoking cessation: a randomized controlled trial. JAMA. 2006;296:47–55. [DOI] [PubMed] [Google Scholar]
- 3.Jorenby DE, Hays JT, Rigotti NA, Azoulay S, Watsky EJ, Williams KE, et al. Efficacy of varenicline, an alpha4beta2 nicotinic acetylcholine receptor partial agonist, vs placebo or sustained-release bupropion for smoking cessation: a randomized controlled trial. JAMA. 2006;296(1):56–63. [DOI] [PubMed] [Google Scholar]
- 4.Piper ME, Smith SS, Schlam TR, Fiore MC, Jorenby DE, Fraser D, et al. A randomized placebo-controlled clinical trial of 5 smoking cessation pharmacotherapies. Arch Gen Psychiatry. 2009;66(11):1253–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Smith SS, McCarthy DE, Japuntich SJ, Christiansen B, Piper ME, Jorenby DE, et al. Comparative effectiveness of 5 smoking cessation pharmacotherapies in primary care clinics. Arch Intern Med. 2009;169(22):2148–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Rose JE, Behm FM. Combination varenicline/bupropion treatment benefits highly dependent smokers in an adaptive smoking cessation paradigm. Nicotine Tob Res. 2017;19(8):999–1002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Baker TB, Piper ME, Stein JH, Smith SS, Bolt DM, Fraser DL, et al. Effects of nicotine patch vs varenicline vs combination nicotine replacement therapy on smoking cessation at 26 weeks: a randomized clinical trial. JAMA. 2016;315(4):371–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.O’Rourke HP, MacKinnon DP. Reasons for testing mediation in the absence of an intervention effect: a research imperative in prevention and intervention research. J Stud Alcohol Drugs. 2018;79(2):171–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Aubin HJ, Bobak A, Britton JR, Oncken C, Billing CB, Jr., Gong J, et al. Varenicline versus transdermal nicotine patch for smoking cessation: Results from a randomised, open-label trial. Thorax. 2008; 63(8):717–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ray LA, Lunny K, Bujarski S, Moallem N, Krull JL, Miotto K. The effects of varenicline on stress-induced and cue-induced craving for cigarettes. Drug Alcohol Depend. 2013;131(1–2):136–42. [DOI] [PubMed] [Google Scholar]
- 11.Hughes JR, Rennard SI, Fingar JR, Talbot SK, Callas PW, Fagerstrom KO. Efficacy of varenicline to prompt quit attempts in smokers not currently trying to quit: a randomized placebo-controlled trial. Nicotine Tob Res. 2011;13(10):955–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Nides M, Oncken C, Gonzales D, Rennard S, Watsky EJ, Anziano R, et al. Smoking cessation with varenicline, a selective alpha4beta2 nicotinic receptor partial agonist: results from a 7-week, randomized, placebo- and bupropion-controlled trial with 1-year follow-up. Arch Intern Med. 2006;166(15):1561–8. [DOI] [PubMed] [Google Scholar]
- 13.Brandon TH, Drobes DJ, Unrod M, Heckman BW, Oliver JA, Roetzheim RC, et al. Varenicline effects on craving, cue reactivity, and smoking reward. Psychopharmacology (Berl). 2011;218(2):391–403. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Fedota JR, Sutherland MT, Salmeron BJ, Ross TJ, Hong LE, Stein EA. Reward anticipation is differentially modulated by varenicline and nicotine in smokers. Neuropsychopharmacology. 2015;40(8):2038–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Ferguson SG, Shiffman S, Gwaltney CJ. Does reducing withdrawal severity mediate nicotine patch efficacy? A randomized clinical trial. J Consult Clin Psychol. 2006;74(6):1153–61. [DOI] [PubMed] [Google Scholar]
- 16.Lu W, Chappell K, Walters JAE, Jacobson GA, Patel R, Schuz N, et al. The effect of varenicline and nicotine patch on smoking rate and satisfaction with smoking: an examination of the mechanism of action of two pre-quit pharmacotherapies. Psychopharmacology (Berl). 2017;234(13):1969–76. [DOI] [PubMed] [Google Scholar]
- 17.Baker TB, Mermelstein R, Collins LM, Piper ME, Jorenby DE, Smith SS, et al. New methods for tobacco dependence treatment research. Ann Behav Med. 2011;41(2):192–207. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Stone AA, Shiffman S. Ecological momentary assessment (EMA) in behavorial medicine. Ann Behav Med. 1994;16(3):199–202. [Google Scholar]
- 19.McCarthy DE, Bolt DM, Baker TB. The importance of how: A call for mechanistic research in tobacco dependence treatment studies. In: Treat T, Bootzin RI, Baker TB, editors. Psychological clinical science: recent advances in theory and practice Integrative perspectives in honor of Richard M McFall. New York: Lawrence Erlbaum Associates; 2007. p. 133–63. [Google Scholar]
- 20.Sayette MA. The role of craving in substance use disorders: theoretical and methodological issues. Annu Rev Clin Psychol. 2016;12:407–33. [DOI] [PubMed] [Google Scholar]
- 21.Hughes JR. Effects of abstinence from tobacco: etiology, animal models, epidemiology, and significance: a subjective review. Nicotine Tob Res. 2007;9(3):329–39. [DOI] [PubMed] [Google Scholar]
- 22.McCarthy DE, Piasecki TM, Fiore MC, Baker TB. Life before and after quitting smoking: an electronic diary study. J Abnorm Psychol. 2006;115(3):454–66. [DOI] [PubMed] [Google Scholar]
- 23.McCarthy DE, Piasecki TM, Lawrence DL, Jorenby DE, Shiffman S, Baker TB. Psychological mediators of bupropion sustained-release treatment for smoking cessation. Addiction. 2008;103(9):1521–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Piper ME, Federman EB, McCarthy DE, Bolt DM, Smith SS, Fiore MC, et al. Using mediational models to explore the nature of tobacco motivation and tobacco treatment effects. J Abnorm Psychol. 2008;117(1):94–105. [DOI] [PubMed] [Google Scholar]
- 25.Hayes AF. Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. 2nd ed. New York: Guilford publications; 2017. [Google Scholar]
- 26.Fiore MC, Jaen CR, Baker TB, Bailey WC, Benowitz N, Curry SJ, et al. Treating tobacco use and dependence: 2008 update. Rockville, MD: U.S. Department of Health and Human Services, U.S. Public Health Service; 2008. [Google Scholar]
- 27.Fagerstrom K Determinants of tobacco use and renaming the FTND to the Fagerstrom Test for Cigarette Dependence. Nicotine Tob Res. 2012;14(1):75–8. [DOI] [PubMed] [Google Scholar]
- 28.Watson D, Clark LA, Tellegen A. Development and validation of brief measures of positive and negative affect: The PANAS scales. J Pers Soc Psychol. 1988;54(6):1063–70. [DOI] [PubMed] [Google Scholar]
- 29.Welsch SK, Smith SS, Wetter DW, Jorenby DE, Fiore MC, Baker TB. Development and validation of the Wisconsin Smoking Withdrawal Scale. Exp Clin Psychopharmacol. 1999;7(4):354–61. [DOI] [PubMed] [Google Scholar]
- 30.Shiffman S, Ferguson SG. The effect of a nicotine patch on cigarette craving over the course of the day: results from two randomized clinical trials. Curr Med Res Opin. 2008;24(10):2795–804. [DOI] [PubMed] [Google Scholar]
- 31.McCarthy DE, Piasecki TM, Jorenby DE, Lawrence DL, Shiffman S, Baker TB. A multi-level analysis of non-significant counseling effects in a randomized smoking cessation trial. Addiction. 2010;105(12):2195–208. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Bryk AS, Raudenbush SW. Hierarchical linear models: Applications and data analysis methods. Newbury Park, CA: Sage; 1992. [Google Scholar]
- 33.Benjamini Y, Hochberg Y. Controlling the false discovery rate: A practical and powerful approach to multiple testing. J Royal Statistical Soc, Series B. 1995; 57:289–300. [Google Scholar]
- 34.Hayes AF, Rockwood NJ. Regression-based statistical mediation and moderation analysis in clinical research: Observations, recommendations, and implementation. Behaviour Res Therapy. 2017;98:39–57. [DOI] [PubMed] [Google Scholar]
- 35.Baker TB, Piper ME, McCarthy DE, Majeskie MR, Fiore MC. Addiction motivation reformulated: an affective processing model of negative reinforcement. Psychol Rev. 2004;111(1):33–51. [DOI] [PubMed] [Google Scholar]
- 36.Hajek P, McRobbie HJ, Myers KE, Stapleton J, Dhanji AR. Use of varenicline for 4 weeks before quitting smoking: decrease in ad lib smoking and increase in smoking cessation rates. Arch Intern Med. 2011;171(8):770–7. [DOI] [PubMed] [Google Scholar]
- 37.Ebbert JO, Hughes JR, West RJ, Rennard SI, Russ C, McRae TD, et al. Effect of varenicline on smoking cessation through smoking reduction: a randomized clinical trial. JAMA. 2015;313(7):687–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Hawk LW Jr, Ashare RL, Lohnes SF, Schlienz NJ, Rhodes JD, Tiffany ST, et al. The effects of extended pre-quit varenicline treatment on smoking behavior and short-term abstinence: a randomized clinical trial. Clin Pharmacol Ther. 2012;91(2):172–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Lindson N, Aveyard P. An updated meta-analysis of nicotine preloading for smoking cessation: investigating mediators of the effect. Psychopharmacology (Berl). 2011;214(3):579–92. [DOI] [PubMed] [Google Scholar]
- 40.Koegelenberg CF, Noor F, Bateman ED, van Zyl-Smit RN, Bruning A, O’Brien JA, et al. Efficacy of varenicline combined with nicotine replacement therapy vs varenicline alone for smoking cessation: a randomized clinical trial. JAMA. 2014; 312(2):155–61. [DOI] [PubMed] [Google Scholar]
- 41.Cahill K, Stevens S, Perera R, Lancaster T. Pharmacological interventions for smoking cessation: an overview and network meta-analysis. Cochrane Database Syst Rev. 2013(5):CD009329. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Stone AA, Shiffman S Capturing momentary, self-report data: A proposal for reporting guidelines. Ann Behav Med. 2002;24:236–243. [DOI] [PubMed] [Google Scholar]
- 43.Rowan PJ, Cofta-Woerpel L, Mazas CA, et al. Evaluating reactivity to ecological momentary assessment during smoking cessation. Exp Clin Psychopharmacol. 2007;15(4):382–389. [DOI] [PubMed] [Google Scholar]
- 44.Kirchner TR, Shiffman S. Spatio-temporal determinants of mental health and well-being: advances in geographically-explicit ecological momentary assessment (GEMA). Soc Psychiatry Psychiatr Epidemiol. 2016;51(9):1211–1223. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Runyan JD, Steinke EG. Virtues, ecological momentary assessment/intervention and smartphone technology. Front Psychol. 2015;6:481. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Hébert ET, Stevens EM, Frank SG, et al. An ecological momentary intervention for smoking cessation: The associations of just-in-time, tailored messages with lapse risk factors. Addict Behav. 2018;78:30–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.