Sex-Specific Mediation of Pre-Quit Smoking Reduction: Secondary Analysis of a Randomized Controlled Trial Extending Varenicline Preloading

Abstract

Introduction

Relative to other pharmacotherapies for smoking cessation, varenicline has significantly greater efficacy in females; however, sex-specific mechanisms have not yet been investigated. We conducted a secondary analysis of ecological momentary assessment data to assess whether reductions in craving, negative affect, and smoking satisfaction/reward/aversion mediate effects of varenicline on next-day smoking to a greater degree in females (n = 179) relative to males (n = 141).

Aims and Methods

Data were from a 3-week medication manipulation period during the pre-quit phase of a double-blind randomized placebo-controlled trial investigating extended preloading (4 weeks) versus standard preloading of varenicline (1 week, preceded by 3 weeks of placebo, NCT03262662). Time-invariant multi-level moderated mediation models and time-varying mediation models were utilized.

Results

A significant time-varying indirect effect through craving that increased in magnitude over the pre-quit period was identified only in females. Exploratory analysis found that decreases in psychological reward and smoking satisfaction mediated the relationship between varenicline and reductions in craving only in females. Time-invariant multi-level models did not evidence a significant indirect effect through candidate mediators in males or females; the index of moderated mediation was not significant in any of the models.

Conclusions

Our findings suggest that the efficacy of varenicline on reductions in pre-quit smoking in females operates through reductions in craving. Furthermore, reductions in craving may be due to decreases in positive subjective experiences of smoking. Augmenting craving coping strategies as well as reducing smoking reward and satisfaction may be a beneficial approach in females.

Implications

This is the first study to investigate sex-specific mediation of varenicline on reductions in pre-quit smoking. Further investigation into varenicline-induced changes in smoking reinforcement and craving is warranted, particularly in females. For example, experimentally manipulating these mediators may inform them as mechanisms for smoking reduction.

Introduction

Cigarette smoking is the leading preventable cause of death in North America.¹ Varenicline is considered the most effective monotherapy for nicotine addiction, with greater rates of sustained abstinence than bupropion or nicotine replacement monotherapy (NRT^2,3). Although females smoke, on average, fewer cigarettes per day (CPD) than males,⁴ they are at greater risk of smoking-related consequences, including cardiorespiratory-related morbidity.^5,6 In addition, despite being more likely to report quit attempts,⁷ females are less likely than males to quit smoking.^8,9 Identifying sex-specific mechanisms of treatment efficacy may inform putative targets for personalizing treatment.

To this end, we examine ecological momentary assessment (EMA) data collected across a 3-week medication manipulation period during which participants were randomized to varenicline versus placebo in a clinical trial¹⁰ investigating extended (4-weeks varenicline) versus standard (3-weeks placebo, 1-week varenicline) preloading before the target quit date (TQD). Notably, pre-quit changes may facilitate end-of-treatment (EOT) outcomes; we recently reported that greater declines in pre-quit smoking significantly mediated greater odds of abstinence at EOT in extended relative to standard preloading¹¹ and this indirect effect was significant in females but not in males.¹¹ In the present paper, we further our examination of the causal chain by focusing on the proximal mediators of declines in pre-quit smoking, in females and males.

Smoking Cessation Pharmacotherapy

Varenicline is a partial agonist at α4β2* nicotinic acetylcholine receptors (nAChRs) such that agonist actions reduce withdrawal and antagonist actions block reinforcement during nicotine exposure.^12,13 Indeed, preclinical models demonstrate dose-dependent reductions in nicotine self-administration and withdrawal symptoms,^12,14,15 and human laboratory studies report that varenicline decreases smoking rate,^16,17 subjective pleasurable effects of smoking,^18–20 and craving.^21–23 Cumulatively, this evidence supports reduced reinforcement, craving, and withdrawal as potential treatment mechanisms. However, there is limited literature investigating mediating pathways of smoking outcomes from intervention studies, with only one study finding an indirect effect of varenicline on abstinence through reductions in craving.²⁴

Studies suggest efficacy of varenicline on smoking cessation may differ by sex. A large network meta-analysis of randomized, double-blind, placebo-controlled smoking cessation pharmacotherapy studies²⁵ found varenicline significantly outperformed both bupropion and NRT for females, whereas there was no significant difference amongst the three pharmacotherapies for males. Moreover, varenicline produced comparable biochemically verified abstinence rates in females and males. Conversely, bupropion and NRTs had significantly lower efficacy in females relative to males, indicating a unique role of partial agonist mechanisms of varenicline in outcomes.

Potential Mediators of Varenicline’s Effects in Females

Sex differences in smoking dependence and cessation may in part explain the efficacy of varenicline in females. Although the literature is mixed,²⁶ some animal models suggest females may be biologically predisposed to experience greater reinforcement from nicotine and non-nicotine stimuli, which may increase craving and negative affect during abstinence, leading to relapse.^27–29 For example, within human studies, female participants reported enhanced subjective responses to low doses of nicotine, relative to males.³⁰ Females also report similar subjective responses to low and moderate nicotine-containing cigarettes,³¹ suggesting greater sensitivity to non-nicotine reinforcement (eg, to olfactory cues, which is additionally supported by denicotinized cigarette responses³²). Specific factors driving increased reinforcement are difficult to disentangle; however, the above findings suggest females may experience greater psychological reward and smoking satisfaction. Studies have also found greater craving and negative affect during smoking reduction/abstinence and greater relief in response to nicotine reinstatement in females.^33–35

Present Study

We conducted secondary analyses to test candidate mediators of varenicline effects on pre-quit smoking reductions during the 3-week medication manipulation phase of a clinical trial when participants were randomized to varenicline or placebo. Specifically, we used moderated multi-level mediation models (MLMs) to assess whether reductions in smoking satisfaction/reward/aversion, negative affect, and/or cigarette craving mediated reductions in next-day CPD, and if these relationships are sex-specific, such that indirect effects are only evident or are stronger in females relative to males.

Following recommended titration guidelines, it takes ~10 days to reach steady state plasma varenicline levels³⁶ with smoking reductions typically occur gradually after the first week of treatment as varenicline levels reach steady state.^17,37 Furthermore, evidence suggests that extinction from reinforcement may require numerous learning “trials” across a range of contexts,^38,39 with smoking behavior gradually declining. One previous study investigating the effect of varenicline on post-quit smoking suggested the mediated effect through craving strengthened over time.⁴⁰ Therefore, time-varying effect mediation models (TVEM) were utilized as a complementary statistical approach to the MLMs.

Materials and Methods

Participants

As depicted in Table 1, participants were N = 320 treatment-seeking individuals who reported smoking 5 + CPD for 6 + months, aged 18–70 years, randomized to extended (4-weeks varenicline, females n = 90; males n = 73) or standard preloading (3-weeks placebo, 1-week varenicline females = 89, males = 68) of varenicline. Exclusion criteria included psychosis, major depressive disorder, suicidality, pregnancy, and other substance use disorders. Full details on sample and inclusion/exclusion criteria, missing data, and adherence may be found elsewhere.¹⁰ The study was reviewed by the State University of New York Institutional Review Board and all participants provided informed consent.

Table 1.

Demographic and Baseline Characteristics of Sample

	Males		Females
	Standard preloading (n = 68)	Extended preloading (n = 73)	Standard preloading (n = 89)	Extended preloading (n = 90)
Demographic characteristics
Age (years)	52.7 (10.9)	53.2 (10.9)	54.2 (9.3)	54.2 (9.8)
Race (%)
Black	14.7	12.3	27.0	26.7
White	80.9	80.8	70.8	70.0
Native American, Alaska Native, mixed-ethnicity, other	4.4	6.8	2.2	3.3
Hispanic (%)	2.9	1.4	2.2	3.3
Post-secondary education (%)a	72.3	67.1	77.0	77.3
FT employment (%)b	52.3	48.5	42.5	43.2
Income > $50k (%)c	63.9	60.6	51.8	49.4
Baseline characteristics
CPD	20.3 (8.1)	19.5 (8.0)	17.0 (6.2)	16.5 (6.1)
FTCDd	5.5 (2.2)	5.5 (1.8)	5.6 (1.9)	5.4 (2.1)
Age started daily smoking	18.7 (3.6)	17.9 (3.6)	18.5 (3.5)	18.8 (3.5)
Duration of daily smoking	33.7 (11.4)	35.1 (11.2)	35.8 (9.5)	35.3 (9.7)

Abbreviations: FT, full time; CPD, Cigarettes Per Day; FTCD, Fagestrom Test of Cigarette Dependence.

Data are mean (SD) unless otherwise indicated.

^aData were missing for 10 of 320 participants (3.1%).

^bData were missing for 12 of 320 participants (3.8%).

^cData were missing for 29 of 320 participants (9.1%).

^dScores range from 0 to 10 with higher scores indicating greater dependence.

Overview

We conducted secondary analyses on data from a 2-group, double-blind, placebo-controlled, randomized, parallel-group design clinical trial (NCT03262662; 10) assessing the effects of extending the preloading period of varenicline from 1 to 4 weeks prior to the TQD, followed by 11 weeks of open-label varenicline. After a phone screen and in-person eligibility assessments, participants completed two laboratory visits, a baseline week of EMA, six clinic visits at week 0, 1, 3, 4 (TQD), 6, and 8, and follow-up visits at Week 15 (EOT) and 28. Varenicline was dispensed following recommended titration guidelines up to 2.0 mg bid. Participants also received brief behavioral counseling sessions at each of six clinic visits; during pre-quit sessions, they were encouraged to smoke as usual without actively trying to cut down or quit. Our parent trial found a significant reduction in pre-quit smoking in the extended relative to the standard preloading group. Although EOT and 6-month abstinence rates were not significantly different between groups,¹⁰ greater reductions in smoking rate during the pre-quit period significantly mediated extended preloading effects on EOT smoking abstinence in females but not males.¹¹ In the current project, we examine EMA-assessed candidate proximal treatment mechanisms of changes in pre-TQD smoking rate during the 3-week medication manipulation period when participants were randomized to varenicline (the extended preloading group) or placebo (the standard preloading group). Note that pill count varenicline adherence did not differ by treatment group, sex, or their interaction (70%–80% of participants in each cell were adherent across the 15-week treatment period; 80% total adherence (see¹⁰ for details).

Ecological Momentary Assessment Procedures

Participants responded to daily questionnaires via the smartphone app mEMA (ilumivu.com) from the Baseline Week through Treatment Week 8 (9 weeks total). Each morning, participants reported how many cigarettes they smoked the previous day (and other assessments not examined here). Following the morning assessment, four random alarms occurred over a 12-hour period tailored to the participant’s usual wake time, with one alarm occurring randomly within each 3-hour block. Assessment content employed planned missingness and depended on whether the participant reported smoking in the past 2 hours. If yes, they were asked how many cigarettes they smoked, and each of the following domains had a 0.50 probability of being assessed: subjective effects of the most recent cigarette, craving, withdrawal, and nausea. If no, each of the following had a 0.67 probability of being assessed: craving, withdrawal, and nausea. Participants had 15 minutes to initiate a response to an alarm and 5 minutes to complete the measures. During the medication manipulation phase, average completion rates for the morning assessment and the 4/day random EMA alarms (when mediators were assessed) were over 80% (~17/21 days) and 55%, respectively (~46/84). Full procedural details can be found in Hawk et al.¹⁰ and details on the EMA methods and compliance can be found in Tonkin et al.⁴¹

Measures

Outcome

Cigarettes Per Day.

Each morning, participants were asked how many cigarettes they smoked the previous day (CPD). We evaluated day-to-day changes in CPD as a continuous outcome over the medication manipulation period.

Candidate Mediators

Candidate mediators were taken from the random alarms of the EMAs. For each mediator, the four random alarm assessments were averaged within a day to form a daily composite.

Craving.

The craving questionnaire was comprised of four items derived from the Questionnaire on Smoking Urges—Brief⁴²: “Nothing would be better than smoking a cigarette right now,” “All I want right now is a cigarette,” “I crave a cigarette right now,” and I have an urge for a cigarette,” each rated 1–5 as in prior work (from⁴³).

Negative Affect.

We assessed negative affect with the irritability, anxiety, and depressed mood items from the Minnesota Tobacco Withdrawal Scale⁴⁴ as recent factor analyses⁴⁵ found that these three items load on a negative affect domain with a good fit.

Subjective Effects of Smoking.

Participants reported subjective experiences of cigarette smoking on the modified Cigarette Evaluation Questionnaire (mCEQ;⁴⁶). Participants were asked 11 questions (one item [whether the cigarette was enjoyed] was accidentally omitted from the mCEQ), each on a seven-point Likert scale. As previous factor analyses on the mCEQ were not conducted with EMA data,⁴⁶ an independent factor analysis was fit with the present data. Separate exploratory and confirmatory factor analyses were conducted for all weeks combined and for each week before the quit date. We found that all items loaded on three subscales with generally good fit: smoking satisfaction (satisfying, tastes good, enjoy sensations, reduce craving), psychological reward (calm, awake, irritability, concentration, hunger), and aversion (dizzy, nauseated). Further details are available from the authors.

Statistical Analysis

This analysis plan was pre-registered on the Open Science Framework (https://osf.io/dc2yp/). Treatment and sex were coded as 0/1 and mediators were grand mean standardized. Correlations between candidate mediators were small to moderate (|r| = .02 −.41), supporting discriminant validity. Additional detail on analysis can be found in Supplementary Material Sections 1 and 2. Moderated mediation MLMs were estimated using the R package “lme4,”⁴⁷ which tested the effect of varenicline on the mediator (a-path) and of the mediator on next-day CPD controlling for treatment condition (b-path), and distinguished the direct effect (c’) from the indirect effect (a*b). In moderated mediation, each coefficient was expressed as a linear function of the moderator (W; 48). We adapted the approach recommended by Bauer et al.⁴⁸ for MLM mediation to fit with fixed slopes (see Supplementary Material for full details). To maintain temporal precedence, we aggregated candidate mediators within the day and the outcome was the total cigarettes smoked the next day. Models were analyzed with and without controlling for same-day smoking and differences are reported. The overall indirect effect was not assessed, given that the goal was to test mechanisms in each sex.

One drawback of the MLM framework is that models are parametric and to assess non-linear relationships, the researcher must specify the shape of growth, substantially increasing model complexity, and introducing additional modeling assumptions. In addition, MLMs provide a single coefficient estimate averaged over time. In contrast, TVEMs are nonparametric, data-driven, and provide a flexible approach to explore day-to-day changes, without having to specify the shape of the data. TVEMs are therefore suitable to evaluate if and when the significance and magnitude of mediated effects, change across time.⁴⁹ Using the R package “tvmediation,” two-step local smoothing methods were used to estimate time-varying coefficients and corresponding standard errors, followed by bootstrapping to obtain confidence intervals for the indirect effect.⁴⁹ Mediators were standardized and outcomes were modeled as percentile rank. Models were implemented following previous publications from package authors.^41,49 Lags for missing data-points were not included. Modeling was consistent with mediation such that the intervention (varenicline) preceded the candidate mediator (eg, craving), which preceded change in outcome (next-day CPD). Nonparametric bootstrapping was sampled with replacement and estimated mediation at each pre-specified time point, creating a confidence interval for each day. Connecting all the time-points yielded a confidence band of the mediation effect across the 21-day intervention period. To assess sex-specific effects, models with each candidate mediator were run separately in males and females, as the “tvmediation” package does not support the formal introduction of a moderator. TVEM outputs are presented using figures for ease of interpretation, anywhere we discuss the results as significant, it is significant per a numerical 95% confidence interval. All analyses were conducted in R v4.3.1⁵⁰, and two-tailed α was = 0.05.

Results

For females in the varenicline group, smoking during the medication manipulation period decreased by 3.4 CPD whereas for females in the placebo group smoking decreased by 1.1 CPD; t₂₇₆₃ = −8.40, p < .001. Similarly, the average smoking rate decreased by 3.7 CPD in males in the varenicline group whereas for males in the placebo group CPD decreased by 2.2 CPD, t₂₁₃₇ = −5.16, p < .001. The interaction of Treatment and Sex on CPD was not significant, t₁₇₆ = 0.90, p = .37.

Invariant Multi-Level Moderated Mediation Models and Model Validation

We summarize the results of the multi-level moderated mediation models with craving as the candidate mediator; similar patterns were observed with other candidate mediators (see Table 2 and further details in Supplementary Material). The index of moderation of the indirect effect for craving was not significant (b = 0.15, 95% CI [−0.08, 0.47]; see Table 2). In females, the indirect effect of craving was not significant, though it was in the predicted direction, (indirect effect, b = −0.08, 95% CI [−0.21, 0.05]). The effect of treatment on lower craving did not reach significance (a-path, b = −0.13, p = .25); however, reduced craving predicted next-day smoking (b-path, b = 0.60, p < .001) and this relationship remained significant after controlling for same-day smoking (b = 0.41, p < .001). In males, the indirect effect was not significant, and it was in the opposite of the predicted direction (indirect effect, b = 0.08, 95% CI [−0.13, 0.36]; see Table 2). The effect of treatment on craving in males was not significant (a-path, b = 0.22, p = .41); lower craving significantly predicted lower next-day smoking (b-path, b = 0.37, p = .04), but not after controlling for same-day smoking (b = 0.14, p = .39). Overall, there was a significant direct effect of varenicline on CPD (b = −1.77, p = .03). Regarding reward, smoking satisfaction, aversion, and negative affect, the index of moderated mediation and indirect effects by sex were not significant (see Table 2).

Table 2.

EMModerated Multi-Level Mediation Models With Each Candidate Mediator

	DV: next-day cigarettes per day
	Mediator
	Craving	Negative affect	Satisfaction	Reward	Aversion
Intercept (m)	0.25 (0.18)	0.12 (0.19)	0.16* (0.10)	0.52** (0.22)	−0.02 (0.08)
Intercept (y)	15.92*** (1.30)	16.09*** (1.31)	14.17*** (0.67)	15.58*** (1.32)	14.29*** (0.66)
A-Path (Female)	−0.13 (0.11)	−0.01 (0.12)	−0.16 (0.14)	−0.26 (0.14)	0.13 (0.12)
A-Path (Male)	0.22 (0.27)	0.18 (0.31)	0.02 (0.15)	0.12 (0.34)	0.0003 (0.13)
Sex	−0.36 (0.26)	−0.38 (0.30)	−0.29** (0.15)	−0.56* (0.33)	0.01 (0.13)
B-Path (Female)	0.60*** (0.06)	0.08 (0.07)	1.08*** (0.10)	0.72*** (0.10)	0.42*** (0.07)
B-Path (Male)	0.37* (0.18)	−0.19 (0.26)	0.89*** (0.14)	1.03*** (0.32)	−0.06 (0.10)
Sex	2.25*** (0.79)	2.31*** (0.80)	2.42*** (0.80)	2.44*** (0.81)	2.21*** (0.80)
C’ (direct effect)	−1.77** (0.79)	−1.87** (0.79)	−1.54* (0.80)	−1.46* (0.80)	−1.55** (0.79)
Sex* A-Path	0.17 (0.17)	0.09 (0.19)	0.18 (0.21)	0.19 (0.21)	−0.12 (0.18)
Sex* B-Path	−0.11 (0.11)	−0.13 (0.14)	−0.19 (0.17)	0.16 (0.18)	0.36*** (0.12)
A*B (indirect effect; Female)	−0.08 (−0.21, 0.05)	0.01 (−0.02, 0.02)	−0.17 (−0.48, 0.10)	−0.19 (−0.40, 0.006)	0.05 (−0.04, 0.16)
A*B (indirect effect; Male)	0.08 (−0.13, 0.36)	−0.03 (−0.30, 0.16)	0.02 (−0.25, 0.29)	0.12 (−0.58, 0.89)	0 (−0.10, 0.10)
Index of moderated mediation	0.15 (−0.08, 0.47)	−0.01 (−0.30, 0.16)	0.19 (−0.21, 0.59)	0.30 (−0.43, 1.12)	−0.05 (−0.17, 0.05)

*p < .05 **p < .01 ***p < .001; data is effect (se) or effect (95% CI).

a-path is treatment group→ mediator, b- path is mediator→ CPD, direct effect is treatment group→ outcome, controlling for indirect effect.

Intraclass coefficients computed from intercept only models: craving (0.51); negative affect (0.61); satisfaction (0.79); reward (0.80); aversion (0.63); cigarettes per day (0.84).

Output reflects results of multi-level 1-1-1 within-person moderated mediation model with slope coefficients depicted at each level of the moderator for a-path and b-path.

Index of moderated mediation represents the difference between the indirect effect at each value of each moderator.

Each mediator was assessed in a separate model.

Bolded cells are significant below the .05 level. Italicized b-path reflects that b-path remained significant after controlling for same-day smoking.

To assist in determining the temporal direction of effects, we reversed the mediator and outcome variables in the MLMs. We found that the results did not support changes in smoking predicting changes in candidate mediators (ie, the opposing temporal direction, see details in Supplementary Material Section 4).

Time-Varying Mediation of Changes in Pre-Quit CPD

The time-varying indirect effects for females are presented in the upper panels of Figures 1–4 in the main text and the time-varying Path A and B effects for females are presented in the upper panels for Supplementary Figures S1–S5; the time-varying effects for males are presented in the lower panels.

Figure 1.

Time-varying indirect effect of varenicline on smoking through craving by sex (solid line; dashed line depicts 95% confidence interval).

Figure 2.

Time-varying indirect effect of varenicline on smoking through smoking satisfaction by sex (solid line; dashed line depicts 95% confidence interval).

Figure 3.

Time-varying indirect effect of varenicline on smoking through reward by sex (dashed line; dashed line depicts 95% confidence interval).

Figure 4.

Left panel: time-varying indirect effect of varenicline on craving through smoking satisfaction by sex. Right panel: time-varying indirect effect of varenicline on craving through reward by sex (solid line; dashed line depicts 95% confidence interval).

Craving

Amongst females, the time-varying indirect effect of craving on CPD was significant on Day 8 (see Figure 1, upper panel). The magnitude of the indirect effect gradually became stronger until Day 10, after which it plateaued. When inspecting the paths separately, a similar pattern was observed in the effects of varenicline on craving (path a) and the relationship between craving and next-day smoking (path b) (Supplementary Figure S1, upper panels). Among males, the indirect effect of craving was not significant at any time point (Figure 1, lower panel). The effect of varenicline on craving was similar in magnitude to the effect found in females and significant after Day 10 (except Day 17–18); however, the relationship between craving and next-day smoking was not significant at any time point (Supplementary Figure S1, lower panels).

Smoking Satisfaction

Among females, the time-varying indirect effect of smoking satisfaction on CPD was significant in females on Days 3, 15–17 (Figure 2, upper panel). Varenicline significantly reduced smoking satisfaction beginning on Day 3, and this effect appeared stronger towards the end of the medication manipulation period (Supplementary Figure S2, upper left panel). The relationship between smoking satisfaction and next-day smoking was mostly non-significant (Supplementary Figure S2, upper right panel). Among males, the indirect effect of smoking satisfaction was not significant (Figure 2, lower panel). The a-path was significant early in the medication period (Day 3–7 and 11–12), whereas the b-path was non-significant. In addition, the magnitude remained relatively consistent (Supplementary Figure S2, lower panel).

Reward

Among females, the time-varying indirect effect through reward was not significant (Figure 3, upper panel). The effect of varenicline on decreases in psychological reward was significant after Day 2 and grew in magnitude over the medication period (Supplementary Figure S3, left upper panel); however, psychological reward was not related to next-day smoking (Supplementary Figure S3, right upper panel). Among males, the indirect effect was significant, but opposite to the direction predicted, on Day 7 and from Day 13–19 (Figure 3, lower panel). The effect of varenicline on psychological reward was significant after Day 2 but the magnitude remained relatively stable (Supplementary Figure S3, left lower panel). However, with respect to path B, towards the end of the medication period lower psychological reward from smoking predicted greater CPD (Supplementary Figure S3, right lower panel).

Negative Affect and Aversion

Additional results from negative affect and aversion time-varying models can be found in Supplementary Material Section 6; the indirect effect of negative affect was not significant in females at any timepoint, and in males the indirect effect of negative affect was significant but in the opposite than predicted direction from Day 6–18. With respect to aversion, the indirect effect was significant from Day 4–6 in females and Day 2, 4–5 in males.

Supplemental Reversed Mediation Models

We reversed the mediator and outcome variables in the time-varying mediation models. None of the indirect effects of smoking on next-day craving, negative affect, smoking satisfaction, psychological reward, or aversion were significant in males or females (data not shown).

Does Reduced Smoking Reward/Satisfaction Mediate Effects of Varenicline on Craving?

Time-varying models suggested that varenicline significantly reduced reward and smoking satisfaction earlier than significant effects on craving but among females only craving mediated varenicline’s effect on next-day smoking. Findings may implicate craving as an intervening variable between reward/smoking satisfaction and reductions in CPD. Due to package limitations, we cannot assess serial mediation with time-varying models; therefore, we opted for a stepped approach.

Among females, but not in males, we found varenicline significantly reduced smoking reward and smoking satisfaction, and this was associated with lower next-day craving (see Figure 4; Supplementary Figures S4 and S5). These effects grew in magnitude over the pre-quit period, as did the time-varying effects of craving on CPD.

A series of MLMs were conducted to evaluate serial mediation component processes in females, given that the time-varying mediation models could not include covariates. Full model results can be found in Supplemental Material Section 5, results supported the proposed serial mediation pathway.

Discussion

This study is the first to investigate sex-specific mechanisms of pre-quit smoking reduction using time-varying models for varenicline treatment. The MLM of the 3-week medication period did not suggest moderation of mediation by sex and did not show significant indirect effects through candidate mediators. However, the planned time-varying mediation model suggested that, among females, once varenicline reached steady state it decreased craving, which reduced next-day smoking. Additionally, exploratory mediation models suggested that the effects of varenicline on craving were mediated by decreases in smoking satisfaction and psychological reward. Comparatively, this was not the case in males. With respect to craving, in females, our findings fit our hypothesis of the time-course of varenicline’s actions, replicating previous findings that craving mediates decreases in smoking during varenicline treatment post-quit^24,51 and extending the findings to the pre-quit period. Notably, other time-varying models did not support a consistent indirect effect across the treatment period for reductions in smoking satisfaction, psychological reward, negative affect, or aversion in mediating reductions in smoking for females nor males.

Post-hoc exploratory analyses suggested that the effects of varenicline on craving may occur through reducing reward and smoking satisfaction in females. Notably, varenicline significantly reduced self-reported reward and smoking satisfaction when controlling for smoking; however, this was not true for craving. Further, craving predicted next-day smoking when controlling for reward, smoking satisfaction, and same-day smoking, whereas the effects of reward and smoking satisfaction on smoking were no longer significant. Finally, when estimating the time-varying effects of varenicline on reward and craving separately, there was a significant effect of varenicline on reward earlier (Day 2) than on craving (Day 7), potentially indicating a prospective relationship. Rapid effects of varenicline on reinforcement are supported by acute reductions in subjective effects of nicotine from laboratory studies.³⁰ Thus, smoking reduction during the pre-quit period in females may occur due to antagonist actions of varenicline on nAChRs that reduce smoking reward and satisfaction,^12,13 in turn modulating craving,^22,23 indicating these are putative treatment targets for females attempting to quit smoking.

Despite significant reductions in pre-quit smoking, we were unable to identify a putative mechanism of varenicline for males. Notably, the results in males were largely inconsistent across time and models. Specifically, component processes were non-significant in males, effects did not change in magnitude across time as was seen in females, and the b-path tended not to be significant when controlling for same-day smoking (with the exception of the reward model). In this context, we hesitate to overinterpret null findings in males. Overall, while pre-quit varenicline reduced smoking comparably among males and females,¹¹ there is little understanding of the psychological mechanisms of that reduction in males.

Sex differences in the pharmacokinetics and pharmacodynamics of nicotine resulting from reciprocal influences with estrogen may explain the unique effects of varenicline on craving in females. For example, nicotine reduces circulating estrogen levels and disrupts estrogen regulation,⁵² and females have higher nicotine metabolism rates (NMR, including in this sample¹⁰), particularly in phases when circulating estrogen is higher (ie, pre-menopause, follicular menstrual phases, when taking oral contraception;^53,54). Faster NMR results in rapid declines in blood nicotine concentrations, contributing to craving, withdrawal, smoking, and lower rates of quitting.⁵⁵ Relatedly, greater circulating estrogen levels are associated with greater craving and nicotine reward in females,⁵⁶ implicating sex hormones in smoking in response to craving. Varenicline has greater efficacy in normal (vs. slower) nicotine metabolizers,⁵⁷ as well as women.⁵⁸ Thus, estrogen may modulate the differential effects of varenicline. Importantly, within this sample, 69% of female participants were over 50 and may have been peri-menopause, in menopause, or post-menopause, possibly affecting results. Additional research should test the effects of sex hormones in cessation outcomes and treatment effects in young adults, as greater sex differences may emerge.

Varenicline produced significantly greater declines in smoking satisfaction and reward in females, indicating varenicline’s antagonist actions on α4β2* may be modulated by sex hormones. Studies suggest that upregulation of β2-nAChRs following smoking acquisition is related to changes in nicotine reward-related behavior and neuronal firing in the midbrain reward pathway in rodents⁵⁹ as well as greater craving during abstinence in laboratory studies.⁶⁰ This process differs by sex, such that females tend not to show upregulation after smoking acquisition,⁶¹ potentially due to greater progesterone⁶² and sex differences in dopamine activation by nicotine.⁶³ Greater nAChR availability in males who smoke may correspond to lower efficacy of varenicline in attenuating reward/smoking satisfaction relative to females and may explain why change in reward does not predict change in craving or smoking in males at the recommended dosage.⁶⁰ Further investigation into dosage optimization may be warranted. Altogether, whereas varenicline may reduce daily self-reported smoking reward and satisfaction in males, the mechanism of pre-quit reductions in smoking have yet to be identified.

Understanding the treatment mechanisms of varenicline in males is an important future endeavor. Given that males have comparable responses to bupropion, NRTs, and varenicline, these medications may share properties that are particularly efficacious in males. For example, treatment expectancies mediated EOT outcomes of varenicline and NRTs⁵¹ and bupropion.⁶⁴ Previous research has found that females may have more negative expectancies of quitting smoking (eg, fear of weight gain or inability to cope with stress;⁶⁵). Thus, positive expectancies for treatment in males may contribute to reductions in smoking irrespective of specific medication actions. This is further supported by findings of greater decreases in smoking in placebo groups in males relative to females,⁶⁶ such that fewer negative expectancies may also benefit males in the placebo group, although to a lesser degree than the treatment group, where reductions in rewarding effects co-occur.

Clinical Implications

The optimal duration of varenicline preloading remains unclear. If preloading effects are dependent on extinction generalizing across multiple scenarios that elicit craving, as is hypothesized, a longer duration may be needed.¹⁰ Notably, our previous work found that in females only, reductions in smoking prior to the TQD significantly mediated the effects of extended preloading on subsequent abstinence.¹¹ Here, time-varying models in females suggested that during the pre-quit period the effect of reward and smoking satisfaction on craving as well as the effect of craving on next-day smoking continued to increase in magnitude. Thus, these findings may support further extending the pre-quit period in females and the role of proximal indirect processes in reducing pre-quit smoking, thereby increasing abstinence.¹¹

Despite varenicline being the most effective monotherapy for smoking cessation, it still only shows approximately 20%–25% efficacy by 6 months.⁶⁷ Mechanisms of smoking reduction are important targets for intervention refinement. Specifically, interventions that address craving, including Acceptance and Commitment Therapy,⁶⁸ may be useful in augmenting the effects of varenicline. In addition, low-content nicotine cigarettes during the preloading period may further reduce smoking reward,⁶⁹ craving, and smoking behavior.⁷⁰

Limitations and Future Directions

There are important limitations to the present work. While we describe the interaction between nicotine and sex hormones as a potential explanation for observed differences in females and males, we did not assess menstrual nor reproductive status, precluding a formal evaluation. Moderation of the indirect effects was not significant in any of the models (non-significant in MLMs; unable to test in the TVEMs), limiting conclusions about sex-specific mediators. Relatedly, given the complexity of the models, they may have been underpowered to detect the moderation of the indirect effect. Specifically, in longitudinal EMA data, the amount of data at any given time point may vary. Thus, confidence bands may have been wider than with larger sample sizes⁴⁹ and this may have had a greater impact on analyses in the male subsample. The unplanned exploratory analysis of mediation of next-day craving by smoking satisfaction and reward should also be interpreted with caution, as we could not control for same-day craving and the TVEM model does not support estimation of a total serial indirect effect. A serial multiple-mediator model could have been assessed using MLMs; however, given that TVEM models suggested the effect was not time-invariant, we elected to assess paths separately using time-varying mediation. Finally, we chose to assess day-to-day changes in CPD to assess smoking as a continuous outcome. Since our random alarm assessments inquired about smoking only in the past 2 hours, they would have necessitated using a binary smoking outcome. Future research may investigate momentary changes that assess continuous smoking, to ascertain whether certain mediators are associated with more granular changes (eg, negative affect, where effects on smoking may be more immediate;⁷¹).

Conclusion

Time-varying models (but not time-invariant MLMs) supported the hypothesis that varenicline indirectly reduces pre-quit smoking through reducing pre-quit craving in females, but not in males, representing a novel contribution to the literature. Additional research is needed both to replicate our findings for females and to better understand the mechanisms through which varenicline reduces pre-quit smoking in males.

Supplementary Material

Supplementary material is available at Nicotine and Tobacco Research online.

Funding

This work was supported primarily by grant R01CA206193 from the NCI (Drs Hawk, Tiffany, and Mahoney); trial infrastructure was supported by grants UL1TR001412 (University of Buffalo) and UL1TR002001 (University of Rochester) from the NCATS/NIH. Pfizer Inc provided free study medication (active varenicline and placebo) at no cost. The sponsors had no role in the design or conduct of the study, collection, management, analysis, and interpretation of the data, or preparation, review, or approval of the manuscript.

Declaration of Interest

Drs. Hawk, Mahoney, and Tiffany reported nonfinancial support from Pfizer Inc during the conduct of the study. Dr. Mahoney also serving as a former speaker/content expert on smoking cessation for Pfizer Inc outside the submitted work. Dr. Ashare reported receiving grant funding from Novo Nordisk, Inc. outside the submitted work. Dr. Brandon reported serving as a paid consultant for the University at Buffalo, State University of New York, during the conduct of the study, and served as an unpaid member of the advisory board of Hava Health, Inc, which was developing a therapeutic e-cigarette for smoking cessation and in which he owned restricted stock. No other disclosures were reported.

Author Contributions

Samantha Johnstone (Conceptualization [equal], Formal analysis [equal], Visualization [equal], Writing—original draft [equal]), Robert Cooper (Conceptualization [equal], Investigation [equal], Methodology [equal], Writing—review & editing [equal]), Jennifer Wray (Investigation [equal], Methodology [equal], Writing—review & editing [Equal]), Sarah Tonkin (Investigation [equal], Methodology [equal], Writing—review & editing [equal]), Craig Colder (Investigation [equal], Methodology [equal], Software [equal], Supervision [equal], Writing—review & editing [equal]), Eugene Maguin (Investigation [equal], Methodology [equal], Software [equal], Writing—review & editing [equal]), Martin Mahoney (Funding acquisition [equal], Investigation [equal], Methodology [equal], Writing—review & editing [equal]), Stephen Tiffany (Funding acquisition [equal], Investigation [equal], Methodology [equal], Writing—review & editing [equal]), Thomas Brandon (Writing—review & editing [equal]), Rebecca Ashare (Writing—review & editing [equal]), and Larry Hawk, Jr. (Conceptualization [equal], Data curation [equal], Funding acquisition [equal], Investigation [equal], Methodology [equal], Project administration [equal], Supervision [equal], Writing—review & editing [equal])

Data Availability

The de-identified primary study data (including baseline characteristics and abstinence outcomes) are archived in the National Addiction and HIV Data Archive Program (NAHDAP; Hawk, 2024) and the data are available via the standard NAHDAP use agreement. The present EMA data will be archived in openICPSR (https://www.openicpsr.org/openicpsr/; DOI not yet available) and linked to the NAHDAP archive. Analysis code will be uploaded to the Open Science Framework Registry (https://osf.io/dc2yp/). Hawk, Larry W., Jr. (2024). EVarQuit: Extinguishing Cigarette Smoking via Extended Pre-Quit Varenicline, Buffalo, New York, 2017-2020. Inter-university Consortium for Political and Social Research [distributor], 2024-06-18. https://doi.org/10.3886/ICPSR39157.