Learning to quit: can reinforcement theories predict the success of smoking cessation attempts using nicotine replacement therapy patches in a general population sample of smokers at 8-weeks and 6-months follow-up?

ABSTRACT Primary, secondary, and tertiary reinforcement contribute to the maintenance of smoking behaviour and may influence the efficacy of different cessation treatments. This analysis examined these relationships in a large general population sample and investigated how previous experiences of the different reinforcement mechanisms impacted future quit attempts. Random digit telephone dialing was used to recruit a sample of Canadian adults who smoked and were interested in being part of a hypothetical program that would provide nicotine replacement therapy (NRT) patches free by mail and half of the eligible participants were randomized to actually receive a five-week supply of NRT patches. During the interviews, reasons for relapse to smoking during previous quit attempts were collected and coded by two reviewers (disagreements were settled by a third reviewer). Binary logistic regression was used to determine if type of reinforcer moderated the intervention effect of the patches. Participants who made cessation attempts in the past year were more likely to report negative (p = .039), secondary (p = .041), and tertiary (p = .010) reinforcers and less likely to report positive reinforcers (p = .016) compared to those who did not attempt to quit. Logistic regressions revealed no significant conditional effects of the intervention on the relationship between reinforcer type and quit attempts or 30-day smoking abstinence. Analysis including all three reinforcers showed negative reinforcers decreased but tertiary reinforcers increased the odds participants reported a cessation attempt before the baseline interview and between baseline and 8-weeks. Understanding the different ways nicotine reinforces smoking behaviour could help guide individuals to more effective treatment options.


Introduction
About 70% of individuals who smoke tobacco report wanting to stop and about 40% make serious attempts, yet only about 3% will successfully quit in a given year (Benowitz, 2010;Burrows et al., 2020).Furthermore, it has been estimated that the average smoker will make 8 to 30 attempts before successfully quitting (Chaiton et al., 2016).This occurs despite the availability of efficacious treatment options that can improve the chances of successful cessation (e.g., psychotherapy, behavioural therapy, pharmacotherapy).While beneficial, the effects of these treatments can be modest and variable, which contributes to relapse remaining common (Burrows et al., 2020).
A number of factors contribute to maintaining smoking behaviour (Jimenez-Ruiz et al., 2014) and nicotine plays a significant role in understanding maintenance, relapse, and treatment success (Brewer et al., 2013).Nicotine is the primary active component of tobacco products and acts on brain regions that regulate cognition, attention, motivation, sleep, memory, developmental processes, rewards, and learning (Sherafat et al., 2021).Activation of these different brain areas is thought to contribute to dependence and the drive to consume more tobacco (Sherafat et al., 2021).In particular, activation of the ventral tegmental area increases dopaminergic effects in the nucleus accumbens, which contributes to the reinforcing effects of nicotine (Sherafat et al., 2021).
Primary reinforcement occurs when, through learning, we come to associate a certain stimulus with a certain response (Coon, 2004).Over time and through repeated pairings, an association forms (Coon, 2004) that increases the likelihood that the individual will engage in the response which in turn further reinforces the behaviour (Burrows et al., 2020).Reinforcement can be positive or negative.Negative reinforcement increases smoking behaviour through the removal of something aversive (Coon, 2004).For example, an individual may feel less stress after smoking and over time when they feel stress they will be more likely to smoke to reduce this aversive state.For positive reinforcement, an individual increases the behaviour in order to feel beneficial effects or pleasure.
A secondary pathway of the reinforcing effects of nicotine is through the association of smoking behaviours with other aspects of the environment.Through repeated pairings, these neutral environmental stimuli become conditioned to cue or trigger smoking behaviours.For example, before conditioning, an ashtray (neutral stimulus) would be unlikely to trigger an individual to want to smoke (unconditioned response); however, through repeated pairings, seeing an ashtray in the future (conditioned stimulus) can trigger the individual to want to smoke (conditioned response).This secondary reinforcing effect can come to be associated with a wide variety of environmental stimuli and create cues that maintain and increase smoking behaviour (Perkins et al., 2017).
A third reinforcing effect of nicotine has been posited in which nicotine acts as an unconditioned stimulus and comes to reinforce another reinforcer (Perkins et al., 2017).This reinforcement-enhancing effect (REE) also works to increase the individual's probability of smoking.For example, socializing with people (unconditioned stimulus) may produce pleasurable feelings (unconditioned response) that through repeated pairings can become a conditioned response, where we seek out our friends (conditioned stimulus) in order to feel pleasure (conditioned response), which increases our likelihood of socializing in the future.Nicotine can enhance this association so that the pleasure experienced during socializing increases (Perkins et al., 2017).Later, when an individual quits smoking, the enhanced effect is gone and the individual feels less pleasure socializing (Perkins et al., 2017).
These three ways nicotine reinforces smoking could have implications for optimizing cessation treatment.With respect to nicotine replacement therapy (NRT) patches, it has been theorized that the slow release of nicotine may be beneficial to individuals experiencing effects from primary negative (US Department of Health and Human Services, 2010) and tertiary reinforcement (Perkins et al., 2017).Conversely, current research does not appear to find a beneficial effect of patches on cravings triggered by secondary reinforcers (US Department of Health and Human Services, 2010) or when smoking behaviour is associated with primary positive reinforcement (Jimenez-Ruiz et al., 2014).
Many of these associations have been tested in clinical settings and we were interested in exploring these relationships in a general population sample.Data collected during a randomized controlled trial (RCT) was used to examine the effect of different reinforcer types on quit attempts and smoking cessation success with NRT patches.Based on the available literature, we hypothesized that more participants who tried to quit smoking in the previous year would report experiencing reinforcements that contributed to relapse, compared to those who had not attempted to quit.We also hypothesized that participants who reported that negative and tertiary reinforcers contributed to a previous relapse would be more likely to attempt to quit if they received NRT patches.

Methods
Details of the original RCT have been reported elsewhere (Cunningham et al., 2016(Cunningham et al., , 2011)).Briefly, between June 2012 and 2014, a sample of Canadian adults who smoked were contacted through random digit dialing.Respondents were asked: if there was a hypothetical program that mailed free NRT patches, would they be interested in quitting smoking?A total of 999 participants were interested and met other eligibility criteria.Half of these were randomly assigned to receive a free, 5-week supply of NRT patches.Verbal consent to participate was collected by telephone and follow-up interviews were conducted 8-weeks and 6-months after the baseline interview.
As part of the baseline, respondents were asked about previous quit attempts.Each participant was asked a list of eight pre-specified reasons, which are commonly cited as contributing to relapses (i.e.cravings, stress, someone in the household smoking, seeing someone smoking, being discharged from hospital, alcohol or drug use).In addition to these reasons, an open-ended question was asked where they could list any reasons not already identified.Reasons were coded by two reviewers independently.Disagreements were settled by a third reviewer.
The research project was approved by the standing research ethics board of the Centre for Addiction and Mental Health.

Analysis plan
Inter-rater reliability was assessed using Cohen's Kappa.Bivariate analyses were used to investigate possible differences in demographics and clinical characteristics between participants who attributed previous relapses to the different types of reinforcers and those who did not.Bivariate analyses were also used to investigate the difference in quit attempts among participants who reported the presence of a reinforcer during a previous quit attempt and was repeated using the 30-day smoking abstinence rate at 6-months, as this was the primary outcome variable in the RCT.Additive moderation analysis was used to examine how the different reinforcers contributed to quit attempts in the previous year.Finally, a series of simple moderation analyses were conducted to examine if the NRT intervention had a conditional effect on the relationship between any of the reinforcer types reported during the baseline interview and the three smoking outcomes (cessation attempts during each follow-up interval -baseline to 8-weeks; 8-weeks to 6-months -and 30-day abstinence at 6-months).The additive moderating effects of all the reinforcer types were also examined.All analyses were conducted using IBM SPSS version 25.

Results
During the baseline survey, 71 participants (7.1%) reported never previously attempting to quit smoking.Those who reported making a cessation attempt in the past were asked about eight specific reasons related to relapse.Participants were able to endorse multiple reasons and the most commonly cited were stress (73.1%) and cravings (64.2% ).The open-ended question resulted in an additional 136 responses at baseline with some of the most commonly reported including habit, boredom, and coffee.These other reasons, as well as the pre-specified list, were reviewed and coded independently by two reviewers as either primary (subdivided into negative and positive), secondary, or tertiary reinforcement.Six reasons from the open-ended responses were not coded because the reviewers felt the responses did not provide enough context to make an evaluation to code the type of reinforcer involved (e.g., 'stupidity').A test of interrater reliability indicated moderate agreement (Kappa = 0.483, n = 145, p < .001;Cohen, 1960).At baseline, over 92% of participants reported that at least one reinforcement-related reason contributed to a previous smoking relapse.Of the 928 participants who reported a quit attempt in the previous year, 93.4% (n = 867) endorsed at least one negative reinforcer, 1.6% (n = 15) endorsed a positive reinforcer, 61.7% (n = 573) endorsed a secondary type, and 48.4% (n = 449) endorsed a tertiary reinforcer.

Demographics and clinical measures
In order to describe the samples reporting each type of reinforcer, demographic and clinical variables were compared using a series of chi-square and t-tests.Participants who were marginally younger were, nonetheless, significantly more likely to report any type of reinforcer (see Table 1).Participants with high school education or less were more likely to report experiencing a negative reinforcer (p = .011)during a quit attempt in the previous year, while those with higher nicotine dependence scores as measured by the Fagerström Test for Nicotine Dependence (FTND; Heatherton et al., 1991) were more likely to report secondary reinforcers (p < .001).Participants who were employed full or part time (p = .011)were more likely to report tertiary reinforcers while females (p < .001),those who were married/common law (p = .041)and smoked fewer cigarettes per day (p = .022)were less likely to report tertiary reinforcement.Results are summarized in Table 1.
Among participants who endorsed each type of reinforcer, bivariate analyses were used to examine differences among those who did and did not make quit attempts or were abstinent at 6-months.At baseline, significantly more participants attempting to quit during the previous year endorsed negative (p = .039),secondary (p = .041),and tertiary reinforcers (p = .010)as contributing to previous relapses in comparison to those who did not make a quit attempt.In contrast, significantly fewer endorsed positive reinforcers (p = .016).There were no other significant differences, except at 8-week follow-up, where more participants who made a quit attempt identified tertiary reinforcers as contributing to previous relapse (p = .031)as compared to those who did not make a quit attempt.Due to the small sample size, responses coded as primary positive reinforcers were not included in subsequent analyses.Results are presented in Table 2.

Moderation analyses
An additive moderation analysis examined the relationship between the three reinforcers and reported attempts to quit smoking and the presence of negative (p = .048)and tertiary (p = .036)reinforcers increased the odds participants reported a quit attempt.Results are presented in Table 3.
A series of simple moderation analyses examined the conditional effect of the intervention on the relationship between each type of reinforcer that contributed to a previous relapse and attempts to quit smoking during the follow-up periods (baseline to 8-weeks; 8-weeks to 6-months) and 30-day smoking abstinence at 6-months follow up.Although all of the models were statistically significant (p < .05),none of the conditional effects indicated that the intervention moderated the relationship between the reinforcer and cessation attempts or smoking abstinence.Of some note, the presence of a negative reinforcer was found to decrease the odds of a quit attempt during the baseline to 8-week follow-up period (p = .012)and approached significance in the model of 30-day smoking abstinence at 6-months (p = .091),such that the presence of a negative reinforcer decreased the odds that a participant would be abstinent at 6-months.Finally, additive moderation analyses examined the effect of all types of reinforcers on attempts to quit smoking during each time interval.Between baseline and 8-weeks, the odds of a quit attempt decreased with the presence of a negative reinforcer (p = .014)and increased with the presence of a tertiary reinforcer (p = .003).Results of all the models are presented in Table 4.

Discussion
These results identified several differences between groups related to those who did and did not report experiencing different types of reinforcers during previous attempts to quit smoking.In this analysis, participants that were somewhat younger (range = 46.6 to 48.9 years versus 50.4 to 52.5 years) were significantly more likely to report a reinforcer of any type as having a role in smoking relapse.Several other demographics also varied but unlike age, these differences were not the same for each type of reinforcer.Understanding how individual demographic differences, such as these, relate to learning and conditioning may be helpful in predicting treatment efficacy and other research has identified some possible relationships.For example, among participants with lower nicotine dependence, .866 varenicline was more effective among those who identified smoking motives related to positive reinforcement while the efficacy of NRT was greater among participants who reported negative reinforcement.The authors further report that more women experienced positive reinforcement from smoking, which may contribute to the challenges women, may experience when quitting.As the authors note, most cessation treatments are more effective in controlling aversive states (negative reinforcers), while few provide positive reinforcement (Jimenez-Ruiz et al., 2014).Indeed, few of the reasons for smoking relapse identified in this sample were coded as positive reinforcers and only one participant who had quit for 30-days at 6-months identified a positive reinforcer as influencing a previous relapse.Unfortunately, such low frequencies resulted in quasi-complete separation of the data and limited our analysis.Nonetheless, positive reinforcement has been theorized to have more influence on smoking initiation and has not previously been identified as a predictor of relapse (US Department of Health and Human Services, 2010).This may partly explain why relatively few of these reasons were coded in the dataset, since we surveyed people who expressed interest in quitting (Cunningham et al., 2011).Research is needed to investigate systematically the effect of positive reinforcement on individual motivation to quit smoking.If positive reinforcement interferes, other interventions may be beneficial before providing treatment.
In the simple moderation analyses, negative reinforcement was the only reinforcer that had a significant impact on any of the outcomes tested.During the baseline to 8-week follow-up interval, experiencing a reinforcer that contributed to a previous relapse decreased the odds of a quit attempt.Similarly, while only tending towards significance (p = .091)the odds of 30-day abstinence at 6-months decreased with the presence of the negative reinforcer.Furthermore, this was one of the only models where the intervention did not significantly increase the odds of a quit attempt or abstinence.This analysis did not find a conditional effect of the intervention; however, this RCT was not designed to investigate the moderating effect of reinforcement on NRT patch effectiveness.Understanding if negative reinforcement impacts NRT effectiveness could be important in providing more personalized treatment options.
The findings in the additive moderation models of quit attempts in the year before the baseline interview and between the baseline and 8-weeks follow-up also merits further investigation.Negative reinforcers were found to decrease the odds of cessation attempts at both time points.Negative reinforcement involves the removal of an aversive state, thus if an individual experienced more stress or cravings during a previous cessation attempt, that memory may play a role in their decision to make future attempts.Despite being in the opposite direction, the finding that tertiary reinforcers increased the odds of a quit attempt appears consistent with the mechanism of REE.In the case of tertiary reinforcement, nicotine acts as a conditioned stimulus that enhances the salience and reward derived from other non-nicotine stimuli (Addicott et al., 2019;Burrows et al., 2020;US Department of Health and Human Services, 2010;Perkins et al., 2017).Using the example of socializing with people, reinforcement can lead to increased socializing behaviour in order to feel the associated pleasure.When nicotine is added to this behaviour pattern, it enhances the pleasurable feelings associated with spending time with friends (Perkins et al., 2017).Later, when an individual quits smoking, the enhanced effect is removed and the individual experiences less pleasure (Perkins et al., 2017).Other research suggests this mechanism partly explains why more lapses to smoking behaviour occur during leisure activities (Addicott et al., 2019;Perkins et al., 2017); however, it may also be relevant to the decision to make a quit attempt.If an individual is not aware that nicotine is enhancing the pleasure they experience, they may not consider this factor in their decision, which appears consistent with our findings that having previously experienced a tertiary reinforcer did not decrease the odds of a future attempt to quit.Furthermore, other research has suggested that REE may occur to some degree during NRT use (Perkins et al., 2019) and is suspected to lessen the loss of pleasure and reward (Coon, 2004).In the case of NRT, the gradual decrease in the provision of nicotine may offset the loss of the REE, reducing some of the challenges connected with a quit attempt, which may contribute favourably to future decisions to try quitting.One limitation of our analysis is that it did not consider change over time.Reinforcers can change and individuals can become less responsive to reinforcers through habituation, or new associations may form (Perkins et al., 2017), which could have an effect on the conditioned behaviour.Another limitation relates to the coding of previous reasons for relapse into different types of reinforcement.As the RCT was not designed for this research question, more detail in participant comments would have aided coding in some instances.Furthermore, the selfreported responses provided by participants may be subject to a number of different biases (e.g., recall, social desirability).Future research also needs to consider potential overlap between the reinforcer types.For instance, as described above, the REE may cause tertiary reinforcement of smoking while socializing with friends; however, seeing friends could also serve as a cue or trigger and therefore be classified as a secondary reinforcer.Furthermore, individual expectancy could also play a role in determining types of reinforcers.For example, on a long drive a person may report smoking to stay awake.If the individual expects the smoking to reduce their boredom, then it should be coded as negative reinforcement.Whereas, if the individual smokes to feel entertained, then it should be coded as positive reinforcement.Future studies investigating reinforcement will need to address systematically such challenges in defining types of reinforcement among reasons for smoking relapses.
Nicotine's role in the maintenance of smoking behaviour is complex and through different mechanisms contributes to reinforcement and learning which increases the likelihood of future smoking.Disentangling these mechanisms and identifying group propensities to different types of reinforcers may help optimize treatment efficacy.This work used a large, general population sample and lends some support to these theories and additional research seems merited.However, at this early stage of research, results need to be interpreted with caution but, it is

Table 1 .
Bivariate comparisons of baseline demographics and clinical characteristics between those who reported a type of reinforcement related to smoking relapse and those who did not (n=928).

Table 2 .
Bivariate comparisons of reinforcement types by attempts to quit smoking at each follow-up.

Table 3 .
Additive moderation analysis of reinforcer types on attempts to quit smoking in the previous year.

Table 4 .
Binomial logistic regression of the presence of each reinforcer type on attempts to quit smoking moderated by the intervention condition.