Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2017 Mar 1.
Published in final edited form as: Psychol Addict Behav. 2016 Jan 14;30(2):229–237. doi: 10.1037/adb0000149

Higher Stimulus Control is Associated with Less Cigarette Intake in Daily Smokers

Stuart G Ferguson 1, Saul Shiffman 2, Michael Dunbar 2, Natalie Schüz 1
PMCID: PMC4801779  NIHMSID: NIHMS743772  PMID: 26766542

Abstract

It is well established that environmental stimuli influence smoking in light, and to a lesser degree, heavy smokers. A two-factor model of dependence suggests that the influence of stimulus control is masked among heavier smokers who primarily smoke for nicotine maintenance. The current study aimed to assess the influence of stimulus control across a range of moderate to heavy daily smokers. Furthermore, as local tobacco control policies may change the role of stimulus control, the study aimed to replicate previous US findings on stimulus control in an Australian setting marked by strong tobacco control policies. In two Ecological Momentary Assessment studies, 420 participants monitored antecedents of smoking and non-smoking situations. In a set of idiographic logistic regression analyses, situational antecedents were used to predict smoking occasions within each individual’s data. Linear regression analysis was used to test for the association between stimulus control and smoking rate, and to test for differences between the two samples. Daily smokers’ smoking was under considerable stimulus control, which was weaker at higher smoking rates. Overall, there was greater stimulus control in the Australian sample. Daily smokers also experience a degree of stimulus control, which is less influential in heavier smokers.

Keywords: smoking, Dependence, addiction, stimulus control

Introduction

Smoking rates remain at unacceptably high levels worldwide (WHO, 2012). New strategies are needed to prevent the uptake of smoking and—importantly for public health in the coming decades—to aid current smokers’ ability to quit (Jha & Chaloupka, 1999). In order to systematically develop and refine such tools the fundamental step is to determine the drivers of smoking behaviour.

In the classical model of addiction (Benowitz, 2008), smokers are thought to smoke primarily due to nicotine dependence with cigarettes spaced over the day in order to maintain nicotine blood values at levels sufficient to ward off withdrawal (Benowitz, 2010). While this model allows for social and/or situational factors to influence smoking during the early stages of dependence, smoking patterns in established smokers are thought to be driven primarily by the desire to avoid withdrawal (so-called through avoidance; Russell, 1971). Notably, the development of nicotine replacement therapy was underpinned by this assumption (Henningfield, Shiffman, Ferguson, & Gritz, 2009).

Researchers have been increasingly interested in the subgroup of non-daily, or intermittent smokers (ITS), in large part because their smoking is inconsistent with this model of nicotine maintenance. ITS smokers make up over a third of all US smokers (U.S. Department of Health and Human Services, 2014) and there is evidence that they are becoming more prevalent (Schane, Glantz, & Ling, 2009). While ITS have smoked thousands of cigarettes (Tindle & Shiffman, 2011) and successfully extracted nicotine from these cigarettes (Shiffman, Fischer, Zettlersegal, & Benowitz, 1990; Shiffman et al., 1992), they do not register as being dependent (Shiffman, Ferguson, Dunbar, & Scholl, 2012), and do not appear to experience withdrawal when they abstain (Shiffman, Dunbar, Tindle, & Ferguson, in press). Rather than smoking because of a desire to avoid withdrawal, ITS smoking appears to be primarily influenced by environmental and social factors; that is, it is under stimulus control. Compared to daily smokers, ITS smoking is more associated with “indulgent activities” (Shiffman & Paty, 2006) such as socializing and alcohol consumption (Krukowski, Solomon, & Naud, 2005; Otsuki, Tinsley, Chao, & Unger, 2008; Shiffman et al., 2014b). Indeed, ITS smoking is under such tight stimulus control, that Shiffman and colleagues were able to use contextual information to differentiate smoking from non-smoking occasions with >90% accuracy (Shiffman, Dunbar, & Ferguson, 2015). Importantly, Shiffman et al. reported that even daily smokers’ smoking was under considerable stimulus control, suggesting that stimulus control plays a role in maintaining smoking even among more dependent smokers (Shiffman, Dunbar, et al., 2015).

The finding that established smoking patterns—both daily and intermittent—are strongly influenced by stimulus control suggests that the classical model of dependence may not be adequate to explain the entire range of observed smoking behaviour. Shiffman and colleagues (Shiffman, Dunbar, et al., 2015) argue that a two-factor model of smoking—one that incorporates both withdrawal avoidance and stimulus control—can more convincingly explain the range of smoking patterns observed among smokers. They argue that all smoking is under stimulus control, but that, as nicotine dependence increases, smoking is increasingly driven by withdrawal avoidance mechanisms. This has the effect of diluting, or muting, the role of stimulus control on smoking (also see: Leventhal & Cleary, 1980). Such a model helps to explain why smokers lapse under particular stimulus exposures (Shiffman, 1989) even when withdrawal symptoms are effectively eliminated (Shiffman, Ferguson, Gwaltney, Balabanis, & Shadel, 2006): once withdrawal is reduced, the underlying stimulus control processes come back to the fore.

So far, support for this model has come from contrasting extreme groups of smokers (specifically, light or intermittent smokers versus daily smokers; e.g., Shiffman, Dunbar, & Ferguson, 2015). What remains to be seen, however, is whether ITS are categorically different from daily smokers—perhaps due to genetic differences (Lerman & Berrettini, 2003)—or whether they are simply at different ends of a continuum. If, as Shiffman et al. proposed, stimulus control does vary along a continuum, we would expect to see predictable variation in the magnitude of stimulus control even within daily smokers. Specifically, we would expect that as smoking rates increase, the degree to which smoking is under stimulus control should decrease. It is this relationship that we test here.

Stimulus control is traditionally assessed in tightly-controlled laboratory settings using the cue reactivity paradigm (Carter & Tiffany, 1999). However, such studies lack ecological validity and as such it is difficult to know how well findings from such studies map to actual smoking patterns (Perkins, 2009; Shiffman, Li, et al., 2015). An alternative strategy is to monitor smoking behaviour in real-world conditions and examine the degree to which the presence or absence of environmental stimuli predict smoking. For example, Shiffman and colleagues (Shiffman, Dunbar, et al., 2015) used Ecological Momentary Assessment (EMA) (Shiffman, Stone, & Hufford, 2008) to monitor smoking in a sample of ITS and daily smokers. Participants were asked to report each cigarette they smoked, and, during a random subset of these events, were asked to report on the mood, location, social setting and activities at the time they decided to smoke. A parallel series of questions was also administered at random times throughout the day to gather information on non-smoking occasions. As learning processes are likely to vary extensively between smokers, analyses that average the influence of situational cues across smokers (e.g., Shiffman et al., 2014b) fail to model the potential variation, and associations that exist within smokers (i.e., idiographic associations) may be missed on the between-person level. For example, one smoker may always smoke when she is with her friends, while another smoker might never smoke with their friends; if we were to average across these participants we would find no association between social setting and smoke despite both smoking patterns being consistent with smoking being under stimulus control. Thus, the researchers used a set of within-subject logistic regression analyses to determine the degree to which different stimulus domains (e.g., current location, mood, social setting, etc.) could successfully differentiate between smoking instances and randomly-timed non-smoking events. If someone’s smoking is primarily driven by withdrawal processes (or other non-stimulus control driven processes for that matter), one would expect that smoking would occur throughout the day regardless of the presence or absence of environmental stimuli; such a relationship would generate an area under the curve for the receiver operating characteristic curve (AUC-ROC) score of close to 0.5 (indicative of a lack of discrimination between smoking and non-smoking events). Conversely, if smoking—or indeed choosing not to smoke—is closely tied to certain stimuli, one would expect that simply knowing whether the smoker is in the presence of such stimuli would be sufficient to reliably predict whether the individual was smoking or not; this would result in AUC-ROC values closer to 1.0 (or perfect prediction) (Hanley & McNeil, 1982).

Here we use a similar analytic procedure to examine the relationship between smoking rate and stimulus control in daily smokers. To achieve a sample with a wide range of smoking intensities, we combine data from two EMA studies of daily smoking, one from the US and one from Australia. This also allows us to examine the consistency of stimulus control across countries differing in tobacco control policies, including smoking restrictions. Smoking restrictions may increase stimulus control over smoking, by forcing smoking into a reduced range of settings.

Methods

Overview

The data for the current analyses are drawn from two EMA studies involving ad lib smoking among established daily cigarette smokers. The studies utilised comparable protocols: participants used handheld diaries to monitor their smoking, activities and social settings in real time and to respond to randomly-timed non-smoking assessments. In Study 1—a randomised controlled trial (RCT) of a smoking cessation behavioural support program—participants were asked to monitor their ad lib smoking for approximately one and a half weeks prior to a study-mandated target quit date; here we use this baseline data to assess the degree to which their smoking was under stimulus control. These data were combined with data from a second EMA study examining smoking patterns among daily smokers not currently trying to quit (Study 2; Shiffman et al., 2014b).

Participants

In Study 1, smokers were recruited via social media websites (Frandsen, Walters, & Ferguson, 2014), flyers and newspaper advertisements to participate in a smoking cessation study. In order to be eligible, smokers had to be ≥18 years old, smoke ≥10 cigarettes per day (CPD) for the last three years, and indicate a high motivation to quit smoking within the next month (scoring above 75 on a 100-point scale) (Shiffman, Patten, et al., 2006). Participants were excluded if they were currently enrolled in another smoking cessation trial (or had been within the past three months) or were planning on using smoking cessation pharmacotherapy as part of their quit attempt. The recruitment strategies and inclusion criteria for Study 2 have been reported in detail elsewhere (Shiffman, Dunbar, et al., 2015). Both studies were approved by relevant institutional ethics boards (Study 1: XXXXXX; Study 2: XXXXXX) and written informed consent was obtained from all participants.

Procedure

The relevant data collection procedures for the two studies were similar, and mirrored those used in other EMA studies of smoking (Begh et al., 2013; Schüz & Ferguson, 2014; Shiffman & Ferguson, 2008). Data were collected with a handheld device running a customised EMA program specifically designed for the study. At the first study visit, participants provided informed consent, completed a baseline questionnaire assessing sociodemographic and smoking characteristics, and received individual training and instructions on how to use the EMA device, the different types of assessments and the study protocol. During the follow-up study visits, EMA data were downloaded, compliance with the study protocol was checked.

Participants were instructed to log every cigarette that they smoked immediately before they smoked it. Every day, about 4–5 of these cigarette logs were randomly-sampled (using the participants baseline smoking rate to determine the sampling probability; see Ferguson & Shiffman, 2011) and followed-up with a series of questions (see below). In addition to event-based cigarette assessments, participants completed random prompt assessments on non-smoking occasions featuring parallel questions to the cigarette assessments. These were announced via an audible signal. Participants were to carry the EMA device with them during their waking hours.

Measures

In both studies, context was assessed during smoking and random occasions and was measured across seven different domains: 1) Food and alcohol consumption (four yes/no items assessing consumption within the 15 minutes preceding an assessment; e.g., food, caffeinated drinks, and/or alcohol); 2) Time of day (logged automatically as a continuous variable and then converted to six exhaustive and mutually exclusive time blocks); 3) Social setting (being alone, or with others [e.g., friends, co-workers and/or partner etc.]); 4) Location (single item, mutually exclusive answers [e.g., home, work, bar etc.]); 5) Activity (single item assessing current activity [e.g., working, leisure, interaction with others etc.]; multiple responses allowed); 6) Smoking context (whether or not smoking was restricted, either by own rule or by law; and 7) Cigarette craving (single item). [Previous studies (e.g., Shiffman, Dunbar, et al., 2015; Shiffman et al., 2014b) also included mood as a domain, however as mood is influenced by withdrawal, and heavier smokers tend to experience stronger withdrawal symptoms, any relationship between mood and smoking rate would be difficult to interpret.] All items have been described in full detail elsewhere (Shiffman, Dunbar, et al., 2015; Shiffman et al., 2014a). The same items were used in both studies, the only differences being that: 1) Study 2 included “casino” as an additional category for location; 2) Study 2 included whether others were smoking as an additional item for smoking context; and 3) Craving was assessed on a scale from 1–7 in Study 1, and 1–100 in Study 2 (see Shiffman, Dunbar, et al., 2015; Shiffman et al., 2014b).

Dataset Construction and Analyses

For Study 1, only EMA monitoring days before the target quit day were analysed. Of the 278 participants recruited and enrolled, 40 individuals (14.39%) were removed from the analyses due to an overall random prompt compliance of <50% (generally due to attrition). Likewise, individual participant days were removed for which random prompt compliance was <50%, resulting in the deletion of 463 days of data (17.47%). Twelve participants did not provide enough data to be included in the analyses (seven participants did not complete any cigarette assessments, five participants did not complete any random prompt assessments). The final dataset consisted of 226 participants who completed 5,525 smoking assessments and 5,321 random prompt assessments across a period of 9.73±4.25 days. A total of 2,198 participant days of observation were available. Random prompt compliance in this sample was 81.53% (SD = 13.19). These data were pooled with 194 daily smokers from Study 2; the underlying dataset and data reduction steps for this study have been reported elsewhere (S. G. Ferguson, Frandsen, Dunbar, & Shiffman, 2015; Shiffman, Dunbar, et al., 2015). Briefly, daily smokers in Study 2 completed 13,761 smoking assessments and 11,640 random prompt assessments across a period of up to three weeks.

The data analyses proceeded in two steps. First, for the idiographic analyses, we performed separate logistic regression analyses for each participant predicting smoking occasion (yes/no) for each domain with the according domain-specific items as predictors, and in an omnibus model including all items. The AUC-ROC values were calculated to determine how accurately the predictors differentiate between smoking and non-smoking occasions for each participant. AUC-ROC values are independent of the direction of the association (unlike odds ratios); this is advantageous in this current paper as we are interested in testing the strength—as opposed to the direction—of the relationship between stimulus control and smoking rate. AUC-ROC scores can range from 0.5 (indicative of a lack of discrimination between smoking and non-smoking events) and 1.0 (perfect prediction). In a second step, in order to test for the relationship between stimulus control (as indexed by the AUC-ROC values) and heaviness of smoking, we performed simple regression analyses predicting AUC-ROC values for the overall model and for each domain with CPD (with study as a covariate; Table 2). We tested for non-linear relationships between AUC-ROC values and CPD up to a cubic trend; all reported relationships were linear. We weighted the individual AUC-ROC values using the inverse of its standard error (Hanley & McNeil, 1982; Shiffman, Dunbar, et al., 2015); this places greater emphasis on individual models that show greater precision. These analyses allowed us to test whether stimulus control loses influence as a function of smoking rate, as well as to test for potential differences in stimulus control between studies (while controlling for dependence).

Table 2.

Standardized linear regression coefficients and t-tests for the regression of Area Under the Curve of the Receiver Operating Curve values by domain on cigarettes per day (CPD) and time to first cigarette (TTFC).

Domain CPD
TTFC
β t(df) p R2 β t(df) p R2
Consumption −0.20 −1.36 (409) .176 .01 −0.27 −1.84 (408) .066 .02
Time 0.19 1.73 (414) .084 .42 0.02 0.17 (413) .864 .42
Social Setting −0.37 −2.82 (413) .005 .18 −0.38 −2.94 (412) .004 .18
Location −0.37 −2.67 (415) .008 .07 −0.36 −2.63 (414) .009 .06
Activity −0.35 −2.67 (414) .008 .17 −0.37 −2.90 (413) .004 .17
Smoking Context −0.39 −2.73 (415) .007 .05 −0.55 −3.91 (414) <.001 .07
Craving −0.60 −4.30 (414) <.001 .06 −0.52 −3.76 (413) <.001 .05
Omnibus −0.59 −5.52 (413) <.001 .35 −0.52 −4.89 (412) <.001 .34
Open in a new tab

Notes. All models control for study.

While the calculation of AUC-ROC values is not dependent on baseline CPD, it could be argued that as the number of CPD one smokes decreases, the opportunity to distribute them differentially increases, which might increase the likelihood of spurious associations between contextual stimuli and smoking. As such, we also replicate the analyses using dependence as measured via time to first cigarette (TTFC; coded as 1 = high dependence [within 30 minutes]/0 = low dependence [>30 minutes]) (Fagerström & Schneider, 1989). All analyses were conducted in JMP Pro (v11.1)(SAS Institute Inc., 1989–2014).

Results

On average, participants (50.89% female) were 41.78±11.92 years old, indicated smoking 17.37±7.19 CPD at baseline; 84.01% reported smoking their first cigarette of the day within 30 minutes of waking (see Table 1).

Table 1.

Smoking Characteristics and Demographics by Study. Data Presented as M (SD) Unless Stated Otherwise

Variable Study 1 (Australia; n = 226) Study 2 (US; n = 194)
Age in years 42.30 (12.53) 41.18 (11.18)
Female gender, % (n) 50.89 (115) 44.85 (87)
FTND score 4.84 (1.96) 5.14 (1.83)
Time to first cigarette >30min, % (n) 19.03 (43) 12.44 (24)
Baseline CPD* 18.50 (7.56) 16.06 (6.51)
Open in a new tab

Note.

*

p < .001.

CPD = cigarettes per day. FTND = Fagerström Test of Nicotine Dependence (Heatherton, Kozlowski, Frecker, & Fagerström, 1991).

Mean AUC-ROC values—both separate for each study and combined—are presented in Figure 1. One sample t-tests showed that the AUC-ROC values for all separate domains and the omnibus model were significantly higher than 0.50 (the null value; all p-values <.001), indicating that all domain models were significantly better than chance in discriminating between smoking and non-smoking events (Cohen’s d effect sizes: Consumption: 0.33, Time: 0.45, Social Setting: 0.50, Location: 0.53, Act: 0.71, Smoking setting: 0.43, Craving: 0.43). The pooled omnibus AUC-ROC (i.e., the model that included all situational predictors at once) was highly discriminative (effect size: 1.66): when all situational variables were used simultaneously one could differentiate between smoking and non-smoking instances with 88% accuracy. Examining individual domains, the highest AUC-ROC values were obtained for the activity domain, followed by craving; the lowest was obtained for consumption.

Figure 1.

Figure 1

Open in a new tab

Weighted Mean Area Under the Curve of the Receiver Operating Curve (AUC-ROC) values for each situational domain and the omnibus model, both overall and by study. All values were significantly greater than 0.5. Error bars indicate 95% confidence intervals. [Asterisks indicate significant difference between studies: ***p < .001, ** p < .01.]

Omnibus AUC-ROC values were significantly higher in Study 1 (Australian sample) compared to Study 2 (US sample), F(1, 413) = 210.17, p < .0001, Cohen’s d = 1.43. Mean AUC-ROC values were also significantly higher for the Australian sample in five of the seven individual domains (Figure 1).

Figure 2 shows a scatter plot of the significant linear relationship between the omnibus model AUC-ROC values and CPD for the two studies. Pooling the data, AUC-ROC values fell as CPD increased (Table 2). Similar patterns were found for the individual domains (Figure 3 and Table 2): there was a significant linear decrease in AUC-ROC values as the number of CPD increased across all domains except the time domain (where the relationship was not significant). The observed linear patterns were similar in both studies (Figure 3).

Figure 2.

Figure 2

Open in a new tab

Scatterplot and study-specific linear regression lines for the relationship between Area Under the Curve of the Receiver Operating Curve (AUC-ROC) values of the omnibus model and cigarettes per day (CPD). Higher AUC-ROC values indicate higher stimulus control.

Figure 3.

Figure 3

Figure 3

Open in a new tab

Scatterplots and study-specific regression lines for the relationship between Area Under the Curve of the Receiver Operating Curve (AUC-ROC) values and cigarettes per day (CPD), by domain: a) consumption, b) time, c) social setting, d) location, e) activity, f) smoking context, and g) craving. All relationships were linear except for the smoking context domain, where there was a quadratic relationship. Higher AUC-ROC values indicate higher stimulus control.

Finally, we examined the relationship between dependence and AUC-ROC values both in the omnibus model, and across the individual domains (while controlling for study effects). Consistent with the CPD data, the AUC-ROC values from the omnibus model were significantly lower among smokers who reported smoking their first cigarette of the day within 30 minutes of waking compared to less dependent smokers (Table 2). Looking at the seven domains separately, TTFC was significantly related to AUC-ROC values in all domains except consumption and time (Table 2).

Discussion

A number of EMA studies (Cronk & Piasecki, 2010; Piasecki, Trela, Hedeker, & Mermelstein, 2014; Shiffman et al., in press; Shiffman, Dunbar, et al., 2015; Shiffman et al., 2013; Shiffman et al., 2014b; Shiffman & Paty, 2006; Thrul, Bühler, & Ferguson, 2014) have now provided evidence that smoking behavior is influenced by stimulus control. The classical model of addiction holds that smoking is maintained primarily through withdrawal avoidance, with smokers smoking regularly across the day in order to maintain nicotine levels. The fact that daily smoking appears to be influenced also by stimulus control would appear at odds with purely pharmacological models in as much as they demonstrate that such models have trouble explaining the range of smoking behaviors observed. Shiffman and colleagues (Shiffman, Dunbar, et al., 2015) have proposed that observed smoking patterns can be better explained by a model of smoking that allows for stimulus control to influence smoking even in established daily smokers. This two factor model of smoking proposes that smoking is driven not just through withdrawal avoidance, but also through stimulus control; rather than stimulus control being displaced by withdrawal processes over time as smokers progress to heavier, daily smoking patterns, they are still influenced by stimulus control, but the influence of such associations are obscured by the more dominant drive to avoid withdrawal. The present study confirms that daily smokers’ smoking is under considerable stimulus control: all domain models could discriminate between smoking and non-smoking events, with effect sizes ranging from moderate (Consumption: 0.33) to large (Omnibus: 1.66). Importantly, we also demonstrated for the first time that stimulus control grows weaker as smoking rate increases.

Previous studies have performed extreme group analyses to contrast stimulus control in ITS and heavy daily smokers (Shiffman, Dunbar, et al., 2015; Shiffman et al., 2014b; Shiffman & Paty, 2006; Thrul et al., 2014). However it remained unclear whether ITS and daily smokers are simply at opposite ends of a continuum, or whether their smoking is maintained by different processes. In line with our hypothesis, the degree to which smoking was under stimulus control—as assessed across a wide range of contextual domains—consistently decreased as smoking rate (and dependence) increased. This finding is also consistent with the predictions of the two-factor model of dependence. Further work is required to trace how stimulus control loses its influence as a smoker progresses to higher levels of dependence, perhaps using multiple waves of monitoring over time (Piasecki et al., 2014).

A secondary aim of our study was to replicate past findings on the role of stimulus control in daily smoking. We know of only two previous real-world studies of stimulus control in adult daily smokers, both of which were conducted in the US (Shiffman et al., 2014b; Shiffman & Paty, 2006). Here we were able to compare the degree of stimulus control in an Australian sample of smokers to that observed in a sample of US smokers. Australia has some of the strictest and most comprehensive tobacco control regulations in the world (World Health Organization, 2013). The regulations should have the effect of limiting the available situations in which a smoker can light up, potentially strengthening the association between contextual stimuli and smoking. Overall, the current study did support this, as AUC-ROC scores were significantly higher in the omnibus model testing all situational predictors for the Australian sample (Study 1) compared to the US sample (Study 2), and the majority of individual domains. A caveat to this comparison, however, is that the two studies also differed in at least one meaningful way: the smokers in Study 1 were enrolled in a cessation study and hence were about to embark on a quit attempt whereas smokers in Study 2 had no immediate plans to quit smoking. This difference is potentially important as it is likely that the motivation to smoke is different in smokers nearing a quit attempt and indeed it is possible that they might actively change their smoking in preparation for quitting, potentially influencing their observed smoking patterns. Further work is necessary to confirm how smoking restrictions influence individual smoking patterns. Importantly for the purposes of this study, however, the association between smoking rate and stimulus control was consistent across the two studies.

The current findings may have implications for smoking cessation treatment for daily smokers, as they suggest that even daily smokers may benefit from strategies to address cue-induced cravings (S. G. Ferguson & Shiffman, 2009). This is consistent with the finding that the efficacy of nicotine patch, which provides a steady input of nicotine, can be improved by adding an acute form of NRT (Stead et al., 2012), which can be used to respond to acute cue-provoked cravings (Stuart G. Ferguson, Shiffman, & Gitchell, 2011).

Strengths of the study include the large sample size and the real-world assessment of stimulus control. It was not, however, without limitations. As noted above, it could be argued that as the number of CPD one smokes decreases, the opportunity to distribute them differentially increases, potentially increasing the likelihood of spurious associations. While we consider this unlikely, to counter this concern we replicate the analyses using TTFC. However, TTFC is highly positively correlated with CPD and as such may be susceptible to the same problem. Future studies should include additional comprehensive—and multidimensional—measures of dependence to fully explore this question. Moving on, while the data generated from EMA studies is ecologically valid, there is a possibility that protocol noncompliance might bias the data. For example, if people are less likely to respond to prompts when engaging in social situations, the data collected might not accurately reflect their behaviour in these circumstances. However, observed high levels of compliance were obtained in both studies and, in any case, if such bias exists it would be unlikely to vary by smoking rate and hence is unlikely to have influenced our findings.

Moving on, while we consider the range of stimuli included to be a key strength of our study, we obviously could not measure all environmental stimuli that might be associated with the smoking behaviour. This incomplete assessment is a limitation for two reasons. Firstly, our assessments focused almost exclusively on external stimuli. It is possible that rather than stimulus control decreasing as smoking intensity increases (as we have proposed here), that the relative strength of stimulus control may shift from external stimuli to internal stimuli. If this were the case, the pattern of results would look similar to those we observed. Data from a number of sources, however, suggests to us that such a transfer of stimulus control from external to internal stimuli is not occurring. Firstly, if such a shift was occurring, we would expect to see that external stimuli become weaker predictors as smoking rate increases, while internal stimuli become stronger predictors. However, as shown in Figure 3 the relationship between craving (an internal stimulus) and cigarette consumption was similar to that observed with the external stimuli examined. Similarly, previous studies of the differences between light and heavy smokers have found that both internal and external stimuli are weaker predictors of smoking in heavier smokers; again, this is contrary to what we would expect if stimulus control was shifting from external to internal stimuli as smoking rate increases (Shiffman, Dunbar, et al., 2015; Shiffman et al., 2014b). Nevertheless, given the incomplete assessment of environmental stimuli—and that the dataset was not longitudinal—, we cannot completely rule out that such a transfer in stimulus control is occurring. Our incomplete assessment of environmental stimuli is also a limitation as it is possible that heavier smokers might respond to a broader array of stimuli than lighter smokers. If this were the case, however, we would have expected to see a trend toward an increasing predictive ability of the omnibus model as smoking rate increased; a pattern that we did not observe.

The present study confirms that daily smokers’ smoking is under considerable stimulus control. In line with previous research showing that ITS smoking is under greater stimulus control than that of daily smokers (Shiffman, Dunbar, et al., 2015), here we demonstrate for the first time that stimulus control weakens at higher smoking rate. Furthermore, this is the first study to provide evidence on the potential influence of tobacco control policies on smoking.

Acknowledgments

The authors would like to thank Jodie Bower, Isabelle Morris, Julie Pellas, Julia Walters and Mai Frandsen (Study 1) and Sarah Scholl (Study 2) for assistance with data collection.

Funding acknowledgement: Study 1 was funded by a grant from the Australian National Health & Medical Research Council awarded to Dr Ferguson (1002874). Study 2 was funded by a grant from the US National Institutes of Health awarded to Dr Shiffman (R01-DA020742).

Footnotes

Conflict of Interests: Saul Shiffman consults to and has an interest in eRT inc., which, under contract, provided electronic diaries for this research.

References

  1. Begh R, Munafo M, Shiffman S, Ferguson S, Nichols L, Mohammed M, Aveyard P. Attentional bias retraining in cigarette smokers attempting smoking cessation (ARTS): Study protocol for a double blind randomised controlled trial. BMC Public Health. 2013;13(1):1176. doi: 10.1186/1471-2458-13-1176. Retrieved from http://www.biomedcentral.com/1471-2458/13/1176. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Benowitz NL. Clinical pharmacology of nicotine: implications for understanding, preventing, and treating tobacco addiction. Clinical Pharmacology and Therapeutics. 2008;83(4):531–541. doi: 10.1038/clpt.2008.3. [DOI] [PubMed] [Google Scholar]
  3. Benowitz NL. Nicotine addiction. New England Journal of Medicine. 2010;362(24):2295–2303. doi: 10.1056/NEJMra0809890. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Carter BL, Tiffany ST. Meta-analysis of cue-reactivity in addiction research. Addiction. 1999;94(3):327–340. Retrieved from http://search.ebscohost.com/login.aspx?direct=true&db=s3h&AN=1743132&site=ehost-live. [PubMed] [Google Scholar]
  5. Cronk NJ, Piasecki TM. Contextual and subjective antecedents of smoking in a college student sample. Nicotine and Tobacco Research. 2010;12(10):997–1004. doi: 10.1093/ntr/ntq136. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Fagerström KO, Schneider NG. Measuring nicotine dependence - A review of the Fagerström Tolerance Questionnaire. Journal of Behavioral Medicine. 1989;12(2):159–182. doi: 10.1007/bf00846549. [DOI] [PubMed] [Google Scholar]
  7. Ferguson SG, Frandsen M, Dunbar M, Shiffman S. Gender and stimulus control of smoking behavior. Nicotine and Tobacco Research. 2015;17(4):431–437. doi: 10.1093/ntr/ntu195. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Ferguson SG, Shiffman S. The relevance and treatment of cue-induced cravings in tobacco dependence. Journal of Substance Abuse Treatment. 2009;36(3):235–243. doi: 10.1016/j.jsat.2008.06.005. [DOI] [PubMed] [Google Scholar]
  9. Ferguson SG, Shiffman S. Using the methods of ecological momentary assessment in substance dependence research--smoking cessation as a case study. Substance Use and Misuse. 2011;46(1):87–95. doi: 10.3109/10826084.2011.521399. [DOI] [PubMed] [Google Scholar]
  10. Ferguson SG, Shiffman S, Gitchell JG. Nicotine replacement therapies: patient safety and persistence. Patient Related Outcome Measures. 2011;2:111–117. doi: 10.2147/PROM.S11545. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Frandsen M, Walters J, Ferguson SG. Exploring the viability of using online social media advertising as a recruitment method for smoking cessation clinical trials. Nicotine and Tobacco Research. 2014;16(2):247–251. doi: 10.1093/ntr/ntt157. [DOI] [PubMed] [Google Scholar]
  12. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982;143(1):29–36. doi: 10.1148/radiology.143.1.7063747. [DOI] [PubMed] [Google Scholar]
  13. Henningfield JE, Shiffman S, Ferguson SG, Gritz ER. Tobacco dependence and withdrawal: Science base, challenges and opportunities for pharmacotherapy. Pharmacology and Therapeutics. 2009;123(1):1–16. doi: 10.1016/j.pharmthera.2009.03.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Jha P, Chaloupka FJ. Curbing the epidemic: governments and the economics of tobacco control. World Bank Publications; 1999. [Google Scholar]
  15. Krukowski R, Solomon L, Naud S. Triggers of heavier and lighter cigarette smoking in college students. Journal of Behavioral Medicine. 2005;28(4):335–345. doi: 10.1007/s10865-005-9003-x. [DOI] [PubMed] [Google Scholar]
  16. Lerman C, Berrettini W. Elucidating the role of genetic factors in smoking behavior and nicotine dependence. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics. 2003;118B(1):48–54. doi: 10.1002/ajmg.b.10003. [DOI] [PubMed] [Google Scholar]
  17. Leventhal H, Cleary PD. The smoking problem: A review of the research and theory in behavioral risk modification. Psychological Bulletin. 1980;88(2):370–405. doi: 10.1037/0033-2909.88.2.370. [DOI] [PubMed] [Google Scholar]
  18. Otsuki M, Tinsley BJ, Chao RK, Unger JB. An ecological perspective on smoking among Asian American college students: The roles of social smoking and smoking motives. Psychology of Addictive Behaviors. 2008;22(4):514–523. doi: 10.1037/a0012964. http://dx.doi.org/10.1037/a0012964. [DOI] [PubMed] [Google Scholar]
  19. Perkins KA. Does smoking cue-induced craving tell us anything important about nicotine dependence? Addiction. 2009;104(10):1610–1616. doi: 10.1111/j.1360-0443.2009.02550.x. [DOI] [PubMed] [Google Scholar]
  20. Piasecki TM, Trela CJ, Hedeker D, Mermelstein RJ. Smoking antecedents: separating between- and within-person effects of tobacco dependence in a multiwave ecological momentary assessment investigation of adolescent smoking. Nicotine and Tobacco Research. 2014;16(Suppl 2):S119–S126. doi: 10.1093/ntr/ntt132. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Russell MAH. Cigarette smoking - natural history of a dependence disorder. British Journal of Medical Psychology. 1971;44(MAR):1. doi: 10.1111/j.2044-8341.1971.tb02141.x. Retrieved from <Go to ISI>://WOS:A1971I648100001. [DOI] [PubMed] [Google Scholar]
  22. SAS Institute Inc. JMP®, Version 11. Cary, NC: SAS Institute Inc; 1989–2014. [Google Scholar]
  23. Schane RE, Glantz SA, Ling PM. Nondaily and social smoking: An increasingly prevalent pattern. Archives of Internal Medicine. 2009;169(19):1742–1744. doi: 10.1001/archinternmed.2009.315. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Schüz N, Ferguson SG. An exploratory examination of the mechanisms through which pre-quit patch use aids smoking cessation. Psychopharmacology. 2014;231(13):2603–2609. doi: 10.1007/s00213-013-3430-0. [DOI] [PubMed] [Google Scholar]
  25. Shiffman S. Trans-situational consistency in smoking relapse. Health Psychology. 1989;8(4):471–481. doi: 10.1037//0278-6133.8.4.471. [DOI] [PubMed] [Google Scholar]
  26. Shiffman S, Dunbar M, Tindle H, Ferguson SG. Non-daily smokers’ experience of craving on days they do not smoke. Journal of Abnormal Psychology. doi: 10.1037/abn0000063. in press. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Shiffman S, Dunbar MS, Ferguson SG. Stimulus control in intermittent and daily smokers. Psychology of Addictive Behaviors, Advance online publication. 2015 doi: 10.1037/adb0000052. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Shiffman S, Dunbar MS, Kirchner TR, Li XX, Tindle HA, Anderson SJ, Ferguson SG. Cue reactivity in non-daily smokers: Effects on craving and on smoking behavior. Psychopharmacology. 2013;226(2):321–333. doi: 10.1007/s00213-012-2909-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Shiffman S, Dunbar MS, Li XX, Scholl SM, Tindle HA, Anderson SJ, Ferguson SG. Craving in intermittent and daily smokers during ad libitum smoking. Nicotine and Tobacco Research. 2014a;16(8):1063–1069. doi: 10.1093/ntr/ntu023. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Shiffman S, Dunbar MS, Li XX, Scholl SM, Tindle HA, Anderson SJ, Ferguson SG. Smoking patterns and stimulus control in intermittent and daily smokers. PLoS ONE. 2014b;9(3):14. doi: 10.1371/journal.pone.0089911. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. 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. Current Medical Research and Opinion. 2008;24(10):2795–2804. doi: 10.1185/03007990802380341. [DOI] [PubMed] [Google Scholar]
  32. Shiffman S, Ferguson SG, Dunbar MS, Scholl SM. Tobacco dependence among intermittent smokers. Nicotine and Tobacco Research. 2012;14(11):1372–1381. doi: 10.1093/ntr/nts097. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Shiffman S, Ferguson SG, Gwaltney CJ, Balabanis MH, Shadel WG. Reduction of abstinence-induced withdrawal and craving using high-dose nicotine replacement therapy. Psychopharmacology. 2006;184(3–4):637–644. doi: 10.1007/s00213-005-0184-3. [DOI] [PubMed] [Google Scholar]
  34. Shiffman S, Fischer LB, Zettlersegal M, Benowitz NL. Nicotine exposure among nondependent smokers. Archives of General Psychiatry. 1990;47(4):333–336. doi: 10.1001/archpsyc.1990.01810160033006. Retrieved from <Go to ISI>://WOS:A1990CX87300004. [DOI] [PubMed] [Google Scholar]
  35. Shiffman S, Li X, Dunbar MS, Tindle HA, Scholl SM, Ferguson SG. Does laboratory cue reactivity correlate with real-world craving and smoking responses to cues? Drug and Alcohol Dependence. 2015 doi: 10.1016/j.drugalcdep.2015.07.673. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Shiffman S, Patten C, Gwaltney C, Paty J, Gnys M, Kassel J, Balabanis M. Natural history of nicotine withdrawal. Addiction. 2006;101(12):1822–1832. doi: 10.1111/j.1360-0443.2006.01635.x. [DOI] [PubMed] [Google Scholar]
  37. Shiffman S, Paty J. Smoking patterns and dependence: Contrasting chippers and heavy smokers. Journal of Abnormal Psychology. 2006;115(3):509–523. doi: 10.1037/0021-843x.115.3.509. [DOI] [PubMed] [Google Scholar]
  38. Shiffman S, Stone AA, Hufford MR. Ecological momentary assessment. Annual Review of Clinical Psychology. 2008;4:1–32. doi: 10.1146/Annurev.Clinpsy.3.022806.091415. [DOI] [PubMed] [Google Scholar]
  39. Shiffman S, Zettler-Segal M, Kassel J, Paty J, Benowitz N, O’Brien G. Nicotine elimination and tolerance in non-dependent cigarette smokers. Psychopharmacology. 1992;109(4):449–456. doi: 10.1007/BF02247722. [DOI] [PubMed] [Google Scholar]
  40. Stead LF, Perera R, Bullen C, Mant D, Hartmann-Boyce J, Cahill K, Lancaster T. Nicotine replacement therapy for smoking cessation. Cochrane Database of Systematic Reviews. 2012;(11) doi: 10.1002/14651858.CD000146.pub4. http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD000146.pub4/abstract. [DOI] [PubMed]
  41. Thrul J, Bühler A, Ferguson SG. Situational and mood factors associated with smoking in young adult light and heavy smokers. Drug and Alcohol Review. 2014;33(4):420–427. doi: 10.1111/dar.12164. [DOI] [PubMed] [Google Scholar]
  42. Tindle HA, Shiffman S. Smoking cessation behavior among intermittent smokers versus daily smokers. American Journal of Public Health. 2011;101(7):e1–e3. doi: 10.2105/AJPH.2011.300186. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. U.S. Department of Health and Human Services. The health consequences of smoking: 50 years of progress. A report of the Surgeon General. Atlanta, GA: 2014. [Google Scholar]
  44. WHO. WHO global report: mortality attributable to tobacco. Geneva: World Health Organisation; 2012. [Google Scholar]
  45. World Health Organization. WHO report on the global tobacco epidemic, 2013: enforcing bans on tobacco advertising, promotion and sponsorship. Geneva: World Health Organization; 2013. [Google Scholar]

RESOURCES