Cue-Induced Cigarette Cravings and Smoking Cessation: The Role of Expectancies

Abstract

Introduction:

Cue-induced cigarette cravings have been oft studied as potentially important predictors of smoking cessation outcomes. The literature on the relationship between cue-induced cravings and cessation, however, remains mixed. One possible explanation for the discrepant results in the literature may be the as-yet untested variability in expectancies of craving. Indeed, as with many interoceptive responses, cravings and their downstream consequences may be influenced by expectancies. To date, no study has examined the influence of expected cravings following smoking cue exposures on actual craving experiences and cessation outcomes. The objective of this study, therefore, was to test the possibility that smokers’ expected craving levels in response to smoking cues would be related to actual cravings following cue exposure and that expected cravings would be related to cessation outcomes.

Methods:

Nicotine-dependent adult smokers (n = 153) were exposed to sets of neutral and smoking cues and completed questionnaires assessing (a) prior to the exposures, the cigarette craving levels they expected to experience following the cue exposures and (b) following the exposures, their actual craving levels. Participants also reported the duration of their most recent quit attempt and their perceived future quit difficulty.

Results:

Findings indicated that expected cravings assessed prior to the cue exposures were significantly related to actual cravings following the exposures. In addition, both expected cravings and actual cravings were related to shorter previous quit duration and higher perceived quit difficulty.

Conclusions:

Study results highlight the importance of considering both expected and actual cravings in cue-induced craving paradigms.

Introduction

Cigarette smoking continues to pose a major public health threat in the United States and abroad. Smoking has been related to significantly increased risk for a number of chronic and deadly illnesses, including cardiovascular disease and cancer (American Cancer Society, 2009). Despite the clear negative health consequences of smoking, smoking cessation success remains elusive for many (Centers for Disease Control and Prevention, 2009). A widely touted explanation for difficulty in quitting smoking is the ubiquitous presence of smoking-related cues in the environment that trigger strong difficult-to-resist cravings to smoke (Tiffany, 1990). A large body of literature has demonstrated that these “cue-induced cravings” can be reliably modeled in the laboratory. In one review (Carter & Tiffany, 1999), the authors conclude that exposure to smoking cues (e.g., handling a cigarette without smoking it) under controlled laboratory conditions consistently gives rise to strong self-reported craving responses.

Less well established, however, is whether these cue-induced cravings are related to smoking cessation outcomes. Indeed, an early review by Niaura et al. (1988) indicated that although reactions to cues tend to be robust, their ability to predict cessation outcomes has not been well established. Since then, few studies have actually examined relations between laboratory cue-induced craving and cessation outcomes and those that did have yielded mixed results. For example, Waters et al. (2004) found that cue-induced craving was predictive of poorer smoking cessation in a prospective study but only among smokers randomized to receive nicotine patch treatment. Our group (Erblich & Bovbjerg, 2004) found in a retrospective study that cravings in response to in vivo, but not imaginal, smoking cues were related to shorter durations of previous quit attempts in a sample of current smokers. Another study (Shadel et al., 1998) found no association between cue-induced cravings and cessation in a group of self-quitters. These disparate results raise the possibility that additional factors need to be considered when evaluating the role of cue-induced cravings in smoking cessation.

Accumulating evidence has pointed to the importance of a construct known as “response expectancies” as a contributor to a variety of nonvolitional outcomes (Kirsch, 1999). Response expectancies refer to an individual’s expectations with regard to how they will respond to a specific stimulus. For example, a cancer patient preparing to undergo chemotherapy may expect to experience nausea, and these expectations may be an important predictor of actual nausea (Montgomery & Bovbjerg, 2001). Research on response expectancies has highlighted the importance of one’s cognitive appraisal of an anticipated experience as related, yet distinct, from one’s actual experiences. In an experimental study of the placebo effect, Montgomery and Kirsch (1997) found that expectancies of pain were predictive of actual pain experienced, but the strength of the relationship differed under various experimental conditions, suggesting that expected and actual responses, even when measured minutes apart, are related yet conceptually distinct constructs.

To our knowledge, no research has been done to characterize the role of response expectancies for cue-induced cravings. The possibility that a smoker’s expectations and appraisals of how he/she would react to exposure to smoking stimuli would relate to actual reactions has never been tested. Moreover, the possibility that the response expectancies for craving may, in turn, influence smoking cessation outcomes had never been addressed.

The notion that cognitive factors can play an important role in cue-induced craving is consistent with early conceptualizations of the phenomenon. For example, Cooney, Gillespie, Baker, & Kaplan (1987) found that alcohol cue exposures led to increased expectations of favorable drinking effects. Marlatt et al. (Larimer, Palmer, & Marlatt, 1999; Marlatt & Gordon, 1985) postulated that strong cue-induced cravings can result in poorer perceived self-efficacy and diminished self-confidence to abstain. The possibility that cognitive factors can influence the cravings themselves, however, has not been tested.

In this study, we evaluated the role of response expectancies in cue-induced craving. Our primary hypothesis was that both actual and expected cravings would be related to smoking cessation variables. We also explored the possibility that actual and expected cravings might be differentially related to cessation variables. In addition, we hypothesized that expected cravings in response to laboratory smoking cue exposures will be related to actual cravings following cue exposures. Finally, we explored the possibility that the absolute levels of expected and actual cravings would be significantly different from one another. Thus, we hypothesized that expected and actual cravings would be related yet conceptually distinct factors that may be important in better understanding smoking cessation.

Method

Participants

Adult smokers (n = 153) were recruited to an ongoing study of smoking behavior by advertisements (for a smoking study) placed in and around an urban medical center in New York City. Participants were a subset of those from an earlier report (Erblich & Bovbjerg, 2004) for whom data on expected cravings were available. All participants qualified for a current DSM-IV diagnosis of nicotine dependence (American Psychiatric Association, 1994) ascertained during initial telephone contact and reascertained in-person upon arrival using a structured clinical interview. To qualify as a smoker, only participants who reported smoking at least 10 cigarettes/day for at least 5 years were included in the study. Participants were excluded if they reported (a) current treatment for smoking cessation, (b) a history of other substance abuse, or (c) a history of hospitalization for mental illness.

Measures

Demographic and Smoking Questionnaire

Participants completed questionnaires assessing basic demographic information (e.g., age, gender, education, income, ethnicity, marital status) and personal smoking history (e.g., age at initiation, cigarettes per day, years smoked). In addition, participants completed a face-valid item assessing the duration of their most recent attempt, as well as a face-valid 0–25 scale assessing how difficult they felt it would be for them to quit smoking ( “perceived quit difficulty”).

Fagerström Test for Nicotine Dependence

The six-item Fagerström Test for Nicotine Dependence (FTND) was used as a measure of the strength of participants’ dependence. The instrument has documented reliability and validity (Heatherton, Kozlowski, Frecker, & Fagerström, 1991) and has been used extensively in the literature.

Strength of Imagery

After each of the two imaginal exposures (neutral and smoking), participants completed a four-item, face-valid (e.g., “How vivid did your images seem?” and “How real did your images seem to you?”) 0–25 scale of imagery strength. This instrument has been used in our previous work (Erblich & Bovbjerg, 2004; Erblich, Lerman, Self, Diaz, & Bovbjerg, 2005) and demonstrated strong internal consistency (Cronbach’s α ranging from .85 to .92 for the two administrations) in the current sample.

Cigarette Craving Questionnaire

Improving on the use of single-item craving assessments and following the recommendations of Kozlowski et al. (Kozlowski & Wilkinson, 1987; Kozlowski, Mann, Wilkinson, & Poulos, 1989) that craving be assessed using multiple descriptors (e.g., crave, urge, desire), this brief, five-item (“I have a desire for a cigarette right now,” “If it were possible I would smoke now,” “All I want is a cigarette right now,” “I have an urge for a cigarette right now,” “I crave a cigarette right now”) 0–100 instrument is designed specifically to make rapid assessments of craving during experimental manipulations. The instrument has been used as an outcome measure in previous reports by us (Erblich, Boyarsky, Spring, Niaura, & Bovbjerg, 2003) and others (Hutchison, LaChance, Niaura, Bryan, & Smolen, 2002). Two versions of the questionnaire were employed: (a) Immediately prior to each stimulus, the instrument was used to assess the degree to which participants expected to crave cigarettes following the stimulus and (b) immediately following each stimulus, the instrument was used to assess the degree to which participants actually craved cigarettes following the stimulus. The instrument evidenced high levels of internal consistency (Cronbach’s α = .96–.97) at all administrations.

Procedures

Respondents to advertisements were screened via telephone to determine eligibility. The study was described over the phone, and if eligible (90% of callers were eligible), participants were scheduled for the study. In particular, prospective participants were informed that they would view and listen to various stimuli and report how it made them feel. Ninety percent of eligible respondents were enrolled in the study. Participants were scheduled for two counterbalanced cue-exposure days (approximately 20 min each); they were exposed to imaginal cues on one day and in vivo cues on the other day. To avoid ceiling effects in cigarette craving, participants smoked one cigarette immediately before beginning the study procedures on each day. Briefly, participants completed demographic, smoking history and cessation questionnaires, and the FTND on the first study day and were exposed either to a set of imaginal or in vivo cues, as described herein. On the second visit, scheduled 1 week later, participants were exposed to the next set of cues, either imaginal or in vivo, depending on the counterbalanced order of administration. For imaginal exposures, participants listened to a 60-s script and were instructed to imagine the scene as vividly as possible, followed by a 30-s silent period, during which they were to continue imagining the scene. Before beginning the exposures, participants were read a practice script (a trip to the grocery store) to familiarize them with the task. The imaginal cues employed have been described elsewhere (Maude-Griffin & Tiffany, 1996); the neutral imagery involved changing a light bulb, and the smoking imagery involved thinking about lighting up a cigarette after a meal.

For the in vivo smoking exposure, participants were presented with their cigarette box, asked to remove one cigarette, and hold it for 90 s (unlit). For the neutral cue, they were presented with a stapler and asked to hold it for the same amount of time. Expected cigarette cravings were measured immediately before each cue induction, and actual cigarette cravings were measured immediately following each cue. For the assessment of expected cravings, participants were described the cues they were about to encounter and told to rate how much they expected to crave cigarettes following exposure to the cues (e.g., How would you expect to feel when asked to imagine smoking a cigarette?).

To avoid possible carryover effects (Heishman, Lee, Taylor, & Singleton, 2010; Sayette, Griffin, & Sayers, 2010) within exposure modality (imaginal/in vivo), on each day, participants were exposed to the neutral cue first, followed by the smoking cue. In addition, participants viewed a nature video for 3 min between each presentation to provide a rest and facilitate a return to baseline. Upon completion of the study, participants were thanked for their participation, offered referrals for smoking cessation interventions, and paid an honorarium for their time.

Data Analysis

In the first set of analyses, we aimed to characterize the relationship between expected and actual cravings. To that end, we performed a three-way repeated measures analysis of variance (ANOVA), including (a) neutral versus smoking cues, (b) imaginal versus in vivo cues, and (c) expected versus actual cravings, as within subjects’ factors and evaluated differences between actual and expected craving levels as the dependent variables. Next, we performed zero-order Pearson correlations between expected and actual cravings in response to the smoking cues as well as partial correlations that controlled for expected and actual cravings in response to the neutral cues.

To address the primary hypothesis that cravings would be related to cessation outcomes, we performed a series of regression analyses, with quit duration and perceived quit difficulty as dependent variables and expected and actual cravings as the independent variables. Demographic and smoking-related variables were tested for possible relationships with dependent variables and if related, were included as covariates. Because the quit duration data were not normally distributed, an iterative Box–Cox transformation procedure was employed to identify the best transform (Box & Cox, 1964), and regression analyses were performed on the transformed values. The transformation was highly effective, reducing skewness from 4.13 to 0.91 and kurtosis from 18.02 to 1.11. The perceived quit difficulty data did not depart significantly from normality and were thus not transformed.

Results

Background Variables

Fifty-two percent of participants (n = 81) were men and 48% (n = 75) were women. The mean age of the sample was 37.63 years (SD = 10.9). Thirty-eight percent of participants reported Black ethnicity, 27% reported Caucasian ethnicity, 26% reported Hispanic ethnicity, and 9% reported other ethnic backgrounds. Twenty-nine percent of participants had a high school education or lower, and 37% of the sample reported annual incomes of $20,000 or less. Participants smoked a mean of 17.14 cigarettes/day (SD = 8.71) for an average of 17.93 years (SD = 10.32). Mean Fagerström score of the sample was 5.28 (SD = 2.17). For the most recent quit attempt, 26% reporting using nicotine replacement therapy, 68% reported using no assistance, and the remaining 6% reported using other approaches. Preliminary analyses revealed that demographic variables were not related to the duration of the most recent quit attempt nor were they related to perceived quit difficulty. Gender, however, was related to expected cravings, with women displaying higher expected cravings (77.81 ± 8.18) in response to imaginal cues than men (63.88 ± 8.57): F(1, 152) = 5.79, p = .009. No other demographic variables were related to expected craving. Gender was included simultaneously with the primary variables in subsequent analyses.

Not surprisingly, higher Fagerström scores were related to a shorter most recent quit duration, F(1, 152) = 7.00, p = .018, as well as higher perceived quit difficulty: F(1, 152) = 4.51, p = .037. The remaining smoking-related variables (i.e., cigarettes per day, number of years having smoked, strength of imagery) were related to neither quit duration nor quit difficulty. In addition, smoking-related variables were not related to expected cravings. Finally, none of the background variables were related to actual cravings (all p > .15).

Relationships Between Expected Cravings and Actual Cravings

Expected and actual craving levels in response to the imaginal and in vivo cues are reported in Table 1. Not surprisingly, a repeated measures ANOVA including three factors (neutral vs. smoking, imaginal vs. in vivo, expected vs. actual cravings) revealed a significant main effect of neutral versus smoking; F(1, 152) = 412.14, p < .0001, with overall mean craving levels being significantly higher following the smoking cue (M = 72.11, S E = 2.57) than following the neutral cue (M = 25.78, S E = 2.61). Interestingly, the Cue × Assessment interaction was not significant, indicating no overall difference in magnitude between expected and actual cravings; F(1, 152) = 2.17, p < .142. There was, however, a significant three-way interaction between the factors (see Figure 1). Further exploring this interaction, pairwise comparisons revealed that actual cravings were slightly higher than expected cravings following imaginal smoking cue exposures but were comparable following in vivo cue exposures (see Tables 1 and 2, as well as Figure 1). Including gender and counterbalancing order as independent variables in these analyses did not alter the results.

Table 1.

Mean (±SE) Expected and Actual Craving Levels Following Cue Exposures

	Imaginal		In vivo
	Expected	Actual	Expected	Actual
Neutral	27.89 (2.51)	23.37 (2.48)	24.22 (2.48)	27.84 (2.93)
Smoking	68.24 (2.83)	72.74 (2.51)	74.19 (2.48)	73.44 (2.72)

Figure 1.

Effects of neutral and smoking cue exposures on expected and actual cravings (M ± SE). Note that the main effect of neutral versus smoking was significant (p < .0001), as well as the three-way interaction (p < .001), with actual cravings following imaginal smoking cues slightly higher than expected cravings.

Table 2.

Statistics for the Analysis of Variance of Expected and Actual Cue-Induced Craving

Effect	F value(df = 1, 152)	p Value	Effect size(partial η2)
Cue (neutral, smoking)	412.14	<0.0001	.73
Modality (imaginal, in vivo)	0.21	.639	.00
Assessment (expected, actual)	3.48	.063	.02
Cue × Modality	0.71	.403	.01
Cue × Assessment	2.17	.142	.01
Modality × Assessment	0.54	.471	.00
Cue × Modality × Assessment	10.53	<.001	.06

Consistent with the study hypotheses, expected cravings were strongly correlated with actual cravings following both the in vivo and imaginal smoking cues. For the in vivo smoking cue, the bivariate correlation coefficient between expected and actual cravings was .70 (p < .0001), and for the imaginal smoking cue, the correlation was .62 (p < .0001). After controlling for expected and actual cravings in response to the neutral cues, the correlation coefficients remained highly significant (p < .0001), though slightly lower (.66 for the in vivo cue and .55 for the imaginal cue).

Relationships Between Expected Cravings and Smoking Cessation Outcome Variables

We then examined the possibility that expected and actual cravings were related to the duration of quit attempts and quit difficulty. Figure 2 displays Box–Cox regression point estimates for the duration of participants’ most recent quit attempts as a function of cravings. Values are back transformed to their original scale for ease of interpretation. Consistent with the study hypotheses, higher levels of expected cravings in response to the in vivo smoking cues were related to shorter durations of recent quit attempts; F(1, 152) = 7.01, R ² = .04, b = −.04, p = .009. Expected cravings in response to the imaginal smoking cues were marginally significantly related to shorter quit durations; F(1, 152) = 2.82, R ² = .02, b = −.02, p = .091. As reported in the earlier study in a larger sample (Erblich & Bovbjerg, 2004), actual cravings following in vivo; F(1, 152) = 7.79, R ² = .05, b = −.04, p = .006, but not imaginal; F(1, 152) = 0.11, R ² = .00008, b = −.001, p = .912, smoking cues were related to shorter quit durations. Covarying the type of quit attempt (e.g., unaided, nicotine replacement) did not change the results.

Figure 2.

Box–Cox regression point estimates of recent quit duration as a function of expected and actual cravings following imaginal and in vivo smoking cue exposures.

Results were similar when examining perceived quit difficulty. As indicated in Figure 3, expected cravings in response to the in vivo smoking cue were related to higher perceived quit difficulty; F(1, 152) = 19.38, R ² = .10, b = .06, p < .0001, and expected cravings in response to the imaginal smoking cue were marginally associated with higher levels of perceived quit difficulty; F(1, 152) = 3.21, R ² = .02, b = .02, p < .067. Similarly, actual craving following in vivo smoking cues; F(1, 152) = 21.73, R ² = .11, b = .06, p < .0001, but not imaginal smoking cues; F(1, 152) = 1.27, R ² = .01, b = .02, p = .252, was associated with greater perceived quit difficulty.

Figure 3.

Regression point estimates of perceived quit difficulty as a function of expected and actual cravings following imaginal and in vivo smoking cue exposures.

Effects of Expected and Actual Cravings in Combined Regression Models

Finally, we explored the effects of actual and expected cravings on cessation variables in combined regression models in which actual and expected cravings were entered simultaneously to provide a head-to-head comparison of the two. In a similar pattern to the independent regression models, expected cravings, but not actual cravings, following the imaginal cues were associated with significantly shorter quit durations; F(1, 152) = 6.23, R ² = .03, b = −.04, p = .007, and higher perceived quit difficulty; F(1, 152) = 4.02, R ² = .02, b = .03, p = .047. Interestingly, the same pattern of results was observed when examining cravings following in vivo cues. Unlike the results of the independent analyses, in the combined regression, expected, but not actual, cravings were significantly associated with both quit duration; F(1, 152) = 3.89, R ² = .03, b = −.02, p = .048 and perceived quit difficulty F(1, 152) = 4.01, R ² = .02, b = .03, p = .049. Because of the correlations between expected and actual cravings reported above, we performed multicollinearity diagnostics, examining the Variance Inflation Factor (VIF) and the Conditioning Index (CI) for expected and actual cravings. Traditionally, VIFs greater than 10 and CIs greater than 30 suggest that multicollinearity may be an issue (Stevens, 2002; Tabachnick & Fidell, 2001). Findings revealed that VIFs ranged from 1.71 to 2.19 and CIs ranged from 6.27 to 9.08, all far below the traditional cutoffs for multicollinearity.

Discussion

The present study aimed to identify a potential role for response expectancies in better understanding reactions to laboratory smoking cues and their relations to smoking cessation. Findings indicated that the absolute levels of actual and expected cravings were comparable, especially in response to in vivo cues. Consistent with the study hypotheses, response expectancies were significantly related to actual craving responses to both imaginal and in vivo smoking cues. In addition, higher response expectancies to in vivo cues were associated with both higher perceived quit difficulty and shorter quit durations during a previous quit attempt. Interestingly, although actual cravings were also related to these cessation outcomes in independent regression models, when tested with expected cravings in the model, actual cravings were no longer significantly related of either quit difficulty or quit duration. This was the case in spite of similarities in the absolute values of expected and actual cravings.

These findings highlight the importance of response expectancies in understanding cigarette craving and raise the possibility that expected cravings may overshadow the importance of actual cravings as predictors of smoking outcomes. To our knowledge, studies have not previously examined response expectancies, and the current data suggest that response expectancies not only are related to craving levels but also may actually play a stronger role in quit success than the actual cue-induced cravings themselves. As indicated above, the literature on relations between cue-induced craving and cessation has been mixed (Perkins, 2009). The present results raise the intriguing possibility that smokers’ cognitive appraisals of their anticipated reactions to environmental smoking cues may be more important than the reactions themselves. If this is the case, then intervention efforts aimed at directly managing the anticipation and expectation of cravings may prove to be more efficacious than classic attempts to address the actual cue-induced craving reactions, such as extinction trials and cue exposure therapy (Ferguson & Shiffman, 2008; Perkins, 2009). These findings may explain why such classic interventions have met with only modest clinical success (Perkins, 2009). It is also possible that higher expected cravings are related to decreased self-efficacy to resist the cravings, which in turn may be related to poor quit success. Indeed, It is well established that self-efficacy is an important component of cessation success (Prochaska & Velicer, 1997). The possible relationship between expected cravings and self-efficacy is an interesting topic for future research.

It is also interesting to note that studies of response expectancies typically examine relationships with actual responses but do not address whether or not they predict clinically relevant outcomes. As one example, while the literature has demonstrated that expected nausea following chemotherapy is predictive of actual nausea (Montgomery & Bovbjerg, 2001), there have been no studies evaluating whether or not expected nausea predicts behavioral outcomes (e.g., use of antiemetic medication, noncompliance with chemotherapy, psychological distress). To our knowledge, this is the first study to identify a link between response expectancies (expected cravings) and psychological/behavioral outcomes (quit difficulty/quit duration). Findings thus suggest that response expectancies, at least those related to cue-induced craving, may have important clinically relevant consequences. Results should be interpreted with caution, however, as effect sizes were in the small range based on Cohen’s (1977) criteria.

A significant limitation of this study is its cross-sectional design. Data on the duration of quit attempts were collected retrospectively from a sample of current smokers. As such, it is conceivable that previous quit failures contributed to stronger expected cravings rather than the reverse. A stronger approach would be to conduct the study using a longitudinal design. Future prospective research on the effects of expected cravings on cessation success are warranted to more definitively elucidate the role of response expectancies in cessation failure. We also found that effects of expected cravings were stronger for in vivo cues than for imaginal cues. It is possible that participants were less sure of what to expect prior to imaginal cue exposure because they were less familiar with imagery than with in vivo cues. This raises the important possibility that effects of expectancies may be stronger when smokers are more aware of how a given cue may affect them. These mixed effects warrant additional attention. Future research could examine the impact of more general craving expectancies on actual cravings that would be relevant both to laboratory-based cues as well as naturalistic triggers in the environment. Finally, the study focused only on smoking cues. The possibility that expected and actual responses to affective cues are important predictors of smoking outcomes also warrants further study. Indeed, studies have demonstrated that affective cues can elicit strong craving reactions (e.g., Niaura et al., 1998).

We also cannot rule out the possibility that assessments of expected and actual cravings in close proximity created a self-fulfilling prophecy, thus inflating their correlation. Nevertheless, our results indicated that the two assessments were differentially related to smoking variables, suggesting that they did tap distinct processes. These findings should be considered preliminary, requiring a follow-up study in which assessments of expected and actual cravings are temporally more distinct. It should also be noted that we employed a multifaceted assessment of craving. Given recent animal research suggesting that cue-induced cravings can have multiple distinct elements (Scott & Hiroi, 2011), it might be important to develop novel assessments for humans that tap these processes more directly.

In addition, the clinical relevance of the relationship between expected cravings and cessation failure needs to be determined. Indeed, the degree to which findings in the current sample of nondeprived smokers not currently in treatment generalizes to deprived smokers trying to quit remains to be determined. Evidence from Sayette et al. (Sayette, Loewenstein, Griffin, & Black, 2008; Sayette, Loewenstein, Kirchner, & Travis, 2005), however, suggests that predictions of later cravings tend to be even more accurate in deprived smokers, suggesting that the present results may be applicable to deprived states as well. Finally, results suggest that it is possible that cognitive interventions aimed at managing expected cravings may prove to be more efficacious than interventions to manage cue-induced cravings per se. Clinical research addressing this interesting possibility is also warranted and may lead to innovations in managing cravings and enhancing quit success.

Funding

This work was funded by grants #K22CA124800, #R21CA118703, and #UL1RR029887, from the National Institutes of Health.

Declaration of Interests

None declared.