Implicit Attitudes and Smoking Behavior in a Smoking Cessation Induction Trial Free

Abstract

Introduction

Although studies have suggested that implicit attitudes may predict smoking-related decisions, evidence that changes in implicit attitudes toward smoking are related to changes in smoking behavior is lacking. Using data from a trial comparing interventions to induce quit attempts among unmotivated smokers, this study examined whether changes in implicit attitudes were associated with quit attempts and cessation after controlling for explicit motivation.

Methods

Daily smokers recruited from the community completed measures of implicit attitudes (Implicit Association Test) and explicit measure of motivation to smoke at baseline, mid-intervention (week 12 [W12]) and follow-up (week 26 [W26]). Quit attempts and cessation were assessed at follow-up, and cessation was biochemically verified.

Results

As hypothesized, Implicit Association Test scores became more negative from baseline to W12, a change that was sustained at follow-up. Logistic regression analyses in which implicit attitudes were used to predict smoking outcomes revealed that negative changes in implicit attitudes from baseline to W12 and from baseline to W26 were significantly related to quit attempts (OR = 0.71, 95% CI [0.52, 0.97], p < .05 for both) independent of explicit motivation. Negative changes in implicit attitudes from baseline to W26 were significantly related to cessation (OR = 0.50, 95% CI [0.25, 1.00], p < .05).

Conclusions

Negative changes in implicit attitudes were associated with positive changes in smoking behavior independent of explicit motivation. This result indicates that smoking cessation interventions may be enhanced by incorporating strategies to change implicit attitudes, and that changes in implicit attitudes are also potentially important intervention outcomes.

Implications

Smoking cessation interventions may be improved by going beyond the current focus on explicit psychological constructs and targeting automatic cognitive processes such as implicit attitudes. The results are encouragement to examine how best to manipulate smokers’ implicit attitudes as well as to determine the effect on their smoking behavior.

Introduction

Decision-making regarding health behaviors such as smoking is understood to be a function of individuals’ positive and negative evaluations of cigarette use and its consequences.¹ Historically, researchers have measured these evaluations, or attitudes, toward smoking with explicit self-report questionnaires.² However, these explicit attitudes only modestly predict the level of smoking and changes in cigarette use over time.^3,4 Recently, researchers have examined implicit attitudes in relationship to health behavior from a dual-process perspective.⁵ According to this model, human behavior including health behavior is guided by two independent systems, explicit and implicit processes. These two motivational processes are thought to have unique contributions to smoking behavior.⁶ Implicit-attitude tests measure fast, parallel, effortless, and uncontrolled processes.^7–9 Implicit attitudes are also less susceptible to socially-desirable responding, which is common in self-reported measurement of stigmatized behaviors such as smoking.^10,11 More importantly, smoking is an addictive behavior, which is thought to be particularly influenced by automatic processes.^12,13

Most studies have found that smokers’ attitudes are significantly less negative than nonsmokers’ attitudes.^14–18 In addition, implicit attitudes toward smoking are more negative in light smokers compared to heavy smokers,¹⁶ are more negative among individuals lower in nicotine dependence,^19,20 and are less negative among smokers during withdrawal.²¹ Implicit attitudes have also been associated with smoking cessation at an 18-month follow-up²² although implicit attitudes assessed before and after smoking cessation have failed to predict relapse.²³ These studies suggest that implicit attitudes can predict smoking behaviors and could be a novel treatment target in cessation interventions.

In order to translate experimental research on implicit attitudes into a clinical context, more evidence is needed of the association between changes in implicit attitudes and smoking behavior in natural contexts. While there is considerable evidence that implicit attitudes are relatively inflexible and difficult to change, especially once they are formed,^24,25 there is also evidence from memory and social cognition research that suggests implicit attitudes can change rather quickly in response to contextual factors and motivational states, with relatively little counterattitudinal information.^26–30 Furthermore, studies have shown that changes in implicit attitudes are not necessarily accompanied by changes in explicit attitudes.^6,26,31

A few studies have observed changes in implicit attitudes due to counterattitudinal messages such as anti-tobacco television advertisements,³² anti-smoking public service announcements,³³ warning labels of cigarette packages,³⁴ and approach-avoidance practice interventions.³⁵ However, long-term changes in implicit attitudes following a smoking cessation intervention have not been investigated. More importantly, no study has related changes in implicit attitudes toward smoking to changes in smoking behaviors, such as quit attempts and cessation. Lastly, existing studies on implicit attitudes toward smoking have predominantly recruited young adult (ie, college-aged), Caucasian, non-daily smokers as participants; thus, these results may not generalize to a broader, more heterogeneous smoking population.

The aim of this study was to determine whether changes in implicit attitudes toward smoking were associated with changes in quit attempts and cessation independently from changes in self-reported or explicitly measured motivation to smoke. Data were drawn from a smoking cessation induction trial involving a diverse sample of community residents. We hypothesized that changes in implicit attitudes toward smoking from baseline to mid-intervention, and from baseline to follow-up, would be associated with quit attempts and cessation. We also hypothesized that the prediction of changes in implicit attitudes on changes in smoking behavior would be independent of changes in explicit motivation to smoke.

Methods

Participants

Daily smokers (N = 255) who were not motivated to quit smoking were recruited from a large Midwestern city for a clinical trial to induce quit attempts.³⁶ Eligible participants were at least 18 years of age, smoked at least one cigarette a day, spoke English, were not pregnant, were not using smoking-cessation medication, and were not motivated to quit smoking (≤6 on a 0 to 10 scale of motivation to quit) or not planning to quit smoking in the next 7 days. The study received Institutional Review Board approval, and informed consent was obtained from all participants.

Measures

Assessments of implicit attitudes and explicit motivation were administered at baseline, mid-intervention (week 12 [W12]) and follow-up (week 26 [W26]). Demographics and smoking characteristics were obtained at baseline, and changes in smoking behavior were reported at W26.

Demographics and Smoking Characteristics

The baseline survey included assessment of demographics and smoking characteristics including number of cigarettes smoked per day and age when the first cigarette was smoked. Nicotine dependence was assessed using the “time to first cigarette” item (ie, “How soon after you wake up do you smoke your first cigarette of the day?”) from the Fagerström Test for Nicotine Dependence.³⁷ This single item has been shown to account for much of the Fagerström Test for Nicotine Dependence’s predictive validity for cessation success,³⁸ and was used to reduce survey length.

Implicit Attitudes Toward Smoking

To maximize consistency with prior research on smoking behavior, implicit attitudes were assessed with the Implicit Association Test (IAT³⁹) using stimuli and procedures from prior smoking studies.^35,40 Stimuli (presented on a 22” computer monitor with a black background) were eight pictures related to smoking (labeled on the screen as “smoking”; eg, a picture of cigarettes next to a lighter, a picture of a lit cigarette), eight picture of shapes (labeled as “shape”; eg, pictures of line-drawn shapes such as rectangles and ovals), eight adjectives with a positive meaning (labeled as “good”; eg, wonderful, nice, pleasant), and eight adjectives with a negative meaning (labeled as “bad”; eg, stupid, disgusting, horrible). The picture size was set to 20% of the height and width of the computer screen. The background and brightness of the smoking and shape pictures were closely matched. The adjective letter size was set to 5% of the height and width of the screen, and adjectives were presented in green letters.

Participants were asked to categorize the stimuli presented in the center of the screen according to the labels presented in the upper right or left side of the screen, as quickly as possible, by pressing the letter e (left) or i (right) on the keyboard. The label letter size was set to 3.5% of the height and width of the screen. The smoking/shapes categories were presented in white letters, while the good/bad categories were presented in green letters. The IAT consisted of seven phases: (1) 20 practice trials categorizing word stimuli into “good” or “bad”; (2) 20 practice trials categorizing picture stimuli into “smoking” or “shapes”; (3) and (4) 20 practice and 40 experimental trials categorizing word and picture stimulus pairs into one of two combined categories (eg, smoking + good vs. shapes + bad); (5) the same 20 practice trials as phase 2 except the right/left locations of category labels (ie, shapes vs. smoking) were switched on the computer screen; and (6) and (7) the same 20 practice and 40 experimental trials as phase 3 and 4 except the right/left locations of the combined categories were switched on the computer screen (eg, shapes + good vs. smoking + bad). On each trial, the stimulus was presented until the participant pressed a letter on the keyboard. Following a correct response, the next stimulus was presented. Following an incorrect response, an X was presented on the screen and the same stimulus was presented again until the correct response was made. The presentation order of paired categories was counterbalanced. Implicit attitudes were indexed using an IAT D score that was calculated for each participant at each assessment point by subtracting the response latencies when smoking pictures were paired with “good” words from the response latencies when smoking pictures were paired with “bad” words.⁴¹ Since longer latencies imply an incongruent association, negative IAT scores indicate longer response latencies in smoking pictures—“good” word pairs and thus negative implicit attitudes toward smoking. Individuals with more than 30% error rates (including too fast trials in which latencies were less than 300 ms and too slow trials in which latencies were above 10 000 ms) in the main IAT task were excluded from data analyses. With this criterion, 27 IAT scores from 18 participants were excluded in the data analyses. For the primary analysis change scores were calculated by subtracting baseline IAT D scores from W12 IAT D scores. Higher scores reflect less of a negative shift in attitudes toward smoking over time. Internal consistency (Chronbach’s α) using difference scores for every experimental trial⁴² was 0.75 for baseline IAT in this study.

Explicit Motivation to Smoke

The 20-item Decisional Balance Scale⁴³ was used to assess self-reported or explicit motivation to smoke. Participants indicated the extent of agreement with a list of items describing the pros (ie, “I am relaxed and therefore more pleasant when smoking”) and cons (ie, “People close to me disapprove of my smoking”) of smoking on a 5-point Likert scale (1 = not important; 5 = extremely important). Explicit motivation to smoke was calculated using a difference score between pro and con items with positive scores indicating positive explicit motivation to smoke. In the primary analysis change scores were calculated by subtracting baseline Decisional Balance Scale scores from W12 scores. Higher scores reflect less of a negative shift in motivation to smoke.

Changes in Smoking Behavior

To assess quit attempts, participants were asked if they had made a serious quit attempt lasting at least 24 hours since the last assessment.^44,45 Reports made at W12 and W26 were aggregated into a single dichotomous variable that indicated whether any quit attempt had been made since baseline. Assessment of smoking status at W26 was based on self-reported 7-day point prevalence abstinence.⁴⁶ Participants were asked if they had smoked “even a puff” in the past 7 days. Reports of cessation were biochemically verified using saliva cotinine.⁴⁷

Procedure

Smokers were recruited through advertisements targeting “smokers” or “smokers not quite ready to quit” and prospective participants were prescreened for eligibility by phone. Participants were told the goal of the study was to learn how best to talk to smokers about their health. Eligible participants were invited to an in-person visit and rescreened. The participants whose eligibility was reconfirmed were asked to complete measures of demographics, implicit attitudes, and explicit motivation. Then, they were randomly assigned to one of three treatment arms (a one-session brief advice to quit arm, a four-session health education arm, and a four-session Motivational Interviewing arm) and received the first intervention session based on treatment group assignment. The single-session brief advice intervention provided smokers with clear, strong, and personalized advice to quit tobacco use. The four-session health education intervention provided education about the risks and costs of smoking, the benefits of quitting, and the potential solutions to the common obstacles to quitting. The four-session Motivational Interviewing intervention used the style and methods of Motivational Interviewing⁴⁵ to foster motivation to quit smoking. Across all intervention arms, persons wanting to quit were provided assistance with behavioral strategies and choosing medication, which was offered free of charge. The three additional health education and Motivational Interviewing sessions were offered every 6 weeks after baseline. At W12 (mid-intervention) and W26 (follow-up), the participants completed measures of implicit attitudes, explicit motivation, and smoking behaviors.

Data Analysis

Analyses of the effects of the treatments on smoking outcomes are reported in detail elsewhere.³⁷ The present study analyses focused on whether changes in smoking behavior were associated with changes in implicit attitudes, regardless of treatment received.

Preliminary analyses used one-sample t tests to examine changes in IAT D scores over time, and Pearson’s correlation coefficients were used to examine associations between IAT D and explicit motivation scores at each time point, between changes in IAT D scores and explicit motivation scores, and between changes in IAT D scores and treatment group. The primary analyses used stepwise logistic regression models in which changes in implicit attitudes (from baseline to W12 or W26) were used to predict any quit attempt and cessation at W26. Change in explicit motivation was entered in the first step and the corresponding change in IAT D scores was entered in the second step. To attempt to confirm that IAT D changes preceded changes in smoking behavior (rather than changes in smoking behavior preceding IAT D change), an additional longitudinal stepwise logistic regression was conducted in which change in implicit attitudes and explicit motivation from baseline to W12 was used to predict any quit attempt from W12 to W26. To reduce small-sample bias, the Firth penalized likelihood approach was used for stepwise logistic regression models predicting cessation. Analyses were also repeated with treatment group as a covariate and with the addition of an implicit attitude X explicit motivation interaction effect; however, because results were unchanged we present only the results without treatment group or interaction effects included.

Results

Preliminary Analyses

To be included in any of the main analyses, the participants needed to have (1) IAT D scores at baseline and W12, or at baseline and W26, and (2) either quit attempt or smoking cessation data at W26. Among the 255 participants, 13 did not meet either criterion; 21 participants did not meet criterion 1; and another 11 did not meet criterion 2. Thus, 45 participants were excluded from analyses. Demographic information for participants included in the analyses is presented in Table 1. The mean age of the participants was 46 years (SD = 10.7), and they were predominantly male (57%) and African American (66%). The mean age at first cigarette use was 16 years (SD = 4.8) and the average number of cigarettes smoked per day was 16 (SD = 9.2). There were no significant differences in age, gender, ethnicity, and smoking history between those who were included in the analyses and those who were not included in the analyses.

The mean IAT D and explicit motivation scores at each time point and the mean IAT D and explicit motivation change scores are presented in Table 2. Overall, the IAT D scores tended to be negative, indicating that the participants held negative implicit attitudes toward smoking. Furthermore, IAT D scores became significantly more negative over time (t(206) = −2.46, p < .05 and t(198) = −2.10, p < .05 for W12 and W26, respectively).

Prior to the main analyses intercorrelations between smoking characteristics, IAT D and explicit motivation scores across time were examined (Table 3). Average change scores for IAT D and explicit motivation were generally not significantly related to baseline motivation to quit and nicotine dependence. Results revealed weak positive associations between average IAT D score and explicit motivation measured at the same time point with only one significant association at W26 (r = 0.16, p < .05). The average change scores for IAT D were not significantly associated with the average change scores for explicit motivation. Also, the average change scores for IAT D were not significantly different across treatment groups at any time interval.

Associations Between Implicit Attitude and Explicit Motivation Changes and Smoking Behavior

Simple correlations revealed that IAT D change scores were negatively related to smoking outcomes but only the IAT changes from baseline to W26 reached significance (Table 3). The pattern of IAT D scores across time for the participants who made any quit attempts compared with participants who did not make any quit attempts is presented in the left panel of Figure 1. The results of the stepwise logistic regression models in which explicit motivation change scores and IAT D change scores were used to predict quit attempts are presented in the top 2 sections of Table 4. In both stepwise models predicting any quit attempts, a comparison of the goodness-of-fit revealed that the addition of IAT D change (baseline to W12 or baseline to W26) in step 2 added significantly to the prediction of quit attempts over and above explicit motivation change entered in step 1 (X² [1, N = 207] = 4.88, p < .05 and X² [1, N = 199] = 4.91, p < .05, respectively). The final models revealed that for both the baseline to W12 and baseline to W26 models IAT D change and explicit motivation change were significant predictors of any quit attempt when both predictors were in the model. The results indicated that negative shifts in IAT D and explicit motivation over time were associated with higher odds of making a quit attempt.

In the longitudinal model predicting quit attempts between W12 and W26 (third section in Table 4) the results were similar although the addition of IAT D change in step 2 did not add significantly to the overall model prediction (X² [1, N = 204] = 3.61, p > .05), and the effect of IAT D change from baseline to W12 was slightly weaker and was not significant.

The pattern of IAT D scores across time for participants who were abstinent at W26 compared to participants who were not abstinent at W26 is presented in the right panel of Figure 1. The results of the stepwise logistic regression models using Firth penalized likelihood approach in which explicit motivation change scores and IAT D change scores were used to predict cessation are presented in the bottom 2 sections of Table 4. In the two stepwise models predicting cessation, a comparison of the goodness-of-fit revealed that the addition of IAT D change from baseline to W26 in step 2 added significantly to the prediction of cessation over and above explicit motivation change entered in step 1 (X² [1, N = 199] = 4.06, p < .05). In the final model examining changes in IAT D and explicit motivation from baseline to W12 as predictors of cessation, neither changes in IAT D nor changes in explicit motivation were significant predictors. In the final model examining changes in IAT D and explicit motivation from baseline to W26 as predictors of cessation, IAT D change was a significant predictor but explicit motivation change was not. In these models a negative shifts in implicit attitudes over time was associated with higher odds of cessation.

Discussion

According to the dual-process theory of cognition, explicit and implicit processes are expected to independently contribute to motivating smoking-related behavior. Until recently, most studies investigating the relationship between cognition and smoking behavior have focused on explicit attitudes.⁵ The focus of this study was on the additional contribution of implicit attitude change to predict changes in smoking behavior over and above explicit motivation change, supporting the dual-process theory of cognition in smoking behavior.

In a large smoking cessation induction trial with predominantly low-income community residents, smokers held somewhat negative implicit attitudes toward smoking, consistent with prior research.^14–18 Results indicated that smokers who had greater negative shifts in implicit attitudes at mid-intervention and at follow-up were more likely to have made quit attempts during the study period. Likewise, greater negative shifts in implicit attitudes toward smoking from baseline to W26 were significantly related to cessation at follow-up while explicit motivation failed to predict cessation. Changes in implicit attitudes from baseline to W12 were not significantly associated with cessation at follow-up, perhaps indicating the effects of implicit attitudes are stronger for quit attempts.

These results provide the first evidence of an association between changes in implicit attitudes toward smoking using the IAT and changes in smoking behavior. A recent study that examined changes in the IAT (from baseline to a few days after the quit day) and 6-month self-reported continuous abstinence among 41 smokers found no significant association.²³ The reason for the different finding is unclear but could include differences related to the timing of the assessment of change in IAT, differences in the outcome measured (ie, induction of quit attempts and 7-day point prevalence cessation), and the sample size.

The potential association between changes in implicit attitudes and smoking behavior is important because implicit attitude is only valuable as a treatment target if changes in implicit attitudes contribute to changes in behavior. Although our results are correlational, they are consistent with a potential causal association and go beyond existing evidence that has associated implicit attitude scores with smoking behaviors.^17–20,22 The present results also extend evidence of short-term effects of counterattitudinal information and practice interventions on implicit attitudes toward smoking^32–35 by showing that long-term changes in implicit attitudes during treatment relate to quit attempts and cessation. This helps translate experimental research into clinical research and practice.

Particularly encouraging from a treatment perspective is that, as hypothesized, the relationships between changes in implicit attitudes and smoking behaviors were independent of explicit motivation. We found only weak associations between implicit attitudes and explicit motivation, and there were no significant correlations between changes in implicit attitudes and explicit motivation, similar to previous studies. Changes in explicit motivation were also associated with outcomes both bivariately and in models with implicit attitude. Thus, our results support previous work that indicates implicit attitudes may be a valuable additional target for facilitating smoking behavior change.

The changes in IAT observed in this study are also consistent with the possibility that existing treatments may lead to shifts in implicit attitude. However, there were no differences in IAT changes related to treatment group and there was no pure attention control group. Shifts in IAT may have been due to other factors or changes in smoking behavior. Nevertheless, these results suggest it might be worthwhile to examine implicit attitudes as an outcome in treatment trials to enhance understanding of treatment effects.

One of the strengths of this study is that the participants were adult smokers from the general community, because prior work has tended to focus on young adults.¹⁸ Cigarette usage and health consequences differ across demographic groups, with African Americans, on average, being less successful at quitting and suffering from higher smoking-related disease mortality than Caucasian smokers.^48,49 Our replication of findings in prior studies in this sample suggests prior implicit attitude findings may be generalizable across diverse groups of smokers.

There are a number of limitations to this study including, as noted, the correlational nature of the study design, which precludes concluding that changes in implicit attitudes caused changes in smoking behavior. We conducted one model that examined if early changes in implicit attitudes were associated with subsequent quit attempts but the effect only approached significance. The reason for this nonsignificant result is unclear but could be due to a weaker effect because of the greater time lag between the occurrence of the attitude change or due to an alternative causal relationship in which quit attempts lead to changes in implicit attitudes. For example, there is evidence that implicit attitudes appear to shift in response to emotional experiences or to maintain cognitive consistency,¹⁰ both of which could be relevant when making quit attempts. To clarify these issues studies are needed that examine the effects of implicit attitude manipulation on long-term smoking cessation.

Another limitation of this study is the use of the IAT to assess implicit attitudes. Although it was selected to allow the results from this treatment trial to be compared with prior experimental work in smoking behavior, questions have been raised about the validity of the original IAT^50,51 and there is evidence that different measures of implicit attitudes are not highly correlated.^17,23 Results should be interpreted with the limitations of the IAT in mind. This study is also limited by the small number of the participants who achieved abstinence. Although this low cessation rate is to be expected in a smoking cessation trial where participants who are unmotivated to quit smoking are recruited, the results with respect to smoking cessation may be subject to some small sample bias. Results may differ in more traditional cessation trials that evaluate treatments among smokers who desire to quit.

In spite of the limitations, this is the first study to examine implicit attitudes in a large treatment trial of diverse adult smokers recruited from the community and to show significant associations between long-term changes in implicit attitudes and smoking behavior. This study provides strong support for continuing to pursue the development of interventions based on implicit attitudes for smoking cessation.

Funding

This study was supported by grant R01CA133068 from the National Cancer Institute of the National Institutes of Health and a University of Kansas Cancer Center Pilot Project Award. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Varenicline (Chantix) was provided by Pfizer through Investigator-Initiated Research Support (No. WS759405).

Declaration of Interests

None declared.

References