Addicted to green: Priming effect of menthol cigarette packaging on brain response to smoking cues

Abstract

Introduction:

Mentholated cigarettes are believed to be more addictive than non-menthol ones. Packaging of most menthol cigarette brands includes distinctive green hues, which may act as conditioned stimuli (i.e. cues) and promote menthol smoking. To examine the cue properties of menthol cigarette packaging, we used a priming paradigm to assess the effect of packaging on the neural substrates of smoking cue-reactivity. We hypothesized that menthol packaging will exert a specific priming effect potentiating smoking cue-reactivity in menthol compared to non-menthol smokers.

Methods:

Forty-two menthol and 33 non-menthol smokers underwent functional magnetic resonance imaging while viewing smoking and neutral cues. The cues were preceded (i.e. primed) by briefly presented images of menthol or non-menthol cigarette packages. Participants reported craving for cigarettes in response to each cue.

Results:

Menthol packaging induced greater frontostriatal and occipital smoking cue-reactivity in menthol smokers than in non-menthol smokers. Menthol packaging also enhanced the mediation by neural activity of the relationship between cue exposure and cigarette craving in menthol but not non-menthol smokers. Dynamic causal modeling showed stronger frontostriatal-occipital connectivity in response to menthol packaging in menthol compared to non-menthol smokers. The effects of non-menthol packaging did not differ between categories of smokers.

Conclusions:

Our findings demonstrate heightened motivational and perceptual salience of the green-hued menthol cigarette packaging that may exacerbate menthol smokers’ susceptibility to smoking cues. These effects could contribute to the greater addiction severity among menthol smokers and could be considered in the development of science-based regulation and legal review of tobacco product marketing practices.

INTRODUCTION

Menthol cigarette smokers have a more severe nicotine addiction and greater difficulty quitting compared to smokers of non-menthol cigarettes [1]. Despite an overall decrease in smoking prevalence, the market share of menthol-flavored cigarettes in the US has continued to grow [2, 3]. Adding menthol to tobacco cigarettes enhances their addictive potential through molecular mechanisms [4]. Tobacco companies promote menthol cigarette consumption through a range of marketing strategies [5].These strategies almost invariably include using green hues on menthol but not non-menthol tobacco product packaging [6, 7]. The “fresh” and “healthy” connotations of the green color may facilitate the initial association between smoking and reward [7, 8]. For an established menthol smoker, menthol packaging provides a specific reminder of the menthol flavor and may intensify the motivation to smoke. Thus, elucidating the neurocognitive effects of menthol cigarette packaging is essential for the science-based regulation of mentholated tobacco products. The critical importance of such data is evidenced by the recent US Food and Drug Administration (FDA) announcement of an intent to ban menthol flavor in cigarettes, which will likely meet with opposition on both legal and scientific grounds [9].

Heightened reactivity to conditioned drug-related stimuli (i.e. cues) is a fundamental phenomenon in addiction [10]. Exposure to visual stimuli that have been repeatedly associated with smoking (e.g., cigarette packaging, ashtrays, cigarette smoke) can elicit conditioned responses such as an urge to smoke, which reinforces smoking behavior and increases the risk of relapse [10]. Cigarette packaging acts as a powerful smoking cue and may be designed to maximize its cue effects by modulating smokers’ expectation of cigarette flavor [6, 7]. The mesocorticolimbic pathways play an important role in cue-reactivity across addictive substances, with variable contributions from other brain systems. Specifically, numerous meta-analyses of functional magnetic resonance imaging (fMRI) research revealed brain regions that are most consistently activated by visual smoking cues, including the medial prefrontal cortex (MPFC), anterior and posterior cingulate cortex (ACC, PCC), anterior insula (AI), nucleus accumbens (NAcc), caudate, and amygdala [11-15]. Cue-induced response in these regions has been shown to predict future smoking behavior and cessation outcomes [16-20], demonstrating greater external validity and prognostic utility than self-report cue-reactivity measures [21, 22]. Circuits formed by these regions may mediate the connection between cigarette packaging effects and smoking motivation. Therefore, fMRI offers a powerful approach to examining smokers’ vulnerability to tobacco-related cues.

The present study used fMRI to investigate the cue properties of menthol cigarette packaging among menthol and non-menthol smokers. An adapted priming paradigm was utilized to examine the capacity of menthol and non-menthol cigarette packaging to modulate brain response to smoking cues compared to neutral stimuli. In this paradigm, priming effects occur when a preceding stimulus (i.e. prime) facilitates the processing of a subsequent stimulus (i.e. target) that is conceptually [23] or affectively [24, 25] congruent with the prime, and suppresses the processing of a target that is incongruent with the prime. For example, alcohol-related primes produce faster response to alcohol-related targets in alcohol-dependent patients [26]. In contrast, anti-smoking and anti-drinking warnings reduce neural response to smoking and drinking cues, respectively [27, 28]. Based on these findings as well as evidence that menthol leads to more severe nicotine addiction, we hypothesized that menthol cigarette packaging would show a greater priming effect on smoking cue-reactivity in menthol compared to non-menthol smokers, while the effect of non-menthol cigarette packaging would not differ between cohorts. We focused on brain responses in a subset of nodes of the putative smoking cue-reactivity circuit identified by prior meta-analyses: MPFC, ACC, PCC, AI, NAcc, caudate and amygdala [11-15]. Effects in other regions were explored by whole-brain analysis. We also sought to investigate the role of brain response in mediating cue-induced craving, and whether the mediated effect would be modulated by exposure to cigarette packaging. Effective connectivity analysis was conducted to explore interregional coupling across experimental conditions and cohorts.

METHODS AND MATERIALS

Participants

Eighty-two adult smokers participated in the study (see the Supplementary Information for sample size calculation and inclusion/exclusion criteria). We excluded five participants due to excessive head motion during scanning (>1 voxel) and 2 due to a high (>90%) proportion of missing behavioral responses during the fMRI task, leaving 75 participants in the final analysis (see Table 1 for demographic characteristics). Participants gave written informed consent to participate in the protocol approved by the University of Pennsylvania Institutional Review Board.

Table 1.

Participant characteristics (mean±SD or N).

	Menthol smokers	Non-menthol smokers
N	42	33
Years of age	33.02±9.51	25.55±8.23
Sex	16 male, 26 female	22 male, 11 female
Race	13 Caucasian, 20 AA 2 Asian, 7 Other	26 Caucasian, 1 AA 3 Asian, 3 Other
Hispanic ethnicity	6	1
Years of education	12.88±1.85	14.70±2.24
CPD	16.00±9.78	10.12±6.93
FTND score	5.88±1.98	3.30±2.11

Abbreviations: CPD, Number of cigarettes smoked per day; FTND, Fagerstrom Test for Nicotine Dependence; AA, African American.

Procedure

Nicotine dependence was evaluated by the Fagerstrom Test for Nicotine Dependence (FTND) [29]. Daily tobacco use was assessed by the Timeline Follow-back (TLFB) interview [30]. The TLFB used a calendar-based format to record participants’ self-reported number of cigarettes smoked per day (CPD), which was averaged across the past seven days. Between 30–45 min before fMRI, participants were escorted outdoors to smoke one of their own cigarettes under observation to achieve a uniform non-deprived state.

The fMRI priming task included smoking-related pictures (smoking cues) and smoking-unrelated pictures (neutral cues), which were preceded by pictures of either green-colored menthol or non-menthol cigarette packages. None of the unflavored cigarette packages included green hues. Cigarette brand names were blurred to minimize brand effect. This resulted in a 2 (Group: menthol vs. non-menthol smoker; between-subjects factor) × 2 (Package: menthol vs. non-menthol package; within-subjects factor) × 2 (Cue: smoking vs. neutral cue; within-subjects factor) study design (see Table S1). On each trial, a package (i.e. prime) was presented for 300 ms followed immediately by a cue (i.e. target) presented for 3000 ms. There were 12 paired-stimulus trials for each of the four possible package-cue combinations, resulting in a total of 48 trials. Inter-trial intervals varied between 1700–4700 ms in duration. Participants indicated how much they wanted to smoke a cigarette (0=not at all, 10=very much) in response to each cue. See the Supplementary Information for details of the task and fMRI data acquisition.

Statistical analyses

Using SPM 12 (Wellcome Trust Centre for Neuroimaging, London, UK), fMRI images were subjected to preprocessed and individual-level statistics that modeled each of the four package-cue pairings. Brain regions that have been most consistently activated by smoking cues, as summarized in prior meta-analyses [11-15], were defined as a priori regions of interest (ROIs): MPFC, ACC, PCC, AI, NAcc, caudate and amygdala (see Figure 1A). Three-way mixed-design analysis of covariance (ANCOVA) [31] tested the effects of Group, Package and Cue on the neural response of each ROI. The repeated measures analysis of variance module of SPSS 25 (IBM Corp, Armonk, NY) was used, with Group entered as a between-subjects factor, and Package and Cue entered as within-subjects factors [32]. The model included the main effects and the two-way and three-way interactions of Group, Package and Cue. Potential confounding variables were included in the ANCOVAs as covariates of no interest (age, age², sex, race, education, CPD, FTND; see Table 1). As seven ROIs were tested, p-values were corrected for false discovery rate (FDR) following the Benjamini–Hochberg procedure [33]. We also conducted exploratory whole-brain search by performing the ANCOVA in a voxelwise manner to investigate brain regions outside the aforementioned a priori ROIs. Significant clusters were identified at corrected p<0.05 using the threshold-free cluster-enhancement (TFCE) algorithm [34]. Additional confirmatory analyses were conducted to verify the above ROI and whole-brain ANCOVAs. First, we performed analysis of variance (ANOVA) on subsamples of menthol and non-menthol smokers that were matched in sample size and all baseline characteristics, thus negating the need to control for covariates (see Table S2). Next, although brand names were blurred in all stimuli, remaining elements of packaging were potentially attributable to a specific brand. Therefore, we repeated the above analysis procedures after excluding trials that presented the participant’s preferred cigarette brand. Details of fMRI data analysis and results of the confirmatory analyses are in the Supplementary Information.

Figure 1.

(A) A priori regions of interest that were anatomically defined using the Neuromorphometrics atlas (neuromorphometrics.com). (B–H) Group×Package×Cue interaction in each region of interest. Mean contrast values of smoking cue minus neutral cue are shown across different levels of Group (menthol vs. non-menthol smoker) and Package (menthol vs. non-menthol package), adjusted for covariates of no interest. Error bars represent standard error of mean. *, p<0.05 corrected for false discovery rate. ns, non-significant. (The figure was created by the authors.)

We performed moderated mediation analysis, adapted from Hayes [35], on brain regions that showed a significant three-way interaction in the above ANCOVA analysis. Specifically, for each region, we tested a moderated mediation model where Cue was the input variable, neural response was the mediation, Group and Package were the moderators, and craving was the outcome variable. We performed dynamic causal modeling (DCM) of the fMRI data to examine interregional effective connectivity. DCM uses Bayesian inference procedures to quantify the causal influences between brain regions and the effects of experimental inputs (e.g., conditions) [36, 37]. Details of the moderated mediation and DCM analyses are included in the Supplementary Information.

RESULTS

Behavioral results

Menthol and non-menthol smokers differed significantly on age (t(73)=3.58, p=6e–4), sex (χ²(1)=6.04, p=0.014), race (χ²(3)=22.57, p=5e–5), education level (t(73)=−3.84, p=3e–4), CPD (t(73)=2.92, p=0.005), and FTND score (t(73)=5.44, p=7e–7), but not Hispanic ethnicity (χ²(1)=2.77, p=0.096) (see Table 1). These unmatched variables (including a quadratic term for age) were included as covariates of no interest in all subsequent analyses. Three-way ANCOVA (Group×Package×Cue) showed higher craving for smoking compared to neutral cues (5.72 vs. 4.63, F(1,64)=25.42, p=4e–5, partial η²=0.28). Other main effects and interactions were not significant (Fs<3.81, ps>0.055, partial η²<0.06).

Regions of interest

ANCOVA on the ROIs (see Figure 1A) showed a significant Group×Package×Cue interaction in the MPFC, ACC, NAcc and caudate (F(1,64)=6.30, 6.82, 5.63 & 6.55, FDR-corrected p=0.034, 0.034, 0.036 & 0.034, partial η²=0.090, 0.10, 0.08 & 0.09) (see Figure 1B-H). Planned contrasts showed that the three-way interaction was driven by a significant Group×Cue interaction in the “menthol package” conditions (F(1,64)=5.34, 10.14, 8.74 & 6.31, FDR-corrected p=0.034, 0.015, 0.015 & 0.034, partial η²=0.08, 0.14, 0.12 & 0.09), such that after being primed with menthol packages, menthol smokers showed greater cue-reactivity (smoking cue vs. neutral cue, F(1,64)=7.91, 23.75, 15.41 & 10.37, FDR-corrected p=0.009, 5e–5, 7e–4 & 0.005, partial η²=0.11, 0.27, 0.19 & 0.14) than non-menthol smokers (Fs<1.54, FDR-corrected ps>0.49, partial η²s<0.02). For the “non-menthol package” conditions, there was no group difference in cue-reactivity (smoking cue vs. neutral cue, Fs<2.59, FDR-corrected ps>0.70, partial η²s<0.04). There was no significant main effect or interaction in the PCC, AI or amygdala (Fs<4.92, FDR-corrected ps>0.07, partial η²s<0.07).

Whole-brain analysis

Whole-brain ANCOVA showed a significant Group×Package×Cue interaction in a frontostriatal cluster encompassing the medial (k=17334 mm³, Z=3.24, x/y/z=−3/53/20) and dorsolateral prefrontal cortex (PFC) (Z=3.86 & 2.99, x/y/z=−24/56/26 & 24/53/32) extending to the adjacent ACC (Z=3.23, x/y/z=−3/44/17) and striatum (Z=3.03, x/y/z=−6/8/−1), and a second cluster in the occipital cortex (k=2673 mm³, Z=3.99, x/y/z=6/−82/−1) (corrected p<0.05; see Figure 2). The three-way interaction was driven by a significant two-way Group×Cue interaction in the “menthol package” conditions (corrected p<0.05). Specifically, menthol smokers showed greater cue-reactivity (smoking cue vs. neutral cue) than non-menthol smokers after being primed with menthol cigarette packages (corrected p<0.05).

Figure 2.

(A) Brain regions showing significant Group×Package×Cue interaction in whole-brain analysis (cluster-level corrected p<0.05; L/R=left/right). (B–C) Significant Group×Package×Cue interaction in the frontostriatal and occipital clusters revealed by whole-brain analysis. Mean contrast values of smoking cue minus neutral cue are shown across different levels of Group (menthol vs. non-menthol smoker) and Package (menthol vs. non-menthol package), adjusted for covariates of no interest. Error bars represent standard error of mean. *, whole-brain cluster-level corrected p<0.05. ns, non-significant. (The figure was created by the authors.)

Moderated mediation

Moderated mediation analysis (Figure 3A-B) was performed on six brain regions that showed a significant Group×Package×Cue interaction, including four a priori ROIs (MPFC, ACC, NAcc and caudate) and two clusters from the whole-brain analysis (the frontostriatal and occipital clusters). There was a significant total effect of Cue on craving (c=1.105, bootSE=0.213, p<0.0001). All six regions showed a significant direct, unmediated effect of Cue on craving, accounting for >99% of the total effect (c’=1.094–1.104, bootSE=0.210–0.213, FDR-corrected ps<0.0001). The neural responses of the a priori MPFC, ACC, NAcc ROIs and the frontostriatal and occipital clusters were significantly correlated with craving (b=0.042–0.065, bootSE=0.013–0.022, FDR-corrected ps<0.021). A significant moderated mediation effect was found for the a priori MPFC, ACC and NAcc ROIs and the frontostriatal and occipital clusters (a₁×b=0.015–0.034, bootSE=0.009–0.018, FDR-corrected ps<0.040). The moderated mediation was driven by a larger indirect effect (i.e. Cue→brain activity→craving) in menthol than non-menthol smokers only when cues were primed by menthol packages (FDR-corrected ps<0.049), but not when they were primed by non-menthol packages (FDR-corrected ps>0.37) (see Figure 3C-H). The a priori caudate ROI did not show significant correlation with craving (b=0.017, bootSE=0.022, FDR-corrected p=0.19) or moderated mediation (a₁×b=0.006, bootSE=0.008, FDR-corrected p=0.19). Additional results from the moderated mediation analysis are reported in the Supplementary Information.

Figure 3.

(A–B) Moderated mediation model presented in the conceptual form and the statistical form, respectively (adapted from Hayes, 2015). (C–H) Indirect effect of Cue (drug cue vs. neutral cue) on craving through cue-induced neural responses. Estimated indirect effects are shown across different levels of Group (menthol vs. non-menthol smoker) and Package (menthol vs. non-menthol package), adjusted for covariates of no interest. Error bars represent bootstrap standard error. ROI, region of interest. */**, p<0.05/0.01 corrected for false discovery rate. ns, non-significant. (The figure was created by the authors.)

Dynamic causal modeling

DCM (Figure 4) showed that across both cohorts of smokers, there was positive, stimulus-independent connectivity from the occipital cluster to the frontostriatal cluster (posterior p=1.00). Smoking cues, regardless of the preceding package category, attenuated the feed-forward connectivity from the occipital cluster to the frontostriatal cluster, while enhancing the top-down connectivity from the frontostriatal cluster to the occipital cluster (posterior ps=1.00). There was also a positive driving effect of smoking cues on the activity of the occipital cluster (posterior ps=1.00). Group comparison showed stronger frontostriatal-occipital connectivity in menthol compared to non-menthol smokers when exposed to smoking cues primed by menthol packages (posterior p=0.99). Menthol smokers also had lower self-inhibitory connectivity (i.e. greater excitability) in both regions than non-menthol smokers did (posterior ps=1.00).

Figure 4.

Results of dynamic causal modeling (DCM). (A) Common effects across both menthol and non-menthol smokers. (B) Left, differential effects between menthol and non-menthol smokers; right, frontostriatal-occipital connectivity (obtained from Bayesian model average) during smoking cue trials as a function of Group (menthol vs. non-menthol smoker) and Package (menthol vs. non-menthol package). Only parameters that had a posterior probability greater than 0.95 were retained (Zeidman et al., 2019). Details of the DCM analysis can be found in the Supplementary Information. (The figure was created by the authors.)

DISCUSSION

We found that priming with menthol cigarette packages led to a greater frontostriatal (e.g., MPFC, ACC, NAcc, caudate) and occipital response to smoking cues in menthol vs. non-menthol smokers. Moreover, neural activity in these brain regions mediated the effect of cues on craving, and such mediation was stronger in menthol compared to non-menthol smokers when exposed to menthol packaging. Priming smoking cues with menthol packaging led to stronger top-down effective connectivity from the frontostriatal area to the occipital cortex in menthol compared to non-menthol smokers. The priming effects of the non-menthol packaging did not differ between cohorts.

Cue-reactivity has been studied extensively using both self-report (e.g., craving) and objective probes (e.g., physiological and neural responses) [38, 39]. The effect of cue exposure on craving intensity is well established. However, the subjective nature of craving limits the robustness of the self-report approach to measuring cue-reactivity [21, 22]. Neuroimaging markers of cue-reactivity complement self-reports and have shown promise as diagnostic [40] and prognostic tools for substance use disorders [39]. The neural substrates of smoking cue-reactivity have been summarized in multiple meta-analyses [11-15]. Among treatment-seeking smokers, prefrontal and limbic response to smoking cues has been shown to predict future relapse in most [16-19] but not all [20] studies. We found that menthol and non-menthol smokers differed in the priming effect of menthol cigarette packaging on neural cue-reactivity. Self-reported craving did not show such an effect, although the validity of our craving provocation task was confirmed by an overall cue effect on craving across cohorts and primes. When we treated neural response as a mediator, being a menthol smoker and exposure to menthol packaging significantly increased the indirect effect of cues on craving. However, the indirect effect only accounted for a small proportion of the total cue effect on craving. As can be seen in Figure 2B-H and Figure 3B-C, smoking cues accounted for less than 1% of the total variance of fMRI signal in any region. In addition, the effect of cues on neural responses was relatively small compared to their effect on craving (see Table S3). Therefore, the weak indirect effects in the moderated mediation analysis may have resulted from the limited sensitivity of neural responses to cues. Another possible interpretation for the small indirect effects is the limited reliability of self-reported craving whose variance may be attributed to unrelated factors such as social desirability bias and poor insight [21, 22]. Objective methods like implicit behavioral measures [41], electrophysiology [27] and fMRI may be better suited to detect priming effects that are usually subtle and can be obscured by self-report and deliberation [42, 43].

The menthol smokers showed heightened priming effect of menthol cigarette packaging in the prefrontal and striatal areas (MPFC, ACC, NAcc, caudate) but not in the PCC, AI and amygdala. These results were corroborated by the whole-brain analysis, which identified a cluster encompassing the four frontostriatal ROIs and extending to the adjacent dorsolateral PFC. It suggests that the frontostriatal section of the smoking cue-reactivity circuit may play a more prominent role in mediating packaging effects. The whole-brain analysis also found a group difference in priming effect in the occipital cortex. Although not typically considered part of the reward system, the occipital cortex is activated by motivationally salient stimuli including visual drug cues [44] and monetary reward cues [45]. The activation likely reflects value-driven attentional capture that biases visual attention towards reward-associated stimuli [46].

The occipital cortex has bidirectional anatomical connections with the PFC [47], which is in turn densely connected to the dorsal and ventral striatum [48]. These connections may serve as the basis for the dopaminergic system to exert top-down, value-driven modulation on visual processing in the occipital cortex [46, 49]. Our effective connectivity analysis showed that while smoking cues increased occipital activity, they attenuated the feed-forward occipital-frontostriatal connectivity and increased top-down frontostriatal-occipital connectivity. Such cue-induced top-down modulation may indicate a shift from stimulus-driven processing to value-driven processing [46]. Most interestingly, compared to non-menthol smokers, menthol smokers showed even stronger frontostriatal-occipital connectivity during presentation of menthol packaging and smoking cues. This finding suggests that menthol packaging facilitates top-down attentional control and enhances value-driven visual processing of smoking cues among menthol smokers.

Colors have long been postulated to have innate significance that is modulated by experience [50]. Commercial advertisers take advantage of color preferences to signal characteristics of product content, causing the color to become a reminder and potentially a conditioned stimulus of the content after repeated consumption [51]. The effects of packaging colors on purchasing behavior have been recognized in tobacco packaging design [6, 7]. In the general culture, green is associated with freshness and health, e.g., organic product labels or names of environmental movements. The color green has been used almost exclusively in the packaging for menthol cigarettes [6, 7]. In a non-smoker or a non-menthol smoker, green-hued packages’ appeal is based on its general cultural meaning, similar to the use of light colors (e.g., silver) to suggest reduced harm in “light” cigarettes [52, 53]. In an addicted menthol smoker, this nonspecific general appeal is likely to be surpassed by the conditioned association between a green-hued package and expected menthol and nicotine reward. Therefore, the degree to which green-hued cigarette packaging could be considered misleading may vary across audiences. This distinction is important because appellate courts have been more likely to uphold regulations on commercial speech where the governmental purpose was to protect consumers from manipulation [54]. Neuroimaging research could help disentangle the mechanisms of packaging appeal in the current vs. potential menthol smokers. Prospective experimental studies could help differentiate between innate, culturally acquired and specifically conditioned associations in product labeling and help define permissible and manipulative commercial speech in product labeling. Until such data are available, banning menthol as a characterizing flavor in cigarettes could be legally more feasible than controlling certain colors in tobacco packaging. However, experience from the menthol flavor ban in Canada suggests that positive response to green hues may extend across mentholated tobacco products and even be transferred to non-menthol products through advertising and marketing methods [55].

Menthol is the only characterizing flavor in cigarettes still allowed by the FDA (21 U.S.C. § 387g). Although not generally considered a psychoactive agent, menthol exacerbates nicotine addiction in animals [56, 57] and humans [1, 58] through a number of mechanisms (reviewed in [4]). For example, menthol increases nicotine-induced dopamine neuron excitability [56], intensifies nicotine withdrawal symptoms [59], and enhances response to conditioned response to nicotine cues [60]. Menthol also activates cold sensitive TRPM8 receptors in the oral and nasal mucosa and reduces the irritation response to cigarette smoke [61]. The presence of green color on a cigarette package almost unambiguously leads to an expectation of menthol. After repeated pairing of menthol packaging and menthol cigarette consumption, it is possible that responses to such packaging resemble the responses to menthol itself. Although it remains unclear whether menthol-related visual cues can reproduce the pharmacological effects of menthol, there is evidence that many of those effects can be induced non-pharmacologically. For example, cold sensation, which is mediated by menthol receptors TRPM8, can be induced by expectation alone, without cold stimulation or cooling agent [62]. Expectation can also potentiate the reinforcing effect of smoking [63] as menthol does [58]. In addition to expectancy, menthol cues may elicit conditioned responses via classical conditioning [64]. Interestingly, the neural correlates of expectation and conditioning largely overlap with the smoking cue-reactivity circuit [45, 65, 66]. It would be interesting for future research to examine whether packaging exposure alone is sufficient to trigger menthol-like pharmacological effects on nicotine reward.

Our study had a number of limitations. First, because we studied chronic smokers with established flavor preference, we could not determine whether the greater priming effect of menthol packaging was the cause or consequence of menthol smoking. This could be addressed by a longitudinal study of non-dependent individuals (e.g., adolescents) at risk for smoking. Second, participants in the study were all current, non-treatment-seeking smokers. Prior studies on treatment-seeking individuals have produced mix findings regarding whether neural cue-reactivity predicts treatment success [20, 67] or failure [17-19]. Future studies on treatment-seeking and former smokers could shed the light on relationship between packaging exposure, craving, and treatment outcomes. Third, as we did not perform an a priori power calculation for the moderated mediation and effective connectivity analyses, these findings should be considered exploratory and require future replication. Fourth, we did not examine other components of cigarette packaging such as descriptors (e.g., “silver”, “smooth”) and warnings (e.g., text, graphic). In addition, we treated the demographic variables between the two cohorts as covariates of no interest and could not elaborate their direct impact on brain response. A more comprehensive analysis of packaging design and smoker characteristics will help determine their joint contribution to tobacco use disorder.

Increased prevalence of menthol smoking among the disadvantaged and high-risk groups (e.g., African Americans [68], adolescents and young adults [69], and pregnant women of low socioeconomic status [70]) underscores the importance of investigation into the neurobiology of the contribution of menthol tobacco packaging to the burden of nicotine addiction. Here we demonstrate that menthol smokers are more vulnerable to menthol cigarette packaging than smokers of non-menthol cigarettes, as evidenced by a stronger priming effect on the frontostriatal and occipital response to smoking cues. Increased frontostriatal-occipital connectivity in menthol smokers exposed to menthol packaging may serve as the underlying mechanism for reward-driven visual attention to menthol tobacco products. Any FDA regulation restricting tobacco product packaging would likely be challenged in court by tobacco companies as a violation of the First Amendment’s protections for commercial speech. Whether such a regulation would survive legal review would depend in part on the quality of the scientific evidence that supports it [54, 71]. Our study demonstrates that neurocognitive research on cigarette packaging may inform science-based regulation of the marketing of tobacco and other products [72].

WHAT THIS PAPER ADDS.

• Our study demonstrates how neurocognitive research could inform science-based regulation of the marketing of tobacco and other products and aid the legal reviews of such regulations.

• Images of green-hued menthol cigarette packages have stronger priming effect on the brain frontostriatal and occipital response to smoking cues in menthol smokers than in non-menthol smokers, and this effect generalizes across menthol cigarette brands.

• The effect of non-menthol cigarette packaging did not differ between menthol and non-menthol smokers.

• Menthol cigarette packages may have greater motivational and perceptual salience in menthol smokers compared to non-menthol smokers.

• Increased frontostriatal-occipital connectivity in menthol smokers exposed to characteristic menthol packaging may serve as the underlying mechanism for reward-driven visual attention to menthol tobacco products.