Perceptions of the Proposed Food and Drug Administration Graphic Cigarette Warnings Among Adults Who Smoke on a Nondaily or Daily Basis: A Within-Subjects Experiment

HIGHLIGHTS

• •In 2019 the Food and Drug Administration proposed new graphic cigarette packaging warnings.
• •These warnings have not been tested among adults smoking nondaily (ASN).
• •Graphic (versus text-only) warnings are more efficacious for ASN and adults smoking daily.
• •Graphic warnings lead to substantial increases in knowledge and believability for ASN.
• •Graphic warnings discourage smoking, especially for ASN versus for adults smoking daily.
• •This is important to improve public health and understand tobacco regulatory impacts.

Abstract

Introduction

People who smoke nondaily represent nearly one third of U.S. adults who smoke and experience significant smoking-related morbidity and mortality, yet they are rarely considered when evaluating the efficacy of graphic cigarette warnings.

Method

In October 2019, U.S. adults who smoked nondaily (n=59) and daily (n=183) evaluated 13 Food and Drug Administration–proposed graphic warnings and 13 equivalent text-only warnings on the extent to which they elicited understanding, perceived new knowledge, worry about smoking, and discouragement to smoke and were believable. Repeated-measures ANOVA and planned comparisons tested for between-subjects (smoking status) and within-subjects (warning type) effects. The strength of within-subjects effects were characterized using Cohen’s d_p and Common Language Effect Size. Finally, repeated-measures ANCOVAs evaluated whether demographic characteristics (e.g., race/ethnicity) that differed between adults smoking nondaily and daily were associated with within-subject evaluations of graphic versus text-only warnings.

Results

Adults who smoked nondaily (versus daily) were younger (p<0.001); more educated (p<0.001); and less likely to be non-Hispanic White (p=0.04), to be female (p=0.04), or to intend to quit within 6 months (p<0.001). Adults smoking nondaily (versus daily) rated graphic warnings as eliciting more discouragement to smoke (p=0.045). Graphic (versus text) warning evaluations were stronger for all outcomes in both groups (p<0.001). The effect sizes for graphic warnings were substantially larger among adults smoking nondaily for discouragement and believability (d_p difference >0.10). Race/ethnicity moderated evaluations of discouragement (p=0.03) and believability (p=0.01) among adults smoking nondaily and daily.

Conclusions

The capacity of graphic warnings to elicit responses associated with motivation to quit smoking may be particularly robust for adults smoking nondaily. That graphic warnings discourage smoking is promising for this underserved group of people who smoke.

INTRODUCTION

People who engage in nondaily smoking represent approximately one third of adults who smoke cigarettes in the U.S.,¹ and this proportion is growing nationally and worldwide.1, 2, 3, 4 On average, adults who smoke nondaily consume 5.5 cigarettes per smoking day, and nearly half (47.2%) smoke on 15 or more days a month.^5,6 Nondaily smoking is prevalent among several vulnerable and underserved populations, including adults with depression,⁷ young adults,^8,9 and Hispanic/Latinx and Black/African American individuals.^1,9,10

The relationship between smoking-related risks and cigarette consumption is not linear, with a jump in risk from nonsmoking to nondaily smoking and then a gradual increase in risk as smoking intensity increases.^11,12 Smoking as few as 6–10 cigarettes a month¹³ is associated with significant morbidity and mortality,^10,13, 14, 15, 16, 17 and those who smoke nondaily are at substantially higher risk of mortality from respiratory disease, heart disease, and cancer and are 1.5–2 times more likely to die from all-cause mortality than adults who do not smoke.^{10,13,14,16,18} Furthermore, a substantial proportion of people who smoke nondaily (18%–33%) are at risk of transitioning to daily smoking,^4,19,20 which further increases health risks.^{10,11,14,15,21} The evidence is clear that individuals who smoke nondaily have much to gain from completely quitting.

Although nondaily smoking has been a recognized public health issue for over a decade,²² adults smoking nondaily (versus daily) are understudied,²³ less likely to perceive themselves as vulnerable to health consequences of smoking,24, 25, 26 and often do not identify as smokers.^25,27, 28, 29, 30, 31 These perceptions may in part be explained by experiencing fewer symptoms of physiologic nicotine dependence (e.g., tolerance, withdrawal) and beliefs that they are not addicted and can easily quit when they choose.27, 28, 29^,32, 33, 34 In addition, people who smoke nondaily tend to perceive fewer benefits of quitting cigarettes³⁵ and are less motivated to quit.^32,36 Given the increased incidence of nondaily smoking, public health efforts to curb cigarette smoking should address this substantial and at-risk subset of adults who smoke.

Graphic cigarette warnings include illustrations or photographs with an accompanying text warning statement. The advantages of graphic (versus text-only) warnings are well established in the general population of adults who smoke.^37,38 Specifically, graphic warnings are associated with accurate assessments of the risks of smoking,39, 40, 41 increased message recall,^42,43 perceived new knowledge,^40,44 and motivation or intention to quit.^39,45,46 Furthermore, an international body of evidence demonstrates that implementing graphic warnings on commercial cigarette packaging is associated with increased knowledge of smoking-related diseases,^47,48 motivation to quit,⁴⁸ quit attempts, smoking cessation, and an overall reduction in smoking prevalence in the general population.49, 50, 51 However, with few exceptions (noted below), adults who smoke nondaily have largely been excluded from investigations of graphic cigarette warnings^39,46 or have not been examined separately from adults who smoke daily.^45,51,52 Given widespread misperceptions around nondaily smoking and reduced motivations to quit among this group, it is important to establish whether graphic warnings could be a particularly powerful tool to inform adults who smoke nondaily about the health harms of smoking and discourage use.

In 2009, the U.S. congress mandated that the Food and Drug Administration (FDA), the entity responsible for tobacco regulatory policy in the U.S., should develop graphic warnings for cigarette packaging. The first set of FDA-proposed graphic warnings was released in 2011. These warnings demonstrated some efficacy for people smoking on a nondaily basis. People who smoked nondaily and daily rated graphic warnings as broadly more effective than text warnings,⁵³ and those who smoked on a nondaily (versus daily) basis were more likely to report that graphic warnings would make them think about not smoking.⁵⁴ After industry litigation,^55,56 the FDA released a second set of proposed graphic warnings in 2019 intended to meet criteria set by congress for implementation on cigarette packaging. These new warnings were designed to enhance new knowledge and understanding of the consequences of smoking and to discourage U.S. adults from wanting to smoke cigarettes (https://www.federalregister.gov/d/2020-05223).

The study team previously established differences in reactions (e.g., understanding the consequences of smoking, perceived new knowledge, discouragement, worry) to these graphic warnings versus text-only warnings among adults who did and did not currently smoke cigarettes.⁴⁴ Yet, the efficacy of these graphic warnings to elicit responses associated with reduced smoking is not well understood among people smoking nondaily, and little work has addressed how such warnings may elicit proximal outcomes associated with intentions to quit such as negative affect (e.g., worry) among this group.⁵¹ Given that individuals may be sceptical of claims of health risks associated with nondaily smoking, it is also important to assess whether adults who smoke nondaily find the warnings believable and to broadly contextualize their potential public health impact. This is especially pressing given that the U.S. Supreme Court declined to hear a tobacco industry challenge to implementation (November 2024),⁵⁷ and the FDA has indicated intention to enforce the rule beginning December 2025.⁵⁸

Study Purpose and Hypotheses

This investigation is a secondary analysis of a study evaluating responses to the 2019 set of proposed FDA graphic warnings (versus text-only warnings) among U.S. adults. The authors focus on describing warning reactions among adults smoking on a nondaily versus daily basis, specifically: understanding the consequences of smoking, perceived new knowledge, worry about health harms of smoking, discouragement to smoke, and warning believability among adults smoking nondaily versus daily. Each of these outcomes has been shown to be associated with motivational changes, such as quit intentions and quit attempts.^39,40,59, 60, 61 The primary aim of this study was to characterize responses to the proposed graphic warnings (versus text-only warnings) among adults smoking nondaily versus daily. Given that in prior work, adults who did not currently smoke (versus those who did smoke) rated graphic warnings as imparting a better understanding of the consequences of smoking and as more discouraging of smoking,^40,44 the authors hypothesized that adults who smoked nondaily, who may not identify as smokers, would rate graphic warnings as more understandable and discouraging than adults who smoked daily. The authors did not have specific hypotheses regarding other graphic warning outcomes or evaluations of text-only warnings.

A secondary aim was to explore the strength of within-group differences of graphic versus text-only warning evaluations of understanding, new knowledge, worry, discouragement, and believability for adults smoking nondaily and daily by examining effects sizes.⁶² The authors expected that consistent with prior work,⁴⁴ graphic (versus text-only) warnings would be more efficacious across the whole sample. Examining the strength of graphic (versus text-only) warning effects offers insight into potential public health implications of a transition to graphic warnings for adults smoking nondaily and daily (i.e., not only is there an effect of graphic versus text warnings, but how meaningful it is). To further explore potential impacts of switching from text-only to graphic warnings, the authors also evaluated whether demographic characteristics that differed between adults who smoked nondaily and daily were associated with within-subject evaluations of graphic and text warnings.

METHODS

Study Sample

The 13 proposed FDA graphic warnings were made public in August 2019. In September 2019, U.S. adults aged ≥18 years (N=412) were recruited through CloudResearch, an online panel platform, and directed to Qualtrics, the hosting platform, for a study evaluating the proposed FDA graphic warnings and equivalent text-only warnings. Participants were compensated $2.50 for survey completion. After giving informed consent, participants were asked, How often do you smoke now, even a puff? Those who did not report current use of combustible cigarettes (i.e., never, I am not a smoker; n=152), who exclusively used E-cigarettes (n=17), or who did not respond (n=1) were excluded, resulting in a final analysis sample of 242. Participants who reported smoking cigarettes at least once a day were classified as engaging in daily smoking (n=183), and those who reported smoking less than once a month or at least once a month but less than daily were classified as engaging in nondaily smoking (n=59).

In this within-subject study, participants viewed each of the 13 FDA-proposed graphic warnings and 13 equivalent text-only warnings in a randomized order generated by Qualtrics. Examples of the graphic and text-only warnings can be viewed in Figure 1.

Figure 1.

Example graphic and text-only warnings.

Participants viewed each warning at their own pace (average viewing time of 26 seconds) and then evaluated each warning on understanding the consequences of smoking, perceived new knowledge, worry, discouragement to smoke, and believability. Detailed recruitment, inclusion criteria, and study procedures can be found elsewhere.⁴⁴ The study protocol was deemed exempt by the Washington State University Human Research Protection Program.

Measures

Basic demographic characteristics (i.e., age, gender, race/ethnicity, education), smoking-related variables, and behavioral intentions were assessed prior to warning exposure. In addition to reporting smoking patterns as described earlier (daily, nondaily), participants were asked, Have you smoked cigarettes 100 or more times in your lifetime? (Yes/No). Behavioral intentions were assessed with Are you planning to quit smoking? (Within the next month, Within the next 6 months, Sometime in the future beyond 6 months, Not planning to quit, I do not smoke). After viewing all warnings, participants indicated previous warning exposure by the question, In the past month, have you seen graphic (pictorial) warnings about the consequences of smoking cigarettes similar to what you have seen here? (Yes/No).

Warning evaluation was assessed after participants viewed each warning. All warning evaluation measures were single items on a 7-point scale (1=not at all, 7=very much/extremely). Understanding was assessed with How much does this label give you better understanding of the consequences of smoking? Perceived new knowledge was assessed with How much did you learn something new from this label that you did not know before? Worry was assessed with How much does this label make you feel worried? Discouragement to smoke was assessed with How much does this warning discourage you from wanting to smoke cigarettes? Believability was assessed with How believable is this message to you? These measures have previously shown sensitivity for warning evaluations.^40,44

Statistical Analysis

Demographics, smoking-related, and behavioral intention variables are described with means, SDs, and proportions. Independent samples t-tests, for continuous variables, and chi-square difference or Fisher’s exact tests, for categorical variables, assessed differences in demographic characteristics and prior exposure among adults who smoked on a nondaily and daily basis. Sensitivity analyses were conducted excluding participants who indicated smoking nondaily or daily but later selected I do not smoke when asked about behavioral intentions (i.e., including only those who self-identified as a smoker). Because there were no substantial differences in demographics or warning ratings (Appendix Tables 1 and 2, available online), full sample outcomes are presented.

Outcome analyses were primarily conducted using the univariate repeated-measures general linear model (GLM) procedure in SPSS 29. First, because within-person evaluations of graphic and text-only warnings had high internal consistency (graphic warning $\alpha$ range=0.92–0.97, text-only warning $\alpha$ range=0.91–0.97), the authors calculated mean ratings of each dependent measure (e.g., discouragement) across all 13 of each warning type (i.e., graphic, text only). Within the GLM procedure, a multivariate repeated-measures ANOVA tested for between-subjects (smoking status) and within-subjects (warning type) effects. Because the aim of this investigation is to examine specific warning evaluations (e.g., new knowledge) between adults who smoke nondaily and daily, for between-subjects effects, planned comparisons explored differences by smoking status (nondaily versus daily) for each response to the graphic and text-only warnings, even in the absence of a significant omnibus test. This approach is recommended when specific comparisons are the outcomes of interest.63, 64, 65

For within-subject effects, the authors then examined main effects of warning type for graphic versus text-only warnings among adults smoking nondaily and daily. Two effect size indicators—Cohen’s d and Common Language Effect Size [CLES]—were calculated to describe the relative strength of significant effects. The authors used the esci procedure⁶⁶ in jamovi 2.6 software, which utilizes calculations developed by Cousineau and Goulet-Pelletier (2021),⁶⁷ to estimate within-subject Cohen’s d_p and associated CIs. Cohen’s d_p more accurately estimates within-subject effects sizes by accounting for the correlation between observations (i.e., within-person ratings) inherent to within-subjects designs.⁶⁷ In line with best-practice recommendations by Cohen⁶⁸ and others,^62,69, 70, 71 the authors interpret these in the context of established benchmarks in psychological research (Funder and Ozer, 2019⁶⁹), which state that ∼ Cohen’s d=0.20 (r=0.10), is “an effect that is small at the level of single events, but potentially more ultimately consequential,” ∼ Cohen’s d=0.41 (r=0.20) is “an effect of medium size that is of some explanatory and practical use even in the short run and therefore even more important,” and ∼ Cohen’s d=0.62 (r=0.30) is “an effect that is large and potentially powerful in both the short and the long run” (p. 166). The authors also calculated CLES,⁷² which describes the probability that for a given dimension (e.g., understanding), a randomly chosen graphic warning would be rated more highly than a randomly chosen text warning. Examining effect sizes can indicate the likely effect of graphic warnings in the real world.⁶²

Finally, within GLM procedures, the authors used repeated-measures ANCOVAs to evaluate whether demographic characteristics that differed between adults smoking nondaily and daily were associated with outcomes at the within-subject level. Following recommendations for the use of ANCOVA in within-subject classification designs,⁷³ the authors interpreted only the within-subject × covariate interactions for each outcome of interest in these models. All demographic characteristics that significantly differed between adults smoking nondaily and daily were included as covariates for each model, and continuous variables (i.e., age) were mean centered prior to inclusion. Follow-up ANOVAs probed significant within-subject × covariate interactions, and these were plotted to reveal the form of the interaction.

RESULTS

Table 1 presents sample characteristics by nondaily and daily smoking status. Adults smoking nondaily were younger (mean age=34.10 years vs daily mean age=42.42 years, p<0.001, d=0.72), less likely to be non-Hispanic White (69.5% vs 82.0%, p=0.041) and to have smoked on more than 100 occasions (83.1% vs 99.5%, p<0.001), and more likely to be male (62.7% vs 46.4%, p=0.036) than adults smoking daily. The majority of adults smoking nondaily were college graduates (64.4%) compared with 29.0% of adults smoking daily (p<0.001). A quarter of the sample (25.2%) reported seeing similar warnings in the past month, and this did not differ by smoking status (p=0.78).

Table 1.

Sample Characteristics for Adults Smoking Nondaily and Daily

Characteristics	Total (N=242)	Nondaily (n=59)	Daily (n=183)	Test
Age, years	40.39 (12.11)	34.10 (8.12)	42.42 (12.50)	t(239)= −5.92, p<0.001, d=0.72
Gender  Female  Male  Nonbinary	118 (48.8%) 122 (50.4 %) 2 (0.8%)	22 (37.30%) 37 (62.7 %) 0 (0.0%)	96 (52.5%) 85 (46.4%) 2 (1.1%)	χ2(1)=4.42, p=0.036
Race/ethnicity  Non-Hispanic White  Black/African American  Asian  Hispanic/Latino  Other/mixed	191 (78.9%) 14 (5.8%) 11 (4.5%) 16 (6.6%) 10 (4.1%)	41 (69.5%) 6 (10.2%) 5 (8.5%) 4 (6.8%) 3 (5.1%)	150 (82.0%) 8 (4.4%) 6 (3.3%) 12 (6.6%) 7 (3.8%)	χ2(1)=4.18, p=0.041
Education  High school or lower  Vocational or Technical school/some college  College graduate or higher	40 (16.5%) 11 (45.9%) 91 (37.6%)	4 (6.8%) 17 (28.8%) 38 (64.4%)	36 (19.7%) 94 (51.4%) 53 (29.0%)	p<0.001a
Planning to quit  Within the next month  Within 6 months  Sometime in the future beyond 6 months  Not planning to quit  I do not smoke	20 (8.3%) 55 (22.7%) 107 (44.2%) 50 (20.7%) 10 (4.1%)	7 (11.9%) 7 (11.9%) 19 (32.2%) 17 (28.8%) 9 (15.3%)	13 (7.1%) 48 (26.2%) 88 (48.1%) 33 (18.0%) 1 (0.5%)	p<0.001a
Smoked $\ge$100 times in lifetime (% yes)	231 (95.9%)	49 (83.1%)	182 (99.5%)	χ2(1)=32.18, p<0.001
Past 30-day warning exposure (% yes)	61 (25.2%)	14 (23.7%)	47 (25.7%)	χ2(1)=0.08, p=0.78

Note: Boldface indicates statistical significance (p<0.05). Values are means (SDs) or percent (count). d is Cohen’s d effect size. The chi-square test for gender compares male with female gender identity only, and the chi-square test for race/ethnicity compares non-Hispanic White with all other races/ethnicities

^aFisher's exact tests were used for comparisons of education and planning to quit with smoking daily vs. nondaily.

Adults smoking nondaily (versus daily) were less likely to intend to quit in the next 6 months (23.8% vs 33.3%), and a greater proportion were not planning to quit at all (28.8% vs 18.0%; χ²[4]=33.67, p<0.001). When asked whether they were planning to quit, 15.3% of those who smoked nondaily reported “I do not smoke,” compared with 0.5% of those who smoked daily.

Table 2^67,72 displays mean warning evaluations and SDs for adults smoking nondaily and daily. The repeated-measures ANOVA revealed no significant omnibus differences in overall warning evaluations between adults smoking nondaily and daily for understanding (p=0.257, η²=0.005), knowledge (p=0.087, η²=0.012), worry (p=0.458, η²=0.002), discouragement (p=0.077, η²=0.013), or believability (p=0.544, η²=0.002). When examining the planned comparisons, there were no significant differences in evaluations of text-only warnings among adults smoking nondaily and daily (all planned comparison p>0.14). However, after viewing graphic warnings, adults smoking nondaily rated the warnings as significantly more discouraging of smoking than adults who smoked daily (mean difference=0.484, SE=0.241, p=0.046, 95% CI=0.009, 0.959). There were no other significant between-group differences on evaluations of graphic warnings (understanding p=0.313, knowledge p=0.075, worry p=0.351, believability p=0.765).

Table 2.

Within-Subjects Evaluations of Text-Only and Graphic Warnings Among Adults Smoking Nondaily and Daily

	Nondaily (n=59)				Daily (n=183)
	Text	Graphic	Cohen’s dp within (95% CIs)	CLES	Text	Graphic	Cohen’s dp within (95% CIs)	CLES
Understanding	4.78 (1.36)	5.24 (1.24)	0.35 (0.19, 0.52)	0.599	4.52 (1.54)	5.02 (1.56)	0.32 (0.24, 0.41)	0.590
New knowledge	3.74 (1.31)	4.23 (1.40)	0.35 (0.20, 0.51)	0.601	3.44 (1.38)	3.83 (1.50)	0.27 (0.18, 0.37)	0.576
Worry	4.38 (1.54)	4.90 (1.45)	0.35 (0.18, 0.52)	0.597	4.26 (1.59)	4.68 (1.55)	0.27 (0.18, 0.36)	0.575
Discouragement	4.53 (1.63)	5.09 (1.41)	0.37 (0.19, 0.55)	0.603	4.17 (1.69)	4.61 (1.67)	0.26 (0.18, 0.34)	0.573
Believability	5.21 (1.11)	5.51 (0.89)	0.29 (0.13, 0.46)	0.584	5.37 (1.30)	5.56 (1.26)	0.15 (0.06, 0.23)	0.542

Note: Outcomes are means (SDs), range 1–7, and all p<0.001. Cohen’s d_p and 95% Cis were calculated with Cousineau and Goulet-Pelletier’s⁶⁷ (2021) adjusted Λ′. CLES (Common Language Effect Size, range 0–1, calculated with Ruscio,⁷² 2008) represents the probability within a group (nondaily/daily) that a randomly chosen graphic warning evaluation for each dimension (e.g., understanding) will be rated more highly than a randomly chosen text warning evaluation.

For within-subject effects, there was a main effect of warning type on all outcomes measured (understanding, new knowledge, worry, discouragement, and believability, p<0.001) (Table 2^67,72). Graphic (versus text-only) warnings were rated as more efficacious by adults smoking nondaily (Cohen’s d_p values=0.29–0.37, average=0.34) and daily (Cohen’s d_p values=0.15–0.32, average=0.25). There was more than a 0.10 difference in point estimates for discouragement (nondaily Cohen’s d_p=0.37, daily=0.26) and believability (nondaily=0.29, daily=0.15). CLES estimates among adults smoking nondaily ranged from 0.58 (believability) to 0.60 (understanding, new knowledge, worry, discouragement), and among adults smoking daily, it ranged from 0.54 (believability) to 0.59 (understanding) (Table 2^67,72).

For the ANCOVA, all significant demographic differences between adults who smoked nondaily and daily were included as potential covariates (i.e., age, gender, race/ethnicity, educational attainment). There were no significant within-subject × covariate effects for understanding consequences (p=0.10–0.87), perceived new knowledge (p=0.20–0.89), or worry (p=0.10–0.73) for adults who smoked nondaily and daily. There was a significant within-subject × race/ethnicity effect for discouragement (F[1, 238]=4.65, p=0.03, η²=0.02) and believability (F[1, 238]=7.54, p=0.01, η²=0.03). There were no other significant interactions with covariates for discouragement (p=0.16–0.35) or believability (p=0.22–0.33).

Follow-up ANOVAs plotted race/ethnicity (non-White, non-Hispanic White) × smoking status (nondaily, daily) for graphic and text-only warning effects on discouragement and believability. Figure 2A and B displays the plots for discouragement. Non-White adults who smoked daily found the graphic warnings more discouraging of smoking than White adults who smoked daily. The pattern was not observed for text-only warnings.

Figure 2.

Mean discouragement to smoke by smoking status and race/ethnicity after viewing graphic or text-only warnings. (A) Graphic warning. (B) Text-only warnings.

Note: Response range is 1–7.

Figure 3A and B displays the plots for believability. A cross-over interaction indicated that White adults who smoked nondaily rated graphic and text-only warnings as less believable than adults of other races/ethnicities who also smoked nondaily. Ratings of believability for graphic warnings were similar among adults who smoked daily regardless of race/ethnicity, whereas believability of text-only warnings was higher among White than among non-White adults who smoked daily.

Figure 3.

Mean warning believability by smoking status and race/ethnicity after viewing graphic or text-only warnings. (A) Graphic warning. (B) Text-only warnings.

Note: Response range is 1–7.

DISCUSSION

This investigation provides empirical evidence on responses associated with motivation to quit among adults smoking on a nondaily (versus daily) basis after viewing the most recently proposed FDA graphic warnings. Consistent with hypotheses, adults smoking nondaily reported that graphic warnings elicited more discouragement to smoke than adults who smoked daily. However, unlike for adults who smoked versus did not,⁴⁴ adults smoking nondaily and daily were equally likely to report gaining understanding of the consequences of smoking.

As expected, for adults smoking nondaily and daily, graphic (versus text-only) warnings elicited significantly more understanding of the consequences of smoking, new knowledge, worry, and discouragement to smoke, and they were rated as more believable. It is probable that any given graphic (versus text) warning would be rated more highly by adults smoking nondaily (58%–60% on all key outcomes) and daily (54%–59% on all key outcomes), indicating the overall efficacy of the FDA’s proposed graphic cigarette warnings among U.S. adults who smoke. These effects may be stronger for specific graphic warnings that elicit the highest ratings (e.g., image showing a neck growth).⁴⁴

The Cohen’s d_p effect sizes for adults smoking nondaily and daily were modest to moderate. The effects for adults smoking nondaily were moderate (on average d_p=0.34, a medium effect), and effects among adults smoking daily were smaller for each dimension (average d_p=0.25). In addition, adults smoking nondaily had stronger responses to graphic warnings (versus text-only warnings) for discouragement to smoke and believability than adults smoking daily. For both of these outcomes, the Cohen’s d_p differences were >0.10, the equivalent of a small step change in effect size.⁶⁹

Thus, in practical terms, for adults smoking nondaily, graphic warnings could have modest immediate impacts, particularly for discouraging smoking. Indeed, many of those who smoked nondaily did not plan to quit, yet for this group especially, the graphic warnings elicited responses associated with quit intentions and cessation.38, 39, 40, 41, 42, 43^,46, 47, 48, 49, 50, 51, 52 For adults smoking daily, the short-term impact of graphic warnings are likely to be limited but could be consequential longer term, given their frequent repeated exposure to graphic warnings on packaging due to daily smoking. Strong reactions to even text-only warnings on cigarette packaging (i.e., concern about the health risks of smoking, believability) are associated with greater odds of quitting or reducing cigarette consumption.⁷⁴ Graphic warnings, which elicit stronger reactions than text-only warnings, are likely to have a stronger influence on outcomes associated with smoking cessation.³⁹

Race/ethnicity may also be an important factor in the interpretation of graphic cigarette warnings, particularly regarding perceptions of warning believability and discouragement among adults smoking nondaily and daily. For example, among those smoking daily, non-White adults found graphic warnings more discouraging of smoking than White adults. In addition, among those smoking nondaily, non-White adults found all warnings more believable than White adults. However, White (versus non-White) adults smoking daily found text-only warnings to be more believable. Thus, graphic (versus text-only) warnings may be particularly efficacious for non-White adults, even for those who smoke daily, and could be an important aid in counteracting the aggressive targeted tobacco industry marketing and high burden of environmental marketing exposure that people of color experience^75,76 by discouraging future smoking. Future research should explore responses to graphic cigarette warnings among adults of different racial and ethnic backgrounds and smoking frequencies to identify best strategies for communicating health harms of cigarette use and to understand tobacco regulatory impacts in these populations.

Although the sample size was small, the proportion of participants who reported nondaily smoking in the sample (24%) was similar to current national estimates and is substantially more than when the first set of FDA-proposed graphic warnings were released in 2011.^1,2,4,20 In addition, in this sample, adults smoking nondaily reflected demographic characteristics similar to those of adults included in U.S. national data collections: they were younger, more educated, and less likely to be non-Hispanic White.^9,10 The proportion of male adults among those smoking nondaily was slightly elevated in the study sample (∼63%) compared with national estimates (∼57%).⁹ The similarity between this sample and general population of U.S. adults smoking nondaily adds confidence in the generalizability of these findings.

Strengths of the study include the within-subjects design, which increases power to detect graphic versus text-only warnings effects⁷⁸ and is likely to produce effect size estimates that are reliable and have low bias⁷⁹ compared with between-subjects designs with similar sample sizes. It also aligns with the goal of providing practical information about how individuals smoking nondaily and daily may differentially react to changes in tobacco product packaging warnings in the U.S. This study is the first, to the authors’ knowledge, to specifically investigate responses to the most recently proposed FDA graphic warnings for adults smoking nondaily. Given the unique characteristics of this group, growth in this use pattern in the U.S., and clear health consequences of nondaily smoking, understanding the potential public health and tobacco control impacts of graphic warnings is crucial.

Future research should address the factors that may impact the efficacy of public health communications among adults smoking nondaily in a diverse sample. For example, repeated exposure to tobacco warnings could enhance or attenuate immediate responses to the cigarette warnings among those who smoke nondaily because they have fewer exposures than those smoking daily overtime.

Limitations

This study was an initial step in evaluating responses to the FDA graphic cigarette warnings among adults smoking nondaily and has important limitations. Several are related to the study design, which aimed to capture within-subjects evaluations of graphic and text-only warnings. First, the authors did not evaluate change in behavioral intentions due to exposure to graphic warnings. Individuals viewed 26 warnings, and it is not possible to determine whether viewing a particular warning would influence motivation or intention to quit for adults smoking nondaily. Second, warning evaluations were assessed using single-item scales, which could result in lower construct validity and reliability than multi-item scales. However, these had the important benefit of reducing participant burden and have demonstrated sensitivity in similar studies evaluating tobacco warnings.^40,77 Third, to reduce participant burden, the authors also used brief assessments of personal characteristics and of smoking behavior, and thus, detailed participant information is not available (e.g., household income), and smoking status misclassification is possible. However, sensitivity analyses revealed that excluding participants who perceived that they did not smoke (and thus did not need to quit) did not substantially change outcomes.

In addition, the study was cross-sectional, and the authors cannot assess the influence of proximal warning responses on subsequent smoking behavior or other longitudinal outcomes. Furthermore, warnings were not presented on cigarette packaging or advertisements; however, results were largely consistent with those of other studies that provided warnings on packages (e.g., Evans et al.,³⁹ 2015). Finally, although the authors examined demographic differences between adults smoking nondaily and daily as covariates of warning ratings, the small sample sizes limited full exploration of race and ethnicity as moderators of smoking status on perceptions of these warnings.

On the basis of these limitations, future work is needed among adults smoking nondaily that (1) is designed to carefully evaluate the efficacy of these warnings to increase motivation and intentions to quit and to promote reduced smoking; (2) includes additional smoking-related metrics such as frequency (e.g., days smoked in past 30 days), nicotine dependence, and past history of daily smoking; and (3) strives to include samples with diverse racial and ethnic backgrounds to extend evaluations of warning perceptions beyond a non-Hispanic White/non-White binary.

CONCLUSIONS

This study offers encouraging evidence that graphic cigarette warnings elicit responses associated with motivation to quit, especially believability and discouragement to smoke, for the understudied and substantial subset of people smoking nondaily and provides timely evidence of the impact of the currently proposed FDA warnings to improve public health because these are likely to be implemented in the near future.⁵⁷ Future proposed tobacco regulatory policy must consider this growing proportion of U.S. adults smoking nondaily to discourage smoking and maximize public health impacts of tobacco regulatory actions.

CRediT authorship contribution statement

Bethany Shorey Fennell: Writing – original draft, Methodology, Data curation, Formal analysis. Renee E. Magnan: Conceptualization, Investigation, Methodology, Writing – review & editing, Supervision. Jennifer I. Vidrine: Writing – review & editing, Resources. Damon J. Vidrine: Writing – review & editing. Linda D. Cameron: Conceptualization, Methodology, Funding acquisition, Writing – review & editing, Supervision.