The Impact of Menthol Cigarette Flavor in the U.S: Cigarette and ENDS Transitions by Sociodemographic Group

Abstract

Introduction:

A better understanding of how menthol cigarette flavoring and ENDS impact smoking initiation, cessation, and transitions between tobacco products could help elucidate the potential impact of a US menthol ban in combustible tobacco products.

Methods:

A multistate transition model was applied to data on 23,232 adults from Waves 1–4 (2013–2017) of the Population Assessment of Tobacco and Health Study (analysis conducted 2020–2021). Transition rates among never, non-current, non-menthol versus menthol cigarette, ENDS, and dual everyday/someday use were estimated, as were transition-specific hazard ratios (HRs) for age, sex, race/ethnicity, education, and income.

Results:

Non-Hispanic Blacks who smoked menthol, compared with non-menthol, discontinued smoking at a much lower rate (HR=0.43, 95% CI=0.29, 0.64), but there was no statistically significant difference in discontinuation rates among non-Hispanic Whites (HR=0.97, 95% CI=0.80, 1.16) or Hispanics (HR=0.81, 95% CI=0.56, 1.16). NHWs who smoked menthol, compared with non-menthol, were more likely to become dual users (HR=1.43, 95% CI=1.14, 1.80). Young adults initiated menthol smoking at a higher rate than older adults (age 18–24 years vs ≥55 years: HR=2.45, 95% CI=1.44, 4.15) but not non-menthol smoking (HR=1.02, 95% CI=0.62, 1.69). There were differences by sex in the impact of menthol flavor on smoking initiation and discontinuation but little difference by education or income.

Conclusions:

Sociodemographic differences in product transitions should be accounted for when estimating the potential impact of a menthol ban.

INTRODUCTION

Menthol cigarettes have been associated with increased youth experimentation and initiation and with decreased adult quitting success,^1–5 and the prevalence of menthol smoking has remained nearly constant in the U.S. over the past few decades.^6,7 Menthol cigarettes are disproportionately used by youth and young adults, women, non-Hispanic Blacks (NHBs), and other minorities.^4,6,8–12

Menthol is the only flavor currently allowed in cigarettes in the U.S. Since the U.S. Food and Drug Administration Tobacco Products Scientific Advisory Committee concluded in 2011 that the removal of menthol cigarettes would result in a public health benefit,¹³ there have been increasing calls for a federal ban on menthol in tobacco products. In 2021, the Food and Drug Administration announced that it would propose tobacco product standards to ban menthol (and other flavors) in combustible tobacco products but not in ENDS.¹⁴ Rulemaking is ongoing, making this a crucial period for improving understanding of the use of menthol cigarettes influences on smoking cessation and product switching in an evolving tobacco product landscape.

The prevalence and sociodemographic profile of those who smoke menthol is well understood, and there is evidence of an impact of menthol flavoring on cigarette initiation and cessation.^1–5 However, very little is generally known about how those who smoke menthol transition to other tobacco products, nor how these transitions are impacted by the increasing prevalence of ENDS use. Developing a fuller picture of menthol’s impact on tobacco use transitions—and how the effect differs for different sociodemographic groups—is needed. To better understand how individual patterns of tobacco use (menthol and non-menthol smoking, ENDS use, and dual use) change over time and how menthol impacts these patterns, this analysis estimates transition rates using a multistate transition model¹⁵ and longitudinal data from the Population Assessment of Tobacco and Health (PATH) Study.^16–18 This approach to analyzing longitudinal data is increasingly being used in the tobacco control field^19–22 because it allows for the estimation of each transition rate simultaneously, accounting for the competing events that are the other potential transitions. Here, transition rates are estimated among cigarette, ENDS, and dual use and whether transition rates depend on menthol cigarette flavoring. This study further investigates whether the impact of menthol on these transitions differs by sociodemographic factors, including age, sex, race/ethnicity, education, and income. Understanding the impact of menthol cigarette flavoring on the rates of various transitions under the current status quo will be critical to assess the potential health and health equity impacts of a ban on menthol in combustible tobacco products in the U.S.

METHODS

Study Sample

The PATH Study is a nationally representative, longitudinal cohort study, described in detail elsewhere.^16–18 Public-use, de-identified data on adults in Waves 1–4 (2013–2017) with a Wave 4 all-waves longitudinal weight were used.¹⁸ The time between waves was approximately 1 year. Analysis was conducted in 2020–2021.

Each PATH participant was assigned to a tobacco use state in each wave based on answers to questions about established use of cigarettes (≥100 ever used) and ENDS (ever fairly regularly used), as well as everyday or someday use of cigarettes and ENDS. Participants were defined as smoking menthol if their regular brand was flavored to taste like menthol or mint.^5,23 Flavor information for ENDS was not included because it was not possible to consistently determine whether a person used mint/menthol-flavored ENDS in PATH. The Appendix provides further information and variable coding.

Measures

Seven tobacco use states were considered: never use (of either product), non-current use (of both products), exclusive use of non-menthol cigarettes, exclusive use of menthol cigarettes, exclusive ENDS use, dual use of ENDS and non-menthol cigarettes, and dual use of ENDS and menthol cigarettes (Figure 1A). Information about other tobacco products (e.g., cigars, cigarillos, smokeless tobacco) was not included, so never/non-current/exclusive use is only in the context of these 2 products. Non-current users were defined as established users (of cigarette or ENDS) who indicated that they did not currently use either product every or some days. Because this definition does not consider how long the user has been abstinent, this study uses the word “discontinuation” (as opposed to “cessation”) to denote transitions to this state. Information on age, sex, race/ethnicity, educational attainment, and income was also incorporated (Appendix). Participants with missing information for any covariate in a given wave (4.3% of remaining participants) and any participants who had only a single tobacco use state observation were removed. In total, 23,232 participants and 8,635 transitions in tobacco use state between waves were analyzed. Descriptive characteristics of the sample and sociodemographic characteristics of the participants in each tobacco product use state are given in the Appendix (Tables A1–A2).

Figure 1.

Tobacco use state definitions (A) and the direct transitions allowed between states in the model (B).

Statistical Analysis

This analysis used a Markov multistate transition model to analyze the underlying transition hazard rates and hazard ratios (HRs).¹⁵ A Markov multistate transition model is a continuous-time, finite-state stochastic process that assumes that transition rates depend only on the current state and not on the length of time in the current state, past states, or the overall transition history.²⁴ In short, a Markov multistate transition model estimates transition hazard rates, that is, the instantaneous risk of transitioning from one state to another. These transition hazard rates collectively define the probabilities of being in each state at any future time, thereby connecting the model to discrete, longitudinal data of the actual observed states. The modeled transitions in this analysis are represented in Figure 1B. Because including rare transitions can interfere with model estimation, this analysis assumed (based on prior analysis¹⁵) that there was no direct, instantaneous transition from never or non-current use to dual use, from dual use to non-current use, or from ENDS to cigarette use. Users could transition between these compartments only by transitioning through another compartment first (e.g., dual to exclusive ENDS to non-current use). Additional details are given in the Appendix.

To confirm that the model accurately represented the data, the modeled 1-wave cumulative transition probabilities were compared to the average 1-wave observed, weighted transition probabilities. Next, transition HRs were determined for each sociodemographic group in univariable and multivariable models, denoting the relative transition intensity across groups. This study used univariable models when transition rate or probability estimates were needed but used the multivariable model when comparing HRs across groups. The results for those transitions estimated to be rare and with highly uncertain sociodemographic HRs were omitted, namely exclusive cigarette (non-menthol or menthol) directly to exclusive ENDS use and switching between dual use with non-menthol cigarettes and dual use with menthol cigarettes. To understand the impact of stratifying by menthol cigarette status, this study also estimated the sociodemographic HRs for a model that did not distinguish between menthol and non-menthol smoking. Finally, the estimated univariable transition rates were used to determine whether transitions differed by menthol cigarette flavoring and how the impact of menthol differed by sociodemographic group.

This analysis accounts for PATH’s complex survey design by incorporating weights.¹⁸ PATH Wave 4 all-wave longitudinal weights were incorporated. Variance estimates were calculated using the 100 replicate weights provided by PATH.¹⁸ These models were implemented in R, version 4.0, adapted from the “msm” R package.^15,25 Example R code is available at: https://tcors.umich.edu/Resources.php.

This study provides 95% CIs for all point estimates. Comparisons of transition probabilities between groups are descriptive and are not accompanied by statistical hypothesis tests. HRs are considered statistically significant if their CIs do not contain 1.

RESULTS

Cigarette use was persistent regardless of flavor, with 89.2% (95% CI=87.3%, 91.0%) of exclusive non-menthol cigarette users and 90.6% (95% CI=88.0%, 93.1%) of exclusive menthol cigarette users remaining cigarette users after 1 wave (adding Columns 3, 4, 6, and 7 across a single row in Figure 2A). By contrast, ENDS use was less persistent, with 71.6% (95% CI=66.2%, 77.0%) of exclusive ENDS users, 52.8% (95% CI=47.2%, 58.5%) of non-menthol dual users, and 53.4% (95% CI=46.4%, 60.3%) of menthol dual users remaining ENDS users after 1 wave (adding Columns 5, 6, and 7 within a row). Many exclusive ENDS users transitioned to non-current use (21.8%, 95% CI=19.1%, 24.5%), and many dual users became exclusive cigarette users (non-menthol: 40.8%, 95% CI=38.0%, 43.5%; menthol: 39.6%, 95% CI=36.1%, 43.0%). The CIs for both the transition rates and probabilities are provided in the Appendix (Table A3).

Figure 2.

One-wave (approximately 1-year) modeled transition probabilities for (A) the adult U.S. population, (B) non-Hispanic White adults, (C) non-Hispanic Black adults, and (D) Hispanic adults.

Excl, exclusive.

Transition probabilities differed substantially by race/ethnicity (Figures 2B–D). Among NHBs, exclusive menthol smoking was more persistent (84.6% in 1 wave, 95% CI=83.0%, 86.2%) than either exclusive non-menthol smoking (83.1%, 95% CI=82.1%, 84.2%) or exclusive menthol smoking (77.5%, 95% CI=75.5%, 79.6%) among non-Hispanic Whites (NHWs), while NHB exclusive non-menthol smoking was less persistent (54.7%, 95% CI=49.7%, 59.6%). NHB exclusive ENDS use was also less persistent (46.9%, 95% CI=35.6%, 58.1%) than NHW exclusive ENDS use (60.9%, 95% CI=57.5%, 64.2%). Persistence of exclusive smoking among Hispanics was lower than the overall population and similar for non-menthol (73.9%; 95% CI=70.9%, 77.0%) and menthol (74.2%, 95% CI=71.0%, 77.4%) cigarettes. Hispanic exclusive ENDS use was less persistent (46.9%, 95% CI=40.5%, 53.5%), comparable to NHBs. Transition probabilities for other sociodemographic groups can be found in the Appendix (Figures A3–A7).

These differences in transition probabilities were driven by differences in the underlying transition hazard rates. After adjusting for other sociodemographic variables (Figure 3, Table A4), NHBs initiated non-menthol smoking at a lower rate than NHWs (HR=0.56, 95% CI=0.34, 0.91) but initiated menthol smoking at a much greater rate than NHWs (HR=4.43, 95% CI=2.73, 7.18). Similarly, NHBs discontinued non-menthol smoking (HR=1.65, 95% CI=1.14, 2.37) at a greater rate than NHWs. NHBs also switched from non-menthol to menthol smoking at a much higher rate than NHWs (HR=18.5, 95% CI=13.5, 25.4) and added ENDS use to menthol smoking at a lower rate than NHWs (HR=0.28, 95% CI=0.18, 0.43). Hispanics initiated non-menthol smoking at a lower rate than NHWs (HR=0.55, 95% CI=0.36, 0.86), but the estimate for the initiation rate of menthol smoking was not statistically significant (HR=2.40, 95% CI=0.93, 6.16). Hispanics had a higher discontinuation rate of non-menthol (HR=1.88, 95% CI=1.52, 2.33) and menthol (HR=1.65, 95% CI=1.18, 2.30) smoking than NHWs. Hispanics also switched cigarette flavors at a greater rate than NHWs.

Figure 3.

Adjusted (multivariable) hazard ratios for sociodemographic groups for selected tobacco use state transitions, comparing the effects on those transitions for non-menthol cigarettes and menthol cigarettes (middle and bottom rows) to those when not distinguishing by menthol status (top row).

Notes: Statistically significant effects are gray, and all others are black. Unadjusted (univariable) hazard ratios are available in the supplementary material (Table A4).

NH, non-Hispanic; HS, high school; BA, bachelor’s degree.

There was increased initiation in younger ages for menthol smoking initiation (age 18–24 years vs ≥55 years: HR=2.44, 95% CI=1.44, 4.15) but not for non-menthol smoking initiation (HR=1.02, 95% CI=0.62, 1.68). Men initiated smoking at a greater rate than women and that the difference was more pronounced for non-menthol smoking initiation (HR=2.24, 95% CI=1.66, 3.02) than for menthol (HR=1.61, 95% CI=1.12, 2.32). There were no significant differences in discontinuation rates by sex for menthol or non-menthol smoking. A comparison of age effects for men and women is in the Appendix (Figures A8–A10).

There was comparatively little difference in the relative impact of education or income when comparing these transitions by menthol versus non-menthol cigarette flavoring, although their impact on initiation (with lower education and income associated with greater initiation rates) was somewhat greater for non-menthol than menthol smoking initiation.

To more directly understand whether menthol cigarette flavoring impacted transition rates, this study compared the relative rates of selected transitions by cigarette flavoring. The direction and magnitude of menthol’s effect differed by sociodemographic group (Table 1). Overall, the rate of initiating menthol smoking was about 20% lower than for non-menthol smoking. However, this ratio differed dramatically between NHBs and NHWs, with NHBs initiating menthol smoking at a much higher rates than non-menthol smoking (HR=2.35, 95% CI=1.49, 3.71) and NHWs initiating menthol smoking at lower rates than non-menthol smoking (HR=0.34, 95% CI=0.24, 0.49). The preference for non-menthol cigarette initiation increased with age and was non-existent for women, unlike for men.

Table 1.

Hazard Ratios for Pairs of Transitions That Differ by Cigarette Flavor by Sociodemographic Group

		Age, years				Sex		Race			Education				Household income
	All	18–24	25–34	35–54	≥55	Male	Female	NH White	NH Black	Hispanic	Less than high school	High school	Some college or AA	BA or higher	<$25,000	$25,000–$50,000	>$50,000
Comparis On	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
Never to menthol smoking versus never to non-menthol smoking	0.80 (0.64, 0.98)	1.13 (0.77, 1.64)	1.32 (0.79, 2.21)	0.62 (0.41, 0.96)	0.46 (0.30, 0.69)	0.98 (0.74, 1.30)	0.67 (0.50, 0.90)	0.34 (0.24, 0.49)	2.35 (1.49, 3.71)	0.93 (0.57, 1.50)	0.59 (0.33, 1.03)	0.81 (0.55, 1.18)	0.70 (0.41, 1.19)	0.64 (0.27, 1.64)	0.96 (0.75, 1.22)	0.49 (0.33, 0.74)	0.70 (0.36, 1.35)
Menthol smoking to non-current versus non-menthol smoking to non-current	0.83 (0.72, 0.96)	1.02 (0.78, 1.34)	0.71 (0.55, 0.92)	0.70 (0.53, 0.92)	0.76 (0.55, 1.05)	0.98 (0.82, 1.17)	0.70 (0.57, 0.87)	0.97 (0.80, 1.16)	0.43 (0.29, 0.64)	0.81 (0.56, 1.16)	0.67 (0.44, 1.00)	0.78 (0.57, 1.08)	0.95 (0.75, 1.20)	0.73 (0.55, 0.99)	0.84 (0.70, 1.01)	0.79 (0.61, 1.03)	0.98 (0.78, 1.24)
Menthol to non-menthol smoking versus non-menthol to menthol smoking	1.68 (1.43, 1.96)	1.36 (1.00, 1.85)	1.06 (0.76, 1.49)	2.02 (1.53, 2.66)	2.08 (1.36, 3.18)	1.43 (1.15, 1.79)	1.92 (1.56, 2.37)	2.63 (2.00, 3.46)	0.22 (0.17, 0.29)	1.10 (0.78, 1.56)	1.56 (1.06, 2.30)	1.66 (1.27, 2.18)	1.59 (1.15, 2.21)	2.17 (1.02, 4.60)	1.45 (1.19, 1.76)	1.45 (1.06, 1.99)	2.57 (1.68, 3.94)
Menthol smoking todual use versus non-menthol smoking to dual use	1.04 (0.86, 1.25)	0.98 (0.68, 1.41)	0.99 (0.69, 1.41)	0.86 (0.65, 1.13)	0.98 (0.59, 1.63)	1.08 (0.85, 1.37)	0.99 (0.74, 1.32)	1.43 (1.14, 1.80)	0.48 (0.23, 1.00)	1.36 (0.79, 2.33)	0.97 (0.54, 1.72)	1.16 (0.81, 1.66)	0.96 (0.71, 1.31)	0.81 (0.44, 1.49)	1.01 (0.78, 1.30)	1.04 (0.75, 1.44)	1.11 (0.76, 1.63)
Menthol dual use to exclusive ENDS versus non-menthol dual use to exclusive ENDS	0.92 (0.64, 1.31)	0.70 (0.38, 1.30)	0.56 (0.27, 1.17)	1.05 (0.49, 2.24)	1.89 (0.44, 8,13)	1.18 (0.68, 2.06)	0.77 (0.48, 1.22)	1.03 (0.70, 1.53)	0.19 (0.02, 1.63)	0.93 (0.34, 2.55)	3.69 (0.56, 24.2)	0.50 (0.15, 1.68)	0.95 (0.50, 1.81)	1.02 (0.42, 2.50)	1.77 (1.09, 2.87)	0.55 (0.23, 1.34)	0.74 (0.39, 1.39)
Menthol dual use to menthol smoking versus non-menthol dual use to non-menthol smoking	0.98 (0.84, 1.14)	0.83 (0.57, 1.19)	1.04 (0.75, 1.47)	0.90 (0.70, 1.15)	1.28 (0.88, 1.87)	1.02 (0.81, 1.29)	0.95 (0.75, 1.19)	1.01 (0.84, 1.21)	0.59 (0.25, 1.41)	1.10 (0.69, 1.74)	1.11 (0.64, 1.95)	1.03 (0.77, 1.37)	1.16 (0.89, 1.50)	0.68 (0.36, 1.28)	0.95 (0.77, 1.17)	0.98 (0.72, 1.34)	1.00 (0.72, 1.40)

Notes: Boldface indicates statistical significance (p<0.05) (95% CIs do not include 1.0).

HR, hazard ratio; NH, non-Hispanic; AA, associate’s degree; BA, bachelor’s degree.

Exclusive menthol cigarette users had a lower discontinuation rate than exclusive non-menthol cigarette users. Again, this ratio differed dramatically between NHBs and NHWs, with NHBs discontinuing menthol smoking at a much lower rate than non-menthol smoking (HR=0.43, 95% CI=0.29, 0.64) but NHWs discontinuing menthol smoking and non-menthol smoking at similar rates (HR=0.97, 95% CI=0.80, 1.16). Older adults and male individuals discontinued menthol smoking at lower rates than non-menthol smoking.

Only NHW cigarette users had a statistically significant difference in the transition rate from menthol smoking versus non-menthol smoking to dual use (i.e., adding ENDS in addition to smoking; 1.43, 95% CI=1.14, 1.80).

Finally, there were few significant differences in the rate of transitioning from dual use with menthol smoking compared with dual use with non-menthol smoking to either exclusive ENDS use or to exclusive smoking of the same cigarette type, although there was less statistical power for these comparisons.

DISCUSSION

This study estimated the impact of menthol cigarette flavoring on transitions between cigarettes, ENDS, and dual use by fitting a multistate transition model to longitudinal data of individuals’ tobacco product use. This approach allowed the investigators to simultaneously estimate transition rates between these products and to understand how they together contributed to observed patterns of use. There were large differences by race/ethnicity in how menthol cigarette flavoring impacted cigarette and ENDS use transitions. There were moderately different impacts of age and sex on initiation and discontinuation of menthol smoking versus non-menthol smoking but comparatively little interaction between menthol cigarette flavoring and education or income.

The impact of menthol cigarette flavoring was different between NHBs and NHWs for almost all tobacco product use transitions. NHBs who smoked menthol cigarettes discontinued smoking at a 60% lower rate than NHBs who smoked non-menthol cigarettes, but there was no difference in discontinuation rates by menthol flavoring for NHWs. By contrast, for Hispanics, there was no significant difference by menthol flavoring for any of the transitions. These findings are consistent with a recent meta-analysis on the association between menthol cigarette flavoring and smoking cessation.⁵ Given the racial/ethnic differences in the impact of menthol on cessation, these results may reflect targeted marketing practices by the tobacco industry, structural barriers resulting in differential access to tobacco products and cessation services, or other factors⁵ rather than menthol’s pharmacologic effects on patterns of smoking topography and dependence.^26–29 One limitation of this study is that race/ethnicity and the other socially constructed sociodemographic variables are only indicators of the underlying causal factors and fail to address intersectional effects or capture the nuance of specific subgroups (e.g., sexual/gender minorities, Hispanic subgroups).

For most sociodemographic groups other than NHWs, menthol smoking was not significantly associated with initiation or discontinuation of ENDS products compared to non-menthol smoking, although the estimates may have been underpowered because dual use was relatively rare. Given the data were from 2013–2017, largely before the widespread use of pod-based and disposable ENDS,^33,34 and as new products will continue to emerge, it is uncertain whether these results will change in future surveys. This analysis also did not account for ENDS flavors; future work will investigate if there are associations of mint/menthol-flavored ENDS with transitions for those who smoke menthol cigarettes.

Young adults initiated menthol smoking at a much higher rate than older adults but their rates of non-menthol smoking initiation were about the same as those of older adults. That menthol smoking initiation is higher in younger adults than older adults is consistent with previous literature,^4,35 but it is surprising that there was no statistically significant increase in non-menthol initiation among young adults compared to older adults. As smoking initiation is not common at older ages, these results imply that menthol smoking was the predominant pathway for young adult smoking initiation during this time period. However, this result may be in part a cohort effect, where decreased initiation with increasing age is offset by decreasing initiation in newer birth cohorts.^36,37 If a cohort effect is reducing the age gradient for non-menthol smoking initiation, then the true age effects for menthol smoking initiation are likely even greater than estimated.

This work has several important implications for the potential impact of a menthol ban in combustible products. First, a menthol ban would likely address racial inequities in tobacco-related harms. NHBs who smoke are disproportionately less likely to discontinue smoking when using menthol cigarettes than non-menthol cigarettes. Thus, a menthol ban could result in substantial smoking cessation among NHBs who would otherwise not quit. Second, this work also suggests that a menthol ban may reduce young adult smoking initiation. Between enhanced cessation and reduced initiation, removing menthol cigarettes from the marketplace would likely result in a lower incidence of smoking overall. This work adds to the growing literature indicating that banning menthol will likely have a public health impact through increased cessation and reduced initiation.^38,39

Limitations

Overall, these results complement and further strengthen the broader literature on the impact of menthol on cessation.⁴ Nonetheless, this study has a few limitations that should be noted when interpreting these results. Discontinuation was defined as “no longer smoking every or some days” rather than long-term quitting, and so the results could be biased if shorter-term periods of abstinence that are not sustained are more common for one product than the other. This analysis also did not explicitly account for other combustible tobacco products, including cigars and cigarillos, which can be flavored and are disproportionately used by NHBs and young adults.^30,31 This analysis also did not distinguish between everyday and someday users.³²

CONCLUSIONS

Given the prevalence and the patterns of transition between menthol and non-menthol smoking and ENDS that were identified, a ban on menthol in combustible tobacco products has the potential to address health disparities in NHBs and other groups. A menthol ban in combustible tobacco products may be the policy tool—short of banning tobacco products entirely—with the most potential for substantial impact on smoking prevalence and thus tobacco-related health outcomes,¹³ particularly if paired with high-quality cessation programs and thoughtful regulation of the ENDS marketplace.