Parsing Intra- and Inter-Individual Covariation Between the Sensory Attributes and Appeal of E-Cigarettes: Associations and Gender Differences

Abstract

Introduction

Perceived sensory attributes of e-cigarettes may associate with their appeal. However, limited studies have accounted for individuals’ variability in sensory attributes or have addressed how associations of sensory attributes with appeal may differ by gender.

Methods

Individuals (n = 119, 32.8% female) who currently used combustible cigarettes and/or e-cigarettes attended one laboratory session in which they completed a standardized e-cigarette puffing procedure according to a 10 Flavor (green apple, strawberry, chocolate, vanilla, menthol, koolada, peppermint, spearmint, subtle tobacco, and full-flavored tobacco) × 2 Nicotine Formulation (free-base, salt) double-blind factorial design. The mean nicotine concentration was 23.4 (SD = 0.9) mg/mL in the nicotine salt formulations and 23.8 (SD = 1.7) mg/mL in the free-base formulations. Following each trial, participants completed ratings of sensory attributes (sweet, smooth, cool, bitter, harsh) and appeal (mean of liking, disliking [reverse-scored], and willingness-to-use-again ratings). Sensory attributes were partitioned into between-person and within-person variables. Gender was tested as a moderator of associations of sensory attributes with appeal.

Results

Sweet, smooth, and cool sensory attributes positively associated with appeal at the between- and within-person level (ps < .001). Bitter and harsh negatively associated with appeal at the between- and within-person level (ps < .001). The associations of between-person sweet, smooth, and cool ratings with appeal was larger in males compared to females. The associations of within-person smooth, bitter, and harsh with appeal was larger in females compared to males.

Conclusions

This study showed important gender differences in associations of sensory attributes and appeal.

Implications

While evidence suggests that sensory attributes may contribute to the appeal of e-cigarettes, there is little experimental evidence accounting for individual variability in sensory attributes and whether sensory attribute-appeal associations differ by gender. The current study provides evidence that average sweet, cool, and smooth ratings positively associated with appeal and that these associations were larger in males. Within-person bitter, harsh, and smooth ratings significantly associated with appeal in both genders, but these associations were larger in females compared to males. Data from the current report reinforces the need for researchers to study gender stratified effects in tobacco regulatory science.

Introduction

Since 2010, the prevalence of e-cigarette use has increased among youth and adults.^1,2 However, estimates from the 2020 National Youth Tobacco Survey show a decline in e-cigarette use in adolescents and young adults.^3,4 Initial observations suggest that the appeal of e-cigarette product characteristics may differ between males and females, including gender differences in appeal and use of flavors.^5–7 As the marketplace of e-cigarettes continues to expand and evolve, it is essential that we proactively understand drivers of e-cigarette product appeal and identify sex/gender differences in appealing e-cigarettes characteristics to ensure that tobacco product regulation equitably protects both males and females.

E-cigarette flavors and nicotine concentrations have been shown to contribute to the appeal of e-cigarettes^8,9 and are important targets for tobacco regulatory policies. One way that flavors and nicotine may contribute to e-cigarette appeal is through alterations in sensory attributes. In samples of individuals who use tobacco products (ie, e-cigarette and/or combustible cigarette), laboratory studies have shown that nicotine increases perceived irritation and bitterness of e-cigarettes.^10–12 Nicotine formulation may also influence sensory properties of e-cigarettes with salt nicotine formulations being rated as sweeter and smoother, and less harsh and bitter than free-base nicotine formulations.¹³ Laboratory studies have noted that tobacco users perceive fruit (eg, cherry) and desert (eg, chocolate) flavored e-cigarettes as sweeter, smoother, and less bitter than tobacco flavors and flavorless.^9,11,12,14 Hence, subjective sensory attributes might be an important proximal indicator of how the user experiences e-cigarettes with different product characteristics, including flavor and nicotine type/concentration.

There is some evidence that variation in sensory attributes of e-cigarettes across different products may be important drivers of an e-cigarette product’s relative appeal. Studies have shown that perceptions of sweetness, coolness, and smoothness associate with increased appeal of flavored e-cigarettes.^9,11,12,14 Perceived bitterness and irritation have been shown to associate with a decrease in appeal.¹¹ Research shows that sensory attributes not only correlate with product appeal, but also mediate differences in user appeal across e-cigarette products with varying nicotine content and flavor.¹² These studies suggest that sensory attributes influence the appeal of e-cigarettes and may be a useful way to characterize meaningful subjective responses to inter-product variation, but important gaps remain in our understanding of associations of sensory attributes and appeal.

Prior studies have not fully disentangled between-person and within-person associations of e-cigarette sensory attributes with appeal. This is important because it is likely that the variance in sensory ratings of e-cigarettes is not explained solely by inter-product differences. Rather, it may be that regardless of an e-cigarette’s product characteristics some individuals may tend to find e-cigarettes as sweeter, smoother, harsher, or more bitter than other individuals. Individual differences in sweet preference have been shown to associate with increased intake of sugary beverages¹⁵ and use of e-cigarettes for weight control.¹⁶ In people who smoke, between-person differences in the 6-n-propylthiouracil (PROP) phenotype, a stable heritable trait that influences between-person variance in taste sensitivity and preferences, has been associated with liking of menthol e-cigarettes¹¹ and is more often reported in people who smoke menthol combustible cigarettes compared to people who smoke nonmenthol combustible cigarettes.¹⁷

Another gap in the current literature is understanding whether associations of sensory attributes and appeal of e-cigarettes are moderated by gender. Evidence indicates that the sensory attributes of cigarette smoking may play a stronger role in the uptake and persistence of smoking in females than males,¹⁸ and this gender difference could extend to e-cigarette use. One study showed that females who smoke cigarettes rated e-cigarettes of their nonpreferred flavor (tobacco, menthol) as less likable compared to males who smoke cigarettes.¹⁹ Similarly, menthol-flavored e-cigarettes had greater appeal/liking than tobacco flavored⁵ and unflavored e-cigarettes²⁰ in young adult females who use e-cigarettes or cigarettes, respectively, but no differences between menthol and tobacco were found in young adult males who use e-cigarettes or cigarettes. Another study found that females (vs. males) who smoke combustible cigarettes reported stronger sensory attributes at lower menthol e-cigarette concentrations.²¹ These studies support the premise that gender may importantly moderate associations of sensory attributes and appeal of e-cigarettes; however, this question has not been tested.

This secondary analysis study examined associations between e-cigarette sensory attributes and appeal from a laboratory experiment in which current tobacco product users rated each of 20 different e-cigarette products in various flavors and nicotine formulations.¹³ The first aim of the current study was to test the associations of between- and within-person sensory attributes with e-cigarette appeal in the overall sample. Based on evidence that individual differences in sweet-liking and taste phenotypes, we hypothesized that individuals with on average higher sweet, smooth, and cool ratings would report greater e-cigarette appeal, and individuals with on average higher bitter and harsh ratings would report lower e-cigarette appeal. Prior evidence has found that products with characterizing flavors that tend to be more sweet, smooth, and cool (eg, fruit, dessert, menthol) are perceived as more appealing than other flavors (eg, tobacco) and the presence of nicotine both of which are liable to increase perceptions of bitterness and harshness are perceived as less appealing.^{9,11,12,14,21,22} Therefore, we hypothesized that within-person variation in which e-cigarette administrations are rated as sweeter, smoother, and cooler than a person’s usual rating would associate with greater appeal and administrations rated as more bitter and harsh would associate with decreased appeal. The second aim of this study was to investigate whether the associations of between- and within-person sensory attributes with appeal were moderated by gender. Because prior research suggests that the associations of sensory attributes of tobacco products may be more important in females who use tobacco products compared to males who use tobacco products,^18,19 we hypothesized that the associations of between- and within-person sensory attributes with appeal would be greater in females than in males.

Methods

Participants

Individuals from the Los Angeles area were recruited to participate in a single-visit randomized clinical trial conducted in an academic medical center outpatient research facility in Southern California.¹³ Inclusion criteria required past 30-day use of e-cigarettes (lifetime e-cigarette duration >2 months; nicotine-containing e-cigarette use >3-day/week over the past 30 days) and/or combustible cigarettes (cigarette use >3-day/ week for >2 years; interest in trying e-cigarettes if not also using e-cigarettes) with NicAlert >10 ng/mL (ie, Level 1–6). Exclusion criteria were planning to quit using tobacco products in the next 30 days, currently or planning to become pregnant/breastfeeding, current daily use of tobacco products other than combustible cigarettes or e-cigarettes, and positive results of breath alcohol test at study visit. Participants provided written informed consent and were enrolled from November 2019 to March 2020. Participant accrual was halted prematurely due to coronavirus pandemic and trial registry information was updated on ClinicalTrials.gov in May 2020. The University of Southern California Institutional Review Board approved the study. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline for randomized clinical trials.²³

Research Design

A within-participant randomized double-blind design was used. The procedure used custom e-cigarette solutions in 10 flavors (green apple, strawberry, chocolate, vanilla, menthol, koolada [nonmenthol cooling agent], peppermint, spearmint, subtle tobacco, and full-flavored tobacco). Each flavor was administered in nicotine salt (benzoic acid added, mean nicotine = 23.4 (SD = 0.9) mg/mL) and free-base formulations (mean nicotine = 23.8 (SD = 1.7) mg/mL) with 50/50 propylene glycol/vegetable glycerin vehicle (Avail Vapors). Each solution was administered via a pod-style e-cigarette (Suorin iShare Pod System Device; 7W; 2.0-Ω resistance; 130-milliamp hour built-in battery).

Procedure

Participants completed an initial telephone eligibility screen. Participants then scheduled a single 3-hour in-person study visit and instructed to abstain from using nicotine or tobacco products for 2 hours before arrival.¹³ Participants completed an informed consent and provided alcohol breath samples (BACtrack S80, BACtrack). Test strips (NicAlert; Jant Pharmacal Corporation) that provide a semiquantitative index of salivary or urine cotinine and exhaled carbon monoxide (Vitalograph) were collected to measure nicotine and combustible tobacco exposure, respectively.

Participants then completed a practice trial (flavorless nicotine-free) and trial blocks spaced at least 10 minutes apart. Questionnaires were completed between blocks. Participant completed a total of 20 trials (10 flavors presented in 2 nicotine formulations) of a controlled guided puffing procedure, which involved a 1-puff cycle (10-second preparation, 4-second inhalation, 1-second hold, and 2-second exhale interval) for each product immediately followed by appeal and sensory attribute ratings.^8,9,12,13 E-cigarette administrations were double blind and presented in a random order. To minimize carryover effects of repeated administrations, participants were given water between each trial.

Measures

Demographics and Tobacco Product Use Characteristics

Participants completed questionnaires that assessed demographics including reporting their gender (“male”, “female”, “other”), gender identity “With which gender identity do you most identify? (Genderqueer can typically be defined as denoting or relating to a person who does not subscribe to conventional gender distinctions but identifies with neither, both, or a combination of male and female genders).” (“Man”, “Woman”, “Transgender man (female-to-male; FTM”, Transgender woman (male-to-female; MTF)”, “Nonbinary/genderqueer”, “Something else (please specify)”, “Decline to respond”), race (“American Indian or Alaskan Native”, “Asian (including the Philippine Islands, Southeast Asia, and India)”, “Black or African American”, “Middle Eastern”, “Pacific Islander (including Hawaii)”, “White”, “Other”, “Decline to respond”), and ethnicity (“Hispanic or Latino”, “Not Hispanic or Latino”). Participants also completed a history of tobacco product use. Individuals who currently used e-cigarettes completed the Penn State Electronic Cigarette Dependence Index,²⁴ which measured current e-cigarette dependence severity (range 0-20). Individuals who currently smoked completed the Fagerström Test of Cigarette dependence,²⁵ which assessed combustible cigarette dependence (range 0–10).

Sensory Attributes

Following each controlled e-cigarette administration, participants rated the sensory attributes of the product they just vaped on a visual analog scale (range 0-100) with rating anchors of not at all to extremely. Sensory attributes included: “how sweet was the e-cigarette”; “how smooth was the e-cigarette”; “how cool or cooling was the e-cigarette”; “how bitter was the e-cigarette”; and “how harsh was the e-cigarette”. Similar sensory attributes have been used in prior work.^8–10

Appeal

After each e-cigarette administration, participants answered three questions that assessed the appeal of the preceding e-cigarette administration: “How much did you like the e-cigarette?”; “How much did you dislike the e-cigarette?”; “Would you use this e-cigarette again?”. All questions were rated on visual analog scales (range: 0 to 100). Like and dislike had the anchors of not at all to extremely. Willingness to use again used anchors of definitely not to definitely. As in prior work,^5,13 dislike was reverse-scored and a composite appeal score was calculated based on the mean of the three ratings for each trial. Liking and disliking were assessed as two unipolar items based on the widely used Drug Effects Questionnaire, which has shown good psychometric support and distinct construct validity of liking and disliking.²⁶

Data Analysis

Sample characteristics and mean sensory ratings across all trials are reported for the full sample and stratified by gender in Table 1. Gender differences in sample characteristics and mean sensory ratings averaged across all 20 trials were assessed using independent samples t-tests for continuous variables and chi-square tests for categorical variables.

Table 1.

Sample Descriptives in the Full Sample and by Gender

	Full sample	Males	Females	Gender difference, p-value
Demographicsa
Age (y), M(SD)	42.05 (14.38)	42.23 (14.52)	41.69 (14.25)	.53
Hispanic ethnicity	23 (19.83%)	15 (19.23%)	8 (21.05%)	.82
Race				.11
Black	46 (39.66%)	32 (41.56%)	14 (35.90%)
White	35 (30.17%)	25 (32.47%)	10 (25.64%)
Asian	9 (7.76%)	4 (5.19%)	5 (12.82%)
Multi-racial	12 (10.34%)	10 (12.99%)	2 (5.13%)
other	14 (12.07%)	6 (7.79%)	8 (20.51%)
Tobacco product useb
Lifetime combustible cigarette	103 (86.55%)	70 (87.50%)	33 (84.62%)	.67
Lifetime e-cigarette	64 (53.78%)	43 (53.75%)	21 (53.85%)	.99
Current tobacco product use				.38
Current exclusive cigarette	55 (48.2%)	37 (48.7%)	18 (47.4%)
Current exclusive e-cigarette	31 (27.2%)	18 (23.7%)	13 (34.2%)
Current dual use	28 (24.6%)	21 (27.6%)	7 (18.4%)
Combustible cigarettesc
FTCD sum, M(SD)	4.46 (2.22)	4.75 (2.22)	3.79 (2.13)	.08
cigarettes/day	11.07 (6.74)	12.14 (6.85)	8.69 (5.94)	.03
Menthol preference	46 (54.76%)	31 (53.45%)	15 (57.69%)	.72
e-cigarettesd
PSECD sum, M(SD)	9.47 (5.03)	8.92 (5.11)	10.55 (4.80)	.24
Puffs/day, M(SD)	3.39 (1.34)	3.28 (1.00)	3.6 (1.85)	.48
E-cigarette flavor used
Unflavored	6 (10.20%)	6 (15.38%)	0 (0.00%)	.06
Tobacco	12 (20.34%)	9 (23.08%)	3 (15.00%)	.47
Menthol	26 (44.07%)	18 (46.15%)	8 (40.00%)	.65
Mint	23 (38.98%)	12 (30.77%)	11 (55.00%)	.07
Candy	13 (22.03%)	10 (25.64%)	3 (15.00%)	.35
Fruit	44 (74.58%)	29 (74.36%)	15 (75.00%)	.96
Chocolate	12 (20.34%)	7 (17.95%)	5 (25.00%)	.52
Other	8 (13.56%)	4 (10.26%)	4 (20.00%)	.30
Nicotine concentration				.18
1–2 mg/ML	2 (3.39%)	2 (5.13%)	0 (0.00%)
3–5 mg/ML	13 (22.03%)	8 (20.51%)	5 (25.00%)
6–8 mg/ML	5 (8.47%)	5 (12.82%)	0 (0.00%)
9–12 mg/ML	2 (3.39%)	2 (5.13%)	0 (0.00%)
12–15 mg/ML	5 (8.47%)	3 (7.69%)	2 (10.00%)
15–25 mg/ML	3 (5.08%)	2 (5.13%)	1 (5.00%)
25+ mg/ML	0 (0.00%)	0 (0.00%)	0 (0.00%)
3–4%	4 (6.78%)	2 (5.13%)	2 (10.00%)
5% or higher	19 (32.30%)	9 (23.08%)	10 (50.00%)
Don’t know	6 (10.17%)	6 (15.38%)	0 (0.00%)
Average sensory effectse
Sweet	47.53 (16.50)	47.99 (17.50)	46.58 (14.40)	.66
Smooth	54.96 (15.29)	54.39 (15.97)	56.11 (13.96)	.58
Cool	40.04 (16.47)	40.76 (17.55)	38.59 (14.20)	.52
Bitter	37.11 (18.64)	39.94 (19.30)	31.30 (15.91)	.02
Harsh	46.38 (16.37)	47.29 (16.77)	44.53 (15.54)	.39

^a116–119 full sample, 77–80 males, 39 in females.

^b115–119 full sample; 77–80 males; 38–39 females.

^cin those reporting current combustible cigarette use full sample 80–84, 55–58 males, 24–26 females.

^din current e-cigarette users full sample 59, 39 male, 20 female.

^eaverage sensory across all trials.

Primary aims were evaluated using multilevel linear models in which repeated assessments were nested within participants. Sensory ratings were partitioned into between-person (Level 2) and within-person (Level 1) variables. Between-person variables were operationalized as the participant’s mean across all completed e-cigarette administrations and were grand-mean-centered. Within-person variables reflected the extent to which a particular e-cigarette administration was the same as, higher than, or lower than that person’s average sensory ratings across the entire study protocol. The statistical modeling approach used here identified intra-individual sources of covariation between sensory attributes and appeal that holds constant inter-individual variation that might contribute to sensory-appeal associations. Model 1 tested the main effects of between-person and within-person sensory attributes controlling for gender to provide information on the associations in the overall sample, holding gender constant. Model 2 included the interaction of within-person sensory attribute × gender and between-person sensory attribute × gender to determine differential associations of sensory attributes with appeal across females and males. Significant interactions were examined using multilevel linear models of sensory attribute-appeal associations, stratified by gender. Separate models were run for each sensory attribute. As the appeal items may have been differentially relevant, supplemental analyses were run to investigate sensory attributes and gender on individual appeal items (ie, use again, like dislike). All models reported in the manuscript used the variable gender (“male”, “female”, “other”). No participants selected “other” gender and models reported in the manuscript reflect binary gender (ie, “male”, “female”). Primary models were also run using gender identity variable limited to those reporting “man” or “woman” and patterns of significance did not differ from the gender analyses (Supplementary Table 1). Data were analyzed using IBM SPSS Statistics Version 27 with alpha set to.05.

Results

Sample Characteristics

See Table 1 for sample characteristics. Of the 119 participants, 80 reported male gender and 39 reported female gender. In terms of gender identity, 71 reported man, 36 reported woman, 1 reported trans man, and 11 selected declined to answer or had missing data. In participants who currently smoked combustible cigarettes, males smoked more cigarettes a day than females (p = .03). There were no gender differences on average ratings of sweet, smooth, cool, and harsh across trials (ps > .05). There were no significant gender differences in the variance of the average sensory attributes ratings (ps > .05). Males reported higher average bitter ratings across trials compared to females (p = .02).

Between-Person Sensory Attributes and Appeal: Associations and Gender Moderation

In the overall sample, between-person sweet, smooth, and cool ratings positively associated with average appeal ratings and between-person bitter and harsh ratings negatively associated with average ratings of appeal (Table 2, Model 1). There was a significant interaction of gender and between-person sweet, smooth, and cool ratings with appeal (Table 2, Model 2). Analyses stratified by gender showed that between-person sweet and cool ratings were significantly associated with appeal in males (Table 2, Males), but not in females (Table 2, Females). Analyses stratified by gender showed that between-person smooth ratings were significantly positively associated with appeal in males and females, but the magnitude of association was 0.55 (SE = 0.16) points larger in males (estimate: 1.01 [SE = 0.09]) compared to females (estimate: 0.46 [SE = 0.14]).

Table 2.

Models of Association of Sensory Ratings with Appeal in the Full Sample and Stratified by Gender

	Overall sample (n = 109)				Stratified by gender
	Model 1		Model 2		Males (n = 80)		Female (n = 39)
Sweet	Estimate (SE)	p	Estimate (SE)	p	Estimate (SE)	p	Estimate (SE)	p
Intercept	51.70 (2.36)	<.001	51.29 (2.28)	<.001	48.48 (1.64)	<.001	51.30 (2.14)	<.001
Malea	–3.16 (2.88)	.27	–2.82 (2.79)	.31
Sweet-BP	0.63 (0.08)	<.001	0.21 (0.16)	.19	0.76 (0.09)	<.001	0.21 (0.15)	.17
Sweet-WP	0.43 (0.02)	<.001	0.40 (0.03)	<.001	0.44 (0.02)	<.001	0.40 (0.04)	<.001
Malea × Sweet-WP			0.03 (0.04)	.40
Malea × Sweet-BP			0.55 (0.18)	.003
Smooth
Intercept	50.18 (2.00)	<.001	50.63 (1.92)	<.001	50.00 (1.36)	<.001	50.64 (1.91)	<.001
Malea	–0.27 (2.46)	.91	–0.64 (2.35)	.79
smooth-BP	0.86 (0.08)	<.001	0.46 (0.14)	.001	1.01 (0.09)	<.001	0.46 (0.14)	.002
smooth-WP	0.58 (0.02)	<.001	0.67 (0.03)	<.001	0.54 (0.02)	<.001	0.67 (0.03)	<.001
Malea × smooth-WP			–0.13 (0.04)	<.001
Malea × smooth-BP			0.55 (0.16)	.001
Cool
Intercept	51.84 (2.58)	<.001	51.17 (2.50)	<.001	48.95 (1.86)	<.001	51.18 (2.19)	<.001
Malea	–2.79 (3.17)	.38	–2.23 (3.06)	.47
cool-BP	0.53 (0.09)	<.001	0.06 (0.18)	.72	0.68 (0.11)	<.001	0.06 (0.16)	.68
cool-WP	0.37 (0.02)	<.001	0.33 (0.03)	<.001	0.39 (0.02)	<.001	0.33 (0.04)	<.001
Malea × cool-WP			0.06 (0.04)	.15
Malea × cool-BP			0.62 (0.20)	.003
Bitter
Intercept	47.77 (2.43)	<.001	49.25 (2.51)	<.001	50.60 (1.76)	<.001	49.26 (2.18)	<.001
Male	2.60 (2.99)	.39	1.34 (3.02)	.66
bitter-BP	–0.55 (0.08)	<.001	–0.30 (0.15)	.051	–0.64 (0.09)	<.001	–0.30 (0.13)	.03
bitter-WP	–0.51 (0.02)	<.001	–0.57 (0.03)	<.001	–0.48 (0.02)	<.001	–0.57 (0.04)	<.001
Malea × bitter-WP			0.09 (0.04)	.04
Malea × bitter-BP			–0.34 (0.17)	.052
Harsh
Intercept	49.56 (2.22)	<.001	50.02 (2.21)	<.001	49.46 (1.65)	<.001	50.03 (1.84)	<.001
Malea	–0.17 (2.71)	.95	–0.56 (2.69)	.84
harsh-BP	–0.71 (0.08)	<.001	–0.48 (0.14)	.001	–0.80 (0.10)	<.001	–0.48 (0.12)	<.001
harsh-WP	–0.42 (0.02)	<.001	–0.48 (0.03)	<.001	–0.40 (0.02)	<.001	–0.48 (0.03)	<.001
Malea × harsh-WP			0.08 (0.04)	.04
Malea × harsh-BP			–0.31 (0.17)	.07

^aGender variable with female gender as the reference category. BP: between-person; WP: within-person.

The main effects of between-person ratings of sensory attributes and individual appeal items showed the same significance patterns as composite appeal (Supplementary Table 2). Similar to composite appeal, there were significant interactions of gender and between-person sweet, smooth, cool ratings with use again, like, and dislike (Supplementary Table 2). There was also a significant interaction of gender and between-person harsh ratings with dislike only (Supplementary Table 2).

Within-Person Sensory Attributes and Appeal: Associations and Gender Moderation

In the overall sample, e-cigarette administrations with higher than average sweet, smooth, and cool ratings were associated with higher ratings of appeal. E-cigarette administrations with higher than average bitter and harsh ratings were associated with lower ratings of appeal (Table 2, Model 1). There were significant interactions of gender and within-person smooth, bitter, and harsh (Table 2, Model 2). Gender stratified analyses showed a significant positive association of within-person smooth and appeal in both genders, but the magnitude of association was 0.13 (SE = 0.04) points larger in females (estimate: 0.67 [SE = 0.03]) compared to males (estimate: 0.54 [SE = 0.02]). Stratified analyses also showed that within-person bitter and harsh negatively associated with appeal in both genders, but these negative associations were more robust in females compared to males (Table 2, Males/Females).

The main associations of within-person sensory attributes and individual appeal outcomes showed the same pattern of significance as composite appeal (Supplementary Table 2). There were significant interactions of gender and within-person smooth for use again, like, and dislike (Supplementary Table 2). There were significant interactions of gender and within-person bitter and harsh for use again only (Supplementary Table 2).

Discussion

Consistent with our hypotheses, we found that higher between-person average sweet, smooth, and cool ratings associated with higher average ratings of appeal, and that higher average bitter and harsh ratings associated with lower ratings of appeal. Prior research has shown that certain sensory attributes (ie, sweet, smooth, cool) associate with increased appeal and other sensory attributes (ie, bitter, harsh) associate with decreased appeal.^9,11,14 This study extends these findings by showing part of the covariance previously demonstrated between e-cigarette sensory attributes and product appeal might be accounted for, in part, by inter-individual differences that generalize across different flavors or nicotine formulations. Future research into such individual differences might explore factors such as tobacco product use history, which affects sensory attributes, or age, which may dampen detection of sensory attributes, each of which might influence sensory perception and its association with product appeal.^20,27–29 Also consistent with our hypotheses, we found within-person associations, indicating that e-cigarette administrations with higher sweet, smooth, or cool ratings compared to one’s average associated with higher appeal. E-cigarette administrations with higher bitter or harsh ratings compared to one’s average associated with lower appeal. Here we show that participants’ self-reported sensory attributes can serve as a proxy for the desirable qualities of e-cigarette products that may vary in their constituents. This result is important given the myriad of different flavored e-cigarettes and nicotine types and strengths on the market,³⁰ which creates difficulties in providing parsimonious characterizations of inter-product differences that are meaningful to the user. Future research on the value of sensory attributes as a unifying factor characterizing different products is merited.

Importantly, the associations of between-person sweet, smooth, and cool ratings with composite appeal were qualified by a significant interaction of sensory attributes and gender. In contrast to our hypotheses, we found that the associations of between-person sweet, cool, and smooth ratings with composite appeal were relatively larger in males compared to females. While we did not find a significant interaction of gender and between-person bitter ratings, on average, males reported higher bitter ratings compared to females. These findings suggest that males who generally experience pleasant e-cigarette sensory attributes may be more likely to find e-cigarettes appealing, but general tendencies to experience pleasant e-cigarette sensory attributes may be less likely to influence appeal in females. The reason for between-person sensory attribute-appeal covariation difference by gender is unknown, but may reflect the differential influence of inter-individual determinants in sensory perception and product preferences. Previous research from the food science literature demonstrates that the influence of personality traits on taste preferences is moderated by gender,³¹ and similar gender-specific processes could hypothetically generalize to the role of sensory attributes in e-cigarette preferences.

There were also significant interactions of within-person smooth, bitter, and harsh ratings and gender with composite appeal. Specifically, we found that associations of smooth, bitter, and harsh ratings were strongly associated with composite appeal in both genders, but these associations were stronger in females compared to males. Prior research suggests that sensory attributes may be more important factor in tobacco product use in females compared to males.^18,19 This study extends these findings by suggesting that product characteristics that increase bitter and harsh ratings may be more likely to reduce appeal of those product characteristics in females. A number of product characteristics (eg, increasing nicotine concentrations, free-base nicotine vs. salt) have been shown to increase bitter and harsh sensory attributes^10,12,13,21 and a prior study showing that the appeal of certain flavors were mediated by inter-product variation in sensory attributes.¹² As such, these findings suggest that regulation of product characteristics that influence sensory attributes of bitter, harsh, and smooth may have a larger impact in females who use e-cigarettes compared to males who use e-cigarettes. Supplemental analyses looking at individual items of composite appeal (ie, use again, like, dislike) showed that the interactions of within-person bitter and harsh and gender were significant for use again, but not like or dislike. These findings suggest that use again may be more relevant to gender differences in within-person sensory attribute associations with appeal.

A notable cross-cutting finding was evidence of the robust role of smoothness in terms of moderation by gender. The effect estimates for the association of smoothness ratings and appeal were larger than the other sensory attributes in the overall sample, and by gender. Smoothness was the only sensory attribute moderated by gender both for within and between-person variation. Furthermore, associations of between-person smooth ratings with appeal were larger for males compared to females, but associations of within-person smooth ratings with appeal were larger for females compared to males. These trends of males having a larger between-person and females having a larger within-person association is consistent with the other sensory attributes. A previous study found that associations of smoothness and product appeal of e-cigarettes were more robust than corresponding associations involving other sensory attributes with smoothness mediating product appeal across multiple flavors (eg, fruit vs. tobacco, menthol vs. tobacco) and nicotine vs. no nicotine.¹² Another analysis from the parent study demonstrated that salt nicotine formulations had higher smooth ratings compared to free-base nicotine and these differences were significantly larger in individuals who never used cigarettes compared to individuals with lifetime cigarette use.¹³ Findings from this study provide further support that smoothness may be a particularly relevant sensory attribute when considering the appeal of e-cigarettes in different populations. A possible reason for the robust role of smoothness is because the subjective meaning of smoothness to the user can refer to a variety of sensory components of the user experience, including the airway effects, taste, and smell. The multi-modal aspects of sensory experiences reflected in smoothness might collectively sum to a larger influence on product appeal, which could increase robustness of associations and sensitivity to important moderation by gender.

This study should be characterized in the context of its limitations. This study included one laboratory visit where participants were administered 20 trials of e-cigarettes that varied by flavor and nicotine formulation. This repeated administration may have resulted in carryover effects. However, the focus on sensory ratings themselves rather than flavor or nicotine formulation provides stronger evidence for these associations. Furthermore, it is unclear if findings generalize to naturalistic e-cigarette behavior. However, studies do support that in adults sweet-flavored e-cigarettes are generally preferred and e-cigarettes that elicit bitterness are less preferred,²² which suggests that laboratory measures of appeal may be indicative of real world behavior. Additionally, in this study, we limited our analyses to binary gender. While we did assess gender identity, we did not have a large enough sample to run analyses. Sexual and gender minorities have higher rates of tobacco product use and are underrepresented in tobacco-related research.^32–38 As such, it is important for future studies to investigate these associations in samples representative of gender diversity.

In this study, we found that sweet, smooth, and cool ratings positively associated with composite appeal at the between- and within-person level. We found that bitter and harsh ratings negatively associated with composite appeal at the between- and within-person level. Importantly, we noted important gender differences in a number of the sensory-appeal associations. First, average ratings of positively associating sensory attributes (ie, sweet, smooth, and cool) with composite appeal were larger in males compared to females. These findings suggest that individual differences in males may result in differential susceptibility to appeal, whereas in females these may be less heterogeneity in these associations. We also noted that on average males reported higher bitter ratings. As bitter e-cigarettes negatively associate with appeal,²² it is possible that females are at a higher risk of increased appeal even though the associations of between-person bitter and appeal are the same between genders. Second, we noted that associations of within-person smooth, bitter, and harsh ratings with appeal were larger in females compared to males. These findings extend prior studies showing that females may be more sensitive to alterations in sensory properties of combustible cigarettes¹⁸ by showing that females may be more sensitive to within-person sensory attributes on the appeal of e-cigarettes. As such, tobacco product regulations that influence bitter, harsh, and smooth sensory attributes may have a larger effect on the appeal of e-cigarettes in females who use e-cigarettes compared to males who use e-cigarettes.

Supplementary Material

A Contributorship Form detailing each author’s specific involvement with this content, as well as any supplementary data, are available online at https://academic.oup.com/ntr.

Funding

This work was supported by Tobacco Centers of Regulatory Science (TCORS) award U54CA180908 from the National Cancer Institute (NCI) and the Food and Drug Administration (FDA), grant K01DA040043 from the National Institute on Drug Abuse (NIDA), and grant K01DK124435 from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NCI, NIDA, NIDDK, or FDA.

Declaration of Interests

The authors have no interests to declare.

Date Availability Statement

The data underlying this article will be shared on reasonable request to the corresponding author.