Can Established Vapers Distinguish Different PG:VG Ratios? A Pilot Study.

Abstract

Objectives:

Using the triangle test, this study explores whether established vapers can distinguish small differences between e-liquid propylene glycol: vegetable glycerin (PG:VG) ratios at fixed levels of nicotine and flavor, in order to examine the extent to which solvent ratios affects sensory experiences.

Methods:

Watermelon flavored e-liquids (16mg/mL nicotine) used for this study differed in the ratio of PG:VG (30:70, 50:50; 70:30). Current vapers were randomized to one of 3 possible study conditions differing by PG:VG ratio. Participants sampled products following the triangle test, which presents 3 blinded products, 2 of which are identical. They were asked to identify and rate the ‘odd’ product.

Results:

Of the 14 participants who completed the study, 34.9% were able to determine the ‘odd’ product. Ratings on the subjective response scales for the ‘odd’ product were quite low. Aversion scores differed significantly by correct identification of the ‘odd’ product (p = .045).

Conclusions:

Established vapers in this study were unable to consistently distinguish PG:VG ratios, even for relatively large differences, and correct identification inconsistently related to subjective effects ratings. These preliminary findings suggest that the PG:VG ratio may not be a salient feature of vaping.

INTRODUCTION

In August 2016, FDA’s authority was extended to emerging tobacco products, including increasingly popular electronic cigarettes (e-cigarettes).^1,2 E-cigarette use has become popular over the last several years, particularly among middle and high school students.^1,2 In 2016, over 2 million middle and high schoolers reported current e-cigarette use.^1,2 On the other hand, daily exclusive use of e-cigarettes may aid adult smokers in achieving cessation.^3,4

Consumer perceptions about a product, which are important drivers of use, are influenced by product design and branding. These can influence sensory response.^5–7 Sensory experiences associated with tobacco products include taste experiences (eg, flavors), throat hit, and physiological reward (eg, nicotine delivery).^5–7 In order to make a product appealing to consumers, a balance of taste experiences, throat hit and physiological reward must be found.^5–7 E-cigarettes have 2 components (a vaporizing device and the e-liquid itself) that can each influence consumer sensory experiences.⁸ E-liquids can differ across a number of dimensions, including nicotine content, flavorings, and the ratio of solvents (vegetable glycerin (VG) and propylene glycol (PG)).^8,9 Changing the ratio of solvents affects mouth and throat sensations.^9,10 A higher proportion of PG supposedly provides consumers with a greater throat hit, while greater VG concentration reportedly is smoother on the throat and creates a larger visible vape ‘cloud’.^9,10 Furthermore, increases in nicotine delivery have been seen at high concentrations of PG.¹¹

Extreme PG:VG ratios (eg, 100% PG) may affect consumers’ sensory experiences substantially,^11–13 but less is known about sensory experiences when comparing differences in PG:VG ratios with the commercially common range (eg, 50:50 or 70:30). This study aimed to evaluate whether established vapers can distinguish a small difference between e-liquid PG:VG ratios at fixed nicotine and flavor, to determine the extent to which these solvents are important to the sensory experience using a specialized method.^14–17 The standardized triangle smoking test procedure, a common industry technique with wide applications in sensory science, was designed to determine sensory differences between 2 similar products with a known difference (eg, a change in ingredients). The procedure asks participants to test 3 products where 2 are identical and one is different, and to identify the odd one.^14–17

METHODS

E-Liquid Analysis

All e-liquids used in this study were watermelon flavored and had a nominal nicotine concentration of 16 mg/mL (The Vapor’s Knoll, Pasadena, MD). Watermelon flavored e-liquid was chosen because almost 50% of exclusive vapers in a recent local study reported a fruit flavored e-liquid, and watermelon was particularly common. The e-liquids differed in their nominal concentration ratios of PG:VG (Liquid A: 50:50, Liquid B: 70:30, Liquid C: 30:70). These ratios were chosen so small to moderate differences within the commercial range of e-liquid solvent compositions could be studied. Although the FDA was granted authority over e-cigarettes in 2016, manufacturers of e-liquids purchased for this study were not yet under regulation. Therefore, nicotine and PG and VG concentrations were confirmed at Roswell Park Comprehensive Cancer Center (RPCCC; Buffalo, NY) along with analysis of emissions of each product.

Nicotine was confirmed in the e-liquid and the vapor, using a modification of the NIOSH 2551 method for determining nicotine in air¹⁸ via gas chromatography with nitrogen-phosphorous detector (GC-NPD; Agilent 7890B Gas Chromatography; Agilent Technologies, Santa Clara, CA). All samples were run triplicate, and vapor samples were collected after 15 puffs. The limit of quantitation for nicotine analysis in e-liquid analysis was 1 mg/mL and 2.5 ng/mL in vapor. A non-targeted analysis of emissions in vapor was performed via by GC-MS (Agilent 7890B Gas Chromatography coupled with Agilent 5977A MSD Detector) using an existing method.¹⁹ Samples were collected in triplicate after 300 puffs. PG and VG confirmation was performed via GC/MS (Agilent Technologies; Santa Clara, CA) with an adaptation of a published method.¹⁹

Participants and Procedures

Participants were recruited from the Buffalo and Niagara Region of Western New York using advertisements in community newspapers, fliers at area colleges, local shops and coffee houses, and postings on CraigsList.com. Eligible participants were current users of e-cigarettes who had used e-cigarettes for a year or more, aged 18 and older, and who smoked no more than 5 tobacco cigarettes per day (by self-report). Participants were randomized to one of 3 possible study conditions (Condition 1: Liquids A vs. B; Condition 2: Liquids A vs. C; Condition 3: Liquids B vs. C). Across 6 sessions, participants were presented with 3 coded cartridges where 2 were filled with the first assigned e-liquid and the third was filled with the second assigned e-liquid (eg, for condition 1, only e-liquids A and B were used). One milliliter of all e-liquids was measured and placed in a V2Pro Series 3 compatible e-liquid cartridge (3.0 ohm) and vaped on a V2Pro Series 3 vape pen (Miami, FL; 130 mAh battery, 3.8 V). This device was chosen because of its design (eg, the e-liquid cartridge sits inside the vape pen, functioning as a partial blind, and the device power and maximum temperature are not user-adjustable. The e-liquid cartridges were primed prior to participant use. Participants were instructed to vape the e-liquids they were assigned in a standardized manner (3 puffs on each product with 30 to 60 seconds between each puff and 2 minutes between sampling each product) and in a predetermined order. The triangle smoking test procedure^14–17 was used to determine if sensory differences exist with various PG:VG ratios. Participants were instructed that 2 e-liquids were identical and one was different, and their ability to identify the e-liquid that was different from the other 2, or the ‘odd’ product, was observed. This process was repeated 3 times at each session, with a 20 minute washout period, following the same predetermined order at each session (eg, for condition 1, Session 1: AAB; Session 2; ABA, Session 3: BAA; Session 4: BBA; Session 5: BAB; Session 6: ABB). When the ‘odd’ product was selected, a product evaluation scale (PES)²⁰ was completed for the selected product whether or not the selection was correct. The PES was adapted from the modified Cigarette Evaluation Questionnaire including items for satisfaction (“Was it satisfying,” “Did it taste good,” “Did you enjoy the sensations in your mouth,” and “Did you enjoy it?”), psychological reward (“Did it calm you down,” “Did it make you feel more awake,” “Did it make you feel less irritable,” “Did it help you concentrate,” and “Did it reduce your hunger for food?”), aversion (“Did it make you dizzy,” “Did it make you nauseous,” “Was it too much nicotine,” and “Were there bothersome side effects?”), and relief (“Did it immediately relieve your craving for an e-cigarette,” “Did it relieve withdrawal symptoms,” “Did it relieve the urge to vape,” “Was it enough nicotine,” and “Did you still have a craving for a cigarette after using the product?”) answered on a 7-point Likert scale (where 1 was ‘not at all’ and 7 was ‘extremely’).^20,21 The 6 sessions were spaced 24–72 hours apart to implement a standardized time frame. Participants were asked not to vape 30 minutes prior to each session.

Statistical Analyses

Data analysis was conducted using Statistical Package for the Social Sciences Version 21.0 (IBM, Armonk, NY). The items for each of the PES subscales were averaged.¹⁷ Generalized estimating equations (GEE) were used to account for repeated responses within-subjects for product identification (binomial distribution, log link function) and PES scores (normal distribution, linear or log link function). Models included session, trial, study condition, and demographics (age, sex, education, and race/ethnicity). This project was approved by the Roswell Park Institutional Review Board and informed consent was obtained from all participants.

RESULTS

Analysis of e-liquids showed analytically-confirmed PG:VG ratios were A = 45:55, B=70:30, C=36:64. The average measured nicotine was 13.5 mg/mL (standard deviation (SD): 1.16). Nicotine delivered in vapor (30 puffs) was 11.2 mg/mL, 13.3 mg/mL, and 13.2 mg/mL for liquids A, B, and C, respectively. Vapor components identified by GC/MS scans in all vaped liquids were glycerin, propylene glycol, nicotine, and ethyl maltol. Liquids A and C were shown to contain benzyl alcohol, potentially because of the higher concentration of VG.

Fourteen participants completed the study (A v. B: N = 5; A v. C: N = 5; B v. C: N = 4). Due to the randomization, demographic and behavioral characteristics did not differ between study conditions. The prototypical participant was a 44 year old (range: 20–67), Non-Hispanic White male who had some college or technical school and made $30,000 or less a year. On average, participants took 345 puffs per day on their e-cigarette. All reported that there was nicotine in the e-liquid, with most (64.3%) reporting a low nicotine concentration between 0.6% and 0.9%. Most of the participants used an e-liquid that was a mixture of PG and VG (78.6%). All but one of the participants currently used an e-cigarette with a tank system, which most bought at a special shop, such as a vapor lounge (64.3%). Reported device brands included Eleaf (N = 3), Aspire (N = 2), Evolv (N = 1), Innokin (N = 1), IPV (N = 1), iTaste (N = 1), Kangertech (N = 1), SmokTech (N = 2), and Volt (N = 1). Most of the participants (85.7%) used an e-liquid that was flavored to taste like menthol, mint, spice, candy, fruit, or alcohol. Participants were asked to write down any flavor they usually use. Reported flavors included artic berry (N = 1), caramel cookie (N = 1), chocolate (N = 2), coffee (N = 1), custard (N = 2), fruit and tobacco (N = 1), mint (N = 1), moose knuckles (N = 1), nutty flavors (N = 1), vanilla (N = 1), strawberry (N = 1), and a mixture of watermelon, strawberry, green apple and a little cotton candy (N = 1). The remaining participants reported tobacco (N = 1) and crunchy black (N = 1) as their usual flavors.

Across the 3 study conditions, a total of 249 triangle tests occurred (A v. B: 87 tests; A v. C: 90 tests; B v. C: 72 tests). Of the 249 triangle tests, 34.9% identified the ‘odd’ product correctly. Between the different study conditions, there were no significant differences in ability to correctly determine the ‘odd’ product in the triangle tests (A vs. B: 34.5%, A vs. C: 33.3%, B vs. C: 37.5%). PES of the ‘odd’ product chosen did not differ in satisfaction (Correct = 1.0–5.5, Incorrect = 1.0–6.5), psychological reward (Correct = 1.0–5.2, Incorrect = 1.0–6.5), relief (Correct = 1.0–6.2, Incorrect = 1.0–6.0) based on correctly identifying the ‘odd’ product PES of the ‘odd’ product chosen. However, it did just reach statistical significance in aversion (Correct = 1.0–3.5, Incorrect = 1.0–5.0; p = .045) based on correctly identifying the ‘odd’ product (Figure 1). Significant predictors of ability to identify the ‘odd’ product correctly included session (p=0.009) and education (p = .017), after adjusting for trial, study condition and demographics. Participants were more likely to identify the ‘odd’ product in the early sessions, and if they had some college or more. Session was also a significant predictor for satisfaction (p = .001), after adjusting for ability to identify the ‘odd’ product, trial, study condition and demographics. Participants saw the ‘odd’ product more satisfying in the early sessions than in later sessions. Study condition, education and race/ethnicity were significant predictors for relief (Study Condition: p = .001, Education: p = .008, and Race/Ethnicity: p = .048), after adjusting for ability to identify the ‘odd’ product, session, trial and demographics. When randomized to study condition B vs. C, participants rated the ‘odd’ product higher on relief. Furthermore, those who were Non-Hispanic White with some college or more found the ‘odd’ product to provide greater relief. Education and race/ethnicity were significant predictors for psychological reward (Education: p < .001, Race/Ethnicity: p = .009), after adjusting for ability to identify the ‘odd’ product, session, trail, study condition and demographics. Those who were not Non-Hispanic White with some college or more had greater psychological reward from the ‘odd’ product. Session, study condition and education were significant predictors for aversion (Session: p < .001, Study Condition: p = .025, and Education: p = .004), after adjusting for ability to identify the ‘odd’ product, trial and demographics.

Figure 1:

Product Evaluation Scale Scores by Study Condition Comparing Those Who Did and Did Not Correctly Identify the ‘Odd’ Product.

DISCUSSION

The established vapers in this study were unable to consistently identify the ‘odd’ product, even with relatively large differences in solvent ratio (30:70 vs 70:30). This finding may suggest that the PG:VG ratio is not as salient a feature of vaping as nicotine or flavoring. Most participants used a flavored, nicotine-containing liquid, many reporting fruit flavors, though most used nicotine levels lower than those used in the experimental sessions. This may have contributed to the relatively low ratings on the subjective response scales for the ‘odd’ product, regardless of whether it was correctly identified – participants may simply have been expressing preference rather than discrimination. The findings seem to indicate that additional sessions provided diminishing returns, as participants were less likely to identify the ‘odd’ product correctly in later sessions. They also found the ‘odd’ product to be less appealing in later sessions despite the fact that they were provided the same products. This potentially could be due to fatigue or adaptation from the repetition over 6 sessions.

There are some limitations to this study. This study only looked at a single e-liquid flavor, and should be replicated using other flavors. This study is limited by small sample size due to low recruitment rates. The e-liquids deviated from their expected solvent ratios and nicotine concentrations by 10–20%, pointing to the need for independent confirmation of labeled/stated values in all research studies. In addition, occasionally participants complained of burnt taste, although the e-liquid cartridges used were new and primed prior to use. Nonetheless, this could have influenced participant choices, as the intended flavor of the product would appear to be altered to the participant. Finally, the nicotine concentrations used in this study were greater than what participants regularly used, though the actual amount of nicotine in vapor is a function of both liquid concentration and device power. This study used a relatively low power device, whereas most participants were likely using more powerful “3^rd generation” devices. For this study, we did not specifically assess the power of users devices).

IMPLICATIONS FOR TOBACCO REGULATION

This pilot study offers initial evidence of perceived sensory equivalence of solvent content among experienced vapers, which should be followed up by larger studies using other flavor and nicotine level combinations. Given emergent regulations around e-cigarettes,²² including new product applications, substantial equivalence determinations, and product standards, research on design elements of devices and liquids that influence user experience and exposure will become increasingly important.