The e-cigarette used in this study, Vype ePen2 (Nicoventures Trading Ltd., Blackburn, United Kingdom), was an updated version of Vype ePen (as tested by Margham et al., 2016). Like the earlier version, it consisted of a rechargeable battery section and a disposable e-liquid containing cartridge (eCap). The battery section comprised a USB-rechargeable battery and an integrated circuit power controller with two voltage settings selectable by the consumer via an external twin-setting surface mounted switch. Device operation commences when the user presses either setting of the power switch, usually in advance of the puff starting, with power operating within the device as long as the button remained pressed. The liquid contained in the eCap was fed to the atomizer through a sintered porous ceramic disk in contact with a silica transport wick. The atomizer comprised a 2.85 Ω nichrome (80% Ni/20% Cr) wire coil heater wrapped around the wick. The updates incorporated into the ePen2 model included physical alterations to the dimensions and appearance of the device and cartridge, and also changes to the electronic features such as micro-USB charging and a reduction in the voltage settings from 3.6–4.0 to 3.5–3.7 V range for the low and high power settings, respectively.

The e-liquids studied were contained in Vype eCaps, that are disposable e-liquid cartridges containing 1.58 ml of e-liquid. In this study, we tested Golden Tobacco (ePen2GT), Dark Cherry (ePen2DC), and Crisp Mint (ePen2CM) flavored e-liquids. All the liquids contained VG, PG, water, flavors, and nicotine. The ingredient specifications for the e-liquids used in this study are shown in Table 1. The ingredients used in the three flavors were toxicologically assessed using the approach of Costigan and Meredith (2015). Also shown in Table 1 are the specifications for a non-commercial, unflavored e-liquid (referred to as “ePen2NF”) that we included in our untargeted aerosol emissions investigations. All the e-liquids had similar specifications (w/w) for water (25.00%) and nicotine (1.86%). The DC, GT, and unflavored e-liquids had the same specifications for VG (48.14%) and similar specifications for PG (23.86–25.00%), while the specifications for the CM e-liquid were 37.64% for VG and 34.73% for PG. The total percentage incorporation of all flavor compounds (the sum of all individual compounds) in the e-liquids was 0.77% for CM, 1.14% for DC, and 0.03% for GT. Commercial manufactured e-cigarettes from a single batch were tested (for both device and eCaps). Quality control checks were conducted on physical characteristics of all products against their manufacturing specifications before conducting the chemical analyses. This e-cigarette has an operating life of over 200 puffs per cartridge, depending on usage patterns, and we conducted these tests at the “High Power” setting (3.7 V).

The tobacco cigarette used for comparison in the current work was the Ky3R4F Kentucky Reference Cigarette, a US-blended king-sized product that has been widely used as a standard test-piece for scientific studies. It has a cellulose acetate filter and a tar yield under ISO 3308 puffing conditions International Organization for Standardization, (2012) of 9.4 mg/cigarette in 9 puffs. Main technical specifications are available on the website of the Kentucky Tobacco Research and Development Center, (2017). The mainstream smoke HPHC yields of the Ky3R4F under both ISO 3308 and ISO-Intense smoking conditions (ISO 20778:2018) have been reported previously (Roemer et al., 2012).

Untargeted analyses were conducted at British American Tobacco R&D laboratories (Southampton, United Kingdom). Targeted analyses were conducted by Labstat International ULC (Kitchener, ON, Canada) using established methods developed, validated, and operated according to ISO17025. In all cases, aerosol, cigarette smoke, or Air/Method Blanks (AMBs) were generated using commercial puffing machines, adapted where necessary to accommodate e-cigarette button activation as part of the puffing cycle.

For untargeted analyses 80 ml puffs, over 3 s, taken twice per minute were produced using a Borgwaldt LX1 automated syringe unit (Borgwaldt KC, Hamburg, Germany), and the generated aerosol collected on conditioned Tenax TA/Sulficarb thermal desorption tubes (Markes International, Llantrisant, United Kingdom). For targeted emission testing, the e-cigarette puffing regime used was the Recommended Method 81 developed by the Cooperation Centre for Scientific Research Relative to Tobacco (CORESTA) for machine puffing of e-cigarettes.

Both of these methods reflect the longer puff duration commonly observed (McAdam et al., 2019) for e-cigarette users as compared with cigarette smokers (CORESTA 2015). The e-cigarette puffing regime specifies a puff volume of 55 ml, and a puff duration of 3 s, taken twice per minute (ISO 20768, 2018a). For products that require button activation to initiate aerosol generation, CORESTA specifies the activation timing parameters. In the current study, the activation button was pressed 1 s before each puff and held down for the duration of the puff (4 s in total for each puff). For this study, we activated the button using robotic, programmed devices synchronized to the puffing engines. Ky3R4F cigarettes were prepared for smoking according to ISO 3402 (International Organization for Standardization, 1999) and smoked under two different smoking regimes, the ISO 3308 International Organization for Standardization, (2012) and the ISO-Intense regimes (Health Canada 1999, ISO 20778, 2018b). The ISO 3308 regime specifies a 35 ml puff volume and 2 s puff duration taken once per minute. The ISO-Intense regime is an internationally standardized version of the smoking regime introduced in 1999 by Health Canada to compensate for potential blocking of the filter ventilation holes during smoking and to reflect the larger puff volumes taken by many smokers. The regime specifies a 55 ml puff volume and a 2 s puff duration taken twice per minute, with all filter ventilation holes blocked. The ISO-Intense smoking regime results in higher smoke yields than the ISO regime. When cigarettes are machine-smoked, the butt length and hence tobacco rod length smoked are predetermined. Under ISO 20768 puffing conditions, the e-cigarette cartridge provides more than 200 puffs of aerosol before the e-liquid becomes exhausted, and the Ky3R4F cigarette yields about 9 puffs under ISO 3308 conditions and 9–12 puffs under ISO-Intense parameters (ISO 20778:2018).

For targeted analyses, emissions data were collected on a per-cigarette basis for Ky3R4F, with the puff number recorded. For the e-cigarettes, the analyses were conducted on the cumulative emissions collected over 100 puffs. In the earlier paper by Margham et al. (2016), emissions were collected and analyzed from two successive 100 puff blocks. Since no significant differences were found between the first and second 100-puff block, it was decided to analyze only the first 100 puffs in the present study. The reported data for the Ky3R4F and the e-cigarette variants each comprises five independent replicates of products sampled at one point in time.

E-cigarette puffing and cigarette smoking were conducted in different dedicated laboratories, to minimise the potential for atmospheric contamination from cigarette smoke on e-cigarette measurements. AMB measurements were conducted to control for potential laboratory background levels of the target analytes (Margham et al., 2016; Wagner et al., 2018). For the e-cigarettes, AMBs were generated by drawing 100 puffs of laboratory air through empty ports of the puffing machine, and samples were analyzed in the same way as aerosol samples. For reference cigarettes, AMB measurements were also taken by drawing puffs through empty ports of the smoking machine under both ISO 3308 and ISO-Intense conditions.

The untargeted aerosol scan was conducted using the semi-quantitative screening method described by Rawlinson et al. (2017). This method detects volatile and semi-volatile organic compounds with volatilities in the range from C₃ hydrocarbons up to C₂₈ hydrocarbons. PAHs with 5 or more rings, as well as other high molecular weight species with low volatilities at 250°C, are not detected by this method. The analysis is semi-quantitative, with compound concentrations estimated in comparison to known quantities of internal standard compounds. The method has been described in detail by Rawlinson et al. (2017) and is summarised in Supplementary File S1.

The emissions of major components (total aerosol and smoke masses, VG, PG, nicotine and water) from tobacco and electronic cigarettes were measured by trapping the generated smoke and aerosol on Cambridge Filter Pads (CFP). The total mass gained by the CFP during cigarette smoke experiments is defined as the TPM. With e-cigarettes, the same approach provides the gravimetric determination of ACM. Chemical analysis of TPM for nicotine and water allows for the calculation of the quantity (“nicotine-free dry particulate matter”) known as cigarette smoke tar. Chemical analysis of the ACM in this study for the major aerosol components PG, VG, water, and nicotine allows estimation of the mass of other unmeasured aerosol components. In both cases the quantity “Balance” was used, which was defined as the difference between either TPM or ACM and the sum of the major measured components.

A total of 98 individual compounds were measured in the emissions from the three flavor variants of the e-cigarette, the Ky3R4F reference cigarette, and respective AMBs. Many of these compounds are on one or more of the regulatory lists of harmful or potentially harmful cigarette smoke components. These include the Health Canada list of 42 toxicants in cigarette smoke (not including tar, nicotine and carbon monoxide) that are required to be measured and reported for cigarettes on the Canadian market (Health Canada 1999a), the FDA’s established list of 93 HPHCs in tobacco products and tobacco smoke (not including nicotine and carbon monoxide) of which 18 have to be reported currently (FDA 2012), and the World Health Organization list of 9 cigarette smoke components for which maximum levels have been proposed (Burns et al., 2008).

In analysing the 98 compounds of regulatory interest 18 different analytical methods were employed. Generally, groups of analytes belonging to the same chemical class were analyzed together in each method. Fewer compounds were tested than in the earlier paper by Margham et al. (2016) as that study showed no evidence for the presence of radioactive elements, polychlorinated dibenzodioxins or dibenzofurans in the e-cigarette aerosol. The concentrations of all these HPHC compounds were previously below the LODs of their measurement methods. The analytical methods used have been reported previously Margham et al. (2016), are summarised in Supplementary File S1 and described briefly here. Nicotine, propylene glycol, menthol, ethylene glycol, diethylene glycol, glycidol, and glycerol were analysed using GC/FID. Nicotine related alkaloids were analyzed by LC-MS/MS. Volatile carbonyls and dicarbonyls were determined by O-(2,3,4,5,6-pentafluorobenzyl)hydroxylamine (PFBHA) derivatization of trapped aerosol, followed by GC-MS analysis. Carbon monoxide was analyzed by non-dispersive infra-red analysis. Nitrogen oxides were analyzed using chemiluminescent techniques following reaction with ozone. Volatile and semi-volatile organic compounds were analyzed using GC-MS methods. Ammonia was analyzed by HPLC and conductivity detection. Hydrogen cyanide was quantified using continuous flow analysis. Phenolic compounds were analyzed by HPLC/FLD analysis. Aromatic amines were analyzed by GC-MS following derivatization by pentafluoropropionic acid anhydride. PAHs were quantified using GC-MS. Tobacco-specific nitrosamines were analyzed by LC-MS/MS. Volatile nitrosamines were analyzed by LC-MS. Metals were determined using inductively coupled plasma-mass spectrometry (ICP-MS) with a H₂/He collision reaction interface.

Numbers of peaks quantified, detected but at levels too low to quantify, or undetected in the targeted analyses are presented in Table 2. These data are also illustrated graphically in Figure 1. Of the 98 compounds analysed, a total of 23 compounds were detected in the e-cigarette AMB. For the e-cigarette aerosols 35 (ePen2GT), 38 (ePen2CM), and 43 (ePen2DC) analytes were detected. For Ky3R4F smoke, under ISO 3308 smoking conditions, 82 were detected. In terms of quantifiable levels of the detected compounds, for the AMB 10 compounds could be quantified. For the e-cigarettes 22 (ePen2CM), 25 (ePen2GT) and 29 (ePen2DC) analytes were quantifiable. 31 of the targeted analytes were measured across all the samples (plus ACM) that could be quantified for at least one of the e-cigarettes. For Ky3R4F smoke 76 analytes were quantifiable.

Findings from the untargeted analyses are summarised in Table 3. The AMB chromatogram contained 22 detected components (row B2), the unflavored e-cigarette (ePen2NF) aerosol had 68, and greater numbers of compounds were detected in the flavored e-cigarette aerosols: 89 for ePen2GT, 94 for ePen2CM, and 128 for ePen2DC. In each case a small number (0–3) were also detected in the AMB at comparable levels to the e-cigarette samples (although greater numbers of compounds were also detectable in the AMB at significantly lower levels than in the e-cigarette samples).

Table 3 also shows further analysis of aerosol constituent numbers. The total numbers of aerosol constituents measured in this study were calculated for each aerosol (Table 3, row H); the total number of compounds from the untargeted analysis were added to those detected (whether quantifiable or not) in the targeted analyses. Those compounds identified in both targeted and untargeted analyses (8–12 for the flavored e-cigarettes) were counted once only. Table 3, row H shows that in total 43 compounds were detected in the AMBs. For the unflavored e-cigarette a total count was estimated because the targeted scans were not conducted on the unflavored variant; the range of values for compounds uniquely found in the targeted analyses of the flavored e-cigarettes suggested a range of 72–79 compounds in the unflavored e-cigarette aerosol. In contrast, greater numbers of compounds were observed in the three flavored e-cigarette aerosols, from 94 to 139, using the combined analytical techniques.

The detected compounds were assigned to estimated sources as shown in Figure 2. Three e-liquid ingredients were detected in the aerosol of the unflavored e-cigarette. Had the targeted analyses been conducted on this sample water would have been detected, and hence number of ingredients would be four for this sample. With the flavored aerosols between 17 and 50 peaks were assigned as ingredients (with DC > CM > GT). Other sources of detected compounds in the aerosol included between 4 and 11 ingredient reaction products, such as acetals/hemiacetals/ketals formed by reaction of carbonyls with PG or VG. Compounds consistent with minor components of ingredients (e.g., minor components of flavors or solvent residues) or deriving from device materials (such as monomer residues) comprised 24 compounds (unflavored) and 26–36 compounds for the flavored aerosols. We also detected 11–13 compounds that were regarded as thermal decomposition products of ingoing ingredients. Subtraction of all these assigned compounds from the total numbers of compounds resulted in totals of 30 compounds for the unflavored aerosol and 25–39 compounds for the flavored aerosols that could not be assigned to a source. The inability to assign a source for these compounds was either because they could not be identified with sufficient confidence (e.g., low library match factor) or there was no clear explanation for their presence in the aerosol.

In the targeted analyses the emissions of 98 aerosol components were quantified. The yields of these components for the three e-cigarette variants, the Ky3R4F cigarette (under both ISO and ISO-Intense smoking conditions), and the matching AMB samples are shown in Supplementary Table S1. As discussed above, 30 analytes were found in one or more of the e-cigarette aerosols and the emissions of these compounds are shown in Supplementary Table S2. These emission values are summarised in the following paragraphs:

It is also of interest to understand the quantities of compounds present in the untargeted analysis. However, full quantitation was not achievable in the untargeted analysis because compound identities were not verified, and moreover where library match factors were low it is possible that the assigned identities were incorrect. Uncertainty over compound identity meant that it was not possible to conduct MS calibrations, which renders the concentration data semi-quantitative at best, providing “order of magnitude” information only. The estimated concentrations of detected compounds in aerosols of the e-cigarette variants that were not common to the AMB and disclosed ingredients were calculated. Of the aerosol components with Mass Spectral Library match factors (MF) ≥75, 65% had concentrations in the estimated range 0–20 ng/puff, and 36% had estimated levels of 0–5 ng/puff. Of the aerosol components with 50 ≤ MF < 75, 79% had estimated levels in the range 0–20 ng/puff, and 21% had estimated concentrations of 0–5 ng/puff. The highest estimated value of around 450 ng/puff was found for a compound eluting at 7.14 min in the chromatogram of the unflavored e-cigarette aerosol. The peak had a poor match factor to the MS library (70%); the best match was a silane–benzeneacetic acid, alpha, 4-bis [(trimethylsilyl)oxy-, trimethylsilyl ester. However, the poor match factor means that its identification should be regarded as tentative at best.

We calculated the composite average percent reductions in toxicant yields from the e-cigarettes compared with those from the Ky3R4F cigarette for different regulatory interest “lists” shown in Supplementary Table S3. Table 6 shows that the WHO TobReg 9 constituents were reduced by 98.5–99.5% in the emissions from all three flavor variants of the e-cigarette when compared on a puff-by-puff basis with smoke from a Ky3R4F cigarette smoked under the ISO-Intense regime. Although not statistically significant, the slightly lower percentage reduction for formaldehyde in CM (>88%) compared with the other flavor variants (96.6 and 92.2%) caused a slightly lower composite percentage reduction for the CM flavor variant. Toxicants on the FDA abbreviated list of 18 compounds generated under ISO-Intense were reduced on average by >97% compared with the Ky3R4F; the % reductions increased to >99% when nicotine is removed from the list.

Figure 3 shows that the sum of measured major e-liquid and aerosol constituents (VG, PG, water and nicotine) accounted on average for 94.2% of the ACM, (rising to 97% using the expected VG value for GT). The calculated difference between ACM and the sum of the major components (“balance”) is clearly sensitive to errors in the determination of the major species. Accurate quantification of water in aerosol streams has traditionally presented significant challenges and may also be associated with relatively large quantification errors in these measurements. In contrast, the comparable balance for cigarette smoke lay between 58 and 76% of the trapped particulate mass. These data also suggest a much more diverse composition of cigarette smoke compared to e-cigarette aerosols.

The carbonyl yields measured in the present study were not significantly different from those found previously by Margham et al. (2016), but the AMB values for formaldehyde, acetone, and MEK were significantly lower in the present study. Such compounds (formaldehyde, acetaldehyde, acetone, propionaldehyde, acrolein, glycolaldehyde, glyoxal, and methylglyoxal), and the compounds propylene oxide and allyl alcohol are thermal decomposition products of the humectants, PG and VG, used as aerosol generators (Stein et al., 1983; Laino et al., 2011; Laino et al., 2012; Sleiman et al., 2016).

Higher levels of aerosol nicotine and tobacco alkaloids were reported in the present study than by Margham et al. (2016) even though the e-liquid nicotine levels were the same in the two studies and ACM emissions were comparable. The tobacco alkaloids other than nicotine are present as low-level constituents of the pharmaceutical nicotine used in e-liquid formulations. Those quantified in the aerosol–nornicotine, myosmine, nicotine-N-oxide, cotinine, and ß-nicotyrine–are naturally present in the tobacco leaf used to produce the pharmaceutical grade nicotine used in e-liquids and some may also be formed through nicotine oxidation in e-liquids (Marion 1950; Kisaki et al., 1978; Martinez et al., 2014).

A number of metals were measured in the e-cigarette emissions. In an earlier paper (Margham et al., 2016) describing the aerosol chemistry of a similar product to that reported in the present study, chromium was quantified in the first 100 of 200 puffs at an average of 0.50 ng/puff but was not quantified (<0.45 ng/puff) in the second 100 puffs. The AMB contained detectable but not quantifiable levels of chromium (>0.13 but <0.45 ng/puff). Some levels of chromium generated by the e-cigarette could therefore not be ruled out. In the present study, levels of iron, zinc, and chromium found in the e-cigarette aerosols and the AMB were not significantly different (Table 5), and we can conclude that the presence of these metals likely arise as artifacts from aerosol collection or other analytical processes. For iron and zinc, Margham et al. (2016) also concluded that their presence in the aerosols was due to laboratory contamination. Williams et al. (2013) using non-standard smoking parameters analyzed 20 metals in the aerosol of a single brand of e-cigarette, including chromium (0.7 ng/puff), iron (52 µg/puff) and zinc (5.8 ng/puff), but none were detected in an AMB. Their results are significantly higher than those of the present study. Tayyarah and Long (2014) reported detectable but non-quantifiable levels of chromium (1–4 ng/puff) in three products they tested as well as the AMB. However, other studies, such as that of Goniewicz et al. (2014) have not detected chromium in e-cigarette emissions.

In the present study, consistent with the findings of Margham et al. (2016), emissions of four aromatic amines were quantified in the e-cigarette aerosols and AMB, with levels not significantly different between the background and aerosol samples. Three PAHs and ammonia were quantified in the e-cigarette emissions and AMB; levels were not significantly different between samples. Emissions of ammonia, chrysene, and naphthalene were higher in the present study for all the e-cigarettes as well as the AMB than found previously by Margham et al. (2016). TSNAs were not detected in any of the e-cigarette samples, even with picogram per puff LODs.

Given the very low levels of many of the toxicants that are now measured in e-cigarette aerosols, combined with the relatively large numbers of puffs taken on e-cigarettes in comparison to tobacco cigarettes, it is essential to understand the contribution to measured values from environmental factors. These include toxicants that may already be present in the laboratory air, in reagents, or that may be introduced by operators or equipment (such as puffing machines) used to generate and collect the aerosol. Hence the importance of the AMB as a means of minimising the possibility of false-positives and overestimates (Tayyarah and Long, 2014; Margham et al., 2016; Wagner et al., 2018; Belushkin et al., 2020). AMB control measurements are widely used in different e-cigarette research areas, such as chemical analysis and indoor air quality studies (Goniewicz et al., 2014; Herrington and Myers, 2015; Marco and Grimault, 2015; Mikheev et al., 2016; Palazzolo et al., 2016; Aherrera et al., 2017; Beauval et al., 2017; Moldoveanu et al., 2017; Olmedo et al., 2018; Halstead et al., 2019).

Table 5 shows that in the present study, there were eight components where AMB values were numerically higher than one or more of the e-cigarette aerosol samples. These were chromium (higher than all 3 e-cigarette samples), iron (higher than 1 sample), naphthalene (1), benzo(a)anthracene (2), chrysene (2), 2-aminonaphthalene (1), 3- and 4-aminobiphenyls (1 sample each). However, ANOVA testing showed that none of these differences were significant at the 95% confidence level. Therefore, in none of these cases was there a significant difference between AMB and the e-cigarette samples. In addition, the ANOVA tests showed that zinc, o-toluidine and ammonia emissions were not significantly different from those found in the AMB (Table 5).

Further insights into the potential presence or concentrations of these toxicants will require greater reductions of chemical background than are currently achievable using established methods. Gaseous and volatile contaminants could be excluded during e-cigarette experiments by use of air-tight assemblies fed by high purity air, as reported by García-Gómez et al. (2016). However, not all of the contamination arises from the laboratory air. Metals such as chromium, iron and zinc appear to arise at least in part from the puffing machines used to generate aerosols and the associated trapping matrices (data not shown); reducing the impact of metal contamination from these sources may represent a way to minimise AMB contamination with these compounds. These approaches represent valuable avenues for future investigation.

AMB experiments for the Ky3R4F cigarette were conducted under two puffing regimes, and contributions to the measured cigarette smoke emissions were found with 29 of the 98 analytes measured in this study. Levels per puff were generally much higher than found with the e-cigarette AMB, due in the main to some elements of environmental tobacco smoke around the smoking engine (generated by the cigarette sidestream smoke as it leaves the burning cigarettes) being pulled into the empty port of the smoking engine during the puffing steps of the AMB experiment. This source does not exist for the e-cigarette AMB experiment. Despite the higher absolute levels measured with the Ky3R4F AMB, their contribution to the measured levels in smoke was generally less due to the relatively high concentrations of toxicants in mainstream cigarette smoke.