Assessing variability of aerosols generated from e-Cigarettes

Abstract

While some in vitro and in vivo experiments have studied the toxic effects of e-cigarette (e-cig) components, the typical aerosol properties released from e-cigarettes have not been well characterized. In the present study, we characterized the variability in mass concentration and particle size distribution associated with the aerosol generation of different devices and e-liquid compositions in an experimental setup. The findings of this study indicate a large inter-day variability in the experiments, likely due to poor quality control in some e-cig devices, pointing to the need for a better understanding of all the factors affecting exposures in in vitro and in vivo experiments, and the development of standardized protocols for generation and measurement of e-cig aerosols.

Introduction

The use and awareness of e-cigarettes (e-cigs) has increased rapidly in the past few years. Although considered less toxic than tobacco cigarettes (McAuley et al. 2012), recent studies have provided evidence that using e-cigs pose a health risk (Wang et al. 2019). The aerosols inhaled by the user are generated by heating a vaping fluid consisting of a solvent, propylene glycol (McAuley et al. 2012; Zhao et al. 2018) and glycerin (Zhao et al. 2018), that contains nicotine and additives for flavoring (McAuley et al. 2012; Zhao et al. 2018). The high temperature needed to aerosolize the vaping fluid has been shown to generate carbonyl compounds, including potentially harmful species such as formaldehyde, acetaldehyde, and acrolein (McAuley et al. 2012; Zhao et al. 2016; Chen et al. 2018). The aerosols also contain a broad range of metals originating from the vaping fluid as well as the heating coils (Olmedo et al. 2018; Zhao et al. 2019, 2020). Therefore, characterizing the chemical and physical composition of the vaping aerosol is critical to assessing the potential health risk of e-cigs (Wang et al. 2019). However, given the wide variety of vaping fluids and e-cig devices available, the physical and chemical composition of the aerosols will vary (Table 1), and likely the health risks. Therefore, mouse and cell exposure studies under controlled conditions are helpful to better assess the health risks of using e-cigs in a model that recapitulates human exposure. The success of this approach depends, in part, on the accuracy and precision of characterizing the exposure. While mouse and cell exposure studies have been reported (Hilpert et al. 2019, Mcgrath-Morrow et al. 2015), the variability in such experimental schemes have not been reported.

Table 1.

Summary of literature highlighting the factors responsible for variability associated with e-cigarette aerosol measurements.

Parameter	Range of values	Reference
Device type	Disposable	Zhao et al. 2018
	Pre-filled ‘pod’	Zhao et al. 2019
	Refillable pod
	Refillable tank
Coil resistance	<1 Ω	Aherrera et al. 2017; Mulder et al. 2020
	1.5, 1.8 Ω
	• 2.2 Ω
Battery voltage	3.9–4.7 V	Zhao et al. 2019; Mulder et al. 2020
Power setting	20–200 W	Zhao et al. 2019
e-liquid formulation	Varying ratios of PG:VG	Zhao et al. 2018; Zervas et al. 2018; Lechasseur et al. 2019; Mulder et al. 2020
	Flavorings
	Nicotine concentration
	Nicotine formulation (salt- vs. free base)

The goal of the present study is to characterize the variability of particle mass, number concentrations, and size distributions of e-cig aerosols from four different fluid compositions, using a popular e-cig model with a re-fillable pod, generated into identical exposure chambers, and develop a system that minimizes this variability.

Materials and methods

The schematic of the setup is shown in Figure 1(a) and the photograph of experimental setup is provided in Figure 1(b).

Figure 1.

(a) Schematic of the Exposure chamber system. (b) Photograph of the experimental setup (Note: four peristaltic pumps are on a moving cart to allow access to the chambers; during normal operation the pumps are directly in front of their respective chamber).

Exposure chambers

E-cig aerosols were generated into four identical exposure chambers of 25-liters each designed for whole-body exposure of mice (CH Technologies, USA). Wang et al. 2019 reviewed e-cig aerosol exposure method for various in-vivo studies. Inhalation exposure in animal studies can be done in two ways: whole-body exposure (Lerner et al. 2015; Mcgrath-Morrow et al. 2015; Sussan et al. 2015; Lauterstein et al. 2016; Zelikoff et al. 2018; Hilpert et al. 2019) or nose-only exposure (Hwang et al. 2016). The advantages of nose-only exposure are that the exposure concentration can be high, there is less waste due to the small chamber volume, and it is feasible to avoid particle intake by other channels and analyze the particle effect just through inhaling. When rats are held immobile in the chamber, the most significant worry is stress. It can also be difficult to maintain the humidity in the chamber which can affect the size of particles being inhaled. Whole-body exposure adds the least stress other than that from the exposure material, making it more suitable for lengthy and repeated inhalation than nose-only exposure. Oyabu et al. 2016 showed that for the size ranges of particles of interest in this paper, the two inhalation methods using the same material under the same exposure conditions resulted in the same particle deposition and histopathology in the lung. In the present study, an exposure characterization method similar to that of Hilpert et al. 2019 is used to have flexibility on the air exchange rate and monitoring the particulate matter level suitable for the range of instruments used. The four chambers are assembled into a system supported on special racks, constructed of T-slot aluminum material and designed to fit inside a standard fume hood. Each individual chamber, composed of polysulfone, can accommodate up to 10 mice or 5 rats each. The special design of each chamber provides an optimal continual mixing of the fresh aerosol coming in, which minimizes the variability of exposures among animals moving about the cage. The inlet of each chamber is located at the bottom, at a height that places it within the breathing zone of the animals. This ensures a constant supply of fresh aerosol. Each chamber inlet is connected to a port of the manifold through a piece of conductive tubing that minimizes aerosol deposition due to electrostatic effects. The inlet tubes of all the cages are equal length to avoid any intra-cage flow and pressure variability. Another identical manifold, with the appropriate number of in-line ports, collects the exhaust from all the chambers and uses a single discharge port to dispose of the exhaust, which in turn is vented inside a fume hood.

Products tested and aerosols generated

Two types of e-cig devices were tested. 1) Suorin Air Plus (SuorinUSA, CA) automatic draw pod-type devices with 0.7 or 1.0 Ohm coil inside refillable cartridges (called pods) at 4.2 V, and 2) button-activated Vinci (Voopoo Inc) (modifiable or ‘mod’ device) with coils available in 0.1–3 Ohms and variable settings that allow voltage selection between 3.2–4.2 V. For experiments described in this paper the Vinci device was used with a 1.0 ohm coil and 3.45 V. The Suorin device was chosen because of its wide use and market availability at the time the experiment was conducted and because of its large pod capacity (2 ml), as one of our aims was to test aerosols from a popular pod device.

Experiments were conducted in two phases. In the first phase, aerosols were simultaneously generated using four Suorin devices each containing different e-liquid compositions to study the independent effects of each e-liquid component on aerosol mass and size distribution. The four e-liquid types were: (i) 30:70 (v/v) solution of propylene glycol and vegetable glycerin (PG:VG) alone, (ii) a solution of 30:70 PG:VG with 5 mg/mL ethyl maltol (EM); (iii) 30:70 PG:VG with 12 mg/mL nicotine salts (Nic), and (iv) 30:70 PG:VG with 5 mg/mL EM plus 12 mg/mL nicotine salts. Pods and coils used from one experiment to the next were kept with their respective device and same e-liquid type. Before each experiment the pods were topped off with e-liquid. Pods were changed every 15 hours per manufacturer recommendations to prevent coil over-use. In the second phase, the Vinci device was used with 30:70 PG:VG alone to verify within- and between-day variability.

Aerosol generation

The e-cig aerosols were continuously generated during three hours using a peristaltic pump (Masterflex, Cole Parmer, IL) with a puffing regime consisting of a 3-second puff followed by a 30 second rest and a 100 mL puff volume at 2 l/min. The weight of pods before and after each experiment was used to calculate the losses to different parts of the system (mass balance). Immediately after each experiment, ~250 μL aerosol from each of the four e-liquid types was collected using a condensing system (Olmedo et al. 2016) to conduct chemical analysis.

Chamber Aerosol characterization

Near real-time particle number and mass distributions were measured 4 times in each chamber for 15 minutes using a Scanning Mobility Particle Sizer (SMPS) (3936 SMPS, TSI Inc, MN) and an Aerosol Particle Sizer (3320 APS) (TSI Inc, MN). Time-resolved PM₁₀ mass concentration was measured in each chamber over 3 hours using a personal DataRAM (pDR-1200, Thermo Scientific, MA) with an inlet cyclone and 37 mm mixed cellulose ester filters (GN-4 Metricel, 0.8 μm, Pall, NY) at 1.5 l/min for gravimetric analysis. Filters were weighed before and after sampling using a microbalance (MT5, Mettler-Toledo, OH) with 1 μg resolution after being equilibrated 48 hours in a clean room maintained at 25 °C and a relative humidity of 40%. The mass of every filter was calculated as the average of three consecutive weightings if they agreed within ±10 μg. All experiments were repeated four times (except PG:VG was only repeated twice) with a 3 hour sampling duration. A complete summary of the experimental design is presented in Table 2. The SMPS was operated at a sheath flow rate of 2 l/min, sample flow rate of 0.2 l/ min, and sampling time resolution of 1 min. These settings measure the size distribution in the range of 16.3 nm to 982.2 nm with a step at 600 nm. The APS measures from 500 nm to 20,000 nm (20 mm) at 1 l/min with a sheath flow of 4 l/min.

Table 2.

Summary of experiments carried out in the present study.

e-liquid composition	Parameter	Number of Repeats (Duration)	Instrument Used
PG:VG	Size Distribution	4 (15 min)	SMPS, APS
	PM10 mass Concentration	4 (5 hrs)	pDR
	Integrated Time Weighted Average Concentration	4 (5 hrs)	Gravimetric
PG:VG + EM	Size Distribution	4 (15 min)	SMPS, APS
	PM10 mass Concentration	4 (5 hrs)	pDR
	Integrated Time Weighted Average Concentration	4 (5 hrs)	Gravimetric
PG:VG + Nic	Size Distribution	4 (15 min)	SMPS, APS
	PM10 mass Concentration	4 (5 hrs)	pDR
	Integrated Time Weighted Average Concentration	4 (5 hrs)	Gravimetric
PG:VG + EM + Nic	Size Distribution	4 (15 min)	SMPS, APS
	PM10 mass Concentration	4 (5 hrs)	pDR
	Integrated Time Weighted Average Concentration	4 (5 hrs)	Gravimetric

Data Analysis

Hatch choate equation

The count median diameter (CMD) from the SMPS was converted into the mass median diameter (MMD) using the Hatch-Choate equation as shown in Equation 1, assuming that the aerosol size distribution was log-normal:

$$\mathrm{M}\mathrm{M}\mathrm{D}=\mathrm{C}\mathrm{M}\mathrm{D}\mathrm{e}\mathrm{x}\mathrm{p}\left(3{\mathrm{l}\mathrm{n}}^2{\mathrm{\sigma }}_{\mathrm{g}}\right)$$ (1)

where MMD is the mass median diameter, CMD is the count median diameter, and ${\mathrm{\sigma }}_{\mathrm{g}}$ is the geometric standard deviation.

Merging of the SMPS-APS data

SMPS and APS use different measurement principles (i.e. electrical mobility versus aerodynamic sizing), and thus the data measured by the two instruments need to be corrected to obtain a single size spectrum (Khlystov et al. 2004). An average ratio of SMPS to APS data was obtained for the size range 600–1000 nm where both instruments overlap. This ratio was used to convert the APS data for the size range where only APS data was available (>1000 nm).

Analysis of variance (ANOVA)

We conducted one-way analysis of variance to look at the differences in the variability of near-real time measurements between and within the experiments conducted on different days.

Results and discussion

Influence of e-liquid components on aerosol size distribution

The influence of e-liquid components on aerosol size distribution when using the Suorin device is shown in Figure 2. The geometric mean (GM) and geometric standard deviation (GSD) for the particle number distributions were 269 ± 5 nm (1.78 ± 0.08) for PG:VG, 258 ± 15 nm (1.86 ± 0.04) for PG:VG + EM, 296 ± 18 nm (1.71 ± 0.08) for PG:VG + Nic, 286 ± 6 nm (1.75 ± 0.08) for PG:VG + EM + Nic respectively. The GM (GSD) determined from APS data were 920 nm (1.45) for PG:VG, 870 nm (1.42) for PG:VG + EM, and 980 nm (1.45) for PG:VG + EM + Nic. The large error bars (Figure 2(a)) are indicative of the low repeatability for PG:VG. One-way ANOVA of the GM and GSD of the SMPS measurements indicate no statistically significant differences (p > 0.05) between aerosols from different e-liquids (Zhao et al. 2016, 2018).

Figure 2.

Mass and number size distribution of aerosols generated using a Suorin pod device from different e-liquids (a) PG:VG (b) PG:VG + EM (c) PG:VG + Nic and (d) PG:VG + EM + Nic measure by SMPS.

The number size distributions measured by SMPS and APS together (Figure 3) indicate a unimodal distribution of particles with a mode size of ~300 nm, irrespective of the e-liquid used. However, several studies have reported a bimodal distribution, with the modes reported at 200 nm and 1 μm (Zhao et al. 2018). The unimodal distribution in our study can be primarily attributed to the evaporation and aging of particles, causing the bimodal distribution to become unimodal as the aerosol ages (Schripp et al. 2013). In the present study, the SMPS and APS measurements were obtained after the aerosol had aged and the concentration had reached steady state (i.e. ~50 min after the start of the experiment). Further, the dilution air provided (1:1 ratio of e-cig aerosol to dilution air) may have facilitated the quenching of the aerosol leading to quicker evaporation.

Figure 3.

A typical particle number concentration plot spanning four orders of magnitude of particle size range using combined SMPS and APS data (Suorin data).

Variability in size distribution and mass concentration

Figure 4 shows average and standard deviations of the real-time mass concentrations measured by pDR for each of the four e-liquid combinations for 4 aerosolizing experiments of 3-hour duration each, using the Suorin pod device. One-way ANOVA of the real-time pDR measurements averaged over 15 min indicate statistically significant differences (p < 0.05) between repeated experiments on different days. Experimentally we aimed to reduce variability as much as possible by keeping pods and coils the same, changing coils at same intervals, and consistently refilling devices at the start of each experiment. However, our data suggests that these devices are not capable of producing consistent aerosol concentrations in a single day or from day to day.

Figure 4.

Time-resolved PM10 mass concentration of aerosols generated with the Suorin pod device from different e-liquids (a) PG:VG (b) PG:VG + EM (c) PG:VG + Nicotine and (d) PG:VG + EM + Nic; measured using four pDR 1200 (one per chamber).

Comparison between time-resolved and gravimetric data

Comparison between time-resolved and gravimetric mass concentration was done to calculate a correction factor, needed for optical PM instruments. Gravimetric PM₁₀ (average±Stdev) mass concentration was 34.3 + 59.5 mg/m³ for PG:VG, 16.2 ± 18.4 mg/m³ for PG:VG + EM, 28.1 ± 39.3 mg/m³ for PG:VG + Nic and 7.4 ± 6.1 mg/m³ for PG:VG + EM + Nic. The filter-based mass loading of PM10 aerosol only contributes 1–5% of the calculated concentration derived from difference in pod weight measurements, indicative of several factors such as evaporation losses, and losses to the sampling setup. On top of this, the viscous nature of the aerosol (droplets may appear larger to the optical sensor) and optical properties of the aerosol that differ significantly from those of the Arizona Road Dust used to calibrate the instruments may cause additional differences. No relative humidity correction was performed since the RH measured inside the chambers was found to be ~15%.

Thus, aerosol dynamics make evaluation between studies done by different investigators using different devices hard to compare. It has been reported that puffing protocol (i.e. duration, frequency and volume), dilution air, and length of time measurement are important parameters that influence the particle dynamics of aerosols generated from e-cigarettes (Ingebrethsen et al. 2012; Mikheev et al. 2016, 2018). Similar observations were also reported by Ruprecht et al. 2014, where they found that nicotine-enriched e-cigarettes produced lower PM levels than similar nicotine-free e-cigarettes. Ruprecht et al. 2014 attributed this phenomenon to the change in the optical properties of the PM such as color, morphology and other physical and chemical parameters. Furthermore, the argument is not limited to optical instruments and may also be applicable to gravimetric measurements. The filter substrate media is not able to capture the semi-volatile aerosols that are present in the environment. Further, adding flavors increases the vapor pressure and thus increases the volatility. Hygroscopicity and volatility of aerosols generated from e-cigs contribute to large extent toward the variability (Ingebrethsen et al. 2012). PG:VG is the main constituent in the e-liquid as well as in aerosol droplet formed in the vaping process (Gillman et al. 2020). Aerosols are generated both in the particle phase as well as the vapor phase (Pankow et al. 2018). For the scope of the present study, aerosols were only measured for the particle phase.

Moreover, the dynamics and changes of the aerosol size distribution are a function, in part, of the retention time in the chamber (Son et al. 2020). In larger chambers or room environments, the existing airborne particles might affect aging, for example, owing to coagulation. Thus, air exchange rate and chamber size are important variables which influence the physical characteristics of aerosols generated from e-cigarettes. The theoretical time taken to reach steady state with the new device is 68.19 min which is close to the initial experimental value of 50 min.

Modified Experimental setup

The protocol that was used in the first part of our study has been described before (Hilpert et al. 2019), but with a mod device instead of a pod. Thus, we selected the Vinci, a mod device, to reduce the variability in our experimental results and made the following changes to the experimental setup shown in Figure 1: 1) The Vinci device can only be activated by pressing a button using an actuator, so one was added; 2) the pump programing was altered to be on for 10 seconds, (although the actuator pressed the Vinci button to maintain 3 second puffs); and 3) the interval between puffs was increased to 99 seconds to account for a denser aerosol generated by the Vinci device. Pump programing was increased to 10 seconds on during the puff time to insure all of the aerosol is pumped from the tubing into the chamber from the 3 second puff. The modified experimental setup is shown in Figure 5.

Figure 5.

Modified setup using two Vinci devices over 3 days (The lab is constructing a 4-way actuator system to be able to generate different aerosols into each of the 4 chambers).

As a result of this new device and modifications to the aerosol generation regime, the variability in the concentration generated in the exposure chamber was significantly reduced as shown in Figures 6 and 7. The influence of e-liquid components on aerosol size distribution when using the Vinci device is shown in Figure 7. The geometric mean (GM) and geometric standard deviation (GSD) for the particle number distributions were 247 ± 22 nm (2.06 ± 0.05) for PG:VG, 337 ± 15 nm (1.81 ± 0.06) for PG:VG + EM, 307 ± 53 nm (1.89 ± 0.16) for PG:VG + Nic, 364 ± 5 nm (1.65 ± 0.03) for PG:VG + EM + Nic respectively. Thus, by using the Vinci mod device we were able to demonstrate significantly reduced variability between two separate devices and three days, suggesting that variability seen in Figures 2–4 with the Suorin devices is due to poor quality control of the devices themselves (battery, charger) and/or the pods (electrical contact between coil and battery, or the coil). The modified design setup is proposed for future studies and for the design and development of standardized protocols for generation of aerosols from e-cigarettes.

Figure 6.

PG:VG PM₁₀ aerosol concentration in the exposure chamber using the pDR device after modification.

Figure 7.

Mass and number size distribution of aerosols generated using a Vinci mod device from different e-liquids (a) PG:VG (b) PG:VG + EM (c) PG:VG + Nic and (d) PG:VG + EM + Nic measure by SMPS.

Conclusions

In the first part of the present study, we found that even controlling all the factors known to introduce variability in aerosol generation, (using the same e-cig model, coil resistance, voltage) the selected device still introduced considerable within day and between day variability. Most of the variability disappeared when using a different device with a more robust configuration. The large variability from pod devices could be due to automatic activation when air is drawn through the mouthpiece, likely due to poor quality control of one or several of the e-cig components. Our results highlight the need for a better understanding of all the factors affecting exposure, especially for in vitro and in vivo experiments, and the need to develop standardized protocols for the generation of aerosols from e-cigarettes.