Open Characterization of Vaping Liquids in Canada: Chemical Profiles and Trends

Currently, there is a lack of comprehensive data on the diversity of chemicals present in vaping liquids. To address this gap, a non-targeted analysis of 825 vaping liquids collected between 2017 and 2019 from Canadian retailers was conducted. Prior to mass spectrometry analysis, samples were diluted 1:500 v/v with methanol or acetonitrile. Chemical compound separation and analysis was carried out using gas chromatography and triple quadrupole mass spectrometry (GC-MS/MS) systems operated in the full scan mode and mass range of 35–450 m/z. Mass spectrum for each sample was obtained in electron ionization at 70 eV and processed. Non-targeted identification workflow included use of automated mass spectral deconvolution and identification system (AMDIS), where required, as well as a number of commercially available spectral libraries. In order to validate identities, an in-house database of expected compounds previously detected in vaping liquids was used along with genuine analytical standards for compounds of interest. This resulted in a dataset of over 1,500 unique detected chemicals. Approximately half of these chemical compounds were detected only once in a single product and not in multiple products analyzed. For any sample analyzed, on average, 40% of the chemical constituents appeared to have flavouring properties. The remainder were nicotine and related alkaloids, processing, degradation or indirect additives, natural extractives and compounds with unknown roles. Data published here from the project on the Open Characterization of vaping liquids is unique as it offers a detailed understanding of products’ flavour chemical profiles, the presence and frequency of chemicals of potential health concern, as well as trends and changes in products’ chemical complexity over a three-year period. Non-targeted chemical surveillance such as this present valuable tools to public health officials and researchers in responding to emergent issues such as vaping associated lung injury or informing chemical based strategies which may be aimed at addressing product safety or appeal.


Supplementary Material
: Chemical compounds of analytical grade or higher purchased from Sigma Aldrich Canada.

Section 1: Open Characterization Project-Assigning roles to detected chemicals in vaping liquids
A wide variety of chemical compounds in vaping products have been detected in the Open Characterization Project. These chemicals have been assigned roles in order to have a better understanding of the part they may play within a vaping liquid. To identify the likely role ( Assigned roles for detected chemicals: 1.) Alkaloids -Class of organic compounds that contain a basic nitrogen atom. In vaping liquids, these most frequently include nicotine and nicotine-related minor alkaloids. Supporting literature used to assign this role are published scientific studies from PubMed, data obtained from PubChem, and also the following data sources: -The Chemical Components of Tobacco and Tobacco Smoke by Alan Rodgman, Thomas A. Perfetti (Second Edition) -Chemistry of alkaloids by P.B. Saxena (2007) 2.) Processing -Various chemical compounds known to be used in the manufacture of vaping and tobacco products, flavours and food industry such as solvents, diluents, and processing agents. In addition to patents, 5.) Indirect Additive/Leaching/Degradation -Chemical compounds previously detected as indirect additives in food contact materials. This category also includes chemicals known to be associated with degradation of various other chemical compounds found to be present in the individual samples, for example, glycidol being a degradation product of glycerin. Sources of information include PubChem, Substances Added to Food (formerly EAFUS) and searching of published studies on degradation, indirect food contact additives and leaching of various components of vaping liquids.
6.) Unknown roles -Two types of chemical compounds are identified with unknown roles: (1) Chemicals whose identities are not known. Chemicals whose identities are unknown are those that are true unknowns, meaning upon use of spectral library and spectral matching no appropriate chemical match was found in any mass spectral library used; National Institute of Standards and Technology (NIST), Wiley or Flavors and Fragrances of Natural and Synthetic Compounds.
Since no International Union of Pure and Applied Chemistry (IUPAC) chemical name or Chemical Abstracts Service (CAS) number could be identified for these chemical compounds their roles could not be determined.
(2) Chemicals for which identities are tentatively identified. There are a number of chemical compounds detected in vaping liquids for which identification is tentatively assigned (chemical name and CAS known) but for which literature synthesis generated none or inconclusive data. For example, a proportion of these chemical compounds with unknown role in vaping liquids did have matches with chemical compounds previously detected in yeast extract. Data matching was accomplished through publicly available informatics platform and data repository Yeast Resource Centre (University of Washington 2018). Yeast extract is a mixture of individual chemical compounds that may be used as food flavouring or enhancer(U.S. Food and Drug Administration 2020). Although a number of yeast related compounds detected in the Open Characterization Project could be originating from the yeast extract used for flavouring, some could be present as a result of product ageing, fermentation or presence of other microbes.

Chemicals with multiple roles
According to literature pertaining to the detected chemicals some may play multiple roles in vaping liquids, therefore more than one role may be assigned to a particular chemical. For example, a chemical may be found in a natural extract as well as have organoleptic properties, thus this chemical would be assigned the roles of natural extract and flavour. Literature used for assigning multiple roles have been discussed under various roles listed above.

Section 2: Open Characterization Project-Analytical method validation and performance
The analytical methods that employ the use of non-targeted analysis are most frequently full-scan methods that do not quantify but rather aim to identify chemical compounds present in the samples analyzed. In operating mass spectrometry detector in a full scan mode concentration levels are not determined, hence the limit of detection or ability to quantify a specific chemical at a prescribed concentration in a consistent manner is not an appropriate measure for method performance. Furthermore, the area under the detected peak for chemical generated in the full scan should not be relied upon to generate a corresponding calibrated concentration level. Instead validation can be performed using the repeated injection of the same sample over a period of time and checking for the detected compounds. In our case we have used laboratory prepared vaping liquid sample consisting of matrix (PG/VG in 50/50 w/w) and nicotine. This sample along with repeat analysis of previously analyzed sample was injected with processing batches or anytime maintenance was performed on the instruments. For example, rough estimates of signal to noise measurements while using full-scan were used for known, laboratory prepared nicotine concentration. Sample of laboratory-prepared vaping liquid containing nicotine, PG/VG and diluted 200 times with methanol were used to simulate and determine at which point nicotine would no longer be detected through varying added concentrations of nicotine and observing signal to noise after each injection. In case of nicotine, limit of detection using full scan was estimated using 3:1 signal to noise and determined at 0.03 mg/mL in this laboratory prepared vaping liquid sample. Such determination for all 1507 chemicals detected is not practical, needed, nor possible given the fact that genuine analytical standards for some infrequently detected chemicals are simply not available for commercial purchase. Further analytical efforts and investments in determining concentrations and limits of detection for non-prioritized compounds would not be justifiable.
Individual limits of detection for analytes of interest have been or will be determined employing the appropriate targeted methods which, depending on the compound, may include chemical extraction prior to analysis and use of labelled internal standards to correct for matrix effects. For example, targeted methodology for quantitation of nicotine has been already developed. Briefly this method employs use of single reaction monitoring (srm) mode while monitoring following ions: Nicotine (m/z) 84 and 162, quantifier and qualifier ions, respectively, Nicotine d7 (m/z) 87 and 169, quantifier and qualifier ions, respectively. Method performance of this method was assessed according to the EPA Regulation 40 CFR part 136 (Appendix B) method (U.S. Environmental Protection Agency 2011). Eight replicates of laboratory prepared vaping liquid using USP grade PG, VG and fortified to a nicotine level of 0.05 mg/mL were put through sample extraction and analyzed. The standard deviation associated with eight replicate analyses of laboratory prepared vaping liquid and processed through the entire analytical procedure was multiplied by the Student's t value of 2.998 (appropriate for a 99% confidence level with 7 degrees of freedom). The method detection limit (MDL) for nicotine was calculated to be 0.002 mg/mL. The limit of quantitation (LOQ) was calculated according to the US EPA method, where the standard deviation associated with the eight replicate analyses of laboratory prepared vaping liquid conducted to obtain the MDL was multiplied by a factor of 10. The LOQ was calculated to be 0.006 mg/mL. The example of nicotine provided here, shows significant increase in detection limits through the use of targeted approach, hence, follow up, targeted methods will include subset of samples that through full scan have been found to contain identified analyte of interest and those that have not.
Another approach to validate non-targeted methodology in our case was through the use of two different GC MS/MS instruments. This idea of validation for non-targeted approach through the use of different instrumental platforms has been studied more extensively and at a larger scale through U.S. EPA's collaboration trial ENTACT (Ulrich et al. 2019) as well as NORMAN network collaborative trial in Europe (Schymanski et al. 2015). Briefly, in Open Characterization study, 15 random samples from various flavour profiles of vaping liquids were analyzed using both, 7000C and Quantum GC, instruments. Although many similar chemicals were reported by the both systems, certain compounds were more likely to be detected by one system than the other in the full scanning mode. On average, Quantum GC was able to detect higher number of compounds in the same product presumably due to the fact that it is a higher-end, more sensitive instrument. However, 7000C was able to detect more frequently, polycyclic aromatic hydrocarbons (PAHs) due to the fact that this instrument is also known as a PAH analyzer. The proportion of identical chemicals detected by the both systems for the same sample analyzed ranged between 41 and 83%, (average 55%). This percent of overlap is significantly higher compared to NORMAN study results of 5.4% overlap among participating laboratories as well as ENTACT trial's preliminary results of 2.9% overlap when comparing GC, LC electro-spray ionization (ESI-), and LC ESI+, based methodologies. Higher overlap in Open Characterization study is likely due to a number of factors. First, vaping liquids are chemically less complex compared to the tested samples in both collaboration trials, raw river water (NORMAN), and, house dust, serum and silicone bands (ENTACT). Moreover, our comparison is between two GC MS/MS systems, run in the same laboratory environment with minimal sample pre-treatment, dilution with solvent only. For ENTACT trial, for example, sample extraction and pre-treatment, as well as analytical instruments differed significantly among participating laboratories. Finally, in our study the list of expected or "suspect" chemicals compiled from published literature was identical for two instruments used.
In the end, in order to complete the project of this magnitude in a timely manner, 810 vaping samples were analyzed using only one of the two available instruments. Samples for each instrument were assigned randomly. In order to account for bias in the subsequent targeted analysis, samples with positive detection as well as those with negative will be analyzed.

Section 3: Open Characterization Project-Automated Mass Spectral Deconvolution and Identification System (AMDIS)
AMDIS is an open source software used to aid in interpretation of complex GC/MS generated spectral data. The process is automated where the software detects background traces, calculates automatically noise levels and analyzes data and ion traces to detect peak maxima and unique traces to yield a clean spectra that can be searched in a spectral library (Stein 1999).
An example of a more complex sample of Creamy Custard with labeled 18mg/mL nicotine and 60/40 (PG/VG proportion) is illustrated in Figure S1. AMDIS has been especially useful in identifying detected compounds that co-elute with a very broad glycerol peak such as this one Figure S2. Figure S1. Chromatogram of Creamy Custard sample with co-eluting unknown peak Figure S2. Example of AMDIS workflow for Creamy Custard sample