An Approach to Flavor Chemical Thermal Degradation Analysis

Toxicological evaluations of flavor chemicals for use in inhalation products that utilize heat for aerosol generation are complicated because of the potential effect heat may have on the flavor chemical. The objective was to develop a thermal degradation technique to screen flavor chemicals as part of a toxicological testing program for their potential use in ENDS formulations. Based upon published data for acetaldehyde, acrolein, and glycidol from ENDS products (common thermal degradants of propylene glycol and glycerin), the pyrolizer temperature was adjusted until a similar ratio of acetaldehyde, acrolein, and glycidol was obtained from a 60/40 ratio (v/v) of glycerin/propylene glycol via GC/MS analysis. For each of 90 flavor chemicals, quantitative measurements of acetaldehyde, acrolein, and glycidol, in addition to semiquantitative non-targeted analysis tentatively identifying chemicals from thermal degradation, were obtained. Twenty flavor chemicals transferred at greater than 99% intact, another 26 transferred at greater than 95% intact, and another 15 flavor chemicals transferred at greater than 90% intact. Most flavor chemicals resulted in fewer than 10–12 tentatively identified thermal degradants. The practical approach to the thermal degradation of flavor chemicals provided useful information as part of the toxicological evaluation of flavor chemicals for potential use in ENDS formulations.


Introduction
Toxicological evaluations of flavor chemicals for use as ingredients in inhalation products that utilize heat to generate an aerosol are complicated because of the potential effect heat may have on the flavor chemical.Sufficient heat may cause pyrolysis or degradation of the flavor chemical, yielding different chemicals.In cigarettes where temperatures can reach 900 • C [1], Dempsey et al. [2] have provided a historical review of these flavor chemical evaluation efforts, which have generally incorporated evaluating the pyrolysisderived by-products [3,4].As reported in the literature, electronic nicotine delivery systems (ENDSs) utilize substantially lower temperatures to generate an aerosol, typically under 300 • C [5][6][7][8][9][10].However, non-temperature-regulated devices may exceed 300 • C, leading to the production of higher levels and different thermal degradation products [11].Although the lower temperature utilized in ENDS products precludes pyrolysis, thermal degradation of flavor chemicals is still possible.Evaluation of the potential thermal degradation products from flavor chemicals used in ENDSs is consistent with a published ENDS ingredient evaluation program [5] and requirements and guidance from regulatory agencies [12][13][14].
ENDS products generally consist of a base formulation consisting of a mixture of glycerin, propylene glycol, and nicotine, although some formulations consist of only glycerin or propylene glycol and nicotine [15][16][17][18].In addition to the base formulation, ENDS products contain various flavor chemicals.In a survey of English language websites conducted in 2014 [14], Zhu and colleagues counted 466 separate brands of e-liquid with 7764 unique e-liquid flavors.More recently, several authors have conducted studies identifying flavor chemicals in ENDS products with the majority but not all being used at low levels [19][20][21][22][23][24][25][26][27].For example, in an examination of 20 refill fluids, the total amount of chemical flavors exceeding 0.1% occurred in only four of the refill fluids [19].In other reports, most flavor chemicals were used at less than 1% in a brand comparison of menthol-and tobacco-flavored ENDSs [24,25].
The objective of the current work was to develop a thermal degradation technique that could be used to screen flavor chemicals as part of a toxicological testing program for their potential use in ENDS formulations.

Materials and Methods
Since the exact temperature (average or maximum) an e-liquid experiences along the coil of an ENDS product throughout a puff cannot be approximated with a single temperature [28,29] in a thermal degradation setup attached to a gas chromatograph/mass spectrometer (GC/MS), a practical approach was used.This practical approach was based upon previous work using propylene glycol showing that although different coil materials can affect the amount of carbonyl production, there appear to be similar thermal degradation processes occurring [30].The practical approach was based upon targeted chemical analysis of several e-liquids using the Juul device [31], which is a temperaturecontrolled ENDS product [8,9,31].Three chemicals (acetaldehyde, acrolein, and glycidol) were selected as indicators of the degradation of glycerin and propylene glycol to set the temperature of the thermal degradation setup attached to the GC/MS.A 60/40 percent ratio (v/v) of USP-grade glycerin and propylene glycol was used to set the analysis temperature in the thermal degradation setup.This ratio of USP-grade glycerin and propylene glycol was consistent with previous reports of commercial ENDS products [9,11,17,18,26,31].The temperature of the GC/MS setup was tested at 275 • C and in 25 • C steps between 350 • C and 575 • C until the ratio of quantifiable acetaldehyde, acrolein, and glycidol was similar to that reported for currently marketed Juul products [31].

Analytical Equipment
All work was performed at a contract research organization (RTI, Laboratories, Livonia, MI, USA).This work used a GC (Agilent 6890)/MS (Agilent 5972) with a Frontier Eco-Cup LF pyrolizer (EGNPY-3030D) and a VF-wax (Agilent CP9206) 30 m × 0.25 mm column with a nominal film thickness of 1 µm and helium carrier gas (0.7 mL/min).Based upon preliminary work monitoring acetaldehyde, acrolein, and glycidol levels from 10 µL samples of USP-grade glycerin and propylene glycol in a 60/40 ratio (v/v), a temperature of 475 • C was selected to conduct the subsequent flavor chemical thermal degradation studies.The desorption time was set to 1 min.Each flavor chemical sample was placed directly in the Eco-cup to avoid volatilization that might occur in an autosampler.Quantitative analysis focused on acetaldehyde, acrolein, and glycidol using calibration curves prepared from analytical standards for each compound.The method performance was verified by acceptable recovery of 5 ng of each compound added to the USP-grade glycerin and propylene glycol mixture as samples.In addition to the quantitative determination of acetaldehyde, acrolein, and glycidol from each flavor chemical, a non-targeted semiquantitative analysis was performed using a greater than or equal to 70% match factor with the GC/MS library (Wiley 7/National Institute of Standards and Technology 05 [NIST]) to tentatively identify other potential thermal degradation products from each flavor chemical.After every 10 flavor chemical samples a quantitative standard was desorbed to demonstrate quantitative accuracy for the targeted compounds.Details of the GC separation method are in Table S1.

Chemical Flavors
A total of 90 flavor chemicals were evaluated using the thermal degradation setup with 5 µL for liquids or 5 µg for solids being used (Table 1).The flavor chemicals were a convenience sample and had been reported in ENDS products [19][20][21][22][23][24][25][26][27].All flavor chemicals were food grade and compliant with Food Chemical Codex, if available, and their purity was provided by the chemical supplier.CAS-Chemical Abstract Services; FEMA-Flavor and Extract Manufacturers Association. 1 Match factor greater than 70%. 2 Thermal degradation also evaluated by Baker and Bishop [3]. 3 Flavor chemicals that had unacceptable chromatography resulting in unusable data. 4Thermal degradation also evaluated by Purkis et al., 2011 [4].

Results
The majority of flavor chemicals tested (46) transferred intact at 95% or greater, with 20 transferring at greater than 99% intact (Table 1).Another 15 flavor chemicals tested transferred at greater than 90% intact (Table 1).Most flavor chemicals resulted in fewer than 10-13 chemicals tentatively identified due to thermal degradation, which represented less than 9% of the estimated mass (Table 1).Thermal degradation analysis of a few flavor chemicals resulted in the tentative identification of more than 30 chemicals.For the few flavor chemicals that had zero percent intact transfer, fewer than nine thermal degradation products were identified for each one (Table 1).Chromatograms for six flavor chemicals (4-vinyl guaiacol, cinnamyl isovalerate, delta-3-carene, ethyl cinnamate, methyl-2-methyl-3-furyl disulfide, and methyl benzoate) were not suitable for analysis due to late elution of the compounds as evidenced by carryover into the next injection (Table 1; Figures S1-S9).
A total of 8 of the 90 flavor chemicals produced measurable amounts of acetaldehyde, acrolein, or glycidol (Table 2).Considering that 5 µL for liquids or 5 mg for solids of each flavor chemical was analyzed, the maximum total amount of acetaldehyde, acrolein, or glycidol measured (0.33 ng/µL acetaldehyde, 1.38 ng/µL acetoin, and 1.052 for glycidol) for any flavor ingredient was a low level of thermal degradation for the targeted compounds and represented approximately 0.02-0.07% of the main intact flavor ingredient peak.

Discussion
In a few cases the percentage of intact transfer was higher than the purity of the flavor chemical (e.g., Alpha Ionone, Cital, etc.) either due to the minimum purity stated by the manufacturer and/or due to some impurities being below the limit of detection.For two flavor chemicals (2,4-Decadienal, purity 91.93%, and Melonal, purity 85.96%) the levels of the targeted chemicals and the number of tentatively identified thermal degradants may have resulted from the impurities or carrier used for these flavor chemicals, rather than from the flavor chemicals.Some of the tentatively identified thermal degradation products from the flavor chemicals (Table S2) could be of toxicological concern (e.g., pulegone identified from butyl anthranilate or ethylene oxide identified from melonal) and would guide the development of targeted chemical analysis to determine the quantitative levels of these tentatively identified thermal degradation products from flavor chemicals.Alternatively, the most prudent approach would be not to use the flavor chemical in ENDS formulation development.The toxicological evaluation of the tentatively identified chemicals or their confirmation for further toxicological evaluation of the flavor chemical is beyond the scope of this work.
One limitation in this study, as evidenced by the chromatograms for six flavor chemicals (4-vinyl guaiacol, cinnamyl isovalerate, ethyl cinnamate, delta-3-carene, methyl-2methyl-3-furyl disulfide, and methyl benzoate) not being interpretable, is that it is likely slightly different analytical methods (e.g., different run time, columns, etc.) are required to screen a large range of flavor chemicals.A second limitation is the use of a 70% match factor to the Wiley/NIST database for chemical identification.A higher match factor (>70%) would increase confidence that the tentatively identified chemical is correct but is likely to reduce the number of tentatively identified chemicals, while using a lower match factor (<70%) would have the opposite effect, more tentatively identified chemicals with less confidence in their correct identification.Another limitation is that our use of a practical approach to set the temperature used in the GC/MS setup in this study was designed for the temperature-controlled JUUL device and would need to be adapted to other ENDS devices that have been reported to reach over 500 • C [11].The practical approach is flexible in that the degradation products of glycerin and propylene glycol can be changed, and their ratios can be adjusted for specific ENDS devices.Finally, single-chemical thermal degradation testing cannot predict what might occur when the flavor chemical is mixed with propylene glycol, glycerin, or other flavor chemicals in an ENDS formulation [31,32], which is why this method is proposed as one screening tool for flavor chemicals, and mechanistic degradation analysis was not performed.To be clear, thermal degradation testing of flavor chemicals is not intended as a substitute for aerosol constituent measurements from the final formulation used in a final ENDS product as has been previously reported [31,33,34].
Baker and Bishop [3] evaluated the pyrolysis of 291 flavor chemicals, and Purkis et al. [4] evaluated the pyrolysis of 91 flavor chemicals, with both ramping temperatures from 300 to 900 • C in their work.The results from the current thermal degradation study were remarkably consistent with this previous pyrolysis work, especially given the dramatic differences in temperatures used between current and previous studies [3,4] (Table 3).For sixteen of the comparable flavor chemicals, the intact transfer percent was within 3% of the intact transfer percent measured in one of the previous studies; however, for the other eight flavor chemicals, differences in the intact transfer percent exceeded 5%, likely due to the temperature differences between the studies.The purity of the flavor chemicals was slightly higher in the current study compared with the purity used in one of the previous studies [3] and slightly lower compared with the other previous study [4], except for D-limonene.As anticipated due to the differences in the used temperatures, there were differences in the number of and specific degradation products identified in the current study compared with the compounds identified from the two previous pyrolysis studies (Supplemental Table S2).As noted in Baker and Bishop [3], slow or constant heat of a chemical can produce different degradation products compared with rapid heating [35].CAS-Chemical Abstract Services; N/A = not applicable. 1 Flavor chemicals that had unacceptable chromatography resulting in unusable data. 2 Combination of E & Z-Citral.

Conclusions
In summary, a practical approach to thermal degradation testing was used as a screening tool for 90 flavor chemicals consistent with an ENDS ingredient evaluation framework [5].Over half of the flavor chemicals transferred intact at ≥95%, with another 16% transferring intact at ≥90%.Only eight flavor chemicals produced detectable levels of acetaldehyde, acrolein, or glycidol resulting from thermal degradation.Most flavor chemicals resulted in fewer than 10-12 thermal degradation products that represented less than 9% of the mass.Overall, the practical approach to the thermal degradation of flavor chemicals provided useful information as part of the toxicological evaluation of flavor chemicals for potential use in ENDS formulations.

Table 1 .
Chemicals evaluated via thermal degradation analysis.

Table 2 .
Amount of acetaldehyde, acrolein, and glycidol resulting from thermal degradation of tested flavor chemicals.
CAS-Chemical Abstract Services.LOQ-limit of quantification.ND-not detected, meaning no peaks were detected.

Table 3 .
Comparison of purity and intact transfer % of selected flavor chemicals.