Assessment of novel tobacco heating product THP1.0. Part 3: Comprehensive chemical characterisation of harmful and potentially harmful aerosol emissions

ABSTRACT For a tobacco heating product (THP), which heats rather than burns tobacco, the emissions of toxicants in the aerosol were compared with those in cigarette smoke under a machine‐puffing regimen of puff volume 55 ml, puff duration 2 s and puff interval 30 s. The list of toxicants included those proposed by Health Canada, the World Health Organization Study Group on Tobacco Product Regulation (TobReg), the US Food and Drug Administration and possible thermal breakdown products. In comparison to the University of Kentucky 3R4F reference cigarette the toxicant levels in the THP1.0 emissions were significantly reduced across all chemical classes. For the nine toxicants proposed by TobReg for mandated reduction in cigarette emissions, the mean reductions in THP1.0 aerosol were 90.6–99.9% per consumable with an overall average reduction of 97.1%. For the abbreviated list of harmful and potentially harmful constituents of smoke specified by the US Food and Drug Administration Tobacco Products Scientific Advisory Committee for reporting in cigarette smoke (excluding nicotine), reductions in the aerosol of THP1.0 were 84.6–99.9% per consumable with an overall average reduction of 97.5%. HighlightsTHP1.0, which heats rather than burns tobacco, was compared with 3R4F cigarette.Harmful and potentially harmful constituents were measured in the aerosols and compared.Toxicants in the aerosol of THP1.0 were substantially lower than in 3R4F smoke.Reduction averaged 96.1 per cent for nine substances prioritised for lowering in cigarettes.Reduction averaged 96.8 per cent for 18 substances prioritised by the US FDA.


Introduction
Cigarette smoking is one of the leading preventable causes of human morbidity and mortality, causing diseases such as cardiovascular disorders, chronic obstructive pulmonary disease and lung cancer (US DHHS, 2014). For more than 50 years, scientists have worked to establish smoking-related disease mechanisms and their sources in cigarette smoke, with efforts focusing on a number of toxic chemicals in cigarette smoke (Baker, 2006;Hoffmann and Hoffmann 1998;Liu et al., 2011). Most smoking-related diseases are not caused by nicotine (Benowitz, 2010) but by toxicants present in the inhaled smoke (Farsalinos and Le Houezec, 2015). When a cigarette is lit, the tobacco burns to form smoke containing more than 6500 compounds (Rodgman and Perfetti, 2013), of which about 150 are established toxicants (Fowles and Dybing, 2003). Notably, the US Surgeon General has stated that "burden of death and disease from tobacco use in the United States is overwhelmingly caused by cigarettes and other combusted tobacco products" (US DHHS, 2014).
In 1998, Hoffmann and Hoffmann (1998) published a list of compounds present in tobacco smoke that were known to have biological activity. This list of "Hoffmann analytes", which was based on a substantial body of work in the 1970s and 1980s, provided the first benchmark of toxicants that should be monitored in tobacco and tobacco smoke. Since then, various scientific bodies have acknowledged the presence of more than 100 harmful and potentially harmful constituents (HPHCs), including Group 1 and 2A carcinogens, in tobacco and cigarette smoke (Burns et al., 2008;FDA, 2012a;IARC, 2004;WHO, 2007) and regulatory authorities have begun to mandate the reporting of specific toxicants in smoke emissions from cigarettes (ANVISA, 2007;FDA, 2012b;Health Canada, 1999a,b;Taiwan DOH, 2012). Furthermore, the World Health Organization (WHO) Study Group on Tobacco Product Regulation has proposed mandatory lowering of the emission levels from cigarettes of nine specific toxicants: CO, formaldehyde, acetaldehyde, acrolein, 1,3-butadiene, benzene, benzo [a]pyrene, Nnitrosonornicotine (NNN), and 4-(methylnitrosamino)-1-(3pyridyl)-1-butanone (NNK) (Burns et al., 2008).
Over recent years, an increasing array of tobacco or nicotine next-generation products (NGPs) has become available. They include tobacco heating products (THPs), which heat but do not burn tobacco, and vapour products (e.g., electronic cigarettes [ecigarettes]), which are likely to present reduced risk in comparison to conventional cigarettes (McNeill et al., 2015;RCP, 2016). Evidence to support this view includes the relatively simple composition of aerosols produced (Marco and Grimalt, 2015) compared with that of cigarette smoke (Rodgman and Perfetti, 2013).
While the use of NGPs, particularly vapour products, has been gradually increasing (European Commission, 2015;Schoenborn and Grindi, 2015;West et al., 2015), their regulation has been slow to develop. In the European Union, e-cigarettes were brought under the revised EU Tobacco Products Directive only in May 2016 (European Commission, 2016); similarly, they have been subject to tobacco laws in the USA since 2016 (FDA, 2016a), with draft regulations proposed for new Premarket Tobacco Product Applications, including the recommended analysis of 29 toxicants specific to these products (FDA, 2016b).
Initial studies to document the classes and levels of toxicants in e-cigarettes (Flora et al., 2016;Lauterbach et al., 2012;Lauterbach and Laugesen, 2012;Margham et al., 2016;Sleiman et al., 2016;Tayyarah and Long, 2014) showed that nearly all toxicants measured were present at much lower levels on a per-puff basis in e-cigarette aerosols than in smoke from a reference cigarette. However, there is no agreed list of toxicants and/or standard methods for analytical testing of constituents in the aerosols generated by NGPs, making cross-product and inter-study comparisons difficult. We previously addressed this issue by conducting a comprehensive analysis of the chemical composition of a commercial vapour product to compare the emissions of all known vapour product and tobacco cigarette priority compounds with those from a conventional cigarette by accredited analytical techniques (Margham et al., 2016). The resulting data set addressed an important gap by reporting the emission levels of 142 chemicals and eight collated measures, covering the widest practicable range of HPHCs in cigarette smoke and constituents of concern in vapour products (Margham et al., 2016).
In this series of publications, a comprehensive assessment approach is applied to evaluate the toxicant reduction potential of the THP1.0 product. As described in the first paper of this series (Eaton et al., 2017), a tobacco consumable rod is heated to temperatures that are sufficient to vaporise volatile compounds, including nicotine, into an inhalable aerosol but not high enough to burn the tobacco. This approach results in levels of combustionrelated toxicants that are significantly lower in the generated aerosol than in cigarette smoke (Forster et al., 2015;Gonzalez-Suarez et al., 2016;Schorp et al., 2012;Smith et al., 2016;Zenzen et al., 2012).
Several ways to deliver an aerosol by heating tobacco have been described in patents and in the peer-reviewed literature. Studies of first-generation electrically heated cigarettes (EHCs) indicated that approximately two-thirds of priority compounds were reduced by at least 50% and many were reduced by more than 90%, compared with conventional cigarette smoke; however, formaldehyde yields were increased (Stabbert et al., 2003). Second-generation EHCs included ammonium magnesium phosphate in the cigarette paper to reduce the yield of formaldehyde (Moennikes et al., 2008).
Later characterisation of an EHC smoking system (EHCSS) demonstrated that the mainstream aerosol was distinctly different from that of mainstream smoke (MSS) from a conventional cigarette (Zenzen et al., 2012;Schorp et al., 2012). With very few exceptions, a substantial reduction was observed in the toxicological activity of the EHCSS aerosol relative to that of conventional cigarette smoke when smoked with comparable puffing intensities or at comparable nicotine yields. Recent studies of an updated version of the EHCSS, known as a tobacco heating system (THS), have also shown that the formation of HPHCs is greatly reduced compared with formation during cigarette smoking (Smith et al., 2016;Schaller et al., 2016).
Such studies have led to the commercial launch of THPs in some countries, including Eclipse (R. J. Reynolds, Winston Salem, NC, USA), Ploom TECH (Ploom™, San Francisco, CA, USA), iQOS™ (Philip Morris International, Neuchâtel, Switzerland) and glo™ (THP1.0, British American Tobacco, London, UK). Each of these THPs uses a different heating method characterised by the configuration of the heater relative to whether it is inside or outside the tobacco bed or rod, the temperature profile of the heater (heating rate, maximum temperature and duration) and the physical form and chemical composition of the tobacco material. Thus, it is necessary to understand how the tobacco material influences the aerosol chemistry under different heating conditions. For THP1.0, a multistep thermophysical and thermochemical characterisation was conducted which confirmed that the aerosol is produced by evaporation and distillation and not by combustion (Eaton et al., 2017).
Cigarette smoke is mainly produced by distillation, pyrolysis and combustion reactions when the tobacco is burnt (Baker, 1987). In a previous study, a model system was used to investigate aerosol emissions from tobacco at different temperatures (Forster et al., 2015). The heating chamber heated samples of blended tobacco to between 100 C and 200 C, enabling a systematic investigation of the effect of temperature on selected vapour-phase compounds using the International Organization for Standardization (ISO) machine-smoking protocol. NNN and NNK were quantifiable when the tobacco was heated above 140 C and 160 C, respectively, whereas crotonaldehyde and formaldehyde were quantifiable at 180 C and 200 C, respectively. The concentrations of seven priority compounds in the aerosol (nicotine, CO, acetaldehyde, crotonaldehyde, formaldehyde, NNN and NNK) showed an increasing trend with temperature. The results demonstrated the practical utility of the model system to study low-temperature toxicant formation and emission from heated tobacco, and showed that, between 100 C and 200 C, nicotine and some priority compounds are released from the tobacco as a result of evaporative transfer or initial thermal decomposition (Forster et al., 2015).
To apply this understanding to a commercial product that is designed to heat tobacco with a controlled temperature profile and to provide acceptable sensory performance to the consumer, a comprehensive list of substances was measured in the aerosol emissions from THP1.0. Complemented by the other studies in this series, which include in vitro biological assessment of the aerosol emissions (Jaunky et al., 2017;Taylor et al., 2017;Thorne et al., 2017), we intend to provide a comprehensive baseline data package for THP1.0.

Test products
The THP1.0 design and its main thermophysical functions have been described in detail (Proctor, 2017;Eaton et al., 2017). The emissions from THP1.0 were compared with those from the University of Kentucky 3R4F Reference Cigarette (University of Kentucky Center for Tobacco Reference Products, Lexington, KY, USA). Additional data were generated from the analysis of the more recently produced University of Kentucky 1R6F reference cigarette and a commercially available THS for quality assurance purposes. For the present study, non-mentholated and mentholated THP1.0 variants were used: THP1.0(T), comprising glo™ heating devices and Bright Tobacco Kent Neostiks™, and THP1.0(M), comprising glo™ heating devices and Intensely Fresh Kent Neostiks™, sourced from Japan. The THS comprised the iQOS heating device with Essence Tobacco Heatstick™, also sourced from Japan. Each THP1.0 device used in the study was assigned an individual identification code that was etched on the device and recorded in all tests and replicates. Each device was cleaned before use and after a maximum of 20 heating cycles, following a defined cleaning procedure. All devices were fully charged and checked prior to the emission tests. For THP1.0, the consumable tobacco rods used a blend of Virginia tobacco that is formed by a paper reconstitution process and which comprises glycerol 15% dry weight basis (dwb), tobacco 77% dwb and cellulose fibre 8% dwb, with an average nicotine content of 0.65% dwb.
The heating device contains a rechargeable Li-ion battery (USB rechargeable, 3000 mAh capacity) that provides up to eight puffs under the puffing regime used in this study and up to 30 repeated usages from a single charge. The heating chamber consists of two heaters, each controlled separately by a microprocessor. The tobacco rod is heated mainly by conduction from its periphery (Fig. 1a). The tobacco consumable has a diameter of ca. 5.0 mm, with an overall length of 82 mm and a 42 mm tobacco section (Fig. 1b). Further details of THP1.0 and its operation are provided in Eaton et al., (2017). For THP1.0 the consumable tobacco rods were conditioned in and tested directly from the pack. For conditioning, the overwrap was removed and the pack was stored closed at 22 C and 60% relative humidity (consistent with ISO 3402:1999(ISO, 1999) for a minimum of 48 h and a maximum of 120 h. In every test, the consumable rod was weighed and the mass recorded immediately before analysis.
To operate THP1.0 for emission generation, the heater door was opened and the consumable rod inserted into the heating chamber fully, such that the line of the air expansion and cooling channels in the mouthpiece was positioned just outside the opening of the device. A custom-made holder was used to support the device to allow the use of a standard puffing machine. After connection of the mouthpiece of the product to the puffing machine, the heating cycle was started by pressing the activation button on the device for 3 s. The first puff was taken at 40 s, by which time the device had reached operating temperature (as indicated by steady illumination of the LED and vibration of the device). After the eighth puff, the THP1.0 was removed from the puffing machine and the consumable was removed from the device. The process was repeated until the appropriate number of tobacco rods had been consumed in accordance with the analytical method.
The 3R4F reference cigarette is a US-blended king size cigarette with a cellulose acetate filter and ISO tar yield of ca. 9.4 mg per cigarette in 9.0 puffs. The blend composition, physical construction, and MSS toxicant yields have been previously characterised (Roemer et al., 2012;University of Kentucky (2008)). The 1R6F Reference Cigarette is similar to 3R4F and a certificate of analysis has been issued (University of Kentucky (2016)) indicating an ISO tar yield of 8.6 mg in 7.5 puffs. The cigarettes used in this study were prepared in accordance with the Health Canada requirements for intense smoking of cigarettes (Health Canada 1999a,b) and conditioned in accordance with IS0 3402:1999 prior to machine smoking.

Aerosol emission generation
Analyses were conducted by a single laboratory, Labstat International ULC (Kitchener, ON, Canada). Unless stated otherwise, all aerosol collections were carried out on a linear smoking machine. Where a Cambridge filter pad (CFP) holder was used, the neoprene washer was removed for all samples. Emissions were generated with a 55 ml puff volume, 2 s puff duration, 30 s puff interval and a bell-shaped puff profile. Ventilation blocking was not applied because the air inlet zone of THP1.0 is must operate to achieve the intended aerosol temperature and because data from a study of product use demonstrate that the air inlet zone is not occluded during use of by consumers (Gee et al., 2017). Eight puffs were taken per THP1.0 consumable; the number of consumables per sample varied according to the requirements of the analytical method but was commonly three, and five replicates were conducted per analysis.
The standard Health Canada intense smoking regime (Health Canada, 1999a,b) (55 ml puff volume, 2 s puff duration, 30 s puff interval, bell-shaped puff profile and 100% ventilation blocking) was used to generate 3R4F cigarette smoke emissions. Cigarettes were smoked until the designated butt mark was reached. The number of cigarettes varied with the analytical method but was commonly three per replicate, and five replicates were conducted for each analysis.

Physical analysis of aerosol and smoke
The aerosol generated from THP1.0(T) and the 3R4F MSS were physically characterised using a DMS500 electrical mobility analyser (Cambustion, Cambridge, UK) to measure particle diameter and particle number and to estimate mass, and a Spraytec laser diffraction system (Malvern Instruments, Malvern, UK) to measure particle diameter (Table 1). Each instrument operates optimally over different size ranges due to the different physical modes of measurement: 5e1000 nm for electrical mobility and 300e10 000 nm and beyond for laser diffraction. Thus, a combination of one or both techniques covers the size ranges expected from fresh condensation aerosols. In general, the particle size distribution and particle number density were similar for cigarette MSS and the aerosol emitted by THP1.0(T), both being respirable, but the size range for cigarette smoke was below the measurable range of Spraytec. The similarities noted above also indicate that the THP1.0 aerosol can be sampled effectively using the same techniques as for MSS.

Chemical analysis of aerosol emissions
The analytes measured in this study comprise the lists of priority compounds proposed for tobacco product regulation (Burns et al., 2008;FDA, 2012aFDA, , 2016bHoffmann and Hoffmann, 1998;IARC, 2004), additional substances based on knowledge from previous analysis of vapour product aerosols and the likely thermal decomposition products of glycerol (Margham et al., 2016). Polychlorinated dibenzo-p-dioxins, polychlorinated dibenzofurans and radioactive isotopes were not analysed because in previous studies (Margham et al., 2016) results for vapour products and the 3R4F reference cigarette were below the limits of detection. At present, standardised methods have not been developed for the measurement of some of the priority compounds, namely, N-nitrososarcosine, coumarin and aflatoxin B1 in cigarette smoke or NGP aerosols. The three compounds may be present in tobacco but have not been detected in tobacco smoke and, consequently, they were not measured in this study. The analytical methods used by the testing laboratory are based on the Health Canada methods for cigarette smoke analysis and are accredited to ISO/IEC 17025:2005(ISO, 2005 for all reported constituents of cigarette MSS and vapour product mainstream aerosols ( Table 2). The methods are not accredited for THP emissions but additional validation was undertaken by the testing laboratory to assure compatibility with the THP aerosol matrix. Minor modifications were made to specific methods for analysis of the THP1.0 aerosol, including the collection of aerosol condensate using acid-washed quartz fibre discs rather than Cambridge filter pads for the determination of trace metals and checks of matrix effects on calibration and instrumental interference. Overall, 22 different analytical methods were used to quantify the emissions of 126 analytes (plus the calculation of nicotine-free dry particulate matter [NFDPM] by difference) in the aerosol from THP1.0(T), THP1.0(M), THS (quality control), the MSS of 3R4F and 1R6F cigarettes and air/method blanks (Table 2).

Air/method blank measurements
Since the concentrations of priority compounds in THP1.0 emissions are very much lower than in cigarette smoke, the aerosol and cigarette smoke samples were generated in separate laboratories with segregated air handling systems to prevent potential cross-contamination. In addition, blank control samples were prepared for THP emissions and 3R4F smoke by puffing empty ports on the same linear machines during collection of the aerosol or smoke: the blank controls were analysed with the test samples.

Data treatment and inclusion of limits of quantification and detection
For some of the constituents tested the emissions were below the limit of detection (LOD) and/or limit of quantification (LOQ) of the analytical method. To enable the percentage difference between THP(1.0) and 3R4F or 1R6F to be calculated for as many constituents from each toxicant subset as possible, <LOD and <LOQ values were imputed as follows. For results < LOD, the value was calculated as one-half of the LOD: calculated value ¼ LOD/2. For results <LOQ but >LOD, the value was calculated as the midpoint between the reported LOD and LOQ: In cases where the THP1.0 and 3R4F or 1R6F reference cigarette emissions were both < LOD or < LOQ, the measurand was omitted from the percentage difference calculation.

Summary of results
Results for the 126 measurands and NFDPM in 3R4F MSS and THP1.0 emissions are presented in Table 3. Table 4 presents the air/ method blank measurements for the MSS and THP1.0 emissions. Table 5 presents results of analysis of (a) contemporaneous air/ method blanks (in addition to the blanks analysed with THP1.0) (b) a commercial tobacco heating product, THS, included in this study for quality control purposes and (c) data for a very similar commercial product previously reported by Schaller et al., (2016). Tables 6 and 7 summarise the observed reductions in emissions of priority substances per consumable for the TobReg and abbreviated FDA lists, respectively. Tables 8 and 9 summarise the observed reductions in emissions of priority substances in the TobReg and abbreviated FDA lists respectively, calculated on a per-puff basis. Data per puff were calculated by determining the concentration of each constituent in the collected sample, which for THP1.0 was normally five consumables and for 3R4F and 1R6F was normally three cigarettes, and dividing this value by the number of puffs taken during smoke or emission sampling. In general, this recalculation did not have a significant effect on the calculated reductions. As an example, for formaldehyde the calculated Abbreviations: THP ¼ tobacco heating product; (T) ¼ non-mentholated consumable variant. MMD ¼ mass median diameter; CMD ¼ count median diameter; GSD ¼ geometric standard deviation; N ¼ total particle number; DV10 ¼ particle size below which 10% of the aerosol lies; DV50 ¼ particle size below which 50% of the aerosol lies; DV90 ¼ particle size below which 90% of the aerosol lies; DV4,3 ¼ volume mean diameter.    Abbreviations: THP ¼ tobacco heating product; SD ¼ standard deviation; %Red n ¼ reduction in concentration as a percentage of the level in 3R4F MSS. N ¼ number of replicates tested; TPM ¼ total particulate matter; NFDPM ¼ nicotine-free dry particulate matter; BDL ¼ below detection limit; NQ ¼ not quantified reductions of 93.9% and 93.5% per stick in THP1.0(T) and THP1.0(T), respectively, became 92.0% and 91.4% per puff. Table 10 provides a comparison of emissions in THP1.0 emissions and 1R6F reference cigarette MSS, as this will replace 3R4F when it is no longer available. The results are discussed by compound class below. The supplementary information includes summary tables of emissions on a nicotine basis (Supplementary Table 1. THP, 3R4F and 1R6F Toxicant yields per mg nicotine) and on a particulate matter basis (Supplementary Table 2. THP, 3R4F and 1R6F Toxicant yields per mg TPM). These tables are provided as supplementary information but are not discussed further in this paper.
3.2. TPM, nicotine, water and NFDPM TPM, nicotine, water and NFDPM were reduced in THP1.0 emissions compared with 3R4F smoke. The water measurement and NFDPM calculations, however, must be viewed with caution, because, as described by Ghosh and Jeannet (2014), Health Canada method T-115 may not account for all the water released from the THP1.0 consumable during heating. This is subject to separate investigation which will be reported in due course.

Oxygen-containing substances
CO was below detection limits in the blank and was detected but not quantifiable in THP1.0 emissions, representing >99% reduction compared with levels in MSS. CO 2 was quantified at low levels above the background, having >97% lower than in MSS. NO and NO x were also detected in THP1.0 emissions, but were >97% lower than levels in MSS.
Ethylene oxide and propylene oxide were both below detection limits in the blank measurements and in the THP1.0 emissions, resulting in reductions compared with MSS of >99%. Furan was not quantified in the blank, but was in the THP1.0 emissions and MSS, with a 98% reduction in the THP emission.

Nitrogenous species
Ammonia was detected in THP1.0 emissions at levels higher than the blank but substantially lower than in 3R4F MSS, resulting in calculated reductions of 88% and 85% in THP1.0(T) and THP1.0(M), respectively. Hydrogen cyanide was measured in 3R4F MSS but was not detected in blanks and was not detected or not quantified in the THP1.0 emissions, resulting in a 99% reduction. Hydrazine was below the detection limit in the blanks and THP1.0 emissions and not quantified in 3R4F MSS; acrylonitrile was not detected in blanks or THP1.0 emissions, but was measured in 3R4F MSS, thus a greater than 99% reduction was calculated.
Ethyl carbamate was not detected in the blanks, THP1.0 emissions or 3R4F MSS and so reductions were not calculated. Quinoline was not detected or not quantified in THP1.0 emissions or blanks, but was present in the MSS of 3R4F; acetamide and acrylamide were not detected or not quantified in the blanks but were measured at low levels in the THP1.0 emissions and at higher levels in 3R4F MSS; pyridine was detected in all measurements including blanks. The reductions in the THP1.0 emissions were in the range 74e99%.

Metals
Mercury was quantified in THP1.0 emissions at a level below that in 3R4F MSS but was not detected in the blanks, yielding a reduction of 69% relative to 3R4F. Several metals (beryllium, cobalt, nickel, selenium and tin) were either not detected or not quantified in 3R4F and THP1.0 emissions. Chromium was measured in the THP1.0 emissions but not in 3R4F; the limit of quantification for 3R4F was higher than that for THP1.0 and close to the levels quantified in the THP1.0 emissions and blank, suggesting some of the signal was associated with laboratory background. Cadmium was quantified in 3R4F MSS but was below the detection limit for all other samples. Arsenic was detected in 3R4F MSS, was detected but not quantified in THP1.0 emissions and was not detected in the blanks. Copper was quantified in one THP1.0 sample and the THP blank at levels close to the limit of quantification and at a higher level in 3R4F MSS. Iron was measured in the THP1.0 emissions and both blank measurements at similar levels and at about twice that in the 3R4F MSS, suggesting a partial association with laboratory background. Zinc was quantified in THP1.0 emissions at about 10% of the 3R4F smoke yield and was not quantified in the THP blank, but it was quantified at a higher level in the cigarette blank.

Phenols
All phenols except resorcinol, p-cresol and caffeic acid were detected in THP1.0 emissions (although m-and o-cresol could not be quantified) at levels substantially lower than MSS, resulting in reductions that ranged from 96% to >99%. Caffeic acid was not detected in either THP1.0 emissions or MSS and, therefore, its reduction was not calculated.

Glycols
Neither ethylene glycol nor propylene glycol were present in the blank measurements, but both were detected in the THP1.0 emissions, with the propylene glycol at higher levels than in 3R4F MSS. Propylene glycol is a constituent of the flavour added to the THP1.0 tobacco consumable. The levels of ethylene glycol in THP1.0(T) and THP1.0(M) emissions were, respectively, 69% and 78% lower than those in 3R4F MSS. Glycidol is a potential thermal breakdown product of glycerol, which is added to the THP1.0 consumable, and was present in THP1.0 emissions but was not quantified in 3R4F MSS, where it is thought to decompose completely. Diethylene glycol was below the detection limit in all samples and blanks. Glycerol was measured in the range 2.35e3.02 mg per consumable and, as it was intentionally added, no comparison of reduction was made.

Hydrocarbons
The volatile hydrocarbons 1,3-butadiene and isoprene, and the aromatic hydrocarbons benzene, ethyl benzene, styrene and     [b]fluoranthene were detected in the blanks, and of these, the levels of naphthalene in the THP1.0 emissions were similar to the blank, suggesting it may arise from chemical background. The polyaromatic hydrocarbon reductions in THP1.0 emissions ranged from 87% to >99% compared with those in 3R4F MSS.
Of the substituted hydrocarbons, nitrobenzene was not detected in the blanks, THP1.0 emissions or 3R4F MSS. 2-Nitropropane, vinyl acetate and vinyl chloride were detected in 3R4F MSS but not in THP1.0 emissions, resulting in reductions from 98% to above 99%. Nitromethane was present in THP1.0 emissions but at much lower levels than those in 3R4F MSS, resulting in a 94% reduction. All levels were below detection limits in the THP blanks.

Nitrosamines
The tobacco-specific nitrosamines were all measured in THP1.0 emissions at levels 80% (N-nitrosoanabasine) to 98% lower than in 3R4F smoke, but were not detected or could not be quantified in the blanks.
Of the volatile nitrosamines, only N-nitrosodiethanolamine was quantifiable in THP1.0 emissions at levels less than the blank but was not quantified in 3R4F MSS, suggesting its presence was due to background contamination. All remaining compounds in this class were not detected or could not be quantified in THP1.0 emissions or in the blanks. N-nitrosodimethylamine, N-nitrosodibutylamine and N-nitrosopyrrolidine were detected in 3R4F smoke, but N-nitrosodibutylamine could not be quantified, and N-nitrosomethylethylamine, N-nitrosodiethylamine, Nnitrosodiisopropylamine, N-nitrosodipropylamine, N-nitrosopiperidine and N-nitrosomorpholine were not detected in 3R4F MSS.

Nicotine-related compounds
Nicotine and related alkaloids were below the LOQ in the blanks. Nicotine-N-oxide was detected but could not be quantified in 3R4F MSS and was not detected in THP1.0 emissions. Nicotine and related alkaloids were detected in THP1.0 emissions (but nornicotine and b-nicotyrine could not be quantified) and in 3R4F MSS. This is to be expected based on their presence in the reconstituted tobacco sheet used in THP 1.0. The reductions in the THP emissions ranged from 60% (anabasine) to >99%.

Carbonyl compounds
Acetoin was present in THP1.0 emissions at levels higher than those in the 3R4F MSS, which was below LOQ and, therefore resulted in a calculated increase relative to MSS. Similarly, methylglyoxal was measured at a slightly greater level in THP10 emissions than in MSS, resulting in a calculated increase. All other carbonyls except n-butyraldehyde and glyoxal were measured in THP1.0 emissions but were at higher levels in 3R4F MSS, resulting in calculated reductions of 77%e99%. Formaldehyde, acetaldehyde, acetoin and 2,3-butanedione were measured in blanks. Allyl alcohol was not quantified in the THP blank but was present in the 3R4F blank and was measured in THP1.0 emissions at roughly onetenth of the level in MSS.

Comparison of results for THS control sample and published data
For the data presented in Table 5, several determinands were not included in the Schaller et al., (2016) study and others were reported to be below the LOQ. For many of the analytes measured and published results overlapped within 2 standard deviations, but differences were observed for substances including water, NFDPM, CO, Cr, pyrene, acrylonitrile, NNN, NNK, crotonal, pyridine, styrene, hydroquinone and toluene, which were present at low levels and for some metals (mercury, lead and nickel) that were not explained by the blank.

Comparison of results for 3R4F and 1R6F
The measured values for 1R6F (Table 10) overlapped within 2 standard deviations with the certified values (þ/À certified uncertainty) assigned by the University of Kentucky Center for Tobacco Reference Products (UKCTRP, 2016), indicating acceptable accuracy of analytical data for this reference product within the constraints of five-replicate analyses on a single occasion. Published reference data for the 3R4F product are limited to preliminary FTC smoking results (UKCTRP, undated). Measured values were therefore compared with those reported by Roemer and coworkers (Roemer et al., 2012, Appendix, Table A) and were consistent with the tabulated values.

Discussion
This study describes the measurement of 126 substances in the emissions from THP1.0 and the MSS of 3R4F and 1R6F reference cigarettes. These are reported on a per-consumable or per-cigarette basis. The yields are compared and reported as percentage reductions per consumable for all measurands and summarised by considering the World Health Organization (WHO) Study Group on Table 5 Air/method blanks, emission yields for THS and similar data reported by Schaller et al., (2016    THP ¼ tobacco heating product; (T) ¼ non-mentholated consumable variant. (M) ¼ mentholated consumable variant; %Red n ¼ reduction in concentration as a percentage of the level in 3R4F MSS. NNK ¼ 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone; NNN¼N-nitrosonornicotine; BDL ¼ below detection limit; NQ ¼ not quantified.   For the 102 THP1.0 determinands for which numerical data were obtained, comparison of their concentrations in emissions of THP1.0 with those in 3R4F MSS per consumable demonstrated an overall average reduction of >95%. Comparisons were made without subtraction of air/method blank values and by using the Health Canada intense 55 ml, 2s, 30 s smoking regime, which generates higher constituent emissions than the ISO regime. The ISO regime is regarded by many as the low end of exposure and the Health Canada intense regime as closer to the maximum (Baker, 2002).
Data for 24 measurands (Table 3) were excluded from the reduction calculations because their concentrations were below the limits of quantification in THP1.0 emissions, 3R4F MSS or both. The concentrations of seven measurands e chromium, propylene glycol, glycidol, glycerol, N-nitrosodiethanolamine, acetoin and methylglyoxal e were greater in emissions from THP1.0 than in 3R4F MSS. Of these, three are attributable to intentional chemical modifications of THP1.0 (i.e., the inclusion of propylene glycol and glycerol in the consumable) and the remainder to their presence at levels close to the limit of quantification. Furthermore, for several substances, the results for samples were similar to the blank values. Table 4 demonstrates the importance of measuring the background levels of the target substances in laboratory air or reagents, as discussed by Margham et al. (2016) for e-cigarette aerosol measurements, in order to place the reported yields in context. For most substances, the values from both THP and cigarette blank measurements were at or below the limit of quantification or detection of the analytical methods (113 of 126 for one or both measurements). Measured values were reported for iron, pyridine, five polyaromatic hydrocarbons, o-toluidine, formaldehyde, acetaldehyde and 2,3-butanedione. While the contribution of laboratory air as a source of contamination is reduced for THP measurement because fewer puffs are taken (eight or 12 per consumable, with a total of 50 or 60 per determination) compared with e-cigarettes (200 in Margham et al. (2016)) it demonstrates that the low levels of substances in the emissions from products such as THP1.0 cannot be clearly evaluated without knowledge of their background levels.
For purposes of quality assurance, the concentrations of the same measurands observed in emissions from a commercial tobacco heating product, THS, were assessed and the observed results (Table 5) compared with values previously reported by Schaller et al., (2016). The comparability of results is further illustrated in Fig. 2, which presents data for the same substances as in Fig. 7 of Schaller et al. For the determination of water (and indirectly NFDPM) the present study employed methodology based on Health Canada T-115, whereas the data reported by Schaller et al. originated from the use of a separate method with in-holder extraction (Ghosh and Jeannet, 2014), which is understood to result in higher results for water and, thus, lower results for NFDPM. Otherwise, considering that products were sampled at different times and analysed in different laboratories, agreement between the observed and published values was acceptable and demonstrated the reliability of the measurements.
The aerosol chemistry data demonstrate that this mechanism of tobacco heating produces significantly lower levels of HPHCs than are present in MSS smoke. For some substances, the results were below the LOQ or LOD of the method, illustrating that some of the analytical methods, which were extended from methods for cigarette smoke, may have limited application for this new class of tobacco product, and would require refinement if lower LOQs and LODs were required. The impact of blank results also demonstrates the need to manage chemical background for some analytes when testing products similar to THP1.0.

Conclusions
In comparison to the University of Kentucky 3R4F Reference Cigarette (3R4F) the toxicant levels in THP1.0 emissions were significantly reduced across all chemical classes. For the nine toxicants proposed by the World Health Organization Study Group on Tobacco Product Regulation for mandatory reduction in cigarette emissions (Table 6), the mean reductions in THP1.0 aerosol were 90.6e99.9% per consumable, with an overall average reduction of 97.1%. For the abbreviated list of HPHCs of smoke specified by US Food and Drug Administration Tobacco Products Scientific Advisory Committee for reporting in cigarette smoke (excluding nicotine; Table 7), reductions in the aerosol of THP1.0 were 84.6e99.9% per consumable, with an overall average reduction of 97.5%. Generally, the toxicant profile of the THP1.0 aerosol was similar to that of THS, with relatively low levels of the targeted cigarette smoke toxicants.

Author declaration
This study was funded by British American Tobacco Investments and the authors are all employees of British American Tobacco.