Heat-not-burn tobacco products: a systematic literature review

Objective To review peer-reviewed evidence on heat-not-burn tobacco products (HnB), their secondhand emissions and use by humans; to identify differences between independent and industry-funded studies. Data sources Medline, Embase, PsycINFO, ProQuest, Scopus and Web of Science databases were searched up to 6 November 2017 for studies on HnB published after December 2009; reference lists were screened and other researchers contacted, yielding 637 records. Study selection Thirty-one publications on HnB secondhand emissions (n=16) or use by humans (n=15) were selected by two reviewers with excellent agreement (k=0.75). Data extraction Data on authors’ affiliations, HnB products, secondhand emissions and human exposure were extracted by one reviewer. Two reviewers assessed the quality of experimental HnB studies using the Effective Public Health Practice Project tool. Data synthesis Twenty out of 31 studies were affiliated with tobacco industry. Studies on secondhand emissions varied by methodology, products and comparators. Compared with cigarettes, HnB delivered up to 83% of nicotine and reduced levels of harmful and potentially harmful toxicants by at least 62% and particulate matter by at least 75%. Experimental HnB use studies were limited to one product, reductions of human exposure to toxicants varied between 42% and 96%. HnB use suppressed urges to smoke, but participants rated HnB less satisfying than cigarettes. While limited by methodological heterogeneity, findings were largely similar for independent and industry-funded studies. Conclusions Studies on HnB secondhand emissions and human use were heterogeneous and largely affiliated with the manufacturers. HnB exposed users and bystanders to toxicants, although at substantially lower levels than cigarettes.


AbsTRACT
Objective To review peer-reviewed evidence on heatnot-burn tobacco products (HnB), their secondhand emissions and use by humans; to identify differences between independent and industry-funded studies. Data sources Medline, Embase, PsycINFO, ProQuest, Scopus and Web of Science databases were searched up to 6 November 2017 for studies on HnB published after December 2009; reference lists were screened and other researchers contacted, yielding 637 records. study selection Thirty-one publications on HnB secondhand emissions (n=16) or use by humans (n=15) were selected by two reviewers with excellent agreement (k=0.75). Data extraction Data on authors' affiliations, HnB products, secondhand emissions and human exposure were extracted by one reviewer. Two reviewers assessed the quality of experimental HnB studies using the Effective Public Health Practice Project tool. Data synthesis Twenty out of 31 studies were affiliated with tobacco industry. Studies on secondhand emissions varied by methodology, products and comparators. Compared with cigarettes, HnB delivered up to 83% of nicotine and reduced levels of harmful and potentially harmful toxicants by at least 62% and particulate matter by at least 75%. Experimental HnB use studies were limited to one product, reductions of human exposure to toxicants varied between 42% and 96%. HnB use suppressed urges to smoke, but participants rated HnB less satisfying than cigarettes. While limited by methodological heterogeneity, findings were largely similar for independent and industry-funded studies. Conclusions Studies on HnB secondhand emissions and human use were heterogeneous and largely affiliated with the manufacturers. HnB exposed users and bystanders to toxicants, although at substantially lower levels than cigarettes.

InTRODuCTIOn
'Heat-not-burn' tobacco products (HnB) are electronic devices that heat processed tobacco instead of combusting it to supposedly deliver an aerosol with fewer toxicants than in cigarette smoke. Commercially available HnB systems like glo (produced by British American Tobacco (BAT)) or IQOS (Philip Morris International (PMI)) include a charger, a holder and tobacco sticks, plugs or capsules. Inserted into the holder, tobacco sticks are heated with an electronically controlled heating element. Other products, like iFuse from BAT or Ploom Tech from Japan Tobacco (JT), produce vapour from a non-tobacco source and pass it through a tobacco plug to absorb flavour and nicotine. 1 HnB products aim for a niche between combustible tobacco smoking and electronic cigarettes (e-cigarettes) that vaporise nicotine suspended in humectants.
'Safer' heated tobacco products that deliver nicotine but limit emissions of tar or carbon monoxide (CO) is a half-century old idea, 2 which had been unsuccessfully market-tested since 1988, first as 'Premier' by the RJ Reynolds Tobacco Company (RJR) and later as 'Eclipse' (RJR) and 'Accord' (PMI). 3 Of the current HnB products, IQOS was launched in several cities in Japan, Italy and Switzerland in 2014, iFuse was released in Romania in 2015 and glo and Ploom Tech were introduced to Japanese cities in 2016. Due to regulations restricting the sale of nicotine-containing e-cigarettes, 4 Japan was a fertile market for HnB producers, 5 suggesting that the products have potential 'for explosive global growth'. 6 By 2017, IQOS was available in 30 countries and was being considered by United States Food and Drug Administration for a reduced-risk product approval, 5 and the UK was one of the first countries to assign a separate taxation category for HnB products. 1 7 Committees advising the UK government carried out a systematic review of HnB studies but excluded research funded by HnB manufacturers, which comprise the majority of evidence published to date. 8 9 Because of the paucity of evidence from independent sources, it is important to look at evidence from tobacco companies 10 and to validate their findings. 11 A recent Public Health England report reviewed evidence related to HnB products, including data from manufacturers. 12 However, expeditiously published new findings call for an update. This review aimed to systematically identify and synthesise evidence from peer-reviewed studies on HnB tobacco products and to answer the following questions: 1. How do the currently researched and marketed HnB products compare with other tobacco and nicotine products: a. in exposure to toxicants and health risks to humans through primary use and secondhand exposure? b. in key performance characteristics (eg, nicotine delivery, use profile and user satisfaction)? 2. What is the population-level uptake of HnB products? 3. Are there any differences between independent and manufacturer-funded studies?

MeThODs search strategy and selection of studies
Medline, Embase, PsycINFO, ProQuest, Scopus and Web of Science databases were searched up to 13 July 2017 and the search was rerun on 6 November 2017. The full search strategy and search outcomes are To assess awareness and use of HnB products in a nationally representative sample.

Review
BAT, British American Tobacco; CHTP, carbon-heated tobacco product; HPHC, harmful and potentially harmful compounds; IT, Imperial Tobacco; JTI, Japan Tobacco International; LLTV, loose-leave tobacco vaporiser; PMI, Philip Morris International; PNTV, Prototype novel tobacco vapour product; RCT, randomised controlled trial; THP, tobacco heating product; THS, tobacco heating system. Cohen's kappa was calculated as a measure of agreement.

Inclusion criteria
The review included peer-reviewed studies that focused on HnB use by humans and the products' health risks associated with use and secondhand exposure to HnB emissions.

exclusion criteria
► Publication was not peer-reviewed or was a conference abstract.
► Published before 2010 or focus was a HnB device that is no longer available (eg, Premier, Eclipse, Accord). ► Focus was not a HnB device (ie, a device did not use tobacco to produce or flavour vapour). ► Publication was not in English, French, German, Lithuanian or Russian (languages known to authors). ► Animal or in vitro study (not directly related to human use). ► Publication presented the same data as earlier publication. ► Study assessed research methodology.

Data extraction and categorisation of included studies
The included studies were reviewed regardless of funding source. However, manufacturers that fund and report findings on their own products are inherently bound by conflict of interests. Throughout the review, funding sources of included studies were reported (table 1), and study outcomes were compared between independent and manufacturer-funded studies where the comparison was possible. Data on authors' affiliations, tested HnB products, methodology, HnB sidestream, mainstream and secondhand emissions and HnB use effects on humans were extracted to a predefined table by one reviewer (ES) and checked by a second reviewer (LSB).

Review
Tobacco products' emissions are categorised into mainstream, sidestream and secondhand smoke. Mainstream smoke is the smoke that a user draws in, 13 and is measured in laboratory using standardised machine smoking regimens to replicate human smoking. Sidestream smoke is emitted from the lit end of a burning tobacco product 13 and is measured in standardised indoor and outdoor environments by recruiting smokers or using machine smoking to test the products. Secondhand smoke is the combination of exhaled mainstream and sidestream smoke. 13 The same categorisation is used throughout the paper for emissions from e-cigarettes and HnB products. Studies were categorised to: i. Nicotine delivery and mainstream, sidestream and secondhand emissions of HnB products. ii. HnB use by humans (experimental, epidemiological and case studies).

Assessment of risk of bias
Quality of experimental studies on HnB use by humans was assessed independently by two authors (ES and LSB) using the Effective Public Health Practice Project tool. 14 The tool evaluates quality of quantitative studies by rating study selection bias, design, confounders, blinding, data collection method and withdrawals and dropouts as weak, moderate or strong. A study is considered of strong quality if no aspect has been rated weak, moderate if one aspect has been rated weak and weak if two or more aspects have been rated weak. 14

Data synthesis
Findings were summarised in a narrative synthesis and quantitative results compared between studies where possible. Studies on HnB emissions used either the International Organisation for Standardisation machine smoking regimen (ISO; 35 mL puff volume, 2 s puff duration, 30 s intervals between puffs, 14 puffs) or the Health Canada Intense regimen (HCI; 55 mL puff volume, 2 s puff duration, 30 s intervals between puffs, 14 puffs). The HCI regimen yields higher levels of harmful and potentially harmful compounds (HPHC), 15 but no machine smoking regimen corresponds to human smoking and exposure, 16 17 and their relevance to HnB use is not tested either. Reference products also differed between studies: the majority used 3R4F tobacco cigarettes (a reference product developed for research 18 ), others used commercially available cigarettes (nicotine and CO yields provided where known) and e-cigarettes (cigalike, pen-style and tank-style).
Levels of nicotine and HPHC 19 delivered per single HnB use to aerosol were calculated and presented as percentage of the levels in smoke from a single reference cigarette. Where data on HnB emissions were provided per puff, a single HnB use was calculated as 14 puffs with reference to ISO and HCI regimens. If independent and manufacturer-funded studies used the same puffing regimen and HnB device, nicotine and HPHC levels were compared post hoc using independent samples t-test. For studies on pharmacokinetic nicotine delivery of HnB, key pharmacokinetic characteristics were compared between HnB products where possible.
Studies on HnB use by humans were grouped by the HnB product (table 1), and findings on levels of exposure to biomarkers of HPHC (see table S1 in Supplementary file 1), nicotine delivery characteristics, human puffing topography, effect on urges to smoke and subjective satisfaction with the products were reported and compared between studies if possible.
Randomised controlled trials (RCT) used a confinement procedure: after randomisation, participants stayed at a trial site Review Table 3 Relative nicotine delivery in mainstream HnB or e-cigarette aerosol in comparison to nicotine delivered to mainstream cigarette smoke  (1.14) for 5 days and were restricted to only using cigarettes or HnB products, or abstained from smoking.
Compared with nicotine in smoke of reference cigarettes, nicotine in mainstream IQOS aerosol ranged from 57% to 83% across studies (table 3). One independent study 22 reported that IQOS delivered more nicotine than a cigalike e-cigarette but less than a pen-style or a tank-style e-cigarette (table 3). A study from glo manufacturers 36 reported that glo delivered 40% of nicotine compared with IQOS and 23% compared with reference cigarette, and a study from the manufacturer of iFuse 30 reported that iFuse per 14 puffs delivered less nicotine than a pen-style e-cigarette (72%) and a reference cigarette (19%) (table 3).
One independent study compared nicotine transfer rates (in this case defined as the ratio of nicotine in mainstream emissions to nicotine in a tobacco stick or a cigarette): nicotine transfer rates were higher for IQOS regular (23.4%) and menthol (23.5%) tobacco sticks than for the 3R4F reference cigarette (11.3%). 23

HPHC in mainstream HnB emissions
Two independent 21 23 and six manufacturer-funded studies 24-26 28-30 reported levels of HPHC in mainstream HnB aerosol compared with cigarette smoke (table 4).
One independent study 21 used data of 50 US cigarette brands 15 to compare levels of polycyclic aromatic hydrocarbons but, as a critique by PMI noted, 37 the authors had inadvertently used reference values obtained under HCI instead of ISO regimen. We provide both the originally published and recalculated ratios (table 4). The study assessing iFuse 30 calculated HPHC yield per 100 3 s puffs on iFuse but followed an HCI regimen for the reference cigarette, creating discrepancies in comparison with other studies (table 4).
One independent 23 and three manufacturer-funded studies 24-26 used the HCI machine-puffing regimen and reported findings on the same HPHC in mainstream IQOS emissions. Levels of CO (t(11)=1.28, p=0.23), water (t(8)=0.43, p=0.68) and total particulate matter (t(8)=1.77, p=0.11) did not differ statistically significantly between independent and manufacturer-funded studies. Compared with manufacturer-funded studies, the independent study reported less tar (9.

Particulate matter and HPHC in sidestream and secondhand HnB emissions
Seven studies, three independent 31-33 and four funded by tobacco manufacturers, 27 34-36 compared HnB sidestream or secondhand emissions with smoke of factory-made 27 31-34 36 or handrolled 31 32 cigarettes, pipes and cigars, 32 aerosol from a nicotine inhalator 35 or e-cigarettes. 31-33 35 A single study 35 funded by Imperial Tobacco company that does not manufacture HnB products 38 explicitly focused on sidestream emissions of a competitor's HnB product (IQOS). The study concluded that in contrast to a cigalike e-cigarette and a nicotine inhalator, IQOS produced sidestream emissions. Similarly, an independent study 32 concluded that higher particulate matter emissions from IQOS than from a pen-style e-cigarette could be explained by sidestream emissions.
Six studies, three independent 31-33 and three manufacturer-funded, 27 34 36 reported on particulate matter in HnB secondhand emissions. One study 31 reported that a pen-style e-cigarette and IQOS emitted 25% of the total particulate matter detected in smoke from a cigarette. Use of the e-cigarette produced higher peak concentration of particles in the air than use of IQOS, but the total amount and time for particles to disperse after use were longer for IQOS and composition of particles was not considered. 32 Most particles emitted by IQOS were <1000 nm, 33 and particles emitted by glo were in the same size range (150-250 nm diameter) as particles in cigarette smoke 36 (table 5). Compared with reference cigarettes, particle mass in emissions from an e-cigarette and IQOS were <2% 33 and from glo was <1%. 36 Two studies by the manufacturer of IQOS 27 34 did not detect particulate matter in IQOS mainstream and secondhand emissions which was at odds with findings from independent studies. 27 34 Three studies, one independent 33 and two manufacturer-funded, 34 36 also reported on HPHC in secondhand emissions (table 5). All studies detected HPHC in air after HnB use; HPHC levels in HnB secondhand emissions were lower than in cigarette smoke, but reported content of the emissions varied (table 5). Methods of the independent and manufacturer-funded studies were heterogeneous to make direct comparisons. However, the independent study 33 detected particulate matter and acrolein in IQOS secondhand emissions when the manufacturer-funded study 34 did not detect these.

studies on hnb use by human participants
The 15 studies on HnB use by humans (5 independent 5 39-42 and 10 manufacturer-funded [43][44][45][46][47][48][49][50][51][52] (table 1) included 5 RCTs (one published in two parts), 44-49 5 cross-over studies 38 39 47 48 50 , 1 case report 39 and 3 epidemiological studies. 5 41 42 Loose-leaf tobacco vaporiser (Pax) An independent study 40 compared Pax, a cigarette and a pen-style e-cigarette on nicotine delivery to blood plasma, expired air CO, suppression of nicotine abstinence symptoms and satisfaction. Based on the research quality rating tool, the study was rated weak (see table S3 in Supplementary file 1). Plasma nicotine levels were 24.4 ng/mL after cigarette use, 14.3 ng/mL after use of Pax and 9.5 ng/mL after use of the e-cigarette. Expired air CO increased up to 16.9 parts per million (ppm) after smoking a cigarette but decreased after HnB and e-cigarette use to 4.5 ppm. Nicotine abstinence symptoms were most effectively suppressed after smoking a cigarette, use of Pax was less effective and e-cigarette use was least effective; no differences were observed between conditions. Study participants found the HnB and the e-cigarette significantly less satisfying than cigarettes. 40

Carbon-heated tobacco product
One manufacturer-funded 5-day confinement RCT of moderate quality 47 compared levels of exposure with HPHC between smokers who were randomised to using carbon-heated tobacco product (CHTP) only, continued smoking or abstinence (see table S3 in Supplementary file 1; table 6).
Compared with participants who continued smoking, on day 5, CHTP users demonstrated less exposure to HPHC, took more frequent and longer puffs that were of higher average and total volume. Differences in CHTP and cigarette use frequency (19.7 vs 18.8 on day 5, respectively, p=0.57), total nicotine equivalents (19.1 vs 17.2 ng/mL) and plasma cotinine for the past 24 hours (319.8 vs 289.8 mg) were not statistically significant.

Tobacco heating system 2.1
Two manufacturer-funded studies of moderate 46

Nicotine delivery
One manufacturer-funded study 50   vs 2.5 hours). Compared with cigarettes, THS 2.1 delivered lower peak levels of nicotine after single (8.4 vs 11.9 ng/mL) and ad libitum use (14.9 vs 24 ng/mL). THS 2.1 users also consumed fewer tobacco sticks per day than smokers smoked cigarettes (10.9 vs 16.7, p<0.001) and perceived THS 2.1 less satisfying than cigarettes: rated it lower on four out of five modified cigarette evaluation scores (mCEQ) 53 subscales (smoking satisfaction, psychological rewards, enjoyment of respiratory tract sensation and craving reduction).

Exposure to HPHC
A manufacturer-funded 5-day confinement RCT 46 compared exposure levels to HPHC in smokers who were randomised to using only THS 2.1 or continued smoking. Exposure to HPHC was lower in the THS 2.1 group (table 6).
The THS 2.1 group used up to 35% more tobacco sticks than the smoking group smoked cigarettes (27.2 and 20.1, respectively, p=0.002) and demonstrated compensatory puffing (increased puff frequency, duration and volume). On day 5, the THS 2.1 group achieved 85% of nicotine and 88% of cotinine levels of the smoking group. Satisfaction with THS 2.1 was again significantly lower on the same four mCEQ subscales.

Nicotine delivery
A study 43 of weak quality (see table S3 in Supplementary file 1) compared nicotine delivery between regular IQOS and regular cigarettes, menthol IQOS and menthol cigarettes and IQOS tobacco sticks and 2 mg nicotine gum. Peak plasma concentrations for both IQOS tobacco sticks and cigarettes were reached in 6 min, actual exposure to nicotine was comparable (IQOS:cigarettes ratio was 96.3% for regular and 98.1% for menthol), as was nicotine half-life (93.1% and 102.3%). Peak nicotine concentration ratio for regular IQOS versus cigarettes was 103.5% and for menthol IQOS versus Review menthol cigarettes, 88.5%. Compared with nicotine gum, the results were less clear: regular IQOS outperformed menthol IQOS for exposure to nicotine (127.2% and 55.9%) and peak nicotine concentration (240.2% and 101.6%). Relative to the gum, nicotine half-life was 87.3% for regular and 92.1% for menthol IQOS tobacco sticks.

Exposure to HPHC
Two 5-day confinement RCTs 44 45 (one strong, one moderate quality; table S3 in Supplementary file 1) assessed exposure to HPHC in smokers randomised to using IQOS, continuing smoking or abstaining from smoking for study period (table 6). Two publications 48 49 (moderate quality, table S3 in Supplementary file 1) reported on the same 5-day confinement RCT followed by 85 days in an unconfined setting. Exposure to HPHC and change in health risk markers were compared between smokers randomised to using menthol IQOS, continuing smoking or abstaining from smoking for study period (table 6).
Across the three studies, exposure to biomarkers of HPHC in the IQOS groups was lower than for smoking groups and approached exposure levels observed in abstinent groups.
Daily product use differed across studies: on day 5 the IQOS group in one study 44 used 20% fewer tobacco sticks than the smokers' group smoked cigarettes (p<0.001); in another study, 45 they used 25% more tobacco sticks than the smoking group smoked cigarettes (p<0.001) and in the third study, 48 consumption did not differ between groups (p=0.63). During an unconfined study period, participants in the IQOS group demonstrated high compliance (89.7%), and dual use of cigarettes and IQOS was low (<0.1 cigarette on average in the IQOS group). 48 Throughout all three studies, IQOS users increased their puffing frequency, duration and number of puffs. IQOS use suppressed urges to smoke similarly to smoking cigarettes but was consistently rated lower on sensory and psychological satisfaction than cigarettes: in two studies, IQOS scored lower on four out of five mCEQ subscales 45 48 and in one study IQOS scored lower on the smoking satisfaction subscale. 44 Over the 85 days following the 5-day confinement period, participants randomised to IQOS use demonstrated reductions in risk markers 54 associated with endothelial dysfunction, oxidative stress, inflammation and high-density lipoprotein cholesterol counts compared with participants randomised to continued smoking. 49 Tobacco heating product 1.0 (glo) A manufacturer-funded randomised cross-over trial 51 (weak quality, table S3 in Supplementary file 1) compared puffing topography, mouth level exposure and daily consumption of glo among cigarette smokers and smokers who dually used IQOS but were naïve to glo. In comparison to cigarette smoking, glo and IQOS users demonstrated significantly higher mean puff volumes (66.7 and 63.5 mL, respectively vs 48.9 mL for a cigarette), shorter puffing intervals (7.4 s and 8.3 s vs 9.7 s), and used glo and IQOS less frequently than cigarettes (12.1 and 13.7 tobacco sticks, respectively vs 16.3 cigarettes). Users of glo and IQOS differed in mean puff volume and average daily consumption: new glo users demonstrated higher mean puff volume (60.9±24.8 vs 55.1±23.9 mL) and used fewer tobacco sticks than the IQOS users (11.2±6.2 vs 13.4±7.8).

Prototype novel tobacco vapour product (Ploom Tech)
A manufacturer-funded randomised cross-over trial 52 (weak quality, table S3 in Supplementary file 1) compared the pharmacokinetic profiles of Ploom (table S2 in Supplementary file 1) and a reference cigarette. They did not differ in the time to reach peak plasma nicotine concentration (median for both 3.8 min) or in nicotine half-life after single use (1.66 for Ploom vs 1.86 hour for a cigarette). However, Ploom delivered significantly lower peak plasma nicotine concentration (45.7% of cigarette) and total exposure to nicotine after single use (68.3%, p=0.002).

epidemiological studies on hnb use
The literature search identified three independently funded surveys on awareness and use of HnB products: two from Japan 5 41 and one from the UK. 42 The studies from Japan reported findings from a nationally representative sample of 8240 respondents aged 15-69 years first surveyed in 2015 41 and followed up in 2016 (follow-up rate 65.6%) and 2017 (52.2%). 5 The data suggest growth in IQOS use: in 2015 0.3% reported using IQOS in the last 30 days, in 2016 this rose to 0.6%, and in 2017 to 3.6%. 5 The use of other HnB products (not mutually exclusive) in 2017 was lower: 1.2% had used Ploom in the last 30 days (a rise from 0.3% in 2015) and 0.8% had used glo in the last 30 days (no data for previous years). 5 The same study reported that among 11.9% of the sample that had been exposed to HnB secondhand emissions, more than a third (37%) experienced at least one symptom (eg, sore throat, eye pain, feeling ill, etc) related to this exposure.
Data from a nationally representative sample of 12 696 adults on awareness and use of HnB products in Great Britain (GB) were collected in March-April 2017 42 ; 9.3% of the adult GB population were aware of HnB and 1.7% had tried or were using the products. Among those who had ever tried HnB, 39% had tried it once or twice and 13% had been using it daily. It should be noted, however, that participants were asked about HnB products prior to answering about e-cigarettes, which could have led to overestimating awareness and use of HnB products.

Case report on hnb use
A single case report 39 of acute eosinophilic pneumonia (AEP) following the use of HnB was identified. The study summarised the case of a Japanese man aged 20 years who used 20 HnB tobacco sticks per day for 6 months and increased to using 40 tobacco sticks 2 weeks before hospitalisation. Based on the relationship between cigarette smoking and AEP, the authors presumed that in this case the rapid increase in the daily use of tobacco sticks caused the onset of AEP. The case report concludes that despite HnB users being exposed to lower levels of HPHC compared with cigarette smokers, they are still susceptible to health risks in general and to acute eosinophilic pneumonia in particular.

DIsCussIOn
The systematic search identified 31 peer-reviewed studies on seven HnB products. Eleven independent studies focused on awareness and use and secondhand emissions of HnB products, while 20 manufacturer-funded studies explored nicotine delivery and mainstream emissions and conducted RCTs assessing exposure to HPHC in HnB users.
By late 2017, awareness and use of HnB products were rising in Japan while in GB HnB use was rare.
Five RCTs demonstrated that switching from smoking cigarettes to using HnB significantly reduces but does not eliminate exposure to HPHC. The evidence, however, was limited to one Review currently available HnB product and came from a single tobacco manufacturer.
Single use of HnB delivered nicotine as quickly as smoking a cigarette but with lower peak concentration and total exposure to nicotine. When used ad libitum, HnB delivered comparable levels of nicotine and weakened urges to smoke similarly to cigarettes, nevertheless, HnB users reported compensatory puffing and consistently rated HnB less rewarding and satisfying than cigarettes.
Studies on machine-generated mainstream HnB emissions generally reported higher proportional reductions in exposure to HPHC than were observed in the RCTs on HnB use by humans. This suggests that machine smoking does not reliably replicate human use; this has been demonstrated for cigarettes 16 and it is likely even less reliable for HnB products.
The tested HnB products were heterogeneous in the way they worked and in the levels of nicotine and HPHC they delivered to mainstream aerosol. Compared with cigarettes, aerosol of IQOS (heating tobacco up to 350°C) contained the highest proportional levels of nicotine and HPHC, followed by glo (250°C) and iFuse (35°C), which produced the least toxicants and delivered the lowest levels of nicotine of the three. Compared with e-cigarettes, IQOS delivered less nicotine than a tank-style but more than a cigalike e-cigarette.
Evidence on HnB secondhand emissions suggested that HnB exposes users and bystanders to substantially lower but measurable levels of particulate matter and HPHC.
Comparisons between findings of independent and manufacturer-funded studies were limited due to heterogeneity in methods measuring mainstream and sidestream emissions or lack of independent evidence on HnB use by humans. Where the comparisons were possible, sample sizes were low and assumptions for t-tests could not be verified. Independent and manufacturer-funded studies reported similar levels of nicotine, CO, water and total particulate matter in mainstream IQOS aerosol, but diverged when reporting on tar (an independent study reported less) and tobacco-specific nitrosamines (manufacturer-funded studies reported less). Inferences from these comparisons are limited, as only two independent studies provided data on a limited number of HPHC. Conclusions on secondhand emissions from HnB devices were at odds between independent and manufacturer-funded studies, with PMI-funded studies reporting no particulate matter in IQOS secondhand emissions.

limitations of the present evidence
Out of 11 trials on HnB use by humans, only 1 was not affiliated with a tobacco manufacturer 40 ; the lack of independent evidence that could validate manufacturer data remains a major limitation.
Other limitations include that none of the trials on HnB use by humans registered a protocol before the enrolment of the first participant as recommended by the International Committee of Medical Journal Editors. 55 Other concerns that were raised in relation to studies funded by the manufacturer of IQOS pertained to the quality of their trials, 56 possible tobacco industry ties of the journal that published most studies on IQOS characteristics 57 58 and non-reporting levels of all 93 HPHC and other potentially harmful constituents in mainstream and sidestream emissions. 59 Some manufacturer-funded publications appeared to overstate conclusions. For instance, studies concluded that IQOS was comparable in satisfaction to smoking 44 48 when it was repeatedly rated lower on four out of five mCEQ subscales. 53 Studies that reported similar findings on HnB secondhand emissions differed in their conclusions: manufacturer-funded studies concluded that HnB use impact on indoor air quality was negligible 34 or that HnB emissions were less harmful than cigarette smoke, 36 while an independent study concluded that despite lower emissions, HnB still pose evident risks through secondhand emissions. 33 Only one trial assessed health effects of long-term HnB (IQOS) use relative to abstinence and continued smoking. 49 However, the validity of the study follow-up results and conclusions are reduced by a lack of validation of self-reported abstinence and adherence to study condition.

limitations of the systematic review
The review included data from manufacturer-funded studies but excluded reports or papers that were not peer-reviewed. As HnB manufacturers not always publish in peer-reviewed journals, this might have limited scope of our study. Nevertheless, this review provides the first comprehensive summary of up-to-date evidence on HnB tobacco products.

Future research
Machine smoking regimens were tailored for testing and comparing emissions from different tobacco cigarettes, and the validity of this method to measure emissions from HnB products is unclear. Future research should clarify whether the existing regimens reliably estimate HnB emissions or need adjusting. Until then, to promote reproducibility and comparisons between studies, research on HnB emissions would benefit from employing standardised protocols: by using the same machine smoking regimen (eg, HCI or other, adjusted for HnB), same reference products (eg, 3R4F tobacco cigarette) and screen for the same list of HPHC.
Current evidence on HnB sidestream emissions comes from a single manufacturer-funded study that is subject to conflict of interests; independent research could disentangle the preliminary disagreement.
Although research on mainstream and sidestream HnB emissions provides valuable preliminary data on HnB characteristics, it is unclear how well these findings represent the actual health risks of HnB use. The discrepancies in exposure to HPHC between machine puffing and human use studies suggest that findings on HnB mainstream emissions underestimate the actual exposure to toxicants. Instead of measuring HPHC levels in mainstream HnB emissions, independent research should prioritise validating manufacturers' findings on exposure to toxicants in HnB use by humans and comparing actual long-term health effects of HnB use with health outcomes of smoking, vaping or using nicotine replacement therapy.
Recently, HnB tobacco products have been introduced to multiple tobacco markets around the world, but only three independent studies from Japan and GB reported on awareness and use. There is a need for future surveillance on the uptake of HnB products and comparisons between countries with different regulatory frameworks for tobacco and nicotine products.
As more HnB products appear on the market, more manufacturer-funded studies are expected. This challenges independent researchers to critically evaluate and validate industry findings. 11 All researchers, whether affiliated with tobacco manufacturers or not, should aim for professional and transparent ways to preregister, conduct and report their findings. 60

COnClusIOn
Peer-reviewed evidence on heated tobacco products indicates that HnB are effective nicotine delivery devices that expose users and bystanders to substantially fewer harmful and potentially harmful Review compounds than smoking cigarettes. The evidence is primarily drawn from tobacco industry data and lacks research on longterm HnB use effects on health. The HnB harm profile needs to be confirmed by independent research and compared with other alternative nicotine products that have reduced health risk exposure profiles.