Heated tobacco products for smoking cessation and reducing smoking prevalence

Abstract Background Heated tobacco products (HTPs) are designed to heat tobacco to a high enough temperature to release aerosol, without burning it or producing smoke. They differ from e‐cigarettes because they heat tobacco leaf/sheet rather than a liquid. Companies who make HTPs claim they produce fewer harmful chemicals than conventional cigarettes. Some people report stopping smoking cigarettes entirely by switching to using HTPs, so clinicians need to know whether they are effective for this purpose and relatively safe. Also, to regulate HTPs appropriately, policymakers should understand their impact on health and on cigarette smoking prevalence. Objectives To evaluate the effectiveness and safety of HTPs for smoking cessation and the impact of HTPs on smoking prevalence. Search methods We searched the Cochrane Tobacco Addiction Group's Specialised Register, CENTRAL, MEDLINE, and six other databases for relevant records to January 2021, together with reference‐checking and contact with study authors and relevant groups. Selection criteria We included randomised controlled trials (RCTs) in which people who smoked cigarettes were randomised to switch to exclusive HTP use or a control condition. Eligible outcomes were smoking cessation, adverse events, and selected biomarkers. RCTs conducted in clinic or in an ambulatory setting were deemed eligible when assessing safety, including those randomising participants to exclusively use HTPs, smoke cigarettes, or attempt abstinence from all tobacco. Time‐series studies were also eligible for inclusion if they examined the population‐level impact of heated tobacco on smoking prevalence or cigarette sales as an indirect measure. Data collection and analysis We followed standard Cochrane methods for screening and data extraction. Our primary outcome measures were abstinence from smoking at the longest follow‐up point available, adverse events, serious adverse events, and changes in smoking prevalence or cigarette sales. Other outcomes included biomarkers of harm and exposure to toxicants/carcinogens (e.g. NNAL and carboxyhaemoglobin (COHb)). We used a random‐effects Mantel‐Haenszel model to calculate risk ratios (RR) with 95% confidence intervals (CIs) for dichotomous outcomes. For continuous outcomes, we calculated mean differences on the log‐transformed scale (LMD) with 95% CIs. We pooled data across studies using meta‐analysis where possible. Main results We included 13 completed studies, of which 11 were RCTs assessing safety (2666 participants) and two were time‐series studies. We judged eight RCTs to be at unclear risk of bias and three at high risk. All RCTs were funded by tobacco companies. Median length of follow‐up was 13 weeks. No studies reported smoking cessation outcomes. There was insufficient evidence for a difference in risk of adverse events between smokers randomised to switch to heated tobacco or continue smoking cigarettes, limited by imprecision and risk of bias (RR 1.03, 95% CI 0.92 to 1.15; I2 = 0%; 6 studies, 1713 participants). There was insufficient evidence to determine whether risk of serious adverse events differed between groups due to very serious imprecision and risk of bias (RR 0.79, 95% CI 0.33 to 1.94; I2 = 0%; 4 studies, 1472 participants). There was moderate‐certainty evidence for lower NNAL and COHb at follow‐up in heated tobacco than cigarette smoking groups, limited by risk of bias (NNAL: LMD −0.81, 95% CI −1.07 to −0.55; I2 = 92%; 10 studies, 1959 participants; COHb: LMD −0.74, 95% CI −0.92 to −0.52; I2 = 96%; 9 studies, 1807 participants). Evidence for additional biomarkers of exposure are reported in the main body of the review. There was insufficient evidence for a difference in risk of adverse events in smokers randomised to switch to heated tobacco or attempt abstinence from all tobacco, limited by risk of bias and imprecision (RR 1.12, 95% CI 0.86 to 1.46; I2 = 0%; 2 studies, 237 participants). Five studies reported that no serious adverse events occurred in either group (533 participants). There was moderate‐certainty evidence, limited by risk of bias, that urine concentrations of NNAL at follow‐up were higher in the heated tobacco use compared with abstinence group (LMD 0.50, 95% CI 0.34 to 0.66; I2 = 0%; 5 studies, 382 participants). In addition, there was very low‐certainty evidence, limited by risk of bias, inconsistency, and imprecision, for higher COHb in the heated tobacco use compared with abstinence group for intention‐to‐treat analyses (LMD 0.69, 95% CI 0.07 to 1.31; 3 studies, 212 participants), but lower COHb in per‐protocol analyses (LMD −0.32, 95% CI −1.04 to 0.39; 2 studies, 170 participants). Evidence concerning additional biomarkers is reported in the main body of the review. Data from two time‐series studies showed that the rate of decline in cigarette sales accelerated following the introduction of heated tobacco to market in Japan. This evidence was of very low‐certainty as there was risk of bias, including possible confounding, and cigarette sales are an indirect measure of smoking prevalence. Authors' conclusions No studies reported on cigarette smoking cessation, so the effectiveness of heated tobacco for this purpose remains uncertain. There was insufficient evidence for differences in risk of adverse or serious adverse events between people randomised to switch to heated tobacco, smoke cigarettes, or attempt tobacco abstinence in the short‐term. There was moderate‐certainty evidence that heated tobacco users have lower exposure to toxicants/carcinogens than cigarette smokers and very low‐ to moderate‐certainty evidence of higher exposure than those attempting abstinence from all tobacco. Independently funded research on the effectiveness and safety of HTPs is needed. The rate of decline in cigarette sales accelerated after the introduction of heated tobacco to market in Japan but, as data were observational, it is possible other factors caused these changes. Moreover, falls in cigarette sales may not translate to declining smoking prevalence, and changes in Japan may not generalise elsewhere. To clarify the impact of rising heated tobacco use on smoking prevalence, there is a need for time‐series studies that examine this association.


Description of the condition
Tobacco use kills 8 million people each year, making it the leading preventable cause of death worldwide (Drope 2018). Approximately 90% of these deaths result from the most harmful form of tobacco consumption: smoking (Drope 2018). Reducing smoking prevalence is therefore one of the most e ective ways of improving population health (Holford 2014).
Although most smokers want to quit, smoking is highly addictive and most who try fail, with less than 10% still abstinent a year a er making a serious attempt to stop (Jackson 2019a). Available treatments such as behavioural support, varenicline, and nicotine replacement therapy (NRT) improve the chance that these attempts will succeed (Cahill 2016;Hartmann-Boyce 2018;Hartmann-Boyce 2019). However, even with these treatments success rates are typically under 25%, and half of those who try to quit do not use any support (Jackson 2019b). There remains an urgent need to identify new, e ective and safer alternatives to cigarettes to reduce smoking prevalence.

Description of the intervention
Heated (or heat-not-burn) tobacco products (HTPs) are handheld devices that heat tobacco to a temperature that is high enough to produce a nicotine-infused aerosol, but too low to cause self-sustaining combustion. Many of the toxic and carcinogenic products of cigarette smoking are formed during combustion. HTPs are marketed as less harmful and as alternatives to conventional cigarettes because they attempt to avoid combustion (Mathers 2017). However, HTPs are relatively new to the market. The extent to which HTPs help people quit smoking is largely unknown, and their impact on youth uptake to smoking is contentious (Czoli 2020). It is therefore unclear what impact HTPs will have on smoking prevalence across the population.
The current HTP market is dominated by four products: 'IQOS' by Philip Morris International, 'glo' and 'iFuse' by British American Tobacco, and 'Ploom Tech' by Japan Tobacco International (WHO 2018). IQOS and glo produce aerosol by directly heating tobacco sticks, which resemble small cigarettes. Conversely, Ploom Tech and iFuse produce aerosol by heating a similar liquid to that found in e-cigarettes. This aerosol is then drawn through a bulb of tobacco to infuse it with flavour. Of these products, IQOS was the first to launch in 2014 in Japan and Italy, and it has since entered markets across Asia, Europe and America. Most recently, in 2019, the US Food and Drug Administration (FDA) permitted the sale of IQOS (FDA 2019) and in 2020 authorised their marketing as a modifiedexposure tobacco product (FDA 2020). At the time of writing, HTPs were most popular in Japan and the Republic of Korea; tobacco sticks for HTPs constituted 15.8% and 8.0% respectively of each country's tobacco market in 2018 (WHO 2018). Market research by Euromonitor estimates that HTPs had an increased share of the retail value of all nicotine or tobacco products between 2017 and 2018, which was similar to e-cigarettes globally (Euromonitor 2020)

How the intervention might work
Nicotine is the primary addictive compound in cigarettes, which causes people who quit smoking to experience withdrawal and cravings (West 2017). Like cigarettes, HTPs contain nicotine. They are expected to aid smoking cessation in a similar way to NRT and e-cigarettes: people can use them to relieve nicotine cravings without smoking cigarettes (Wadgave 2016). HTPs may also provide certain advantages over NRT. A limitation of NRT is that it poorly addresses the behavioural and sensory cues associated with cigarette smoking, such as repeated hand-to-mouth actions and the 'scratch' at the back of the throat when inhaling smoke. Evidence shows that denicotinised cigarettes reduce cravings and withdrawal symptoms among abstinent smokers, despite containing negligible levels of nicotine (Rose 2006). This suggests that these cues probably contribute to cigarette dependence. HTPs may more closely replicate these cues than NRT. Because HTP aerosol is delivered to the throat and lungs, nicotine absorption occurs more rapidly than from patches, gum or lozenges, which are absorbed through the skin or buccal mucosa (Simonavicius 2018). The speed with which nicotine is absorbed may be one of the key determinants of dependence (Benowitz 2009), so HTPs may provide a better replacement for cigarette smoking than NRT. Ecigarettes also deliver nicotine rapidly to the throat and possibly lungs (Wagener 2017; Hajek 2020) and, like HTPs, they mimic the hand-to-mouth actions of cigarette smoking. But only HTPs contain tobacco leaf, so their flavour may more closely resemble cigarette smoke (Poynton 2017), which may make them more attractive to smokers (Tompkins 2020). Moreover, in some countries, the sale of nicotine e-cigarettes is banned or heavily restricted (Dyer 2019). In such environments, HTPs may be the only consumer product available that delivers nicotine rapidly through a potentially less harmful medium than tobacco smoke.
A substantial proportion of people who use HTPs for smoking cessation may continue using these products for at least a year a er they quit smoking, as is the case with e-cigarettes (Hajek 2019; Simonavicius 2020). We will refer to this as 'switching'. Encouraging people to switch from smoking cigarettes to using HTPs would only be beneficial if HTPs are less harmful to health or if HTPs eventually help people taper o nicotine entirely. The safety of HTPs to users depends on both the acute harm, measured by adverse and serious adverse events, and the long-term harm of repeated inhalation of damaging compounds in HTP aerosol. Biomarkers can be used to measure exposure to these harmful toxicants and carcinogens. Manufacturers of HTPs claim that the aerosol they produce contains significantly lower levels of toxicants than cigarette smoke and, as a result, that they have reduced risk potential or are less harmful (PMI 2018; BAT 2020). Two systematic reviews supported claims about lower toxicant levels, but found that most research into HTPs was funded through sources a iliated with the tobacco industry (Simonavicius 2018;Jankowski 2019). In addition, reduced exposure does not necessarily indicate reduced harm. The US FDA judged that there was su icient evidence that iQOS reduced exposure to harmful chemicals (FDA 2020), but insu icient evidence on whether switching from smoking to HTPs reduces harm, such as pulmonary function or biomarkers linked to smoking-related harm (Glantz 2018; Moazed 2018). It is also the case that safety, especially of longer-term use, cannot be addressed with certainty until long-term cohort studies have collected su icient data.

Why it is important to do this review
There is substantial variation between countries in their regulatory approaches to HTPs, and within countries across di erent nicotine products. In order for policymakers to regulate HTPs e ectively and proportionately, there is a need for evidence to inform a judgement Library Trusted evidence. Informed decisions. Better health.
Cochrane Database of Systematic Reviews on their likely public health impact. The net impact of HTPs on public health will depend on a variety of factors. Three influential elements that could result in HTPs benefiting public health are if they increase smoking cessation, decrease smoking prevalence, and are less harmful than cigarette smoking. Conversely, HTPs could damage public health if they decrease smoking cessation, increase smoking prevalence, or are just as harmful as cigarette smoking.
The e ect of HTP use on smoking prevalence will depend on whether they influence rates of attempted quitting among cigarette smokers, the proportion of these attempts that are successful, cigarette uptake among non-smokers, and relapse among people who had previously quit smoking. We are therefore not only interested in studies that report individual-level e ects of HTPs on smoking cessation, but also those that estimate their populationlevel e ects on smoking prevalence. This review will investigate upto-date evidence for both, using appropriate study types.
The growing popularity of HTPs means that people who smoke are likely to seek advice from practitioners who need to know whether HTPs are e ective for smoking cessation and how their safety compares with cigarettes and other alternative nicotine products. If HTPs are found to be safe and e ective for smoking cessation, they could add a novel treatment for cigarette addiction. Moreover, if there is evidence that an increase in HTP use prevalence is associated with a decrease in smoking prevalence, then this will help to guide the regulation of HTPs.

O B J E C T I V E S
To evaluate the e ectiveness and safety of HTPs for smoking cessation and the impact of HTPs on smoking prevalence.

Types of studies
We have divided the methods into the three subsections, representing the di erent objectives of the review:

E ectiveness for smoking cessation
We will include individual-level and cluster-RCTs to examine the e ectiveness of HTPs for tobacco smoking cessation.

Safety
We will include individual-level, randomised cross-over and cluster-RCTs to explore adverse and serious adverse events and biomarkers of toxicant and carcinogen exposure.

Smoking prevalence
We will include interrupted and multiple time-series studies to examine the population-level e ect of HTPs on cigarettesmoking prevalence. Smoking cessation interventions may not be representative of the way most people use HTPs, without support from a researcher or trained specialist. Moreover, even if HTPs encourage smoking cessation among those trying to quit, their impact on smoking prevalence depends on how they a ect smoking initiation and the number of people who make a quit attempt. We will use time-series studies to assess how changes in HTP prevalence are associated with changes in smoking prevalence, with the limitation that associations might not reflect causal e ects.
We will include studies regardless of language or status of publication.

E ectiveness and safety
We will include adults, defined as current cigarette smokers by the study, at the time of enrolment.

Smoking prevalence
We will not restrict by participant characteristics, as we are interested in population-level data. We will focus on any individuals who report on their smoking status or consumption and HTP use or consumption, measured by survey or by record of sales.

Types of interventions
HTPs, defined as hand-held devices that heat tobacco to a temperature that is high enough to produce a nicotine-infused aerosol but too low to cause self-sustaining combustion. HTPs di er from e-cigarettes in that they produce aerosol by heating compressed tobacco leaf rather than a liquid that is infused with nicotine concentrate.

E ectiveness and safety
We are interested in studies that compare HTPs, or the addition of HTPs, to no treatment (i.e. continued tobacco smoking), placebo or any other smoking cessation treatment, including nicotine replacement therapy (NRT), electronic cigarettes, varenicline, bupropion, and behavioural support. HTPs can be provided in addition to any other smoking cessation treatment, providing there is equivalent provision for the control group. We will only include studies where participants were instructed to stop smoking combustible cigarettes.

Smoking prevalence
We are interested in studies that estimate the extent to which changes in the prevalence of HTP use are associated with changes in the prevalence of cigarette smoking, a er adjusting for other influences that could a ect changes in the prevalence of smoking at the population level.

E ectiveness
• Tobacco smoking cessation at the longest follow-up point available, using intention-to-treat and biochemically-verified abstinence where possible. While HTPs contain tobacco, they are designed to heat the tobacco and attempt to avoid or minimise combustion and smoke. HTP use will therefore not be classified as tobacco smoking. We will restrict the review to studies which report abstinence at four-week follow-up or longer. We will use the strictest definition of abstinence recorded, i.e. prolonged or continuous abstinence over point prevalence, and biochemically-verified over selfreported abstinence.Typically, Cochrane Tobacco Addiction Group reviews only include data on smoking cessation at six months or longer. We will include short-term outcomes in this review because we anticipate a paucity of longer-term data. In future updates, as and when more data become available, we may change the inclusion criteria accordingly.

Safety
• Number of people reporting adverse events (AEs) and serious adverse events (SAEs).

Smoking prevalence
• Change in the prevalence of cigarette smoking, measured as the proportion of people in a given locality that regularly smoke cigarettes or other combustible tobacco products, over a defined time period. We will include cigarette sales as a proxy for prevalence and measured as the number of cigarettes sold in a given locality over a given time period.

Secondary outcomes
All secondary outcomes are measures of safety. We will only include studies which report safety outcomes at one-week followup or longer.
• Biomarkers of toxicant and carcinogen exposure. We will include measures of exposure to tobacco-specific N-nitrosamines, polycyclic aromatic hydrocarbons, volatile organic compounds, and carbon monoxide (see Appendix 1 for details). • Biomarkers of harm. We will include measures of lung function (FEV1, FVC and FEV1/FVC), blood pressure, heart rate, heart rate variability, and blood oxygen saturation.

Search methods for identification of studies Electronic searches
We will search the following databases: We will restrict the search to studies published since 2008, three years before the first internet searches for HTPs began (Google Trends 2020).
The search terms will be: heated tobacco OR carbon-heated tobacco OR heat-not-burn OR heat not burn OR tobacco heating system$ OR tobacco heating device$ OR tobacco heating product$ OR tobacco vapor product $ OR tobacco vapour product$. We will also search for the term smoking AND (iqos OR glo OR ploom OR ifuse OR fuse OR pulse OR teeps OR pax OR mok OR lil OR iuoc OR htp OR thp OR ths OR chtp).
As we are only interested in studies that use human subjects, we will exclude those with the terms animal$ OR mice OR rat$ OR in vitro OR in silico OR in vivo in their title.

Searching other resources
We will search the reference lists of eligible studies found in the literature search.
In order to identify government reports and in-press or unpublished studies, we will contact tobacco control charities and authors of published research or trial protocols. We will use the searches of Clinicaltrials.gov and the ICTRP detailed above to identify trial registry records.

Selection of studies
Two review authors will independently pre-screen titles and abstracts of articles identified in the search, using a screening checklist. We will resolve disagreements through discussion or referral to a third review author. We will conduct screening using Covidence so ware (Covidence).
Two review authors will independently screen the full text of articles that pass pre-screening, with a third review author to resolve any disagreements that are not resolved through discussion.

Data extraction and management
We will produce two custom data extraction forms: one for e ectiveness and safety, and the other for smoking prevalence. Both will include: • Biomarkers of exposure to the volatile organic compounds acrolein, lead, cadmium, and butadiene (3-HPMA, lead, cadmium, and MHBMA3 respectively) at baseline and longest follow-up point available; • Biomarkers of tobacco-specific N-nitrosamine (TSNA) exposure (NNAL) at baseline and longest follow-up point available; • Lung function (measured using FEV1, FVC and FEV1/FVC); • Blood pressure; • Heart rate; • Blood oxygen saturation.
Smoking prevalence forms will also include: • Coe icient and standard error for change in the trend following intervention in prevalence or sales; • Coe icient and standard error for step-level change in prevalence or sales; • Coe icient and standard error for changes between cigarette prevalence or sales and heated tobacco product prevalence or sales; • Statistical method used; • Covariates included in model; • Temporal granularity (e.g. weekly, monthly, annual); • Time when heated tobacco products entered the market; Two review authors will independently extract data from included studies. When discrepancies cannot be resolved through discussion, we will refer to a third review author. We will contact authors of included studies if additional information is needed.

E ectiveness and safety
Two review authors will independently assess risks of bias for all included RCTs using the Cochrane 'Risk of bias' tool (ROB). The version we use (1 or 2) will depend on so ware availability at the time of data extraction. Regardless of which tool we use, we will follow the guidance as set out in the Cochrane Handbook for Systematic Reviews of Iterventions (Handbook) to evaluate the appropriate domains. For ROB 1 these are sequence generation, allocation concealment, blinding of participants, personnel and outcome assessment, incomplete outcome data, and selective reporting, as well as other sources of bias (Higgins 2011). However, if version 2 is deemed feasible, we will evaluate studies within the following domains: bias arising from the randomisation process; bias due to deviations from intended interventions; bias due to missing outcome data; bias in measurement of the outcome; bias in selection of the reported result (Higgins 2020).

Smoking prevalence
Two review authors will independently assess risks of bias for included time-series studies using the ROBINS-I tool (Sterne 2016).

E ectiveness and safety
We will calculate risk ratios (RRs) and 95% confidence intervals (CIs) for binary outcomes, such as smoking cessation.
We will calculate mean di erences (MDs) and the corresponding 95% CIs, between the intervention and control groups in the mean change from baseline to follow-up for continuous safety data.
For smoking cessation, we will use the longest follow-up data reported, and for safety outcomes we will assess data at all time points. In both cases we will calculate treatment e ects on an intention-to-treat basis.

Smoking prevalence
For interrupted time-series studies, the treatment e ect will be reflected by the step change and the change in trends in smoking prevalence or cigarette sales following the introduction of HTPs to the market, a er controlling for confounding variables.
For multiple time-series studies, the treatment e ect will be the association between HTP prevalence and smoking prevalence or cigarette sales, a er controlling for confounding variables. It is likely that the variables will have been log-transformed to address non-stationarity and for easier interpretation. The resulting coe icient would describe the percentage change in cigarette smoking prevalence associated with a 1% change in HTP prevalence.

E ectiveness and safety
For RCTs with more than two intervention arms, we will combine data from all relevant intervention conditions where HTPs were o ered, where appropriate. If it is not appropriate to pool the intervention arms then we will split the control arm to act as a comparator to each separate intervention arm. For RCTs with more than two control arms, we will not combine data from each of these arms, and we will choose the most appropriate comparator. For cluster-RCTs, we will attempt to extract an estimate of the e ect that accounts for the cluster design of the study. Where this is not reported, we will attempt to perform the correct analysis if required data are available.

E ectiveness
When assessing smoking cessation, we will assume that people with missing data at follow-up have not stopped smoking, as is common in the field.

Safety
When assessing AEs and SAEs, we will calculate the proportion of those available at follow-up who experienced an event, rather than the proportion of people who were randomised. When assessing biomarkers, we will remove participants with missing follow-up data from the analysis.

Library
Trusted evidence. Informed decisions. Better health.

Smoking prevalence
We do not anticipate problems with missing data in reported timeseries studies.

Assessment of heterogeneity
To assess whether to conduct meta-analyses, we will consider the characteristics of included studies to identify substantial clinical or methodological heterogeneity. If we deem the data to be su iciently homogeneous to combined them meaningfully, we will assess statistical heterogeneity using the I 2 statistic. If I 2 is greater than 50%, we will report substantial heterogeneity.

Assessment of reporting biases
If we deem it appropriate to pool 10 or more studies in any analysis we will assess reporting bias using funnel plots. The greater the asymmetry in the plots, the higher the risk of reporting bias.

E ectiveness
The primary outcome of smoking cessation provides dichotomous data. Following the standard methods of the Cochrane Tobacco Addiction Group, we will combine RRs and 95% CIs from individual studies using a Mantel-Haenszel random-e ects model, to calculate pooled overall RRs with 95% CIs.

Safety
For dichotomous safety outcomes (i.e. AEs and SAEs) we will combine RRs and 95% CIs from individual studies using a Mantel-Haenszel random-e ects model, to calculate pooled overall RRs with 95% CIs.
For continuous safety outcomes measuring biomarkers, we will pool the MDs or standardised mean di erences (SMDs) and measures of variance of individual studies using a generic inverse variance random-e ects model.

Smoking prevalence
We will calculate pooled estimates and their standard errors using a random-e ects model for each of three coe icients: step change in smoking prevalence or cigarette sales following the introduction of HTPs; change in these trends a er the introduction; and changes associated with changes in prevalence or sale of HTPs. For timeseries studies with notably di erent time periods (for example, weekly versus annual), we will be unable to pool between studies.

Subgroup analysis and investigation of heterogeneity
We plan to undertake subgroup analyses to investigate di erences by: • Intensity of behavioural support provided; • characteristics of HTP device (e.g. model used).

Sensitivity analysis
We will carry out sensitivity analyses where we remove studies: • judged to be at high risk of bias for at least one domain; • where the study is funded by (or authors have received funding from) the tobacco industry; • with a minimum length of follow-up less than four weeks (safety outcomes only).
We will also carry out sensitivity analyses where we: • only include participants who exclusively used the product they were assigned to (safety outcomes only); • only classify participants as HTP users if they use their product daily (smoking prevalence only).

'Summary of findings' tables and GRADE
We will create a 'Summary of findings' table using GradePro for all primary outcomes (GRADEpro GDT), following the guidelines in Cochrane Handbook of Systematic Reviews of Interventions 6.1 (Schünemann 2020). We will use the five GRADE considerations (risk of bias, inconsistency, imprecision, indirectness and publication bias) to assess the certainty of the body of evidence for each of these outcomes.