Trends in prevalence and patterns of use of a heated tobacco product (IQOS^™) in Japan: A three-year repeated cross-sectional study

Setting

Cross-sectional surveys in representative samples of the Japanese general adult population (JGAP) and, separately, in samples of Japanese adult IQOS users (JAIQOS) registered in the IQOS owner database of Philip Morris International (PMI)’s affiliate in Japan were initiated in December 2016 and repeated annually over three full calendar years from 2016/2017 to 2018/2019 (Figure 1). Considering that IQOS^TM was relatively new on the Japanese TNP market in 2016, the IQOS use prevalence in the JGAP was expected to be low. Therefore, additional surveys among JAIQOS samples were conducted alongside the JGAP surveys to obtain reliable estimates of IQOS use patterns among Japanese adult IQOS users²³.

Figure 1. Overall survey study design for the JGAP and JAIQOS samples in each wave of the 3-year study period (2016/2017, 2017/2018, and 2018/2019) of the survey fielding.

Abbreviations: ICF, informed consent form; JAIQOS, sample of adult Japanese IQOS users from PMI’s IQOS™ owner database in Japan; JGAP, representative sample of the Japanese general adult population; PMI, Philip Morris International.

Participants and study design

Study participants

To be included in the JGAP or JAIQOS samples, individuals had to be of legal age for purchasing TNPs in Japan (i.e., ≥20 years), current residents of Japan, and fluent in Japanese. Those included in the JAIQOS samples also had to have used >100 HEETS™/HeatSticks™ in their lifetime²⁴, be a current user of IQOS™ with HEETS/HeatSticks, have access to the internet, and not be currently employed by PMI or its affiliates.

JGAP — Sampling, sample size, and survey mode

The JGAP samples were obtained via a syndicated (Omnibus) survey overseen and coordinated by an independent global consumer research company. The fieldwork provider in Japan was Central Research Services Inc (Tokyo, Japan). The Omnibus surveys employed a three-stage stratified proportional random sampling strategy that included the whole country. In stage 1, sampling points in the 12 Japanese administrative regions were allocated on the basis of their share of the population²⁵. Households within each sampling point were identified in Stage 2 by using an electronic residential map, from which about 40 households were randomly selected. In the final stage 3, participants who met the inclusion criteria were selected from within the sampled households. Within each sampling point, quotas on age and sex were set to ensure the representativeness of the Japanese population.

The annual JGAP sampling consisted of four approximately equal-sized waves spaced throughout each study year to account for potential seasonal differences (Figure 2). A sample size of 5,000 participants per year was sufficient to estimate an IQOS™ use prevalence of 5.0% with 95% confidence and a precision of ±0.6% units. In the third year, six survey waves (7,000 participants) were conducted to increase the sample size and improve the accuracy of the estimates.

Figure 2. Allocation of the JGAP and JAIQOS survey waves over the 3-year study period (2016/2017, 2017/2018, and 2018/2019).

Abbreviations: JAIQOS, sample of adult Japanese IQOS™ users from PMI’s IQOS owner database in Japan; JGAP, representative sample of the Japanese general adult population; PMI, Philip Morris International; W, survey wave.

The JGAP surveys were conducted at participants’ homes through in-person face-to-face pen-and-paper interviews. However, to avoid any bias on basis of social desirability of their response regarding their personal TNP use, the participants were handed the “Tobacco Use Prevalence” questionnaire section for self-completion. For completing the Omnibus questionnaires, each participant was given a coupon for JPY 500 (approximately USD 4).

JAIQOS — Sampling, sample size, and survey mode

Upon purchasing an IQOS™ device, users were invited to register in the PMI Japan IQOS owner database, which included about 350,000 adult IQOS owners in July 2017 and reached close to six million in 2019. Considering the demographic age-sex distribution of the database, individuals were randomly selected from the database and invited by email to participate in the survey for each wave.

A sample size of 2,000 participants per year was sufficient to estimate a 50% proportion of exclusive IQOS™ use with 95% confidence and a precision of ±2.19% units. Each annual IQOS user sample consisted of four approximately equal-sized waves spaced throughout each study year to account for potential seasonal differences (Figure 2), with the aim of recruiting 500 adult participants per wave. The JAIQOS surveys were conducted entirely online through computer-assisted self-interviewing. For completing the online survey, participants were given a gift code valued at JPY 500 (about USD 4). The existing standard TNP use questions are available in the literature.

Survey questionnaires

For the present study, the “Tobacco Use Prevalence” questionnaire was developed on the basis of several existing standard TNP use questions available in the literature to capture information about TNP use such as the (i) CDC Adult Tobacco Use Questions of the National Health Interview Survey (NHIS), (ii) WHO/CDC/CPHA Questions of the Global Adult Tobacco Survey (GATS), (iii) NIH/FDA PATH Study questionnaires, and (iv) MHLW Tobacco Use Questions of the Japan National Health and Nutrition Survey. The questionnaire was not specifically validated. In both the JGAP and JAIQOS samples, the same questionnaire was used. However, while for the JGAP samples a pen-and-paper self-completion survey mode was used, for the JAIQOS samples an online mode was applied. The survey questions can be found in the Extended data.

Ethical conduct of the study

All subjects gave informed consent for inclusion in the study prior to participation. The participants of the JGAP samples gave verbal consent that was recorded by the interviewers as part of the Omnibus interviews, while the participants of the JAIQOS samples provided written consent. The study was conducted in accordance with the ethical principles that have their origin in the Declaration of Helsinki and were consistent with Good Epidemiological Practice (GEP)²⁶. The study protocol, including the procedures of providing informed consent, were approved by the Hakata Clinic Institutional Review Board (Reference ID: J-186) in Fukuoka, Japan.

Analytical methods

Analyses were conducted using SAS v9.4 (or higher; SAS Inc., Cary, North Carolina, USA). For both the JGAP and JAIQOS samples, data were analyzed and summarized descriptively for each study year. For participant characteristics and outcome measures, continuous data are presented as mean and standard deviation (SD) or 95% confidence intervals (CI) and categorical data as number and percentage (95% CI) for the total samples and/or stratified by age and sex. Missing data were not included in the statistical analyses.

The following definitions were applied: “Use/never use” of cigarettes or IQOS™ with HEETS™/HeatSticks™: having/not having used 100 cigarettes or 100 HEETS/HeatSticks in the lifetime, to differentiate established cigarette or IQOS users from triers or experimenters²⁴. “Current use”: daily or nondaily use of a TNP at the time of the survey. “Exclusive”, “dual”, and “poly” use: current use of only one type, two types, or three or more types of TNPs, respectively. “Initiation”: the time point at which a participant started established use/smoking of a TNP. “Initiation rate”: proportion of initiation in the last 12 months among never TNP users. “Relapse” and “reinitiation”: restarting TNP use following a period of quitting all TNPs for ≤12 months and >12 months, respectively.

Prevalence of current TNP use for overall TNPs or by TNP category (cigarettes, IQOS™, e-cigarettes, etc.) was calculated in the JGAP samples. For both JGAP and JAIQOS samples, the following was calculated: response rates, sample characteristics, and patterns of TNP use (JGAP: exclusive, dual, and poly use; JAIQOS: exclusive IQOS use and IQOS use with combustible or smoke-free TNP) as well as frequency (past 30-day use), intensity (average daily consumption), and history (JGAP: initiation, relapse, and reinitiation with IQOS; JAIQOS: previous cigarette smoking history before starting IQOS use) of TNP use.

Survey dispositions and outcome rates

Regarding survey dispositions and outcome rates (Figure 3), the JGAP samples had a response rate of >30% in each of the three study years (Y1-Y3), which resulted in sample sizes of 4,878, 4,791, and 7,236 for Y1-Y3 of the Omnibus survey, respectively. In the Y1-Y3 JAIQOS samples, response rates of 19.4%, 4.7%, and 2.0% yielded sample sizes of 2,000, 2,044, and 2,013, respectively.

Figure 3. Flow diagram of study samples and survey dispositions in the JGAP and JAIQOS samples.

Abbreviations: JAIQOS, sample of adult Japanese IQOS™ users from PMI’s IQOS owner database in Japan; JGAP, representative sample of the Japanese general adult population; n1-n3, sample sizes for study years 1–3, respectively; PMI, Philip Morris International.

Sample characteristics

JGAP samples

Overall, the demographic characteristics of the JGAP samples were similar across Y1-Y3 (Table 1) and comparable with those of the Japanese adult population²⁵. The mean (±SD) ages of the Y1-Y3 samples were 53.8 (±17.9), 54.5 (±17.6), and 54.8 (±17.8) years, respectively, and each of the samples included slightly more women (Y1-Y3: 51.9%, 53.3% and 53.2%) than men (Y1-Y3: 48.1%, 46.7%, and 46.8%), mirroring the female skew in the actual Japanese population²⁵.

In each of Y1-Y3, a larger proportion of the sample was based in a major city (Y1-Y3: 27.4%, 28.1%, and 28.6%) than in rural areas (Y1-Y3: 10.0%, 9.3%, and 8.8%). Across Y1-Y3 (Table 1), most of the samples reported high school (Y1-Y3: 49.1%, 50.8%, and 49.8%) or college/university (Y1-Y3: 40.6%, 40.0%, and 41.0%) as the highest level of education, and the most common occupations were homemaker (Y1-Y3: 24.8%, 24.5%, and 24.8%), manual employee (Y1-Y3: 21.8%, 22.8%, and 21.6%), and clerical employee (Y1-Y3: 19.0%, 17.6%, and 17.8%).

JAIQOS samples

Overall, the demographic characteristics of the JAIQOS samples were similar across Y1-Y3 (Table 1). The mean (±SD) ages of the Y1-Y3 samples were 38.5 (±9.7), 39.7 (±10.1), and 39.9 (±9.9) years, respectively, and in each of the samples there were more men (Y1-Y3: 81.6%, 80.3%, and 79.9%) than women (Y1-Y3: 18.4%, 19.7%, and 20.1%).

Across Y1-Y3 (Table 1), most of the participants reported completing college/university (Y1-Y3: 56.8%, 54.4%, and 53.9%) or high school (Y1-Y3: 36.3%, 36.8%, and 37.0%), and the most common occupations were manager (Y1-Y3: 20.7%, 21.1%, and 19.8%) and self-employed/small business owner (Y1-Y3: 16.5%, 17.0%, and 17.6%).

TNP use in JGAP samples

Prevalence of overall TNP use

Across Y1-Y3, the prevalence of overall current (Y1-Y3: 18.5%, 18.9%, 18.2%) former (Y1-Y3: 18.7%, 16.3%, 16.9%), and never (Y1-Y3: 62.9%, 64.8%, 64.9%) TNP use as well as of TNP use by age and sex were similar (Table 2).

Prevalence of individual TNP use

Cigarette smoking prevalence decreased from 17.6% in Y1 to 16.0% in Y3, while the use prevalence of other TNPs, including HTPs and e-cigarettes, increased (Table 3). The use prevalence of all HTP brands (i.e., IQOS™, Ploom/Ploom Tech, and glo) increased across Y1-Y3, and, of all HTP brands surveyed, IQOS had the highest use prevalence (Y1-Y3: 1.8%, 3.2%, 3.3%). The use prevalence of e-cigarettes increased from 0.7% to 1.6% to 2.0% during Y1-Y3.

In each of Y1-Y3, cigarette smoking was more prevalent among men (Y1-Y3: 28.2%, 28.4%, and 26.3%) than women (Y1-Y3: 7.9%, 7.7%, and 7.0%) and was highest among 40–49-year-olds (Y1-Y3: 24.6%, 22.2%, and 20.4%). The IQOS™ use prevalence in each of Y1-Y3 was higher among men (Y1-Y3: 3.0%, 5.1%, and 5.4%) than women (Y1-Y3: 0.6%, 1.5%, and 1.5%) and was highest among 20–29-year-olds (3.8%) in Y1, but shifted to be highest among 30–39-year-olds in Y2 (8.7%) and Y3 (9.0%). In both Y1 (1.7%) and Y2 (3.0%), e-cigarette use prevalence was highest among 20–29-year-olds, but in Y3 shifted to be highest among 30–39-year-olds (3.9%; Table 3).

Patterns of TNP use

Across Y1-Y3 (Table 4), exclusive use of only one type of TNP decreased (Y1-Y3: 82.3%, 75.0%, and 70.4%), while dual use of two types of TNPs increased (Y1-Y3: 14.3%, 17.2%, and 16.7%) and poly-TNP use increased markedly (Y1-Y3: 2.1%, 6.1%, and 10.0%).

Across Y1-Y3 (Table 4), the greatest proportion, although declining, of participants who reported TNP use were exclusive cigarette smokers (Y1-Y3: 79.5%, 68.1%, and 63.0%), while conversely, the proportion of exclusive IQOS users increased (Y1-Y3: 2.5%, 4.8%, and 5.3%), and that of exclusive e-cigarette users remained low (Y1-Y3: 0.3%, 1.1%, and 0.5%).

Frequency and Intensity of TNP use

Among the participants in the JGAP samples in Y1-Y3 who were currently using cigarettes (Y1-Y3: n=852; n=825; and n=1,150), the average number of cigarettes smoked per day (over the last 30 days) appeared to be stable (Y1-Y3: 16.0, 15.7, and 15.5) (Table 5).

TNP initiation/relapse/reinitiation

Among the participants who were never TNP users 12 months prior to the survey (Y1-Y3: n=3,066; n=3,109; and n=4,685), TNP use initiation with cigarettes in the preceding 12 months was considerably higher (Y1-Y3: 0.2%, 0.3%, and 0.2%) than initiation with IQOS™ (Y1-Y3: 0.03%, 0.1%, and 0.1%) (Table 6).

Among current TNP users in Y1-Y3 (Y1-Y3: n=894; n=900; and n=1,304), in each year only one participant reinitiated TNP use with IQOS™ (Y1-Y3: 0.1%, 0.1%, and 0.07%). No relapse to IQOS use was reported in any of the three study years (Table 7).

TNP use in the JAIQOS samples

Patterns of TNP Use

Across Y1-Y3, a decreasing majority of participants in the samples used IQOS™ exclusively (Y1-Y3: 63.4%, 52.3%, and 49.4%), while the proportion who used IQOS together with other smoke-free TNPs increased (Y1-Y3: 7.6%, 17.7%, and 27.0%) and the proportion who used IQOS together with combustible TNPs decreased (Y1-Y3: 28.4%, 25.4%, and 23.6%). Consequently, by Y3, a greater proportion of participants used IQOS together with other smoke-free TNPs than IQOS together with combustible TNPs (Table 8).

Frequency and intensity of TNP use

In each of Y1-Y3 (Table 5), the average number of days of IQOS™ use in the last 30 days (Y1-Y3: 29.1, 28.9, and 28.8) and the average number of HEETS™/HeatSticks™ used on the days of IQOS use in the last 30 days (Y1-Y3: 16.2, 16.5, and 15.9) were relatively stable. Thus, the average daily HEETS/HeatSticks consumption (over the last 30 days) across Y1-Y3 was similarly stable (Y1-Y3: 15.9, 16.1, and 15.5).

History of TNP use

In each of Y1-Y3, the majority of the JAIQOS sample participants had a smoking history before starting IQOS™ use (Y1-Y3: 98.0%, 98.7%, and 99.3%), while only a few were never smokers (Y1-Y3: 2.0%, 1.3%, and 0.7%) before starting IQOS use.

Strengths and limitations

Major strengths of this study, among those previously described²³, include the annual repeated data collection using the same sampling framework and methods, face-to-face interviews, and nationally representative samples.

The limitations of the current study include all biases typically associated with self-reported measures, such as recall, social desirability, or response bias as well as sampling and selection bias. To overcome these limitations, particularly in the JAIQOS sample, the response rates were monitored and compared against preestablished quotas on the basis of age and sex.

The limitations of the current study include foremost the cross-sectional and observational design that does not allow for cause-effect inference or investigation of switching/transition behaviors over time. Moreover, the study may have suffered from all biases typically associated with self-reported measures, such as response bias including i.a., recall, social desirability, or order effect bias, as well as sampling and selection bias. To overcome possible sampling and selection bias, particularly in the JAIQOS samples, the response rates were monitored and compared against pre-established quotas on the basis of age and sex. Furthermore, the JAIQOS samples, which we used to obtain reliable estimates for investigating the patterns of IQOS™ use, were not representative of the IQOS users in the Japanese general adult population, but only of the IQOS users registered in the large database of PMI’s affiliate in Japan. Similarly, because IQOS was the first HTP and had the highest use prevalence in Japan, IQOS use behavior may not have represented the use behavior of other HTPs available in Japan. Finally, the present study neither addresses quitting behavior, e.g., investigation of whether IQOS™/HTP use prevents those TNP users who are willing to quit all TNPs to do so, nor does the study cover IQOS/HTP underage use, which are critical aspects for assessing the impact of IQOS/HTP use on tobacco harm reduction.