Article Text

other Versions

Download PDFPDF
‘Enter at your own risk’: a multimethod study of air quality and biological measures in Canadian waterpipe cafes
  1. Bo Zhang1,
  2. Farzana Haji1,
  3. Pamela Kaufman1,2,
  4. Sarah Muir1,
  5. Roberta Ferrence1,2
  1. 1Ontario Tobacco Research Unit, Dalla Lana School of Public Health, University of Toronto, Toronto, Canada
  2. 2Center for Addiction and Mental Health, Toronto, Canada
  1. Correspondence to Dr Bo Zhang, Ontario Tobacco Research Unit, Dalla Lana School of Public Health, University of Toronto, Room: T521, OTRU, 33 Russell Street, Toronto, ON, Canada M5S 2S1; bo.zhang{at}


Background Tobacco and non-tobacco-based waterpipe smoking has increased exponentially in many countries in recent decades, particularly among youth and young adults. Although tobacco smoking is banned in many indoor public places, waterpipe smoking, ostensibly non-tobacco, continues in Ontario and other jurisdictions where only tobacco smoking is prohibited. This study assessed air quality and exposure in waterpipe cafes using multiple methods and markers.

Methods Indoor (n=12) and outdoor (n=5) air quality was assessed in Toronto, Canada waterpipe cafes from 30 August to 11 October 2012. Real-time measurements of air nicotine, fine particulate matter less than 2.5 microns in diameter (PM2.5) and ambient carbon monoxide (CO) were collected in 2 h sessions. Levels of CO in breath were collected in non-smoking field staff before entering and upon leaving venues. Observations of occupant behaviour, environmental changes and venue characteristics were also recorded.

Results In indoor venues, mean values were 1419 µg/m3 for PM2.5, 17.7 ppm for ambient CO, and 3.3 µg/m3 for air nicotine. Levels increased with increasing number of active waterpipes. On outdoor patios, mean values were 80.5 µg/m3 for PM2.5, 0.5 ppm for ambient CO, and 0.6 µg/m3 for air nicotine. Air quality levels in indoor waterpipe cafes are hazardous for human health. Outdoor waterpipe cafes showed less harmful particulate levels than indoors, but mean PM2.5 levels (80.5 µg/m3) were still ‘poor’.

Conclusions Staff and patrons of waterpipe cafes are exposed to air quality levels considered hazardous to human health. Results support eliminating waterpipe smoking in hospitality venues indoors and out.

  • Nicotine
  • Non-cigarette tobacco products
  • Public policy
  • Secondhand smoke
  • Environment

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Secondhand smoke causes death and disease in never-smokers.1 ,2 The 2006 Report of the US Surgeon General concluded there is no risk-free level of exposure to tobacco smoke.1 To limit exposure, many jurisdictions, including Canada, Australia, Italy, Arizona, California and New York, have banned smoking in indoor public places and workplaces. However, legislation is often limited to tobacco products, exempting smoking of non-tobacco-based products, sometimes referred to as ‘herbal’, or of waterpipe altogether.3–5

Waterpipe smoking (also known as hookah, shisha, narghile and arguileh) is a form of tobacco consumption traditionally used in Iran, South Asia and Middle Eastern countries.6 The tobacco mixture (known as ma'assel) is indirectly heated using charcoal on perforated aluminium foil that sits on top of the tobacco. The resulting smoke is drawn through the body of the waterpipe and cooled by the water at the base.7 The smoker inhales the smoke through the mouthpiece at the end of the hose attached to the waterpipe.

A new epidemic of flavoured waterpipe use began in the 1990s and has spread throughout the Middle East and several Western countries, particularly among youth.8–15 Contrary to the popular belief that waterpipe smoking is less harmful than cigarette smoking,16 recent studies show that waterpipe smoking is linked to exposures, health risks and dependence comparable to those of cigarette smoking.17–21 A recent systematic review found that waterpipe tobacco smoking was significantly associated with lung cancer, respiratory illness, low birth weight, carbon monoxide (CO) poisoning, adverse cardiac events and periodontal disease.19

Some countries, including Lebanon, Kazakhstan, Turkey, parts of India and Saudi Arabia where waterpipe has been widely used, have recently banned its use in indoor public places. Several jurisdictions in the USA and Canada have introduced similar legislation. In Ontario, the Smoke-Free Ontario Act banned the smoking of ‘lighted tobacco’ in indoor public places and workplaces in 2006, but the provincial government concluded that waterpipe smoking using non-tobacco-based products was exempt. This has led to enforcement issues in waterpipe cafes because of the expense and difficulty associated with establishing the presence of tobacco use. The aim of this study was to assess real-world indoor and outdoor ambient air concentrations of three major components of waterpipe smoke: ultrafine particulates (PM2.5), CO and nicotine, as well as a biomarker for exposure among field staff.

Materials and methods

Study sample

Toronto Public Health provided a list (n=30) of waterpipe cafes in the City of Toronto, which excluded venues under investigation for illegal sales or other infractions (considered unsafe for data collection). Of these, 13 were excluded (seven closed, five not offering waterpipe anymore, and one with safety concerns). In this paper, we report measurements from 12 indoor waterpipe cafes and five patio venues, located across the city.

Data collection

Data were collected between 30 August and 11 October 2012. Most air quality sampling took place on busy days for the establishments on Thursdays, Fridays and Saturdays, 19:00–23:30.

Real-time measurements of PM2.5 were collected using TSI AM510 SidePak (TSI) personal aerosol monitors, following the indoor air monitoring protocol developed for the Global Air Monitoring Study.22 The monitor was fitted with a 2.5 µm impactor, and the flow rate was set at 1.7 L/min to measure the concentration of PM2.5 at 30 s log intervals. A standard calibration adjustment factor of 0.32 was applied to the raw data to correct for the properties of secondhand smoke.

Air nicotine was measured by clipping an active nicotine monitor to an upper outside pocket of the clothing of one of our field staff for at least 2 h in each venue (one filter per venue). Active nicotine monitors include a small motor that filters air at a higher speed than with passive nicotine monitors. Active monitors require a minimum of 2 h to adequately measure air nicotine (personal communication, Charles Perrino, UC Berkeley). Filters are treated with sodium bisulfate so that nicotine in the air is absorbed when air passes through. The nicotine is then desorbed from the filters and analysed by gas chromatography with nitrogen-sensitive detection.23–25 Nicotine monitors and filters were obtained from and processed at the Hammond laboratory at UC Berkeley.

The TSI's Q-Trak multifunction indoor air quality monitor (model 7575, TSI) was used to measure ambient CO, carbon dioxide (CO2), temperature, and relative humidity (RH) at 30 s log intervals.26 The monitor was previously calibrated for CO and CO2 by running a span gas with a known concentration (CO at 35 ppm and CO2 at 1000 ppm) and a zero gas (nitrogen) through the monitor by the TSI distributor.

Venues were visited unannounced by a team of at least two trained field staff. They identified the busiest area of the cafe where waterpipe smoke was active, and positioned themselves as far as possible from kitchen areas and open windows to reduce contamination from other sources of particulates and CO. To obtain unobtrusive measurements, monitors were carried in a handbag or backpack. A length of tubing was attached to the inlet of the monitors, with the other end protruding slightly outside the bag, but not visible to bystanders. Monitors were placed on a table or seat to sample ambient air. Air quality data were collected continuously for at least 120 min at each venue. To obtain background readings, field staff took continuous readings for at least 5 min in a neighbouring park, courtyard or street with no nearby smokers.

Field staff used personal height and pacing to estimate the length, width and height of the venues and the distance between monitors to the nearest active waterpipe, with training, were able to produce estimates comparable to those obtained from standard measurement tools. Personal height and pacing measures have been used in other air quality studies.27 ,28 This approach was employed because laser meters are likely to be noticed by cafe staff, possibly jeopardising the study. Our non-smoking staff also measured their own breath levels of CO using a piCO+ monitor before entering and just after leaving the venue.29 Breath CO was measured by field staff at 11 indoor and five patio venues, and by one staff person at one indoor venue (total of 23 indoor and 10 patio breath CO measurements). The breath CO monitor was previously calibrated according to manufacturer's instructions.

Number of active waterpipes, distance from the nearest active waterpipe to the monitors, number of people at the venue, and number of door and/or window openings were recorded at 5 min intervals throughout each 2-h measurement session in a time-log, synchronised with the monitors. Any cigarette smoking on patios was also recorded every 5 min during data collection.


Descriptive analysis was conducted for air quality measures (PM2.5, ambient CO, and air nicotine), breath CO, and venue characteristics (number of patrons, number of active waterpipes, volume of venues, CO2, temperature and RH). Pearson correlation coefficient was calculated for breath CO levels of the two field staff. Paired t Tests were used to compare breath CO readings before entering and after leaving the venue. t Tests were used to compare differences in air-nicotine levels among background, patio and indoor locations. Ordinary Linear Regression was used to examine the impact of the number of active waterpipes on air-nicotine level. Mixed-effects linear models, using autoregressive covariance structure and restricted maximum likelihood method, which accounts for the repeated-measures design, were used to model the effect of waterpipe smoking on different air quality measures. Two types of mixed-model analyses were employed, one treating the number of active waterpipes as categorical data, with the background measures as the referent group, and the other treating them as continuous data and excluding background measures. The analysis including background measures was adjusted for temperature, RH and CO2. The linear trend analysis was adjusted for venue volume, distance (metres) from the nearest active waterpipe to the monitors, and door/window openings, in addition to temperature, RH and CO2. Natural log-transformed data for ambient PM2.5 and CO were used to obtain the normal distribution for mixed model analysis. All analyses were conducted using SAS V.9.3 (SAS, Cary, North Carolina, USA).

No measurements were made on human subjects other than our field staff, and venues were not identified. Field staff provided verbal consent to participate in the study and were advised to leave venues if they felt uncomfortable due to smoke exposure. Waiver of requirement for full ethical review was obtained from the Research Ethics Board, University of Toronto.


The 12 indoor waterpipe cafes were located in the downtown core (n=3), inner suburbs (n=7), and suburban areas (n=2) across Toronto; room volume ranged from 88 to 2039 m3. Indoor waterpipe smoking was observed in all cafes, and there was no visual evidence of indoor cigarette smoking. On average, there were 25 patrons and 10 active waterpipes inside each cafe during the 2 h measurement session; the active waterpipe smoker density was 3.8 per 100 m3 (table 1). The five patio venues were located in the downtown core (n=1), and in suburban areas (n=4). Patio venues had fewer patrons and active waterpipe smokers compared to indoor venues, with areas ranging from 6.9 to 139 m2.

Table 1

Venue characteristics, ambient and breath air quality measures by location, August–October 2012, Toronto Waterpipe Study

The air quality inside waterpipe cafes was much poorer than background levels (outdoor spaces with no nearby smoking). On average, PM2.5 was 69 times higher (1419.4 vs 20.7 µg/m3), and ambient CO was 89 times higher (17.7 vs 0.20 ppm) inside waterpipe cafes, compared with background measures (p<0.001). Temperature and CO2 were higher indoors, and RH was lower compared with outdoors. Breath CO among the non-smoking field staff was eight times higher, after 120 min inside venues (p=0.05) (table 1), and readings for the two staff were highly correlated (r=0.95). PM2.5 was almost four times higher on patios than background, whereas ambient CO levels on patios were largely unchanged.

We compared selected air quality measures by nicotine level (<1 vs ≥1 µg/m3) (table 2). Five waterpipe cafes had nicotine levels <1 µg/m3 (mean nicotine=0.70 µg/m3) and seven venues had levels ≥1 µg/m3 (mean nicotine=5.67 µg/m3). PM2.5 was nine times higher (p<0.05) in venues with high levels of air nicotine; ambient CO was five times higher (p<0.05), and breath CO of field staff was six times higher (p<0.001) than venues with low levels of air nicotine. This is at least partly due to the smaller mean number of active waterpipes in cafes with low air-nicotine levels (n=4) than venues with high air-nicotine levels (n=14).

Table 2

Selected air quality measures by nicotine level in indoor venues, August–October 2012, Toronto Waterpipe Study

For almost half the measurement time (45%), there were 10 or more active waterpipes in all indoor venues. We observed only a small proportion (4%) of measurement time (1449 min for all 12 indoor venues) with no active waterpipes in four indoor venues (less than 20 min in each venue). With no active waterpipes, air quality measures were only marginally higher than background measures. With 10 or more active waterpipes, PM2.5 was 102 times higher than background (mean=2602.1 vs 25.4 µg/m3), ambient CO was 65 times higher than background (mean=26.1 vs 0.4 ppm), and air nicotine was eight times higher (mean=4.65 vs 0.57 µg/m3), than with 1–4 active waterpipes (table 3).

Table 3

Air quality measures by number of active waterpipes in indoor venues (n=12), August–October 2012, Toronto Waterpipe Study

Results from adjusted analyses confirmed findings in the crude analyses presented in table 3 (table 4). PM2.5, ambient CO and nicotine levels by number of active waterpipes increased with increments of active waterpipes, after adjusting for temperature, RH, CO2, venue volume, number of door and window openings and distance from the nearest active waterpipe to the monitor (the first three factors in model one, and all factors in model two); however, the trend analysis for ambient CO and air nicotine was marginally significant.

Table 4

Adjusted analysis of air quality measures by number of active waterpipes, August–September 2012, Toronto Waterpipe Study

To illustrate, figure 1 shows the real-time air quality measures of PM2.5 and ambient CO in the indoor waterpipe cafe with the highest mean levels of ambient CO and air nicotine, and the fifth highest mean for PM2.5 among all 12 indoor venues. Levels of air quality measures increased dramatically as field staff entered the venue and as number of active waterpipes increased. Levels dropped immediately upon exiting the venue. During the 2 h measurement session, PM2.5 and ambient CO levels were almost always greater (95% and 91% of measurement time, respectively) than the hazardous level based on the air quality index (AQI) of the Environmental Protection Agency (EPA)30 (AQI >301, corresponding to PM2.5 >250 µg/m3 and ambient CO >30 ppm). The 2 h average level of PM2.5 in this venue was 2039 µg/m3, 67 ppm for ambient CO, and 7.91 µg/m3 for air nicotine. The mean number of active waterpipes was 9.4, and the volume of the room was 294 m3. Breath CO of field staff was 2 ppm before entering the venue and 70 ppm after 2 h of exposure. (This very high level is no doubt related to the high level of ambient CO).

Figure 1

Real-time air quality measures of PM2.5 (µg/m3) and carbon monoxide (CO) (ppm) for one indoor waterpipe cafe with the highest mean value of CO, measured on 21 September 2012, Toronto Waterpipe Study. Mean values were 67 ppm for CO, 2039 µg/m3 for PM2.5, 7.9 µg/m3 for air nicotine, and 9.4 for active waterpipes, during 123 min measurement session.

On average, employees of Toronto waterpipe cafes who worked 8-h shifts were exposed to hazardous levels of air quality, based on the EPA AQI. Even if they were exposed to only background levels found in this study for the remaining 16 h per day, their PM2.5 over 24 h would equal 487 µg/m3. This is much higher than the ‘hazardous’ level of 250 µg/m3 for PM2.5 indicated by EPA AQI. Visits of only 2 h for non-smoking patrons, assuming background levels of exposure at other times, would lead to a mean exposure of PM2.5 over 24 h of 137 µg/m3, corresponding to the ‘unhealthy’ level, and the ‘very unhealthy’ level of 236 µg/m3, if there were 10 or more active waterpipes. The highest mean PM2.5 measured was 5318 µg/m3 (mean active waterpipes=22). As little as 15 min of exposure to PM2.5 at this venue would be at the ‘unhealthy’ level. Although PM2.5 levels on patios were much lower than indoors, the average of 81 µg/m3 was at the ‘poor’ air quality level according to the Ontario Ministry of the Environment (46–90 µg/m3 for PM2.5 over 3 h).31


To our knowledge, this is the first study to measure air quality at waterpipe cafes using multiple ambient air measures and a biomarker. In this study, we measured PM2.5, ambient CO, air nicotine and breath CO, as well as factors that may affect these measures, including number of active waterpipes, distance from monitors to the nearest active waterpipe, CO2, temperature, RH, venue volume and door and window openings.

At indoor cafes, employees and patrons were exposed to a mean level of PM2.5 of 1419 µg/m3 (median=535 µg/m3) during the 2 h measurement session. This level is much higher than that reported for smoking rooms of restaurants and bars in a recent European study (median PM2.5=120.5 µg/m3).32 If we assume that waterpipe cafe employees work 8 h shifts, and that patrons stay at least 2 h (our field staff observed that many patrons, especially university students, stayed for two or more hours while studying or using the internet), and that both groups are exposed to the mean background level of PM2.5 for the remainder of the period, employees’ average exposure over 24 h would correspond to the ‘hazardous’ level (487 µg/m3, AQI>301), and patrons would be exposed to ‘unhealthy’ levels (137 µg/m3, AQI 151–200) based on the EPA AQI.30 In some circumstances, exposure to PM2.5 in these waterpipe cafes for as little as 15 min would be considered ‘unhealthy’.

Health Canada's Residential Indoor Air Quality Guideline recommends maximum limits for ambient CO exposure of 25 ppm, based on a 1 h average, and 10 ppm based on a 24 h average.33 We found that the average ambient CO level during a 2 h session was over 25 ppm in three of 12 venues. Pickett and Bell recommend leaving the environment or admitting fresh air if the ambient CO level is greater than 20 ppm based on a 1 h average.34 The ambient CO levels measured in our study indicate that employees and patrons in Toronto waterpipe cafes face serious health risks from CO exposure alone, without accounting for particulate and nicotine exposure.

The average air-nicotine level was 3.27 µg/m3 across all indoor cafes. A recent European study reported a median air-nicotine level of 3.69 µg/m3 in cigarette smoking venues, and 0.48 µg/m3 in non-smoking restaurants and bars.32 The mean air-nicotine level in our low-nicotine venues, where apparently few patrons were using tobacco products was 0.70 µg/m3, comparable with the nicotine level in non-smoking venues in the European study. (Our field staff observed a traditional tobacco product being served even in the venue with the lowest air-nicotine reading.) However, the PM2.5 levels in these low-nicotine venues were still hazardous (mean=223.4 µg/m3; cf. Shihadeh et al35). If employees work in these venues for 8 h shifts and are exposed to mean background levels for the remaining 16 h of their day, their average PM2.5 exposure over 24 h would be 88.3 µg/m3, corresponding to the ‘unhealthy’ level of the EPA AQI.

The mean breath CO level of 15.1 ppm for our two non-smoking field staff (measured across all indoor venues for 2 h) was midway between mean CO levels for 90 smokers who consumed 1–10 cigarettes (mean=11.9 ppm, 95% CI 8.7 to 15.0) or 11–20 cigarettes per day (mean=17.7 ppm, 95% CI 14.3 to 21.0) in a study at John Radcliffe Hospital, Oxford (mean age=34, 62% male and 14% students).36 The maximum breath CO of 70 ppm reported in our study is very high compared with a double-blind clinical trial37 that found 84% of participants who smoked waterpipe for at least 45 min had breath CO levels ≤35 ppm. However, that study used controlled ventilated laboratory conditions, and the mean peak ambient CO level was only 4.0 ppm. Higher levels of breath CO were reported among 59 waterpipe smokers with high waterpipe dependence scores (mean breath CO=62.3 ppm), after 30 min of waterpipe smoking.38 In our study, CO level in the breath was measured after a much longer period of exposure in a very smoky waterpipe cafe (mean and median active waterpipes: 9.4 and 10, respectively). The waterpipe cafe was small (volume: 294 m3) with no ventilation; the corresponding ambient CO was high (mean, median and maximum: 67, 69 and 120 ppm, respectively), and the indoor ambient CO level in the last hour was above 70 ppm. Although our sample size for breath CO was small, the high correlation coefficient (0.95) suggests that a larger sample would have produced similar results.

These findings indicate that non-smoker exposure to waterpipe smoke in Toronto waterpipe cafes for just 2 h is approximately equivalent to smoking 10 cigarettes per day or to smoking waterpipe for 15 min. Moreover, non-smoking employees and patrons at outdoor waterpipe cafes may be exposed to ‘poor’ levels of air quality based on the Ontario Ministry of the Environment AQI (PM2.5 between 46 and 90 µg/m3 over 3 h).31 The mean PM2.5 (81 µg/m3) on patios from waterpipe smoke in this study was also much higher than the annual outdoor average (10 µg/m3) that WHO sets as the lowest cut-off at which lung cancer and cardiopulmonary deaths are likely to increase.39 This is a particular concern for chronically exposed hospitality workers.

Cobb et al3 measured PM2.5 in Virginia waterpipe cafes and found a mean level of 374 µg/m3 in waterpipe cafe smoking rooms, much lower than the mean level of 1419 µg/m3 found in our study. Although the average active waterpipe smoker density was higher (4.8 active waterpipes per 100 m3; calculation based on published results) than in our study (3.8 active waterpipes per 100 m3), we measured PM2.5 for 120 min compared to their 30 min period. In 2010, Fiala et al40 also found unhealthy to hazardous levels of PM2.5 in 10 hookah lounges in Oregon with 30 min of measurement. Particulate levels were even lower, averaging 198 µg/m3, and only one reading surpassed 220 µg/m3. An earlier experimental study showed cumulative exposure levels to PM2.5 and CO increased with exposure time.41 Other factors that might explain the difference include ventilation of the venues (perhaps poorer in our study) and differences in waterpipe smoking products. Unlike Cobb et al,3 we found that levels of PM2.5 increased significantly with number of active waterpipes, even after adjusting for other environmental factors (temperature, RH, CO2 and venue volume) that may potentially confound the measurement concentration of PM2.5, ambient CO and nicotine. These factors indicate that our findings of high levels of PM2.5 are robust.

Ambient nicotine is also potentially hazardous: its presence in ostensibly tobacco-free venues indicates that patrons may be unknowingly served nicotine, which can lead to addiction and long-term use of other tobacco products. Maternal exposure to nicotine, even at low levels, is associated with a range of serious health problems in laboratory animals, including second-generation effects, and probably in humans.42 Nicotine is now known to combine with nitrous acid (HONO) producing tobacco-specific nitrosamines that circulate indoors for several hours exposing occupants.43 Taken as a whole, this evidence suggests that any indoor exposure to nicotine is potentially hazardous.

Although not measured in this study, new research shows high levels of benzene, a potent carcinogen, in waterpipe smoke,20 and there is now evidence for genetic damage from thirdhand smoke that off-gases in indoor environments.44

Although our study sample of waterpipe cafes in Toronto is relatively small and not random, venues were widely dispersed in the city, and included almost all waterpipe cafes that were not under investigation for illegal sales or other infractions. Since waterpipe cafes under investigation are more likely to serve tobacco-based waterpipe, we expect that our results underestimate actual exposure to nicotine.

In conclusion, this study provides evidence that levels of particulates and CO found in Toronto waterpipe cafes are hazardous to the health of smokers and non-smokers, and the cafe staff in particular. Nicotine exposure also poses serious risks. We provide new evidence that outdoor waterpipe smoke exposure can be substantial and concerning. Additionally, continued social exposure to smoking in any form normalises the use of tobacco and alternative non-tobacco products. Public health action in the form of a ban on waterpipe smoking in all indoor and outdoor public places is warranted to protect employees and patrons, as well as family members, including children, who may live in adjacent spaces.

What this paper adds

  • This is the first field study to employ several real-time air quality and biological measures, including PM2.5, ambient carbon monoxide (CO), air nicotine and breath CO for a 2 h period in waterpipe cafes. Unlike other studies that measured 30 min exposures, this longer measurement period approximates real-life exposure.

  • Uniquely, we assessed air quality in 12 indoor venues and on five outdoor patios, and found hazardous and poor air quality, respectively.

  • Levels of ambient CO approached or exceeded recommended exposure thresholds for industrial settings and pose a serious risk to those exposed.

  • Mean levels of air nicotine in indoor venues that claimed to serve ‘tobacco-free’ waterpipe products were comparable with levels of air nicotine found in smoky bars, suggesting that exemptions for non-tobacco-based (‘herbal’) products are ill-advised.

  • Our findings provide important evidence that waterpipe cafes, indoors and outside on patios, are toxic environments for staff and patrons.


We are grateful to David Baines of Toronto Public Health for providing the list of potential venues. We also thank Charles Perrino of the UC Berkeley Lab for his very helpful counsel.



  • Contributors All authors contributed substantially to conception and design of the study. FH and SM carried out data collection. BZ conducted data analysis and drafted the manuscript. RF conceived of the project, obtained funding, and oversaw all aspects of the study. All authors contributed substantially to resolution of field work issues and interpretation of data. They also critically reviewed the manuscript, made revisions for important intellectual content and gave final approval of the submitted version.

  • Funding This research was undertaken at the Ontario Tobacco Research Unit, Dalla Lana School of Public Health, University of Toronto, which receives funding from the Ontario Ministry of Health and Long-term Care.

  • Competing interests All authors receive funding from the Ontario Ministry of Health and Long-term Care through a core grant for the Ontario Tobacco Research Unit. All authors have declared no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years, and have no other relationships or activities that could appear to have influenced the submitted work.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement At this point, our data are not available to other researchers. We plan to do further analyses.