Comparative Indoor Pollution from Glo, Iqos, and Juul, Using Traditional Combustion Cigarettes as Benchmark: Evidence from the Randomized SUR-VAPES AIR Trial

Modified risk products (MRP) such as electronic vaping cigarettes (EVC) and heat-not-burn cigarettes (HNBC) are appealing alternatives to combustion cigarettes. Limited between- and within-device comparative data are available on MRP. We aimed at comparing indoor particulate matter (PM) emissions measured in a randomized trial enforcing standardized smoking sessions, testing different devices and flavors of MRP, using traditional combustion cigarettes (TCC) as benchmark. Overall, MRP yielded significantly lower levels of indoor PM in comparison to TCC (with median PM levels during smoking for MRP < 100 μg/m3, and for TCC > 1000 μg/m3). Despite this, significant differences among MRP were found, with Iqos appearing associated with a significantly lower burden of emissions for all the monitored fractions of PM, including total PM (all p < 0.05). Precisely, during use, PM ≤1 µm (PM1) emissions were 28 (16; 28) μg/m3 for Glo, 25 (15; 57) μg/m3 for Iqos, and 73 (15; 559) μg/m3 for Juul (p < 0.001 for Glo vs. Iqos, p < 0.001 for Glo vs. Juul, and p = 0.045 for Iqos vs. Juul). Exploratory within-MRP analyses suggested significant differences between flavors, favoring, for instance, Ultramarine for Glo, Bronze for Iqos, and Mango for Juul, even if results varied substantially according to individual smoker. In conclusion, leading MRP have significantly less intense and persistent effects on indoor pollution in comparison to TCC. Yet, when focusing solely on MRP, between-product and between-flavor differences appear, with quantitative estimates suggesting lower polluting effects with Iqos. These results, if confirmed externally, could be used to individualize product and flavor choice to minimize the untoward effects of EVC and HNBC on indoor pollution.


Introduction
Smoking of traditional combustion cigarettes (TCC) is associated with major burdens of mortality, morbidity, and cost [1,2]. Recently, novel smoking approaches have been introduced, with very favorable market penetration, including electronic vaping cigarettes (EVC) and heat-not-burn cigarettes (HNBC) [3,4]. The popularity of EVC has considerably increased during the past decades in the United States [5] and in European countries [6]. Similarly, the use of HNBC has rapidly increased worldwide, particularly among young individuals [7]. Whether these novel modified risk products (MRP) may actually prove as reduced risk products (RRP) remains still uncertain, despite several reports suggesting that both EVC and HNBC may prove less detrimental than TCC in several dimensions [8,9].
Indeed, by minimizing combustion (HNBC) or avoiding it altogether (EVC), MRP may profoundly reduce the risks associated with smoking, for instance, by reducing emissions of particulate matter (PM) and carbon monoxide (CO) [10][11][12][13][14], despite still delivering nicotine, with its established detrimental clinical effects [15]. Intriguingly, while EVC would theoretically appear less toxic than HNBC given the vaporization methods, the possible presence of volatile toxic agents may undermine such purported benefits [10,16]. This might hold even truer when specific flavors are used, given the potential to increase exponentially the release of potentially hazardous agents, putting particularly some individuals at risk who could be more susceptible than others [17]. Accordingly, several reports have highlighted that MRP cannot be considered risk-free, given the multidimensional hazards associated with their use, chronic as well as acute [18][19][20].
Despite an expanding evidence base on the reduced risk of indoor pollution and ensuing passive smoking associated with MRP, comparative analyses focusing on different MRP are limited in scope and detail, as well as methodology [16,21,22]. In particular, most studies focusing on MRP published to date had one or more of the following methodological weaknesses: being non-randomized, lacking comprehensive panels of different MRP, or lacking state-of-the-art multilevel statistical models for data analysis. Moreover, few if any reports have compared MRP at large, and then, focused on MRP flavors, which may impact on short and long-term safety [23]. Recently, we have described results of a randomized trial which showed significantly lower indoor emissions of PM when using Glo (British American Tobacco, London, UK), or Iqos (Altria, Richmond, VA, USA) as HNBC and Juul (Juul Labs, San Francisco, CA, USA) as EVC, in comparison to TCC [24]. PM, a well-known toxic and carcinogen independently from its physical status (solid or liquid) and chemical composition [25], was used in such study as a global indicator of indoor air pollution. We hereby aim at providing a more poignant comparison of aggregate MRP, as well as different flavors of each MRP type, in order to expand and confirm prior findings, while capitalizing on modern state-of-the-art statistical modeling tools [26,27].
Emissions of PM with diameter ≤ 10 µm (PM 10 ), ≤ 4 µm (PM 4 ), ≤ 2.5 µm (PM 2.5 ), and ≤ 1 µm (PM 1 ) were continuously measured in real use conditions 5 min before, during, and 5 min after smoking each product in a room of 53 m 3 , with temperature and relative humidity ranging between 20 and 23 • C and 36 and 40%, respectively ( Figure S1). The air exchange rate (λ = 0.69 h −1 ) of the test room was calculated using the CO 2 tracer gas technique, as previously reported [24,[29][30][31]. During each experiment, windows and doors were maintained closed. Conversely, they were opened after each experiment until room conditions were again at initial levels of PM. The aerosol concentrations (µg/m 3 ) for PM 10 , PM 4 , PM 2.5 , and PM 1 were measured with 3 s time resolution, using a portable, laser-operated aerosol mass analyzer (Dusttrak II Aerosol Monitor, model 8530, TSI, 0.1-10 µm particle size range, TSI, Shoreview, MN, USA). The measurements were carried out in "cumulative" mode, including the mass of all particles smaller than or equal to the defined size. The aerosol was sampled directly through the entry of the instrument without using any tube or collector. The instrument was placed approximately 1.5 m above the floor level and approximately 1.5 m from the smoker, accruing hundreds of measurements for each smoking session ( Figure S1). Twelve puffs were made for each session that lasted about 5.5 min (1 puff each thirty seconds), since the common way of smoking typically consists of 10-12 puffs of a cigarette, for a period of about 5-6 min [24].

Analysis
Descriptive analysis was based on median, 1st quartile, and 3rd quartile. Inferential analysis was based, for exploratory purposes, on a mixed linear model with Gaussian likelihood using as fixed effects, timing (before, during, or after smoking) and session identifier, and as random effects, subject identifier. Specifically, this analytical approach expands, in terms of scope and precision, the previous findings as reported by Protano et al. [24]. Exploiting such a refined analytical framework, which takes into account all measurements, recognizing in the multilevel model the individual, the session, the phase (before, during and after smoking), the MRP subtype, and the MRP type (and thus, capable of simultaneously capturing within-subject, within-session, within-MRP subtype, and within-MRP type effects, as well as between-subject, between-session, between-MRP subtype, and between-MRP type effects), we focused first on the comparison between Glo, Iqos, and Juul, using TCC as the benchmark, after log10 transformation. Then, we performed within-MRP comparisons of flavors (e.g., comparing the 6 different flavors of Iqos). Finally, we explored between-subject variability in PM emissions. Statistical significance was set at the 2-tailed 0.05 level, without multiplicity adjustments. Computations and visualizations were performed with Stata 13 (StataCorp, College Station, TX, USA) with the meglm package, and R 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria) with the ggplot2 package.

Result
Overall, MRP yielded significantly lower levels of indoor PM in comparison to TCC (Table 1). In particular, during smoking sessions, median PM levels rose to >1000 µg/m 3 , with further increases due to exhalation and redistribution in the after smoking phases, at odds with median PM levels always lower than 100 µg/m 3 with MRP, irrespective of the flavor or smoking phase (Tables 1-4; Table  S1).    Notwithstanding the limited overall indoor contamination associated with MRP in comparison to TCC, between-and within-MRP comparative analysis showed significant differences in PM emissions (Table 1, Figure 1). Specifically, Iqos appeared associated with a significantly lower burden of emissions for all classes of PM, including total PM emissions (all p < 0.05), even if differences were of relatively small magnitude, and substantial variability and skewness were evident. In addition, PM concentrations quickly decreased after use with all MRP. Precisely, total PM concentrations measured in the test room during use were 39 (24; 127) µg/m 3 for Glo, 31 (20; 63) µg/m 3 for Iqos, and 76 (20; 565) µg/m 3 for Juul (p < 0.001 for Glo vs. Iqos, p < 0.001 for Glo vs. Juul, and p = 0.021 for Iqos vs. Juul), actually profiling Iqos as less polluting and Glo as more polluting. Similar effects were found for PM 10 , PM 4 , PM 2.5 , and PM 1 (all p < 0.05).
Within-MRP comparisons provided evidence that flavorings may impact PM indoor emissions, either because of smoke features or because of different smoking patterns (e.g., puff frequency and depth, or nasal vs. oral expiration), for all types of MRP under investigation. In particular, different Glo flavors were associated with significant differences in PM emissions (Table 2, Figure 2), with Ultramarine being associated with the lowest levels of PM of any size, at odds, for instance, with Aegean (e.g., for PM 10 concentrations during use, which were, respectively, 33 (22; 59) vs. 82 (31; 277) µg/m 3 , p = 0.027).
Even for Juul, between-flavor differences appeared significant, at least in terms of nominal statistical thresholds (Table 4, Figure 4). In particular, across the four flavors tested, Mango was associated with the lowest emissions of PM, at odds in particular with Golden Tobacco, which appeared associated with almost twice larger emissions (e.g., PM 4 levels during use were, respectively, 17 (11; 199) vs. 214 (61; 1270) µg/m 3 , p = 0.042).   Finally, exploratory analysis for between-smoker differences highlighted significant differences and variability in patterns of emissions, with some smokers generating lower PM concentrations, and others, higher PM concentrations, in some cases, with significant variability and skewness (Table S1, p < 0.001).

Discussion
The present work, building upon a randomized trial comparing different leading MRP and focusing on indoor pollution in a standardized experimental setting, has the following major implications. First, the MRP under investigation in the present study, i.e., Glo, Iqos, and Juul, are associated with trivial increases in indoor pollution in comparison to TCC. Second, notwithstanding the limited impact of indoor MRP use on indoor PM concentrations, statistically significant differences between Glo, Iqos, and Juul in terms of indoor pollution appear evident, with Iqos appearing less polluting and Glo more polluting, at least in relative terms. Third, within each MRP type, flavor may impact on polluting effects, either because of smoke features or because of indirect effects mediated by smoking patterns (e.g., nasal vs. oral expiration). Fourth, there remains substantial individual variability, such that indoor pollution may be high when MRP are used by some smokers, and low when used by others. Fifth, we cannot provide a consistent explanation concerning between-flavor variability such as in the case of the less polluting effects of Iqos Bronze. We can speculate that, as in the case of TCC, menthol, by having local anesthetic properties, could conceal the negative sensations of smoking, due to desensitizing receptors [32]. Accordingly, smokers could hold their breath more and thus, reduce emissions during expiration. Irrespective of these potential confounding effects, our findings have important implications in the sense that may help, if confirmed externally in larger series, smokers wishing to use MRP as a risk reduction strategy (i.e., to quit altogether TCC), to pick the one which is less likely to be harmful, at least in terms of indoor pollution [33].
The evidence base on MRP is expanding exponentially, and it is clear that, despite ongoing efforts at regulating their use (e.g., with increased taxation), MRP usage will continue to grow [3,8,9,34]. Accordingly, it is paramount to expand the evidence concerning these products, with details worth being sought ranging from cardiovascular effects to additivity and polluting impact. Indeed, while the dramatic and persisting indoor polluting effect of TCC is very clear, MRP also adversely modify indoor ambient air [20], for instance, by releasing PM and other established or potentially toxic agents [35]. This holds true for HNBC as well as EVC, despite their evident inherent differences [35,36]. In particular, it is established that even single usage of MRP increases indoor PM levels significantly [36]. Our work builds upon such findings and provides, to date, the first thorough comparison between two leading HNBC, Glo and Iqos, with a leading EVC, Juul. Interestingly, MRP appear unequal in terms of indoor polluting effects, in the sense that in our study, Iqos was associated with less indoor polluting effects than Glo and, albeit to a lesser extent, Juul. Accordingly, indoor pollution associated with MRP is not MRP-type specific, but rather, device-specific.
Furthermore, building upon accruing evidence suggesting the impact of different flavors of a given MRP on its risk profile, we explored within-device (i.e., between-flavor) polluting effects, finding indeed that some flavors are associated with more indoor pollution than others [37][38][39][40][41][42][43][44][45]. Notwithstanding the evident impact of smoker features (which may range from habitus to cardiopulmonary physiology or smoking style), it is clear that MRP should not be considered all identical, and even specific flavors may be more or less hazardous than others [38,41]. Indeed, differences in exhaled aerosols from different smokers depend on many factors, including individual characteristics (i.e., age, gender, lung capacity) and status (stress, anxiety, time since the last cigarette, nicotine addiction degree, etc.), in addition to the specific way of smoking (puff frequency, intensity, volume, and duration at different stages of cigarette consumption, as well as breath hold) [46]. Thus, different smokers/vapers generated different shapes of exhaled aerosol, but this is a common limitation of all experiments performed by humans. Despite these interesting results, further research is required to expand our findings. In particular, a larger sample is required for external validation. In addition, other dimensions of toxicity should be appraised (e.g., aromatic cyclic compounds or heavy metals). Similarly, mitigating factors will need careful appraisal, such as ambient volume, impact of aeration, and so forth. Finally, the association between passive and active smoking effects remains the focus of intense research [47], especially when considering also other established and cardiovascular risk factors [48].
An important issue of our work is the applicability of our findings on Iqos to other HNBC, and, similarly, on Juul to other EVC. In fact, other HNBC devices are available in selected markets, with underlying mechanisms to heat tobacco that differ, at least in part, from Glo and Iqos. Given the evident differences between Glo and Iqos hereby described, we can expect that PM emissions may be specific to each HNBC type, and even its flavors. Similar arguments may apply to EVC. Indeed, we chose Juul for this trial because, on top of being an established market player, it is characterized by default standardization of aerosol emissions. Many EVC can, however, let users customize exposure, in terms of both solution components, concentration, and volume. Accordingly, we cannot safely recommend extrapolating our results on Juul towards other EVC. Accordingly, most likely some EVC may prove significantly more harmful than Juul in terms of toxic emissions [49,50].
This work has several drawbacks, which range from the small sample size to the focus on only three MRP. In addition, apparently healthy volunteers participated in the trial, and thus, our results cannot be considered immediately applicable to patients with cardiopulmonary disease. Notably, differences in indoor polluting effects, while often statistically significant, were of modest magnitude betweenand within-MRP, especially when compared to the much greater impact of TCC, and thus, their clinical impact remains to be determined [51]. In addition, we focused our measurements only on PM, but the potential scope of toxic agents released by MRP is very wide, and thus, additional studies with a multidimensional measurement scope are required. Moreover, in vitro experimental studies should be performed to assess whether the observed indoor PM concentrations could increase inflammation and oxidative stress. Furthermore, it is evident that Juul mostly emits liquid droplets, whereas PM produced by HNBC consists of largely solid material. Finally, different analytical approaches can yield different comparative and inferential estimates, given the clear impact of focusing only on smoking sessions vs. more comprehensive pre-, during-and post-smoking sessions, as clearly showed by the differences between the present work and Protano et al. [24]. Yet, it remains undisputed that MRP cannot be considered equal in terms of PM indoor emissions, in terms of both between-and within-MRP comparisons.

Conclusions
Leading MRP such as Glo, Iqos, and Juul have significantly less intense and persistent effects on indoor pollution in comparison to TCC. Yet, when focusing solely on MRP, between-product and between-flavor differences appear, with quantitative estimates suggesting lower polluting effects with Iqos. These results, if confirmed externally, could be used to individualize product and flavor choice to minimize the untoward effects of EVC and HNBC.
Supplementary Materials: The following are available online at http://www.mdpi.com/1660-4601/17/17/6029/s1, Figure S1. Example of measurements obtained during a single-modified risk product subtype smoking session by an individual smoker, totalling 262 values. Left panel highlights measurements before smoking, middle panel measurements during smoking, and right panel measurements after smoking, for a total time of 13 min. In the present case, a Glo Aegean was used. PM 10 = PM with diameter ≤ 10 µm; PM 4 = PM with diameter ≤ 4 µm; PM 2.5 = PM with diameter ≤ 2.5 µm; PM 1 = PM with diameter ≤ 1 µm., Table S1. Levels of particulate matter (PM), expressed as µg/m3, comparing different smokers.