Critical Appraisal of Exposure Studies of E-Cigarette Aerosol Generated by High-Powered Devices*

We had to follow an atypical process for searching the literature, since we aimed at searching for papers whose experiments relied on specific aerosol generation conditions (explained in Section 2), a background material that is only mentioned and/or described in the materials and methods section of each article. Google Scholar appeared to be the most appropriate search engine for this purpose, since it can look for specific words inside the articles. However, filtering the search output must be done by hand because Google Scholar lacks an inbuilt capacity to filter the information. To illustrate our approach, we provide below a flow chart of a step-by-step description of the criteria applied to filter the articles. The search query we used as input (Google Scholar 11/29/2023) was “(electronic cigarette(s) OR electronic-cigarette(s) OR e-cigarette(s) OR e-cigarettes OR vaping OR “Electronic Nicotine Delivery System”) AND “SCIREQ”. The results were extracted and after suppression of duplicates the list was finally composed of 263 references.

• The first step was to remove abstracts of conferences, reviews, books, thesis or other reports.

• The next step was identifying (by title) and removing them from the list articles not dealing with tobacco products (heated or combusted), since the Google Search does not restrict the titles, and many selected articles failed this search criterion even if the term “electronic cigarette” was mentioned in their introduction

• We removed from the list articles that did not contain keywords linked to vaping.

• At this point in the selection all papers mentioned the SCIREQ equipment, but it was not evident that they used its exposure chamber or the aerosol generator. We had to look manually in each article to find the specific equipment or instrument that was used, removing those articles that did not use the SCIREQ manufactured aerosol generation equipment.

• Once completing the above-mentioned selection, the final step was to select for more specific conditions of aerosol generation, since the SCIREQ equipment has been used with several EC devices that are outside the scope of our review, such as Juul® or Blu® pods.

However, as previously explained, aerosol generation conditions are often ill described, with authors often providing incomplete and/or missing information (18, 19). Since we are searching for experiments conducted with the 0.15 Ω coil at high powers, we extracted articles that contained basic descriptive terms of the device, which at least would allow for a suspicion of our targeted usage, such as “third generation device” or “ECX E-Vic VTC”, the name of the device marketed by SCIREQ. At this step, we also kept articles containing the term “device” but providing no further information.

Additionally, the Subtank uses several coils that are available, and we also assumed that not all the testing conditions should be under an Overheating Regime.

So we applied a filter based on the electric resistance, selecting articles mentioning the 0.15 Ω or 0.2 Ω coils. We suspect that articles reporting a 0.2 Ω coil were really using a 0.15 Ω coil rounded to 0.2 Ω, since SCIREQ mainly provides three coils with resistances: 0.15 Ω, 0.5 Ω and 1.5 Ω (12, 13).

The last step in filtering the literature is linked to the power supplied. In most of the articles, this fundamental information was missing. Figure 7 displays the PRISMA chart associated with the literature search we have described.

Figure 7.

Prisma chart associated with our literature search (see Section 3).

The resulting 41 papers that met the search criteria previously described are classified in Table 1. Aerosol generation parameters of the 41 revised studies using the InExpose. Power tested as declared by authors allegedly from display in the instrument screen of the box mod.

Six articles used nicotine concentrations above 30 mg/mL, 14 studies between 20 mg/mL and 30 mg/mL, 9 studies between 10 mg/mL and 20 mg/mL, 12 studies between 3 mg/mL and 10 mg/mL, while 11 studies did not use nicotine.

In 20 articles nicotine concentrations above the Tobacco Product Directive requirement of 20 mg/mL were reported. We have a high level of confidence in our ability to identify aerosol generation conditions in experimental aerosol generation procedures that are potentially associated with overheating, as described in Section 2.

• “Certain” means that the device, coil and power or voltage were fully identified.

• “Almost certain” means that one item of information was not clearly described or was missing, but the available information is sufficient to evaluate aerosol generation.

• “Suspicious” means incomplete minimal information that is insufficient to fully assess aerosol generation conditions. However, based on the significant similarity in experimental design and the materials and methods between these studies and the fact that operating the InExpose requires training, all of which makes it likely that the authors followed the aerosol generation procedures of the “almost certain” and “certain” categories stated above.

• “Uncertain” means that information is so restricted that it prevents the discussion of aerosol generation conditions. In all cases the usage of InExpose or the term “SCIREQ” is mentioned, but there is no information on supplied power, and the device is at best partially identified. As argued in the previous point, it is very likely that complete information might have led to the same operating conditions as studies with more information.

• “Unknown” means a complete lack of information on aerosol generation parameters (either the device, the coil and supplied power, only the SCIREQ equipment is mentioned). In this case, aerosol generation conditions are unknown, which renders the studies completely irreroducible.

In the 41 studies listed in Table 1, we can immediately identify the first top-down 14 studies (“Certain”, “Almost Certain” and “Suspicious”) that provided at least basic information on their aerosol generation procedures. The lack of information on one parameter (for example power or coil resistance) can be reasonably inferred by the remaining parameters and the constraints from the usage of the InExpose (also from the similarity of the studies). Of the remaining 27 studies, the 16 (“Uncertain”) provided even less information, but their mention of terms such as “E-Vic”, or “0.15 Ω coil”, or “KangerTech” is sufficient to infer that they followed similar aerosol methodology as the first 14 studies, which is consistent with the usage of the InExpose, as explained in the instruction video in (13). The remaining 10 studies (marked “Unknown”) lack even minimal information and thus inference on their methodology becomes much more difficult. Consequently, these studies are completely non-reproducible (a serious methodological flaw).

The authors (35) considered three coil resistances (0.15, 0.5 and 1.5 Ω) and three nominal voltages (2.8, 3.8 and 4. V) in the quantification of nicotine and the main aldehydes (formaldehyde, acetaldehyde, acrolein). They reported a significant increase in aldehyde yields while puffing the 0.15 Ω coil with various voltages. Airflow rate was 1.1 L/min. Aldehyde yields were low for 2.8 V (comparable to those obtained with the 1.5 Ω coil). However, rising the voltage to 3.8 V and then to 4.8 V in the 0.15 Ω coil produced an abrupt increase of these yields, reaching respectively 136, 273, 232 times the yields measured with respect to the 1.5 Ω coil using the Butterflavour e-liquid.

The large increase of aldehyde yields for the voltages 3.8 V and 4.8 V and the 0.15 Ω coil bears evidence of an aerosol generated under an Overheating Regime. This fact follows from the results of our voltage calibration test displayed in Figures 4 (with Figures 2 and 3 as reference), showing that 3.8 V and 4.8 V correspond to values of measured power of 38 W and 44 W, which are clearly above the upper end (above 30 W) of the Optimal Regime (Figure 4), well into the Overheating Regime for 1.1 L/min airflow. As a contrast, measured supplied power is in the Optimal Regime (below 30 W) for 2.8 V with 0.15 Ω and for all tested voltages with the 1.5 Ω coil.

Figure 4.

Calibration curve of the E-Vic VTC modes using a 0.20 Ω coil (left y-axis) voltage calibration and (right y-axis) resulting wattage supplied in wattage mode. Gray boxes mark the voltages (4.1 V and 4.2 V) reported by three studies listed in Table 1.

Another problematic issue with NoËl et al. is the excessively high nicotine concentration of 36 mg/mL in e-liquids puffed with the high-powered E-Vic Mini box (see Table 1). We comment on this issue in Section 5.2. The authors justify this as an effort to “mimic nicotine exposure by heavy smokers”, but this combination of high nicotine with high power strongly misrepresents consumer usage in which high nicotine levels are used with low powered devices. This fact emerges from observations carried out by Soule et al. (67), who found a strong correlation between high-powered devices (even at 30 W) and low nicotine concentrations (below 6 mg/mL). Besides this point, such high nicotine levels also produce an unrealistic overexposure to nicotine, in in vitro and in vivo systems. This problem is shared by other studies that we review further ahead (29, 31).

The study by Cahill et al. (31) likely used the same cinnamon fireball flavoured e-liquid (reported as 50/50 PG/VG). The authors used the same device and puffing protocols as (35), but supplied only 4.2 V, which corresponds to 40 W–45 W measured power (see Figure 4). Therefore, Cahill et al. also generated aerosol under certain overheating conditions. The authors report yields of 0.021 μg and 0.386 μg per puff, respectively for formaldehyde and acetaldehyde. Although these levels are lower than those of (35), they signal very likely overheating conditions, since formaldehyde mass yield is always higher than that of acetaldehyde under optimal conditions.

Theoretically, the dehydration reaction leads to the formation of formaldehyde and to vinyl alcohol, which itself rearranges to into acetaldehyde (68, 69). This explains why acetaldehyde levels cannot be higher than formaldehyde levels under optimal conditions (a situation mainly observed in the literature). However, acetaldehyde is also released by the pyrolysis of the cellulose (70) suggesting that the reported extra acetaldehyde was released under overheating conditions, which led to the pyrolysis of the wick.

The study exhibits other minor irregularities: the authors claim that 0.165 mg/puff of e-liquid was vaporized but the “mg” unit they used is almost certainly erroneous, since even a low power device like the Juul vaporizes at least 1 mg of e-liquid per puff (71). The e-liquids had 3.3% nicotine concentration, which is also unrealistic for usage in a high-powered device nicotine concentration (67) as commented before on NoËl et al. (35).

The authors (34) examined the impact of e-liquid composition in testing in vitro systems, with the same aerosol generation parameters as Cahill et al.: the 0.15 Ω coil with 4.2 V (40 W measured, see Figures 2, 3 and 4) and 1 L/min airflow.

As a preliminary step of their investigation, the authors generated aerosols from non-commercial e-liquids only composed of PG and VG. The study quantified around 7 μg of formaldehyde (333 times higher than (31)) and above 2 μg of acetaldehyde (5 times higher than (30)). When commenting on these high yields the authors claim that the aerosols “were not produced under ‘dry puff’ conditions”, a statement that without further explanation must be regarded as merely an assumption, probably based on e-liquid not being depleted. However, having used the same experimental set up as (35) and having obtained similar aldehyde yields, clearly points to overheating conditions, which can occur without e-liquid depletion.

The authors then used commercial e-liquids with their most common PG/VG ratio and two nicotine concentrations: a 70/30 strawberry flavoured e-liquid, a 50/50 Catalan cream flavoured e-liquid and a 30/70 vanilla flavoured e-liquid with 12 or 18 mg/mL of nicotine concentration. Of the 6 tested conditions, two (Vanilla 12 and Catalan cream 18) produced significant quantities of formaldehyde (above 3 μg/puff). Two others produced significant quantities of acetaldehyde, higher than formaldehyde (strawberry 12 and strawberry 18) and the two last combinations (Vanilla 18 and Catalan cream with 12 mg/mL) are too low for comparison.

Finally, the authors also tested pod devices. While pods are out of the scope of our present review, the comparison between fourth generation and third generation devices is interesting, as the tested e-liquids have close solvent composition (at least the solvents themselves are the same). Removing the Puff bar OMG 5% nicotine salt which is suspicious regarding the other results, a rough approximation leads to 0.01 μg of formaldehyde released for 2 mg of e-liquid vaporized (MEV) per puff (5ng of formaldehyde per mg of e-liquid vaporized). Without providing the MEV that results from the non-commercial e-liquid, the ratio of yields cannot be investigated. However, reaching 7 μg of formaldehyde by puff suggests that 1400 mg of e-liquid per puff was vaporized (assuming the same ratio and indirectly the same condition). This enormous ratio suggests possible conditions significantly hotter than usual for the tested pods. Since vaporization occurs under the same conditions in high-powered devices and pods under an optimal vaping regime, this suggests a high likelihood that overheated aerosol was also generated by the fourth-generation pods.

The authors (29) also used an airflow rate of 1.1 L/min and the Joyetech E-Vic Mini VTC mod with 0.15 Ω coil with 3 V. No justification was provided for setting this voltage, which would correspond to 60 W from Ohm’s law, though our calibration tests (Figure 2, Figure 3 and Figure 4 in Section 2) determine that it corresponds to measured 30 W (the upper limit of the Optimal Regime). The tank was filled with a Crème-Brûlée flavoured e-liquid (30/70 PG/VG and 30 mg/mL of nicotine) or a Menthol flavoured e-liquid (40/60 PG/VG and 36 mg/mL of nicotine). Cells were exposed to both flavoured e-liquids during 1h on one day but an additional experiment was done with Crème Brûlée where the protocol was replicated during three consecutive days.

Both flavoured aerosols were generated and chemically characterized. Nicotine yields were around0.028 μg/puff with the Menthol flavoured e-liquid (40PG/60VG) and 0.026 μg/puff for the Crème-Brûlée one (30PG/70VG). Formaldehyde, acetaldehyde and acrolein were quantified respectively at around 2.6 μg/puff, 1 μg/puff and 0.02 μg/puff for the menthol e-liquid and very high levels of 13 μg/puff, 4 μg/puff and 9 μg/puff of acrolein for the Crème-Brûlée flavoured e-liquid.

High levels of aldehydes are not expected at 30 W (calibrated power from 3 V), a power level marking the onset of the overheating region (see Figure 5 and (14)). Nicotine levels were also excessive for a sub-ohm device.

The authors (27) used the 0.15 Ω coil at 220 °C or 80 W, which also corresponds to overheating conditions characterized by measured values of 45 W and close to 300 °C (Figures 2, 3 and 4). They tested two commercial e-liquids containing menthol or tobacco flavour with 0 or 6 mg/mL of nicotine and one laboratory liquid with 1:1 PG/VG ratio. Interestingly, each flavoured liquid was chemically analyzed in both liquid and vapor phase. The resulting lists were compared for each phase. Of the 98 compounds found in the tobacco liquid, 33 were also available in the 91 ones detected in the menthol liquid. However, in the aerosol, 82 compounds are common and only 2 and 7 are specific to the tobacco and menthol liquids, respectively.

The authors did not examine the comparison between the e-liquid and the aerosol. The tobacco flavour had only one common compound (ethane, 1,1,1-trichloro-) in both analyses, whereas in the menthol case there are four (ethane, 1,1,1-trichloro-; α and β-pinene; acetaldehyde). The high quantified levels of some compounds (ethanol) in the aerosol but not in the liquid suggests problems in the e-liquid analysis, as well as the formation of byproducts during the vaporization, such as acrolein resulting from glycerol degradation. It should be noted that the absence of formaldehyde and acetaldehyde, with the quantification of the propionaldehyde, raises questions on the methodology used to sample or to analyse the aerosol.

Of the 82 molecules found in the aerosol from both liquids, 35 contain chlorine atoms, 4 contain bromine atoms (1,2-dibromoethane, bromofluorobenzene, bromomethane and bromoform) and 1 contains sulfur (carbon disulfide). The list of compounds in both liquids does not show the presence of Br and S, while chlorine (Cl) is only present in one molecule (ethane, 1,1,1-trichloro-). Based on the points argued before, the quantification of chlorine molecules raise concern on the possible availability of sucralose in the e-liquids (a molecule that is difficult to measure using Gas Chromatography Mass Spectrometry (GC-MS) due to its low volatility). This sweetener is documented for its pyrolysis and the releasing of chlorinated byproducts and for enhancing overheating conditions due to its caramelization on the coil. The lack of information on the specific commercial e-liquids that were analyzed prevents a deeper investigation on these inconsistencies.

Finally, the authors used passive sensors to evaluate the air quality in the mouse exposure chamber, measuring carbon monoxide (CO) and Volatile Organic Compounds (VOCs). Carbon monoxide was quantified in the five experiments from 0.5 ppm, using e-liquid containing menthol and nicotine, to 33 ppm only using PG/VG mixture. These average values limit the interpretation of the measurements. Indeed, the concentrations are real-time values, and the curves would have probably shown abrupt change in some of these concen-211 trations (assumption made based on the deviations equivalent to the average values in CO). Besides the presence of these quantified values, the presence of CO is a clear indication of cotton pyrolysis and supports the conclusion that the aerosol was generated under overheating conditions.

While the 41 selected studies provide detailed information on the characteristics of cell lines and rodents and on the methodology of biological and toxicological procedures, most of the studies failed to provide sufficient information on their aerosol generating procedures. As discussed in Sections 2 and 3 and depicted in Table 1, only 9 studies provide sufficient information to analyze their aerosol generation with full or almost full certainty (the ones denoted as “certain” and “almost certain”). Six studies (the “suspicious”) provided minimal sufficient information. The remaining 27 articles failed in providing this information at different levels, with 10 of them providing no information besides having used the InExpose.

Since these studies are focused on the effects on biological systems from aerosol exposure, the generation of the aerosols constitutes an important technical issue that must be fully described in their Materials and Methods section. The vacuum of information on aerosol generation procedures renders these studies unreproducible and/or impossible to replicate, a serious methodological flaw in experimental research that certainly hinders the quality of these studies, irrespective of the possible impeccable quality of the biological procedures. Evidently, we cannot claim full certainty in our evaluation of aerosol generation in the 27 studies that provided insufficient information. Hence, we only reviewed (Section 4) in detail 5 studies that provided this information in full, also reporting aldehyde yields and nicotine concentrations. However, given the thematic similarities of all the studies listed in Table 1 and the common usage of the InExpose that requires training and a learning curve to operate, we can claim with high likelihood that the 16 studies that reported the 0.15 Ω coil and usage of SCIREQ equipment followed the methodology of the 14 studies that did provide at least a minimally acceptable degree of this information (which includes the 5 studies reviewed individually). Besides the 10 irreproducible studies with practically no information, in at least 31 of the 41 selected studies we have sufficient direct and indirect evaluation elements to assert that cell lines and rodents were exposed to overheated and aldehyde loaded aerosols, generated by the combination of high power with low airflow and low resistance, a combination that is also at odds with consumer usage of these devices for the ‘direct to lung’ inhalation.

Three studies (29–31) generated aerosol from e-liquids with high nicotine concentrations (30, 36 and 50 mg/mL). The authors justify this as an effort to “mimic nicotine exposure by heavy smokers”, but this combination of high nicotine with high power strongly misrepresents consumer usage in which high nicotine levels are used with low powered devices. This fact emerges from observations carried out by Soule et al. (67), who found a strong correlation between high-powered devices (even at 30 W) and low nicotine concentrations below 6 mg/mL. Besides this point, such high nicotine levels also produce an unrealistic overexposure to nicotine in in vitro and in vivo systems.

The high levels of quantified aldehydes found in (29, 34, 35) supports the main criticism of the present literature review: overheating accompanied with an exponential production of aldehydes occurs when puffing a high-powered device like the E-Vic Mini with a 0.15 Ω coil with an airflow of 1.1 L/min (see Figure 6). The authors of (29, 34, 35) reported yields that (at least) signal early stages of the overheating conditions (masses above 1 or 2 μg per puff), with higher levels (7–8 μg per puff) denoting more advanced stages of overheating that can be associated with the onset of highly energetic pyrolysis close to ignition, a highly likely development supported by levels of acetaldehyde comparable or higher than formaldehyde. To highlight overheating conditions, it is useful to compare these high aldehyde yields with the formaldehyde yields from the Juul in (29). The Juul released 5 ng of formaldehyde per mg of e-liquid vaporized, hence the same ratio of mass per puff means that 7 μg of formaldehyde would have to be generated by 1.400 mg of e-liquid by puff (assuming indirectly the same condition), an enormous e-liquid quantity that questions the conditions used in these studies to generate the aerosol. Table 2 summarizes the aldehydes results graphically extracted from each reference.

The quantification of CO by Muthumalage et al. (27) is an important outcome supporting the main criticism presented in this review. Also, they systematically quantified CO inside the exposure chamber and the maximal level reached 66 times higher than those of the lowest experiment. To assess the toxicological relevance of these outcomes on biological systems, we remark that they constitute a clear signal of advanced overheating through energetic pyrolysis in the onset of oxidizing processes acting on the cotton element of the wick (and its subsequent structures) (74). The presence of CO also means that other cotton byproducts must have been also released, thus polluting the aerosol generated and increasing its toxicity. The literature on cotton pyrolysis shows that during various stages of pyrolysis the cellulose (in the wick) releases highly toxic furans and polycyclic aromatic hydrocarbons (PAHs). In fact, the results of previously published articles that looked at CO in vaping aerosols highlight a high-correlation with excessive supplied power with an insufficient airflow (i.e. conditions above the power ranges of the Optimal Regime) (72, 73, 75).

The EC aerosol puffed by the box of the E-Vic Mini is transported by a pump at an airflow of 2 L/min to the exposure chamber through a tubular conduct. It is well known from filtration and sampling in aerosol physics that this transport involves, besides air dilution of the generated aerosol, modifications of its particle/gas partition, mostly affecting the particulates (liquid droplets) with larger diameters that either impact or settle on the conduct walls or (depending on environmental variables) might coagulate, condense or evaporate. However, this air dilution and aerosol physics phenomena will not remove nor decrease the toxicity levels of an aerosol generated under overheating conditions. An empiric proof of this is furnished by Muthumalage and Rahman (27), who found high levels of CO and VOCs in their chemical analysis of the aerosol in the exposure chamber (after its passage through the conducts).

Considering the results of our own laboratory tests summarized in Section 2 (published in (14)) and our review of 5 individual studies in Section 4, it is clear that overheating conditions increases the toxicity of the generated aerosol as specified by high levels of aldehyde yields and signals of energetic pyrolysis of the organic wick material (specially CO). Therefore, it is not surprising that deleterious effects in cellular physiology were detected in cells and animal tissues resulting (at least in part) from the molecular presence of these toxic byproducts.

The study by Muthumalage and Rahman (27) that quantified concentrations of CO and VOCs of similar magnitude provides an illustrative example of how inappropriate laboratory conditions that lead to overheating can induce misleading physiological effects. As recounted in their Figure 9, summarizing the altered proteins in lung tissue, these authors report that a pure solvent PGVG mixture leads to the highest number of altered proteins (around 240 proteins). Adding tobacco flavour leads to 156 alterations with 88 in common with those of the PGVG mixture. Adding nicotine leads to 118 alterations with 52 in common with both PGVG only and tobacco flavour. PGVG exposure has the highest overheating condition observable with the maximal CO followed by the tobacco + nicotine and the tobacco experiments.

Applying the same reasoning to menthol liquids leads to 174 alterations and 107 alterations without and with nicotine where 49 and 51 are in common with PGVG alterations. The reductions in common alterations with PGVG compared to tobacco experiments can be due to lower overheating conditions, which is observable by the CO concentrations closer to 0 ppm in menthol experiments whereas it is above 15 ppm in tobacco experiments. Despite it can be a coincidence, the absence of a PGVG testing under normal conditions means that the contribution of abnormal testing and flavouring agent or nicotine adding in the alteration process cannot be investigated.

The JoyeTech® E-Vic Mini was released in 2015 as one of the early devices allowing for a temperature-controlled mode. Currently, this device is difficult to find and its usage is marginal. This follows from the evolution of consumer patterns. Between 2015 and 2019 usage of third generation tank devices became very popular, including devices operating at high power (> 30 W) with sub-ohm resistances (< 1 Ω).The demographic study by Jiang et al. (76) shows the market evolution of third generation devices in the USA between 2015 and 2019. These devices are subdivided in “tanks” and “mods”, with “mods” describing the bulky devices that are used at high power settings. While tanks became very popular, reaching in 2017 over 70% of preferences in all ages, their usage declined to 30–35% in 2019, “mods” have remained low in preference for all ages, decreasing in the period 2015 to 2019 to 6.3% (youth) and 9.5% (young adults). It is very likely that these percentages of “mod” usage are currently even smaller with usage of low powered devices, including cartridge-based pods and disposables, becoming dominant. Preference of low powered devices is even more marked among adolescents and young adults (77, 78).

Given these developments in consumer preferences, it becomes hard to justify the usage of a device like the E-Vic Mini in current preclinical studies that aim to assess the toxicological profile of generic EC aerosols. At best, if such toxicological assessment is conducted under appropriate laboratory conditions, it will be only applicable to a reduced minority of consumers (which can still be useful).At worse, an assessment based on inappropriate laboratory conditions (high power with insufficient airflow) unrepresentative of consumer usage is necessarily misleading and has little utility to all consumers and stakeholders.

Usage in the laboratory of any arbitrary combination of e-liquids, coils and devices does not necessarily qualify as testing under “normal” conditions simply because all these combinations are commercially available. The case in point that we have stressed in the present review is laboratory usage of high-powered devices (for example operating well above 30 W). This testing cannot be justified as normal usage under any arbitrary airflow or nicotine levels simply because devices and e-liquids with these characteristics are available to consumers in the vaping market (by the same token, driving at 160 km/h or faster cannot be regarded as generic “normal driving”, simply because such high speeds are accessible in many commercial automobiles). For laboratory testing of high-powered sub-ohm devices to be objective and useful, authors must acknowledge their representative and overwhelming majority usage with low nicotine concentrations and deep inhalation (i.e., large airflow) needed to cool and condense efficiently the air diluted aerosol (up to 50 mg/puff) in large puff volumes (~ 500 mL) produced by such devices (whose design with wide mouthpieces to minimize air resistance is consistent with this usage).

It would be very useful and would enhance the quality of pre-clinical research to pay more attention to consumer behavior. Consumer patterns are not only reported in consumer forums and magazines (79) and in peer reviewed publications sampling social network films (80), but are described by manuals elaborated by manufacturers and they are also reported in published peer reviewed literature (16, 81, 82), including observational studies (67). The different ways different devices are used can also be understood as a result of the trial and error self-training guided by sensorial effects that vary from user to user, but practically all users conduct this self-training when they begin vaping. A naïve user may try inhaling a powerful device as if puffing a Juul and will receive a hot and unpleasant aerosol, since the low airflow appropriate for a Juul is insufficient to evacuate and cool the large amount of vaporized e-liquid. But users learn and adapt.

When reporting high aldehyde yields, Nöel and Ghosh (34) comment that the aerosols “were not produced under ‘dry puff’ conditions” presumably because the e-liquid in the InExpose tank was not depleted. Therefore, these authors mistakenly assumed that the aldehyde yields occurred under normal usage conditions without a dry puff. This comment reveals misunderstanding in the assumption that the repellent “dry puff” phenomenon is only tied to a single event marked by the depletion of e-liquid in the atomizer, thus normal usage occurs as long as e-liquid is not depleted. The same misconception was expressed by Beard et al. (83) who also puffed a high-powered device with insufficient airflow to examine cytotoxicity from “dry hits”. These authors classified as normal “standard vaping” all usage without e-liquid depletion. These assessments by Nöel and Ghosh (34) and Beard et al. (83) fail to understand that e-liquid depletion only marks the endpoint of an overheating process characterized by critical thermal phenomena such as film boiling, when the coil temperature goes above the boiling point of the liquid mixture (24). This endpoint is the advanced manifestation of overheating and is unequivocally accompanied by energetic pyrolysis of the organic material of the wick and rapid increase of thermal degradation byproducts.

Beard et al. (83) argued that abnormal vaping conditions should also be studied. We fully agree, abnormal and critical conditions are researched in many issues, for example simulating automobile accidents at high velocities in order to improve safety measures. However, this type of simulations or experiments should not convey (not even hint) the notion that the tests somehow describe “standard”driving conditions. The same applies to cytotoxicity tests on dry puff conditions. High toxicity levels from overheating conditions need to be investigated to improve our knowledge on this phenomenon and to achieve a comprehensive perspective of vaping products. However, studying these conditions is misleading without an explicit acknowledgement that they are abnormal and deviate from normal consumer usage.

Just as automobile drivers recognize an unpleasant risk of excessively high velocities, there is solid peer reviewed published observational evidence that users perceive sensorially the onset of overheating well before e-liquid depletion. The organoleptic experiments by Visser et al. (25) proved that increasing pyrolysis is perceived by users in gradual progressive stages, increasing from 0 (no perception) to 1 (full dry hit), a perception that was correlated to results of laboratory testing (even with filled tank). Another clear signal of overheating of aerosol generated by the InExpose is in the introductory video in NoËl Table (13), showing the initially colorless e-liquid before the experiment became brownish at the end. This brownish color is a clear signal of compounds produced by cotton pyrolysis (84).

The overwhelming majority of users of ECs (vapers) are currently still adult smokers replacing cigarettes with the much safer nicotine deliver through the EC aerosol not generated by combustion. Therefore, the comparison with cigarette smoke is still highly desirable for toxicological and preclinicalstudies to best contribute to advance public health goals.

A low priority in comparison with tobacco smoke can be based on the broad scientific consensus already existent that supports harm reduction for adult smokers who switch to vaping. An additional argument supporting this low priority follows from the surge of vaping among teenagers and young adults who have never smoked (or have smoked very infrequently). An assessment of these arguments is beyond the scope of the present review, but even if taking them at face value, given the overwhelming preference of adolescents and young adults for low powered devices (77, 78), there is no justification or public health benefit in assessing EC aerosol toxicity under inappropriate laboratory conditions and/or testing high-powered devices whose usage has become marginal.

While the authors can argue that using outdated devices in pre-clinical studies is inevitable, since these studies are expensive and require a long time frame, from assuring sufficient grant funding, to preparing the materials and analyzing the outcomes. However, this claim is not consistent with the fact that the consumers shift to low powered devices was already in full swing when most of the studies we reviewed were published: only 9 were published between 2016 and 2020, 6 in 2021, while 20 (50% of the studies) were published in 2022 and 2023 (10 each year). Therefore, this delay of publication really applies only (at most) to 15 studies published before 2022 (and to only one of the 5 studies reviewed individually).

The present review is evidently limited by the various degrees of insufficiency in providing information on aerosol generating procedures in 27 of the 41 selected studies. As argued in Section 5.2, this vacuum of information leads to various degrees of uncertainty in assessing the generation of EC aerosols under overheating conditions in these 27 studies. However, in 17 of these 27 studies there is sufficient indirect evidence to infer these conditions. Another limitation (common to all research on ECs) is the difficulty of considering the full effect of the wide diversity and of individual usage patterns of the devices, further complicated by the rapid changes in vaping technology and of the regulatory landscape, as well as the capacity of users to modify behavior to adapt to all these developments. Nevertheless, in dealing with complexity there is no better option than to resort to the best assessment from general tendencies supported by observation and by solid theoretical and experimental facts.