"Third Hand" vaping in animal models

Summary of a useless research

Background

I have discussed the possible existence of “third hand” vaping (and “third hand” smoking) in previous posts (here and here).

The conclusion from these posts is straightforward: under the usual environmental conditions in which vaping takes place, there is no indoor aging process in environmental vaping aerosol (EVA) analogous to the indoor aging process of “second hand” smoke (SHS) (which I will denote as environmental tobacco smoke ETS) that forms the physical and chemical phenomena known as “third hand” smoke (THS).

I have criticized in Pub Peer comments (here and here) studies that try to simulate “third hand” vaping in animal models (mice). The narrative of these studies is very hostile towards vaping, a narrative expressed exuding overconfidence (as usual in these studies). Authors claim that “third hand vaping” exists and that their mouse studies prove it is as harmful as THS (produces various deleterious physiological effects).

While the authors of all these studies conducted virtuous and correct biological analysis (as far as I can tell), their exposure protocols are extremely unrealistic, overexposing mice to toxic aerosols that would even be harmful for humans (more so for 25-30 gram animals). This is a serious flaw in studies probing the effects on mice from exposure to aerosols.

In this post I will criticize this study:

• Umphres, S. S., Alarabi, A. B., Ali, H. E., Qadri, S., Khasawneh, F. T., & Alshbool, F. Z. (2026). “The Prothrombotic Phenotype of Thirdhand Electronic Cigarette Exposure is Sex Independent and Involves Systemic Mediated Effects on Platelet Function: Evidence from a Mouse Model.” Cardiovascular Toxicology, 26(3), 30
  https://doi.org/10.1007/s12012-026-10093-z.

Summary

The authors examine the exposure of laboratory mice to (what they call) a “third hand e-cigarette” (THEC) aerosol, a phenomenon they assume to be factual and generic of vaping aerosols, as an accurate analogy to the way third hand smoke (THS) relates to tobacco smoke. There are several serious shortcomings in this study (as people can verify in my two previous posts: here and here):

The authors’ assumption that “THEC” is a good analogue of THS is mistaken. THS forms from residues of aged ETS, but this is possible because

1. ETS contains abundant nicotine together with many other reactive gas phase volatile and semi-volatile organic compounds (VOCs and SVOCs) and

2. ETS remains airborne for long times.

The environmental vaping aerosol (EVA) (the “second hand” of vaping) does not comply with these properties (under normal environmental conditions):

1. nicotine is the only abundant reactive SVOC in EVA

2. EVA has a short timed environmental presence, it is volatile and diluted, so it evaporates rapidly.

Nicotine adsorption in indoor surfaces is the only THS phenomenon that can be reproduced in vaping, but it occurs with much less intensity and efficiency than in smoking.

1. Nicotine surface densities from THS are significantly measurable in normal environments (homes, cars, hospitality venues)

2. Nicotine surface densities from vaping are only significantly measurable under abnormal conditions: aerosols confined to small laboratory chambers or in indoor environments where many vapers vape simultaneously (vape shops, vape fests),

The authors claim that their protocol simulates “real world exposure conditions” in vaping. This assumption is mistaken. To generate “THEC”, household samples in their study were exposed to an excessive mass of a toxin laden aerosol from 2000 puffs (400 daily) generated by a high powered vaping device. This massive overexposure of samples renders the authors’ study completely unrealistic and irrelevant.

Animal models of THS (third hand smoke) exposure

The authors apply to e-cigarette emissions an animal study protocol previously applied by Martins-Green [1,2] to study the exposure of mice to artificially generated THS (third hand smoke). The methodology followed by [1] is illustrated by figure 8 of Jacob III et all [2], which we reproduce below:

Protocol of animal model simulation of THS.

The exposure procedure is explained by [1,2]:

• “common household fabrics (curtain material, upholstery and carpets) are placed in empty mouse cages and subjected to ETS (second hand smoke SHS) exposure”.

A simulation of ETS was machine generated from the smoke of two packs of 3R4F research cigarettes smoked each day, 5 days/week. We quote from [1]

• “The smoke was routed to a mixing compartment and distributed between two exposure chambers, each containing 4 cages with the materials”. I quote below [1,2]:

• “The materials were always exposed to the same level of ETS by adjusting the machine to deliver the same total particulate matter (TPM) to the chambers containing the cages with the materials.

• The levels of TPM were adjusted to fall within those found by the EPA in the homes of smokers (in the homes of smokers, 15−35 μg/m^3; in our machine, 30± 5 μg/m^3​​)”

• “At the end of each week, cages were removed from the exposure chamber, bagged, and transported to the vivarium where mice were placed into the cages. For the next week, an identical set of cages and fabric was then prepared and exposed to smoke in the same way as that described above.

• Using two sets of cages and material, each of which was exposed on alternate weeks, ensured that mice always inhabited cages containing fabric that had been aged and with fresh THS at any given week (Figure 8)“

Notice: how the THS exposure protocol is careful to mimic (as best as possible) realistic exposure conditions of fabrics to ETS.

The authors’ “THEC” model

The authors adapted the THS model of [1,2] to “third hand” vaping that they denote as THEC “third hand e-cigarette”. Their protocol is roughly illustrated in the following graphic abstract (of this study)

Adaptation to vaping

• Samples of household material (3 unspecified fabrics, 2 carpets and 1 upholstery) were hung in racks inside a single inhalation chamber of volume 3081 in^3​​} (0.05 m^3​​).

• Instead of exposing samples to simulated ETS, the authors exposed them to primary e-cigarette aerosol along the following protocol:

Aerosol was generated by a custom e-vapeTM vapor inhalation system from La Jolla Alcohol Research equipped with a TFV8 Big Baby tank, set with 5 V and resistance 0.4 Ω (power 62 W), with menthol favored e-liquid, PG/VG ratio 30/70 and nicotine concentration 18 mg/ml.

The La Jolla mice exposure equipment T1 (it costs $18,250 USD)

Each day during 5 days the samples were exposed to 400 puffs, 3 s duration, 30 s between puffs, puff volume 50 ml and airflow rate 1 L/min. The same 400 daily puffs procedure was applied in subsequent weeks to assure that mice inside their cages would be exposed to a fresh set of materials previously exposed to e-cigarette aerosol in the exposure chamber, with materials re-exposed every week. Male and female mice exposed to THEC and to clean air were housed in cages of volume 458.6 in^3​​ (0.007 m^3​​) for a duration of 4 months.

Differences with the THS protocol

As opposed to the THS protocols in [1,2], the aerosol in the authors’ study was not routed to a mixing compartment but delivered directly to the exposure chamber containing the fabrics.

However, there is a more striking difference: while the THS study in [1] carefully monitored that samples were exposed to an ETS concentration of 30± 5 μg/m^3​​ that matches concentrations measured by the US EPA inside homes of smokers, the authors exposed each one of the samples to a barbaric amount of 2000 puffs (400 daily), without providing any justification for the exposure of this enormous aerosol injection in a small chamber of 0.05 m^3​​.

We cannot use TPM to compare the authors’ mouse study with the THS protocol because we do not know the TPM mass of the aerosol they generated. But since nicotine is common to both ETS and e-cigarette aerosol, a comparison can be made in terms of their nicotine content of the two emissions. We make this comparison further ahead.

The authors’ aerosol (mass & nicotine per puff)

The authors did not measure TPM or nicotine concentrations in the aerosol they generated to expose the samples. However, a very accurate indirect inference can be made from this study by Lalo et al (2020) [3] that precisely used almost the same aerosol generation parameters of the present authors: atomizer, coil, nicotine level, PG/VG mixture and puffing protocol: TFV8 Big Baby tank,resistance 0.4 Ω, 30/70 PG/VG ratio, voltage 5 V, nicotine level 18 mg/ml, 3 s puffs every 30 s, airflow 1 L/min.

The only differences are minor: Lalo et al did not use the battery of the La Jolla Alcohol Research equipment, but a commercial mod and they generated the aerosol at 50 W (while the present authors did it 62.5 W). Lalo et al obtained the following results displayed in their table 2

• Aerosol mass per puff 25.5 mg

• Nicotine mass per puff 230 µg

Aerosol mass per puff was estimated from precision weight after and before each puff, while nicotine was estimated from the cascade impactor and the chemical analysis. These authors also conclude (despite only 59% of aerosol recovery) that almost all nicotine was in the particulate phase, a result consistent with other laboratory studies on similar vaping devices (in particular, Uchiyama et al, 2020 [4]).

The real (unmeasured) numbers of the authors’ mouse model should be slightly higher because they puffed the device at slightly higher power (62.5 vs 50 W), but we take the values at 50 W as conservative estimates to be on the safe side.

The authors’ aerosol: excess aldehydes

The device used by the authors: TFV8 Big Baby tank,resistance 0.4 Ω puffed at 62.5 W is a 3rd generation tank model used for “direct to lung” vaping, which involves deep inhalation (6-10 L/min or more). We have shown in various articles and literature reviews (here, here, here and here) that devices puffed with the combination of high power and low airflow unequivocally generate overheated aerosols with a heavy load of organic byproducts and metals.

Yet the authors puffed a device at 62.5 W with an airflow of 1 L/min that is only appropriate for devices below 20 W. These conditions unequivocally identify an overheating regime. This has practical consequences: a naive user tying to puff a sub-ohm device at 62 W with the small airflow that would be used to puff a Juul would get a horrible load of hot aerosol in a burning mouthpiece. No consumer does this, but vaping machines just keep puffing.

However, besides exposing mice to samples previously exposed to an aerosol that no human consumer would inhale, this aerosol is necessarily loaded with toxins. To estimate how much is the load, I looked at the data from the 2020 study of Uchiyama et al [4] that tested sub-ohm devices at various powers with the same puffing parameters as the authors (3 s puffs, 55 mL puff volume, airflow rate 1 L/min). These were their results for formaldehyde, acetaldehyde, acrolein and nicotine at 62.5 W (in µg per 15 puffs, red curve is gas phase, blue is particulate phase)

Aldehydes and nicotine, the dotted lines correspond to 62.5 W, data from Uchiyama et al 2020

To get quantities per puff we divide the numbers by 15. These quantities must be multiplied times 2000, the number of puffs in the mouse study. The following table summarizes the results from Lalo et al and Uchiyama et al that are reasonable proxies for the aerosol in the authors’ study:

Notice that if Uchiyama et al had injected nicotine in 2000 puffs the total nicotine would be 360 mg, while in the mouse study it is 460 mg. The difference is because Uchiyama et al used e-liquids with 6 mg/mL nicotine while the authors used 18 mg/mL. The lower nicotine level means that the values of aldehydes measured by Uchiyama et al likely underestimate the values that would have been measured in the aerosol of the mouse study. Nevertheless, we can take them as a conservative estimate, though even as approximations these numbers are enormous even in human scale.

An excessive and unrealistic exposure of samples and mice

The quantities listed in the tables above reveal that the samples placed in a 0.05 m^3 chamber were exposed to an enormous overbearing injection of 51.2 grams of aerosol, 0.46 grams of nicotine and 0.26 grams of 3 toxic aldehydes. No wonder deleterious biological effects would be recorded in mice exposed to these aerosol soaked samples. Evidently, the authors’ claim of having simulated “real life” conditions is ludicrous. We discuss the human scale implications further ahead.

Comparison with the THS protocol is also illustrative. The THS protocol used the smoke produced by 2 packs of 3R4F cigarettes which emit 0.6 to 1.79 mg/cigarette of nicotine (for the ISO and CIR, reference here), leading to a total of between 24-71.6 mg of injected nicotine. Therefore, in the authors’ mouse study injected between 6.5 and 10 times the amount of nicotine injected in the THS model.

What happens with the nicotine?

Nicotine (either in ETS or in EVA) adsorbs (i.e. attaches forming thin liquid films) on indoor surfaces. We call nicotine the “adsorbate”, while “adsorbents” are the sample materials on whose surfaces nicotine adsorbs. Nicotine is a very efficient adsorbate. There is experimental evidence [5] that almost 100% of nicotine in ETS injected into a stainless steel chamber becomes adsorbed into the walls after a few days.

In the authors’ experiment nicotine is the main injected adsorbate acting on two adsorbent sinks: one exposed fabric and the chamber walls, presumably made of glass or pyrex. We can ignore the adsorption/desorption of propylene glycol and glycerol that are much less efficient adsorbates than nicotine. We can also assume that after the delivery of 2000 puffs in 5 days, the overwhelming majority (or almost all) of the 460 mg of injected nicotine have been adsorbed into the walls and into the fabric samples hung inside the authors’ exposure chamber.

The proportion of nicotine adsorbed on the samples and on the walls can be obtained by a kinetic model that requires knowing the adsorption and desorption coefficients that are obtained experimentally. Therefore, to design a qualitative model approximation to the authors’ experiment we can use instead data from another experiment that measured adsorption of nicotine on two distinct adsorbents occupying the same area. This is important because adsorption/desorption rates depend on the ratio of areas of the adsorbents (in this case, the fabric and the walls).

The results of the laboratory study by Marcham et al [6] are useful, since the study measured nicotine adsorption in glass and cotton in a laboratory chamber. They found the mean amount of adsorbed nicotine density to be 0.75 µg/cm2​​​ in the cotton sample, six times the density 0.125 µg/cm2​​​ in the glass sample. Both sinks were separately placed in Petri dishes of the same surface (58 cm2​​​).

This study provides valuable comparative information: we can assume that the surface density of nicotine adsorbed by the cotton sample is on average 6 times the density adsorbed by the glass walls. Since , Marcham et al tested in the same surface, this density ratio only depends on the properties of the adsorbate (nicotine) and the adsorbents (cotton and glass). We can adapt this data to the authors’ mouse study considering the upholstery sample of 0.027 m2​​​ made of cotton that makes a fraction of 0.03 of the total area 0.81 m2. We have the following variables

The unknowns are Mg and Mc, the masses of adsorbed nicotine on the glass and on the cotton. We also have the following conditions

Gathering this information, we can find the unknowns (Mg and Mc) in terms of known variables through the following equations

Substitution of numerical values leads to the mass of adsorbed nicotine in the cotton and in the glass: M​c​​ =71.97 mg and M​g​​ = 388.02 mg (84% of the nicotine adsorbed in the glass and 16% in the cotton fabric). The surface density we are interested is that of the cotton sample: Mc/Sc

This is evidently an enormous superlative nicotine density 3-4 orders of magnitude larger than the largest nicotine ever measured in a fabric placed in any environment inhabited by humans.

We need to compare the nicotine surface density on the sample of the authors’ study with measured nicotine surface densities obtained from vaping in vape shops. This data is found in two studies: Khachatoorian et al 2019 [7] and Son et al 2020 [8]:

Higher nicotine surface densities are expected in vape shops, since in the years these studies were published (2018-2020) they allowed vaping in their premises and thus had intense abnormal vaping activity. But they are only sporadically visited for short periods, they are not normal vaping environments. On top of this, the authors (Khachatoorian et al and Son et al) artificially enhanced these densities by bringing and placing inside the vape shops external fabrics made of very efficient adsorbents.

The only realistic natural case of nicotine surface density on surfaces are the ones measured by Son et al on wipes placed on surfaces inherent of the vape shops (not transplanted externally). Therefore, we have the following result:

The exposure of the sample in the authors’ mouse study produced a nicotine density on a cotton sample : 2665 mg/m2, that is 4,953.4 times higher than the maximal natural density 0.536 mg/m2 from Son et al.

The authors’ nicotine density on the sample is still 24 times larger than the highest density of 108 mg/m2 measured by Khachatoorian et al on an implanted terrycloth towel after 3 months exposure.

The mouse study authors could get the 108 mg/m2 nicotine surface density in their sample with much fewer injected nicotine and much fewer puffs

While this is a totally artificial measurement (placing a terrycloth towel in a vape shop is far from natural), at least it is an actual measurement. For the natural density 0.536 mg/m2 we obtain M =0.08 mg = 80 µg, which is the nicotine yield per puff of a pod device.

The density 0.536 mg/m2 is not artificial, but it was obtained in a fixed surface of vape shop, which is an unrepresentative environment of normal vaping. Nicotine densities in more normal conditions should be much lower. For example (Melstrom et al, 2017) measured 6 µg/m^2 in a small room after 2 vapers vaped al libitum for 2 hours. Therefore:

It might not be possible to simulate in an animal model the low level of nicotine surface densities that should produce vaping in the environments were it normally takes place.

The human scale

Mice in the authors’ study weigh 30 g. Using the weight area conversion of Khachatoorian et al: 1 g for 26.25 cm2 of cotton fabric and the area of the sample used by the authors (0.027 m2), we obtain a sample weight of 10.3 g. The sample was exposed to 51.2 g of aerosol, 0.46 g of nicotine and 0.126 g of aldehydes.

For an average (male/female) human adult of 60 kg. This would be exposure to a fabric weighing 20.4 kg, previously exposed to 102 kg of aerosol, containing 0.92 kg of nicotine and 262 g of aldehydes. Just to have a robot generating 102 kg of aerosol at (say) 5000 daily puffs (in 24 hours this would be a puff every 17.3 s) at 25.5 mg/puff = 0.0000255 kg/puff, or 0.1275 kg/day, it would take puffing full stop 24 hours every day for 800 days.

Epilogue

The study (by Umphres et al, 2026) analyzed biological effects on platelet function of mice exposed to samples previously exposed to a massive overdose of a completely unrealistic (and toxic) aerosol. Exposure to samples subjected to this massive aerosol injection would be harmful for exposed humans. Therefore, the biological effects reported by the authors have no relevance to assess risks from vaping (either active vaping or environmental).

Future stuff:

I found 6 Australian mouse studies published between 2018 and 2025 trying to simulate “third hand vaping”. They exhibit the same shortcomings as Umphres et al, but their exposure conditions are not so extreme,

Also, I will post the complete dissection of a flawed risk analysis

Stay tuned.

References.

[1] Jacob III, P., Benowitz, N. L., Destaillats, H., Gundel, L., Hang, B., Martins-Green, M., ... & Whitehead, T. P. (2017). Thirdhand smoke: new evidence, challenges, and future directions. Chemical research in toxicology, 30(1), 270-294.

[2] Martins-Green, M., Adhami, N., Frankos, M., Valdez, M., Goodwin, B., Lyubovitsky, J., ... & Curras-Collazo, M. (2014). Cigarette smoke toxins deposited on surfaces: implications for human health. PloS one, 9(1), e86391.

[3] Lalo, H., Leclerc, L., Sorin, J., & Pourchez, J. (2020). Aerosol droplet-size distribution and airborne nicotine portioning in particle and gas phases emitted by electronic cigarettes. Scientific Reports, 10(1), 21707.

[4] Uchiyama, S.; Noguchi, M.; Sato, A.; Ishitsuka, M.; Inaba, Y.; Kunugita, N. Determination of thermal decomposition products generated from E-cigarettes. Chem. Res. Toxicol. 2020, 33, 576–583

[5] Van Loy, M. D., Lee, V. C., Gundel, L. A., Daisey, J. M., Sextro, R. G., & Nazaroff, W. W. (1997). Dynamic behavior of semivolatile organic compounds in indoor air. 1. Nicotine in a stainless steel chamber. Environmental science & technology, 31(9), 2554-2561

[6] Marcham, C. L., Floyd, E. L., Wood, B. L., Arnold, S., & Johnson, D. L. (2019). E-cigarette nicotine deposition and persistence on glass and cotton surfaces. Journal of occupational and environmental hygiene, 16(5), 349-354.

[7] Khachatoorian, C., Jacob III, P., Sen, A., Zhu, Y., Benowitz, N. L., & Talbot, P. (2019). Identification and quantification of electronic cigarette exhaled aerosol residue chemicals in field sites. Environmental research, 170, 351-358.

[8] Son, Y., Giovenco, D. P., Delnevo, C., Khlystov, A., Samburova, V., & Meng, Q. (2020). Indoor air quality and passive e-cigarette aerosol exposures in vape-shops. Nicotine & Tobacco Research: Official Journal of the Society for Research on Nicotine and Tobacco, 22(10), 1772–1779. https://doi.org/10.1093/ntr/ntaa094