Cancer and non-cancer risk analysis of vaping. Part II
Correcting a flawed risk analysis
Summary of Part I
Part I showed the main methodological flaws of the risk analysis of this article
Zhao, S., Zhang, X., Wang, J. et al. Carcinogenic and non-carcinogenic health risk assessment of organic compounds and heavy metals in electronic cigarettes. Sci Rep 13, 16046 (2023) https://doi.org/10.1038/s41598-023-43112-y
The authors chose as target chemicals for the analysis 4 aldehydes (formaldehyde, acetaldehyde, acrolein, acetone) and 7 metals (As,Cd,Mn,Cu,Cr, Ni, Pb). As mentioned in Part I, they concluded that cancer risks (CRs) for inhalation of nickel and chrome and non-cancer risks (Health Quotient HQ) for acrolein were very high, 2-3 orders of magnitude higher than comparative safety thresholds (CR = 10^(-5) and HQ = 1). The authors also concluded that, in particular, inhalation of acrolein should cause serious harm.
However, a thorough revision of the article revealed two serious methodological flaws, (1) assuming a continuous 24/day exposure when exposure in vaping is short timed and intermittent and (2) assuming a chronic exposure duration to last the full life time, as if chronic users started from birth, not as adolescents or young adults.
Once these shortcomings are corrected, the cancer and non-cancer risks of all target chemicals (aldehydes and metals) computed by the authors decrease by 2-3 orders of magnitude (factor 0.0096). For the average input values risks become acceptable, below the comparative indicators. Although acrolein still remains with HQ> 1, it is no longer the toxic monster.
I reproduce below two tables from Part I showing the dramatic reduction of numerical values of CR and HQ for all the chemicals considered as ranges (minimum to maximum) and as averages taken from published emission studies. Numerical entries underlined in red indicate values above safety references
The tables show that the correction factor 0.0096 places all mean values of cancer risk below the comparative safety thresholds CR = 10^(-5) and (except acrolein) are below HQ = 1. However, CR > 10^(-5) and HQ >1 still hold for the maximal values of the ranges of formaldehyde, nickel and chromium remain above 10^(-5). I will examine these issues in this post.
Contents of Part II
I show in this post that cancer and non-cancer risks can be further reduced by two extra corrections to the authors’ risk model focused on:
The acrolein entries in the authors’ study (their tables 6, 7 and 8) were miscalculated (this error is besides the correction factor 0.0096).
The highest values in the ranges of all chemicals (aldehydes and metals) were measured in experiments conducted under extreme overheating conditions, with direct evidence of a “dry puff”. These measurements should not be taken as input in a risk evaluation.
Once these two extra shortcomings are corrected, then all risks are below safety thresholds. I will also discuss at the end whether the safety of vaping should be assessed using only strict safety references protective for the whole population.
NOTE: I am not displaying the full data from the 21 published emission studies on aldehydes and metals used as input for the risk formulae by the authors of the risk analysis. This data makes 4 large tables and thus displaying them would surpass the length limit of the Substack. Please contact me if you wish to see this data.
The case of acrolein
The input data for acrolein is listed in the authors’ table 4 (see further ahead )
Input data. Range: 0 to 1.66 µg/puff; Mean value 0.242 µg/puff
From this data the authors claim to have obtained (using equations (3) and (4) of Part I) the following values of HQ
HQ (authors). Range: 0 to 1670; Mean value: 152
Notice that a HQ = 1670 is a really huge HQ (it is 10 times the HQ for tobacco smoke, see this study). No wonder their risk analysis lead the authors to believe that acrolein is a super toxic monstrous chemical. However, as I show below. They miscalculated these values.
From the authors input data values above we obtain the Daily Mass Intake (DMI) by multiplying them times 163 puffs/day. The result is:
DMI. Range: 0 to 270.58 µg/day; Mean value: 39.45 µg/day
The air concentration Cair for the range and mean value of acrolein is obtained by substitution of these DMI values in this formula (equation (4) of Part I expressed in µg, not in mg)
To compute the HQ (non-cancer risk) we divide this concentration by the RfC (the safety reference of USEPA) of acrolein, which is RfC = 0.02 µg/m3. The result is
HQ. Range: 0 to 676.45; Mean value: 98.62
These are not the HQ values reported by the authors as entries in their Table 8 (left hand side of our Table 2). The authors badly miscalculated their acrolein entries: they reported HQ = 0 to 1670 µg/day, when under their own formulae the correct value is HQ = 0 to 676.
However, this is a correction before the correction factor 0.0096. Once we apply this second correction HQ becomes
HQ (corrected). Range: 0 to 6.04; Mean value: 0.94
where now the mean value is below HQ = 1, though the upper end of the range is still HQ > 1. The reason for this large value is the extremely small RfC = 0.02 µg/m3 that makes the quotient HQ = Cair/RfC very large. Using the other HQ safety reference: the REL = 0.35 µg/m3 from OEHHA (which is also protective) leads to an HQ = 0 to 0.5 for the range and HQ = 0.074 for the mean value.
Therefore: the authors’ claim that acrolein is the super toxic monster in vaping aerosols has been debunked.
On the highest values of the ranges of aldehyde and metals.
Even incorporating the correction factor 0.0096 the authors’ risk analysis still overestimates cancer and non-cancer risks. This overestimation comes from taking as input for their risk formulae the maximal values µg/puff in the ranges reported in the literature for each chemical. The references of this literature are listed in one of the supplementary files of the authors. I am not displaying this data but can show it by request.
The authors report in their Table 4 (see below) the values µg/puff (intake mass per puff) used as inputs of their risk analysis, The table shows these values in terms of “concentration range” (minimal to maximal µg/puff), a “Mean value” and Standard Deviation. These inputs come from 12 studies on aldehydes and 9 studies on metals. We reproduce bellow the authors’ Table 4
The values underlined in red in the table above are the maximal µg/puff values that lead to unacceptable cancer risks (CR > 10^(-5)) for formaldehyde, nickel and chrome and unacceptable non-cancer risks (HQ > 1) for acrolein and nickel (see right hand side of our Tables 1 and 2).
The authors took these µg/puff values from 12 emission studies that measured aldehydes yields and 9 studies measuring metals. The emission studies are listed in a supplementary file. An examination of the data from these 21 studies shows that these upper end µg/puff values were obtained as measurements from problematic experiments whose results should not be included as inputs in a risk analysis (I am not displaying the full data from these studies but can show it by request).
Maximal values of aldehydes. Out of the 56 devices tested in the 12 studies on aldehydes, the maximal values of the 4 aldehydes shown underlined in the authors’ table 4 (above) came only from a single vaping device of the 5 devices tested by Gillman et al (2016).
This e-cigarette was their “Device 1”, a CE4 “top-coil” tank-style (Vision, Shenzhen, China) e-cigarette whose aerosol had very high formaldehyde and acrolein µg/puff comparable to yields from tobacco smoke. The device was tested at 4 power levels with µg/puff values as shown below
Notice that the averages in the table above for formaldehyde and acetaldehyde (28.12 µg/puff and 22.47 µg/puff) exactly coincide the authors’ maximal range values in the authors’ Table 4 displayed before.
However, Gillman et al described that this device had serious malfunctions
“At the conclusion of this study, the coil for Device 1 was examined and found to be charred, an indication of thermal decomposition. The charred coil, the observed decrease in yield in mg/watt production at the highest power level, and the elevated levels of aldehydes and acrolein, all indicate that the results for Device 1 may not represent typical usage of this device, we hypothesize, and a typical user might experience noxious dry-puff effects and discontinue use at that power setting”
Gillman et al also explain that Device 1 had a deficient design
“It should be noted that this style of atomizer is largely out of favor now in the vaping community, due to the difficulty of wicking with some liquids, and the propensity for dry-puff to occur.”
Top Coil devices are no longer used. Yet, the risk analysis under criticism used the high aldehyde yields from this defective device as legitimate input, which overestimates the mean values in their Table 4 and leads to high cancer risks despite the correction factor 0.0096. The authors’ risk analysis assumed as valid an extremely unrealistic situation: that users will puff for their full lifetime an old malfunctioning device that has become a historical relic.
Maximal values of metals. The same problem described above for aldehydes occurs with metals. Looking at the 9 emission studies focusing on metals (not displayed,but can be shown by request), all the large metallic yields in the authors’ Table 4 (above) that lead to high risks from nickel and chromium came from testing a single device in a single study by Zhao D et al (2019)
Zhao D et al puffed one of the e-cigarettes tested by (the “OD2”) at 40 W, 120 W and 200 W. It was a SMOK (Smoktech, Shenzhen) device with a stainless steel coil with a 0.6 Ω resistance. Information from the manufacturer (smok-vape.com) identifies the coil as the TFV8-baby-Q2-coil, with recommended usage in the range 20-50 W. It is evident (and scandalous) to puff in the laboratory a device recommended for 20-50 W at 120 W and 200 W, assuming it is “normal” usage.
It is absolutely certain that Zhao D et al generated an aerosol under extreme overheating conditions that should have charred and even ignited the coil. Hence, the enormous yields of all metals, specially Cu, Ni and Pb. We revised and criticized this deficient study in this review, but it is still cited as “reference” for metals in vape aerosols. It is regrettable that the authors of the risk analysis under critique took the experimental results of Zhao D et al as legitimate input, when just for minimal consistency such results must not be used as input for a lifetime risk analysis.
Conclusion on maximal yields. Looking at the 21 studies on aldehydes and metals (not displayed, but can be shown by request), it is obvious that enormous concerning yields of toxins were only detected in two devices out of about 100 tested devices. Recalculating cancer and non-cancer risks after removing these two inappropriate and unrepresentative experimental results decreases CR and HQ by an order of magnitude, leaving CR < 10^(-5) and HQ < 1 for all chemicals.
The fact the authors of this risk analysis mistakenly took these values as valid inputs shows ignorance of most current researchers (especially health professionals) on basic notions of e-cigarette operation (how can researchers accept as representative to puff at 200 W a device recommended for 20-50 W?). It also shows a careless and uncritical approach to research on vaping, with most researchers, reviewers and editors simply taking any published result as valid.
Is the smallest comparative safety references the most “protective”?
In the case of acrolein non-cancer risks are high (despite corrections) because its chronic exposure RfC is a very small 0.02 µg/m3 and thus HQ = Cair/RfC is large. This RfC is a really small concentration making only 1% of the ppb (parts per billion) of acrolein (2.29 µg/m3).
The utility of this RfC seems to be questionable, since data from the USEPA shows that many households and public spaces in the US register environmental concentrations of acrolein 10-100 times its RfC. Other comparative safety references for chronic exposure are perhaps more realistic: the REL from OEHHA (0.35 µg/m3) or the Minimal Risk Levels (MRL) of ATSDR (0.9 µg/m3) based on the No Observed Adverse Effect Level (NOAEL).
In the case of acrolein both references (RfC and REL) were available but the risk analysis authors chose the RfC, likely for being the smaller one (0.02 µg/m2 < 0.35 µg/m3) and for easily leading to HQ > 1, probably believing that a smaller reference is more protective. However, I have not found any reference from the agencies (USEPA, OEHHA, ATSDR) justifying this notion. A study comparing the RfC with the MRL explains how the differences between them depend on the methodology used to derive them and on the specific risk approach of the agencies. There is no mention that “smaller is universally more protective”.
However, none of these comparative safety references are meant to be absolute safety indicators, they admit an inherent uncertainty and methodological limitations. This is recognized by the US EPA when addressing the RfC = 0.02 µg/m3 of acrolein (emphasis mine):
It is based on squamous metaplasia and neutrophilic infiltration of nasal epithelium in rats. The RfC is an estimate (with uncertainty spanning perhaps an order of magnitude) of a continuous inhalation exposure to the human population (including sensitive subgroups) that is likely to be without appreciable risk of deleterious noncancer effects during a lifetime. It is not a direct estimator of risk but rather a reference point to gauge the potential effects. At exposures increasingly greater than the RfC, the potential for adverse health effects increases. Lifetime exposure above the RfC does not imply that an adverse health effect would necessarily occur.
Similar explanations are found for the chronic REL and MRL, they are also protective for the general population but recognize uncertainties (see summary). Therefore, in the case of acrolein the inherent uncertainty of the RfC might very well encompass the higher values of the REL and MRL.
Are the strictest comparative safety references necessary for vaping?
E-cigarettes are intended (and largely used) by adults, specially adult smokers and ex-smokers. Strict comparative safety references (RfC, REL, MRL) are protective for the whole population, including children and vulnerable individuals, but it is practically certain that vaping aerosols are not inhaled by children and vulnerable individuals with delicate health conditions. Therefore, I believe it is a legitimate question to ask if it is absolutely necessary to evaluate the safety of e-cigarettes by means of safety references that are protective also for sub-populations that do not use them.
It is true that a minority of vapers are adolescents aged 12-17 whose usage is unintended, though they mostly vape infrequently with few among them declaring in surveys usage 20 days or more per month. Healthy adolescents are not (necessarily) more vulnerable than healthy adults with respect to risks of inhalation of toxic and carcinogenic compounds in vaping aerosols.
Since healthy adolescents are not children and are not particularly vulnerable, strict protective standards justified for children and adults with delicate health conditions perhaps should not apply to them. The main concern around adolescent vaping is its potential for nicotine addiction, but this concern does not seem to require protective references like the RfC, REL and MRL conceived for inhalation of chemicals.
OEL (ocupational exposure limits) are alternative comparative safety reference, conceived as time-weighted average (TWA) concentrations of airborne substances to which a healthy worker may be exposed without any health impairment during 8 hours 5 days a week shifts throughout a working lifetime.
However, quoting the document by the Division of Nonclinical Science (DNSC) of the FDA (“Use of Reference Values in the Toxicological Evaluation of Inhaled Tobacco Products”), the FDA (and I suspect the USEPA and all US health institutions) recommend the toxicological evaluation of regulatory applications and usage of “tobacco products” (that include vapes) to be conducted only through the selection and use of toxicity reference values protective for the general population, which obviously does not include OELs, which are biased towards the healthy worker and are not applicable to the general population.
The document emphasizes the exclusion of OELs for the evaluation of carcinogenic tobacco product constituents, for which any increase in exposure must be associated with an increase in risk (the notion that cancer risks lack thresholds). OELs may only be acceptable to inform the toxicity evaluation for non-cancer effects. The document emphasizes a two-part approach separating cancer and non-cancer effects as the current paradigm based on significant differences in the risk assessment methods for these effects.
Further, the document justifies why OELs are inappropriate for the toxicological evaluation of usage of “tobacco products” (the emphasis is mine)
OELs are typically time-weighted averages representing repeated sampling of workplace chemical concentrations that can remain relatively constant for a defined period during a work shift, whereas tobacco product inhalation exposures are the summation of several intense short duration exposures that are repeated throughout the day and over a chronic time period
Consequently, inhalation exposure to smoke constituents that occur during the use of a tobacco product is likely to be substantially different from exposure in an occupational setting.
This justification for the need to exclude OELs seems to be only appropriate to smoking, not for vaping (hence only “smoke constituents” are mentioned in the second paragraph). In smoking the “short duration intense exposures” last longer because of the side stream emission and they are made of abundant very reactive and toxic compounds that linger in the environment. Therefore, the OEL in an occupational scenario is really insufficient for a toxicological evaluation of a heavy smoker, either working in this scenario or in an other scenario.
As a contrast, for a vaper in a given occupational scenario these “short duration intense exposures” are of real short duration and just involve a very limited level of extra toxicity in comparison with a non-vaping worker in the same scenario. This reasoning does not necessarily imply that OELs are appropriate for a toxicological evaluation of vaping, but that they could roughly apply equally to vapers and non-vapers in same occupational scenarios.
The FDA document I cited and quoted above was released in 2017, 9 years ago. I do not know (and could not find) if there is a recent update, or if the FDA has modified its criteria of toxicological evaluation of “tobacco products”.
Conclusion.
I believe I have shown in these 2 posts that cancer and non-cancer risks remain below concern when evaluated for inhalation of toxic and carcinogenic byproducts in vaping aerosols, as long as the input for the risk analysis excludes experimental results conducted under methodologically flawed conditions hat are also unrepresentative of normal consumer usage.
Notice that CR and HQ remain acceptable even considering strict comparative safety references (RfC, REL, CVP) that are protective to the general population. Besides the reduced presence of toxic and carcinogenic byproducts in vaping aerosol, another important factor that sustains the safety of vaping is its intermittent short time exposure.
The methodological flaws I encountered in criticizing the risk analysis in these two posts occur in other published articles. This study by Salazar et al (2025) from the University of California, Davis, evaluated metallic yields in 7 unregulated disposable, evaluating also cancer and non-cancer risks. The study also assumed a 24 hours/day exposure, hence their risk values are overestimated by 2-3 orders of magnitude (see a critique of Salazar et al and correction to their risk analysis in our paper soon to be published).
The review by Fowles et al (2020) on cancer and non-cancer risks from metallic contents in vaping aerosols also included as input the study by Zhao D et al (that puffed at 200 W a device recommended for 20-50 W). Evidently, the conclusion of this review on vaping safety is very negative. Unfortunately, the perception that vaping is harmful (or at least not sufficiently safe) is fed and enhanced by the ongoing publication of reviews and articles that use as input experiments conducted under inappropriate and unrepresentative conditions. The authors of these articles keep citing each other and (so far) ignoring our critical reviews.