Cancer and non-cancer risk analysis of vaping. Part I

Correcting a flawed risk analysis

Motivation: understanding risk models in vaping

It caught my attention a recently published 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

presenting an analysis of cancer and non-cancer risks in vaping (aerosol inhalation and ingestion/dermal exposure from e-liquids). The article was written by academics of medical institutions of China. As we have seen in most of the literature on vaping, the authors’ repeat the usual hostile narrative towards vaping. It was specially annoying to read how the authors refer to vaping as “e-cigarette smoking” and vapers as “smokers”. More than a bad nomenclature, this is a veiled attempt to play down the differences between vaping and smoking.

The authors adapt to vaping the Exposure Assessment Tools of the US Environmental Protection Agency (US EPA) and the Guidance Manual to calculate Non-Cancer and Cancer Risk Estimates by the Agency for Toxic Substances and Disease Registry (ATSDR). The experimental input for these risk models comes from reported yields of the target chemicals in published research literature.

The authors considered as target chemicals 4 aldehydes: (formaldehyde, acetaldehyde, acrolein, acetone) and 7 metals: (As, Cd, Mn, Pb, Cu, Ni, Cr). To use them as input for their risk formulae they considered for each chemical ranges (minimum to maximum values) and averages taken from the experimental results reported in 24 studies published between 2003 and 2025.

The results of the authors’ analysis seem to be quite shocking: while cancer and non-cancer risks are acceptable for ingestion/dermal routes in all chemicals. For inhalation, cancer and non-cancer risks were acceptable only for acetone and 3 metals, As, Cd, Mn, with cancer and non-cancer risks of the remaining chemicals resulting well above comparative safety thresholds. Cancer risks for inhalation of nickel and chrome and non-cancer risks for acrolein are reported as being sky high, 2-3 orders of magnitude higher than comparative safety thresholds. In particular, the authors report acrolein as a sort of monster chemical whose inhalation should cause serious harm.

After just a cursory look into the inhalation risks of the study, I detected some errors in the authors’ risk model. In particular, it seemed that they took as valid input values experimental results in some of the 24 studies that tested e-cigarettes aerosols under inappropriate experimental procedures (hence they report high yields of metals or organic toxins in the aerosol). This is a common error in literature reviews that take at face value all reported laboratory results, even those obtained in methodologically flawed studies. In particular, the notion that acrolein is such a toxic monster seemed very odd (in many emission studies it is not even detected).

SPOILER. After a thorough revision of the article, I found it was much worse that what I thought originally. The authors’ risk analysis is mired by several serious methodological shortcomings. Once these shortcomings are corrected, the cancer and non-cancer risks of all target chemicals (aldehydes and metals) are acceptable, below the comparative indicators. Acrolein is no longer the toxic monster. In the present post I explain the reasoning and arguments justifying my assessment.

Why this matters? Because risk modeling (admitting its limitations) is important to assess the safety of vaping, but reported sky high risks from a deficient analysis contribute to disinformation and confusion and aids a narrative that downplays THR (“… how come you say vaping is safer than smoking when cancer risks for inhaling nickel or formaldehyde are comparable or larger?).

Safety comparative indicators

The risk formulae of the USEPA and ATSDR evaluate each chemical in comparative reference with safety thresholds. Numerical values of these indicators are provided by all agencies, including The California Office of Environmental Health Hazard Assessment (OEHHA).

• For cancer risks the threshold is the number CR = 10^(-5) = 1/100,000, computed by the product of an exposure dose (mg/kg day) and a cancer reference indicator: the USEPA Inhalation Unit Risk (IUR) or the Cancer Value Potency (CVP) of OEHHA. Why this number 10^(-5)? it is a conservative estimate suggesting that exposure to a substance above this mark might cause one more case of cancer in every 100,000 persons in a given jurisdiction.

• Non-cancer risks follow from the Health Quotient HQ, with acceptable risks when HQ < 1. The HQ is computed for each chemical by dividing the concentration associated with the daily exposure by a daily concentration deemed to be safe. For inhalation these indicators are the USEPA Reference Concentration (RfC), the OEHHA Reference Exposure Levels (REL) or the ATSDR Minimal Risk Levels (MRL). Concentrations are defined accordingly for the exposure route: inhalation (air) and ingestion/dermal exposure (food, water, soil).

For many chemicals and the three exposure routes numerical values of the comparative thresholds for non-cancer are provided by toxicological agencies. In he US: USEPA, ATSDR and OEHHA. Similar indicators are provided by agencies in the EU and other countries. However, safety indicators are only available for a small minority of known chemicals. The authors of the study I am reviewing (Zhao S et al) provide numerical values for these indicators for inhalation in their Table 1

The safety CPV, RfC, REL and MRL indicators are very strict risk thresholds, protective for the general population, including children and vulnerable frail individuals. They contemplate voluntary and involuntary exposure. Since vaping is a voluntary activity aimed at adults, it is worth arguing whether its safety should be evaluated in reference to thresholds protective for the whole population.

There are less strict safety indicators for occupational exposure by healthy adults on 8 hours daily working shifts: the Threshold Limit Values (TLVs) published by the American Conference of Governmental Indiustrial Hygienists (ACIGIH), the Occupational Safety and Health Administration (OSHA) and The National Institute for Occupational Safety and Health (NIOSH). The NIOSH publishes the TLVs for hundreds of chemicals and other information in its Pocket Guide.

Safety indicators are obtained for specified exposure periods (intermittent, daily, yearly and lifetime chronic). Their numerical values follow, mostly, from experimental results of animal models, but subjected to a careful extrapolation to human scale based on: differences among species, specific disease end points, models of organ specific kinetic absorption and metabolism and other physiological effects. This is one the main tasks of Toxicology. See here a discussion on toxicological characteristics of various chemicals.

In what follows, I will deal only with the inhalation route, hence the input for the risk formulae comes from emission studies on the aerosols. In what follows I dissect the study by Zhao S et al.

The authors’ risk model

The authors introduce in their equations (1), (5) and (8) the formulae for chronic daily inhalation exposure “DD”, Cancer Risk “CR” and Health Quotient HQ (non-cancer risks). The basic equation is for DD (CR and HQ follow from it)

where P (mg/puff) is the mass per puff of a chemical measured by emission studies taken as input (the authors provide these values in their Table 4 as µg/puff), T = 163 is the users’ average number daily puffs (from a demographic study), EF is the exposure frequency (days/year), ED is exposure duration (years), BW is body weight in kg, AT is averaged exposure time (in days). Since P X T leads to the daily mass intake DMI of the chemical (mg/day), we can rewrite (1) as

The authors define CR and HQ as

where CPV is the OEHHA Cancer Potency Values, R is either the USEPA Reference Concentration (RfC) or the OEHHA Reference Exposure Levels (REL), while Cair is the air concentration based on the authors’ assumption of 20 m3/daily air volume breathed by adults:

The authors also assumed daily exposure EF (365 days/year) = 1, chronic lifetime exposure ED = 70 years, BW = 70 kg and AT = 70 X 365 days = 25,550 days. These assumptions reduce DD and CR in (2) and (3) to

The authors use (4) and (5) to list the values of DD, CR and HQ for all the chemicals as entries in their tables 6, 7 and 8, from the inputs of all chemicals listed in their table 4. I verified this carefully.

The USEPA and ATSDR exposure formulae

The authors provide no citation or source for their exposure and risk formulae, but it is evident they used the chronic daily exposure formulae of the USEPA and CR and HQ from the ATSDR. Hence, I compare their formulae from these formulae. The equivalent of the authors’ “Chronic Daily Exposure“ DD in (1) and (2) is the Average Daily Dose ADD in same units (mg/kg day)

where InhR is the inhalation rate and ET is the exposure duration (for intermittent or short times exposures lasting less than 24 hours/day). Notice that the Daily Mass Intake (DMI in (6)) is

For the DMI in (7) to coincide with the authors’ DMI in (4) the following relation must hold

whose only solution is (0.8333 is the air volume breathed per hour for 20 m3in 24 hours)

MAJOR ERROR IDENTIFIED. As a consequence of ET =1 and inhalation rate of 20 m3/day, the authors’ risk model implies a continuous 24 hours/day exposure. In other words, the chemical is in contact with the respiratory system every second during 24 hours per day. Evidently, this assumption is incompatible with the intermittent exposure of vaping when chemicals are only in contact with the respiratory system a much sorter time while the user inhales and exhales.

Corrections to the authors’ risk model

To correct the authors’ risk analysis following the guidelines of USEPA and ATSDR, we need to incorporate non-trivial values for the exposure factors InhR and ET. The authors assumed 20 m3/day as the daily inhalation rate of adults, but following the USEPA Exposure Factors Handbook, Chapter 6 Inhalation Rates, daily inhalation rate for men and women aged 21-31 undertaking light activity is 15.7 m3/day, with the 95 percentiles at 21.3 m3/day. Hence, for consistency with the authors’ choice of 20 m3/day, we use the 95 percentiles for this age group and light activity, leading to the inhalation rate of InhR = 0.016 m3/minute.

Since the exhaled aerosol is diluted because users retain between 80-90% of the inhaled aerosol (see St Helen et al and Hua et al) and its chemical composition is made of volatile compounds (Zhao T et al, van Drooge et al), it evaporates rapidly. Therefore, chemicals in the aerosol are only in contact with the user’s respiratory system during the inhalation/exhalation cycle at each puff. Assuming this cycle to last 6 s per puff (see Gupta et al), the total daily exposure time in 163 puffs is 978 seconds, or 16.3 minutes.

Substitution of InhR = 0.016 m3/minute and ET = 16.3 min/24h in (6) and (7) introduces a correction factor of 0.013,

For the lifetime chronic scenario: 70 years lifetime exposure: EF = 365 days/year, ED = 70 years, BW = 70 kg, AT = 70 X 365 = 25,550 days, we obtain a decrease of the authors’ DD by 2 orders of magnitude

However, the authors’ risk model exhibits another problem: it considers ED = AT = 70 years, which implies an unrealistic lifetime usage of an e-cigarette during 70 years from birth, when a lifetime user will typically start as a young adult of age, which we can set at 18 years old. Hence, (8) and (9) must be modified by setting ED = 52 years, which implies multiplying (8) by a factor 52/70 = 0.7428, which multiplied times 0.013 leads to a final correction factor of 0.0096:

Consequently, the values of exposure daily dose DD obtained by the authors must be corrected by multiplying them times the exposure factor 0.0096 to account for the intermittent exposure of vaping and for initiating to vape at age 18. Since ADD is used to compute cancer risk CR and the Health Quotient (HQ) is also based on the volume of the intermittent exposure, the avoidance of this exposure factor is a systemic error that transmits to all the authors’ calculations listed in Tables 6, 7 and 8 and in their statistical analysis.

Corrected risk estimates

The authors’ scary risk estimates are over-estimations by 2-3 orders of magnitude. Once we incorporate the correction factor in equation (12), these sky high risks collapse. I reproduce in Tables 1 and 2 below the inhalation estimates in the authors’ Tables 7 and 8 (cancer and non-cancer risks), placing next to the authors’ estimated values with the corrected estimates. All entries in Tables 1 and 2 with values CR > 10^(-5) and HQ > 1 are underlined in red.

Tables 1 and 2 show that the correction factor 0.0096 in equation (12) places all mean values of cancer risk below the safety thresholds CR = 10^(-5) and all mean non-cancer risks (except acrolein) are below HQ = 1, but the maximal values of the ranges of formaldehyde, nickel and chromium remain above 10^(-5). The upper end values of the ranges of acrolein, nickel and copper had also HQ > 1.

I will show in Part II that these end values were taken from experiments that tested devices under extreme overheating conditions (almost certain “dry puff”). The results of these experiments should not be taken as input in a risk analysis or in a literature review. Unfortunately, these results are often used and cited.

Non-cancer risks from acrolein remains high (HQ > 1), even after the correction of the exposure factor in equation (12). However, the authors miscalculated the entries of acroleine in their Tables 6 and 8 from their paper. I will deal with this in Part II.

Conclusion and preamble to Part II

The very scary and worrying risks reported by this published study are not real. They arise because the authors made mistaken assumptions: a 24 hours/day exposure to the chemicals emitted by vaping (instead of 16 minutes/day) and lifetime usage from birth (instead of starting at, say, age 18). We reported in an article (soon to be published) the same error of 24 hours/day exposure in a University of California study of metals in 7 disposables.

This is the message to take home: the safety of vaping is robust, not only because of the lack of combustion, but also IMPORTANTLY because exposure is intermittent. This means that toxic chemicals are in contact with our respiratory system ONLY while we inhale and exhale (6 seconds cycle done 100-300 every day making at most 30 minutes/day). Smoking is NOT intermittent because smokers are exposed to the continuous side-stream emission from the smoldering tip of the lit cigarette. Exposure to cooking aerosols lasts for hours and exposure to normal indoor pollution is continuous. This makes vaping safety very competitive.

What’s next in PART II? I will show that acrolein is not the super-toxic monster described by the article under critique. I will also show that its authors included as input experiments that tested devices in sheer “dry puff” conditions. Finally, I would like to criticize the institutional obsession to assess vaping safety ONLY with the strictest possible indicators (like assessing the safety of whisky in terms of its possble effects on toddlers).

---