Texas underestimates cancer risks of ethylene oxide

I know something about ethylene oxide. During my time at the US Environmental Protection Agency (EPA), I spent over 10 years on the team developing EPA’s cancer assessment of ethylene oxide, and for the last 5 of those years I was the project manager for the assessment.  Published in 2016, after careful analysis and extensive internal and external peer review and public comment, EPA’s cancer assessment concluded that inhalation of ethylene oxide is “carcinogenic to humans” and provided a cancer risk estimate.

So when the Texas Commission on Environmental Quality (TCEQ) proposed a cancer risk estimate for ethylene oxide far below EPA’s, I was interested to understand why the estimates were so different.  On behalf of UCSF PRHE, I reviewed the science and methods used by TCEQ and found that its approaches do not reflect the data and dramatically underestimate the potential cancer risks for ethylene oxide.  I discuss two major problems with the TCEQ proposal below.  More detail on these and other problems with the TCEQ proposal can be found in UCSF PRHE’s comments to TCEQ.

Ethylene oxide is a gas used to manufacture other chemicals and as a sterilizing agent, notably for medical equipment. People who work in jobs involving the use of ethylene oxide have the highest exposures.  EPA’s 2014 National Air Toxics Assessment (NATA) also identified a number of communities with ethylene oxide air concentrations of public health concern.  Thus, these assessments have real implications for the health of workers and people living near ethylene oxide sources.

EPA’s cancer hazard conclusion was based on increased risks of lymphoid cancer and female breast cancer found in studies of sterilizer workers exposed to ethylene oxide, as well as increased cancer risks in rodent studies and evidence that ethylene oxide damages DNA.  EPA derived a unit risk estimate of 5.0 x 10-3 per ug/m3 for ethylene oxide based on lymphoid and breast cancers—this is an upper bound estimate of extra cancer risk (above background risk) per set amount of ethylene oxide breathed in (see Table).

Table.  Comparison of TCEQ and EPA unit risk estimates for ethylene oxide. TCEQ’s estimate for total cancer is orders of magnitude below EPA’s.

Source: Table extracted from the UCSF PHRE 2019 comments to TCEQ.
1 Total cancer based on human data for breast and lymphoid cancers from EPA. (2016) Evaluation of the Inhalation Carcinogenicity of Ethylene Oxide (Final Report). EPA/635/R-16/350F.
Lymphoid cancers based on human data from EPA. (2016) Evaluation of the Inhalation Carcinogenicity of Ethylene Oxide (Final Report). EPA/635/R-16/350F.
Total cancer, but only includes lymphoid cancers, based on human data from TCEQ (2019) draft Ethylene Oxide Carcinogenic Dose-Response Assessment.

Reviewing TCEQ’s proposal, the first major problem I found is that TCEQ excludes breast cancer from its cancer risk estimate. EPA’s conclusion of a breast cancer hazard from exposure to ethylene oxide and the mathematical model it used to derive a risk estimate from the breast cancer data were supported by the EPA’s independent Science Advisory Board (SAB).  TCEQ also acknowledges a potential breast cancer hazard but, nevertheless, excludes the endpoint from its cancer risk estimate and mistakenly claims that its cancer risk estimate based solely on lymphoid cancer would also be protective of females.  EPA’s 2016 unit risk estimate for breast cancer incidence is almost 25% of its total cancer unit risk estimate and almost 3000x the TCEQ’s total cancer estimate.

The second major problem I found is that to estimate lymphoid cancer risk, TCEQ used a model that does not reflect the data. To derive unit risk estimates for lymphoid cancer, both EPA and TCEQ used data on lymphoid cancer from a study by the US National Institute for Occupational Safety and Health (NIOSH) of workers exposed to ethylene oxide in a number of sterilizing plants.

Consistent with EPA guidance and current risk assessment practice, EPA first developed a mathematical model to represent the data in the range of exposures encountered by the workers.  To select the most suitable model, EPA used statistical procedures and criteria recommended by its SAB.  Then, EPA extrapolated to lower environmental exposures using a linear extrapolation, which is appropriate for ethylene oxide because of its ability to damage DNA.

The TCEQ used a model rejected by EPA and its SAB, a model that provides a poor reflection of the data (see Figure). While the EPA model (orange line) more closely follows the data from the NIOSH study (black dots), TCEQ’s model (dashed blue line) does not reflect the shape of the exposure-response relationship and underestimates the slope at low exposures. Based on this inappropriate model selection, the TCEQ derived a unit risk estimate for lymphoid cancer over 3000 times less than EPA’s.

Figure: EPA and TCEQ models compared to categorical data (quartiles) from the NIOSH study. Source: extracted from EPA’s 2016 assessment, Figure 4-3.

Together, ignoring breast cancer risk and using a poorly fitting model for lymphoid cancers results in TCEQ’s unit risk estimate drastically underestimating the risks of ethylene oxide exposure. Use of this value would lead to inadequate protection of the public health of Texans. Texans would be better served if TCEQ adopted the unit risk estimate derived by EPA, which was the result of a rigorous assessment process and extensive peer review.

About the author

Jennifer Jinot is a former U. S. EPA staff scientist. Jennifer has over 25 years of experience in environmental health risk assessment, and she contributed to EPA’s 2016 cancer assessment of ethylene oxide, as well as EPA’s health assessments of secondhand smoke and trichloroethylene (TCE), among others. Jennifer has a multi-disciplinary background, including risk assessment work in toxicology, epidemiology, and the derivation of quantitative risk estimates, with a specialization in risk estimates based on human data.