Ethylene Oxide Development Support Document (DSD)

After its release in 2016, the TCEQ conducted a thorough scientific review of USEPA’s ethylene oxide (EtO) cancer dose-response assessment. TCEQ’s review identified serious scientific issues surrounding USEPA’s assessment. As a result, the TCEQ evaluated all relevant EtO data and conducted a cancer dose-response assessment. In doing so, the TCEQ was able to address the various scientific shortcomings of USEPA’s 2016 assessment and consider new data and analyses by the TCEQ or appearing in the scientific peer-reviewed literature since 2016.

Prior to conducting any assessments, as part of the TCEQ Guidelines to Develop Toxicity Factors, the TCEQ provides an opportunity to outside parties to submit any chemical-specific information they would like the Agency to consider. On August 16, 2017, the TCEQ publicly announced plans for the EtO assessment and requested any interested parties (industry, academia, NGOs, private citizens, etc.) provide any relevant information through our listserv and via our website. Any information submitted through this process is taken into consideration during the literature review phase of the assessment.

Subsequently, the TCEQ conducted extensive analyses for the EtO dose-response assessment; the results of the analyses guided the TCEQ’s decisions and approach to the assessment. The proposed Development Support Document (DSD) provides transparent documentation of those analyses, allowing readers the opportunity to see how and why decisions were made. The first draft EtO DSD was proposed for public comment on June 28, 2019, and the public comment period ended September 26, 2019. The agency received numerous comments on the proposed DSD from diverse groups (industry, academia, NGOs, and private citizens). Scientifically substantive comments were reviewed and fully addressed by the TCEQ, which has resulted in a revised and improved draft EtO DSD (dated January 31, 2020) that is currently undergoing external expert peer review.

The external expert scientific peer review is expected to conclude within the next few months. This thorough and extensive scientific peer review process will culminate in the final TCEQ DSD. The final DSD for EtO will incorporate the best available science in deriving an inhalation unit risk factor and an associated long-term effects screening level, which will be used by TCEQ's air permitting program.

• Both the TCEQ draft and USEPA final EtO cancer dose-response assessments are ultimately based on the same NIOSH cohort of workers; results from those workers have been adjusted for the general U.S. population.

• The TCEQ cancer dose-response assessment is based on the same carcinogenic mode of action (MOA) as USEPA’s assessment (i.e., a mutagenic MOA); consequently, TCEQ’s assessment does not assume a threshold but rather that carcinogenic risk may be estimated no matter how small the EtO inhalation dose, all the way down to zero dose.

• In the TCEQ’s revised assessment, adjustment for the greater susceptibility of children was conducted by applying age-dependent adjustment factors (ADAFs) using USEPA’s preferred method.

• The USEPA acknowledges that human data are insufficient to demonstrate that EtO causes any cancer in humans (e.g., lymphoid or breast cancer), even in workers exposed to levels up to millions of times higher than environmental levels to which the general U.S. population may be exposed. Nevertheless, the TCEQ assessment evaluates both lymphoid cancer and breast cancer as candidate endpoints, ultimately utilizing lymphoid cancer as the cancer endpoint with the strongest (although still insufficient) human evidence (see Appendix 7 of the TCEQ DSD).

• The TCEQ assesses lymphoid cancer risk to both males alone and males + females combined. However, because males appear more susceptible than females to EtO-induced lymphoid cancer, it is important to understand that TCEQ’s application of male dose-response results to the general population (including females) results in a lower health-based air concentration for everyone (i.e., the health-based air concentration would actually be higher if male + female results were used).

• The USEPA used an unconventional and overall supra-linear model that is statistically significantly over-predictive of the lymphoid cancer data when compared to the data that was used to develop the model. USEPA’s model selection process was flawed because the agency misinterpreted Science Advisory Board advice on model parameter values and miscalculated model fit criteria (e.g., Akaike information criteria (AIC) and model fit p-values). USEPA’s error was estimating model parameter values (i.e., “knots”) based on model fit to the data and then not counting them as fitted values in the AIC and p-value calculations.

• The TCEQ used a widely accepted dose-response model, the Cox proportional hazards model. This standard dose-response model neither statistically over- or under-estimates the number of lymphoid cancer mortalities when compared to the data used to develop the model, but rather is reasonably accurate; the same cannot be said for USEPA’s model. Furthermore, the AIC and p-value model fit criteria calculated by the TCEQ support using the widely accepted Cox proportional hazards model over the unconventional model used by USEPA. The Cox proportional hazards model has a lower/better AIC for model fit (low AIC is an indication of better model fit), and the TCEQ’s selection of the Cox proportional hazards model is also strongly supported by other considerations, such as the model predictions mentioned above.

• It is important to note that the human body naturally produces appreciable levels of EtO (a.k.a. endogenous EtO). In contrast to TCEQ’s risk-based results, USEPA’s risk-based air concentrations correspond to doses that are orders of magnitude below even the 1st percentile of the normal endogenous range of EtO in the nonsmoking population. In other words, the human body naturally produces EtO at levels that are orders of magnitude higher than doses corresponding to USEPA’s calculated risk-based air concentrations. As a result, USEPA’s assessment appears to overpredict the actual measured rate of lymphoid cancers in the general U.S. population.

• Some public comments suggest, because of the “Healthy Worker Effect” (workers may be healthier than the general public and therefore may be less sensitive to the effects of chemicals), that the TCEQ’s model prediction calculations of the lymphoid cancer data used to develop the models are not appropriate. However, the 95% confidence interval (CI) on the Standard Mortality Ratio (SMR) for unexposed NIOSH workers includes 1, which indicates that the mortality rate in the unexposed workers and the U.S. population mortality rate are not statistically significantly different. Similar results are obtained for the male NIOSH workers that drive lymphoid cancer risk (i.e., the lymphoid cancer SMR in unexposed NIOSH males is 1.03 (6/5.8) with a 95% CI of 0.38, 2.25). Thus, it is demonstrated that in fact, there is no Healthy Worker Effect for this critical cancer endpoint in this key cohort (i.e., TCEQ calculations are appropriate).

Because ethylene oxide is emitted in Texas and has been determined to be a carcinogen, TCEQ undertook an inhalation carcinogenic dose-response assessment to derive a unit risk factor and an effect screening level for this chemical. These numbers are not alternatives to EPA’s IRIS value; they are derived by TCEQ to be used in the TCEQ air permitting program. TCEQ has similarly derived toxicity factors for thousands of chemicals for use in Texas programs (for more information: https://www.tceq.texas.gov/toxicology/esl).

TCEQ bases its ethylene oxide number on the same studies that the EPA uses. However, the primary difference between the two assessments is that the TCEQ uses a different modeling method for estimating cancer risks from that data. TCEQ has shown the EPA assessment, which rejects the standard Cox proportional hazards model in favor of an unconventional overall supra-linear model, to be scientifically flawed. For example, not only does the EPA assessment rely on incorrectly calculated p-values and Akaike’s Information Criteria values in selection of their model (discussed in TCEQ’s assessment), but TCEQ has demonstrated the EPA model to be over-predictive of the very lymphoid cancer data that drives EPA’s unit risk factor (described below). The EPA’s Science Advisory Board did not specifically review the issue about the p-values and AIC values that the EPA calculated, but the EPA’s calculations were inconsistent with the SAB’s comments on fixed parameters not estimated from the data, and the SAB has never reviewed TCEQ’s calculations of p-values or Akaike’s Information Criteria values. Furthermore, EPA’s maximum acceptable concentration corresponds to a dose that is about 1/40^th of even the 1^st percentile of the distribution of ethylene oxide doses that your body naturally produces (endogenous); such a miniscule additive dose is not biologically plausible to have actual significance given the range of endogenous doses.

TCEQ assessment will undergo independent external expert peer review in the next six months.

TCEQ makes judgements based on the weight of scientific evidence. The weight of evidence for ethylene oxide-induced breast cancer is detailed in the proposed assessment and will be further detailed in the version revised after public and peer review comments.

The choice of how to estimate the change in cancer risks with increasing concentrations of a chemical depend on how the chemical does damage to the body (called the mode of action). Ethylene oxide directly interacts with DNA to cause damage that can lead to mutations and cancer. TCEQ and the EPA agree that this is ethylene oxide’s mode of action. TCEQ then used the standard approach for a chemical with this type of mode of action, which predicts a linear increase in cancer risks with increasing ethylene oxide exposure. You are correct that this method assumes that only higher levels of the chemical are dangerous, which is consistent with the vast majority of substances in our lives. The EPA’s approach of assuming that lower concentrations have higher risk (called a supra-linear model) is not consistent with ethylene oxide’s mode of action, nor could EPA provide any justification for using that type of concentration model based on ethylene oxide mechanistic information. In fact, the EPA’s SAB advised the EPA to use a model with a dose-response form that is both biologically plausible and consistent with the observed data. However, the EPA does not provide biological plausibility (i.e. mode of action) for a supra-linear model, nor does that model accurately predict the observed data (as discussed in our DSD and in responses to questions below).

In no place in the DSD does TCEQ discount childhood exposure (i.e., EPA age-dependent adjustment factors for childhood exposure are applied to the values), and it is not clear to us where you got this information.

Public health is best protected using the best available science. That is, science that realistically predicts risk so that priorities to mitigate exposures are properly set and efforts are expended on reducing exposures to chemicals commensurate with the risks that they truly represent to public health. The primary benefactors of decisions based on sound science are ultimately the public. Based on TCEQ’s ethylene oxide assessment, public health can be protected both against unacceptable environmental levels, and with the use of ethylene oxide-sterilized medical devices. TCEQ is on the side of the best available science.

When EPA IRIS toxicity factors or toxicity factors from other agencies are promulgated, the TCEQ performs a quick, expedited review of them to determine if we want to adopt those values, or alternative values, on an interim basis. At the time (March 2017) TCEQ adopted an interim animal data-based unit risk factor for ethylene oxide to be used until a thorough multi-year systematic review and dose-response assessment could be conducted under the extensive TCEQ toxicity factor guidelines. As part of TCEQ’s standard process for conducting a thorough review of the literature for our ethylene oxide cancer dose-response assessment, we sent out a request for ethylene oxide information on Aug. 16, 2017. In response to the request for information, any interested parties (e.g., industry, academia, NGOs, private citizens) can submit whatever they think is relevant information (e.g., studies, presentations) for the agency’s evaluation. Such parties can request a meeting with TCEQ to present information, such as the American Chemistry Council (ACC) did. Similarly, after the TCEQ released the draft DSD for public comment, the Environmental Defense Fund requested and received a meeting with TCEQ to discuss their comments. While the agency greatly appreciates the submittal of information and comments by all parties, we ultimately evaluate and weigh all relevant scientific information on our own and based on our guidelines to arrive at our own decisions. TCEQ has conducted extensive analyses of multiple lines of evidence, each of which could have either supported a given approach or not. The results of each set of analyses have guided the TCEQ’s decisions and approach, as is clearly and transparently documented in the scientifically reasoned DSD.

We are assuming that you are referencing the Oct. 18, 2019, memorandum from Kristina Thayer of the EPA’s Office of Research and Development. That memo does not refer to Texas or TCEQ at all, and it describes the modeling choices that the EPA made in their 2016 ethylene oxide assessment. However, the EPA reiterating their modeling conclusions does not provide any additional evidence that those conclusions are correct. The TCEQ has shown in our DSD that the EPA assessment, which rejects the standard Cox proportional hazards model in favor of an unconventional overall supra-linear model, is scientifically flawed (e.g., see the first response and TCEQ’s document).

The proposed DSD does not indicate that the IRIS factor overestimates by 1,179 deaths because that figure was an interim placeholder figure while documentation of the relationship between the upper and lower splines of EPA’s two-piece spline model could be located in EPA’s document. It was known at the time that the EPA model overestimated but not by exactly how much. As detailed in the document, TCEQ has shown EPA’s selected model assessment to be over-predictive for lymphoid cancer in the NIOSH cohort as a whole (comparing the predicted 141 cancer cases to the actual 53 cancer cases), for every exposure group, and for the US general population (when factoring in endogenous production of EtO).

The human body cannot tell the difference between ethylene oxide that is produced by industrial sources, versus small amounts that are naturally present in the air, versus the ethylene oxide that is produced internally by our own metabolism. It is the total amount of ethylene oxide that comes into contact with the DNA in cells that will lead to an increase in cancer, regardless of the source – this is a basic toxicological principle that states that equal internal doses give rise to equal risk.

The EPA’s 2016 dose-response assessment was based on occupational exposures of workers who, let’s assume, all had the same low background exposure to ethylene oxide (from ambient air and internal production). Therefore, as EPA states, the risk estimates produced by those studies would be the risk caused by occupational exposure to ethylene oxide in addition to the background exposure. From analysis of this data, the EPA ultimately concludes in their dose-response assessment that air concentrations higher than 10 ppt will cause an unreasonably high cancer risk (higher than 1 in 10,000 excess risk). However, the EPA has recently measured ethylene oxide concentrations of 110-220 ppt in areas with no nearby ethylene oxide sources. Based on the conclusion about risk in addition to background, one could conclude then that the EPA would choose an acceptable air concentration of 110-220 + 10 ppt (let’s say 120 ppt). However, the EPA does not do this, they use 10 ppt as the acceptable concentration limit, meaning that they are not considering risk levels above background.

As noted in our DSD, estimated average internally-produced concentrations of ethylene oxide are about 1.9 ppb in non-smokers and about 18.8 ppb in smokers. Accordingly, at measured background levels of ethylene oxide, the EPA’s unit risk factor for lymphoid cancer (7.1E-03 per ppb, ADAF adjusted) would predict a population-weighted lymphoid cancer incidence rate of ≈3.7% (in the absence of any exogenous ethylene oxide ambient air exposure or other potential causes of lymphoid cancer). By contrast, the EPA-cited lymphoid cancer background incidence rate (which would have many contributing factors, not just a single chemical) is 3%, demonstrating that EPA’s unit risk factor overestimates observable lymphoid cancer risk based on endogenous/background levels of ethylene oxide alone.

Misinformed and misplaced environmental priorities; medical device shortages https://www.fda.gov/news-events/press-announcements/statement-concerns-medical-device-availability-due-certain-sterilization-facility-closures

After the Sept. 26 deadline, TCEQ will consider the public comments, revise the assessment as needed, and then post the assessment as final, along with responses to comments. Once finalized, TCEQ will use the ethylene oxide effect screening level in the review of air permits for new facilities in Texas. The newly derived ethylene oxide long-term ESL will replace the one that TCEQ is currently using, which is set at 1 part per billion. TCEQ’s proposed long-term ESL is 4 ppb.

To derive the ethylene oxide cancer dose-response assessments, both EPA and TCEQ used data from a United States-based group of workers who were exposed to very high concentrations of ethylene oxide for many years and who experienced an increased rate of lymphoid cancers. From this data, both TCEQ and EPA had to estimate what the risk would be to a person who was exposed to typical environmental concentrations of ethylene oxide, which can be millions of times lower than the occupational levels the workers had been exposed to. The first step in this extrapolation is to determine how the chemical could cause cancer: In this case, ethylene oxide can cause cancer by causing damage to DNA. Based on that mechanism, the standard and conventional risk assessment method is to use a mathematical dose-response model that essentially draws a best-fitting straight line from the high dose data (from the worker exposure study) down to low doses (so it is applicable to ambient exposures). This is the standard method that TCEQ used, and using that method, agency toxicologists were able to accurately predict the number of cancers that were observed in the worker study. In contrast, instead of using the standard straight-line risk model, the EPA chose to assume that low doses of ethylene oxide are more potent than high doses for causing cancer (this is called a supra-linear model, and is the unconventional model that TCEQ referred to). EPA’s model was shown by TCEQ to significantly over-predict the number of cancers that were observed in the worker study, which is how we mathematically demonstrate that the EPA’s method over-predicts cancer risk.

In addition, the human body naturally produces low levels of ethylene oxide, with background levels being higher in smokers. Using the EPA’s risk assessment, the background levels of ethylene oxide in the population would be predicted to cause more lymphoid cancer than is actually observed in the general population (and ignoring any other potential cause of lymphoid cancer). In this way, we also know that EPA’s model over-estimates the cancer potency of ethylene oxide.

If finalized, this new effect screening level for ethylene oxide to be used in air permitting of new facilities would be similar to the one that TCEQ is currently using (see the response to your first question).

As noted in the response to your second question, TCEQ’s model accurately predicts the lymphoma cancer risk that was documented in the underlying U.S. worker study, while the EPA’s model was mathematically demonstrated by TCEQ to significantly over-predict risk.

True, progress in the protection of public health cannot be made based on demonstrably flawed science but rather must be made on accurate assessments of the risks posed by chemicals. Only then can chemical exposures of the public be appropriately prioritized for mitigation to achieve the greatest reductions in real health risk. Otherwise, public, government, and industry efforts and resources are wasted on addressing unrealistic risks created by flawed science. The TCEQ assessment, which more accurately predicts cancer risk than EPA’s (as discussed above), will allow Texas to better assess potential health risks posed by ethylene oxide-emitting facilities in Texas and act accordingly. The TCEQ assessment is not based on a males-only study but rather the same NIOSH cohort used by EPA, which is composed of about half female workers.

The TCEQ assessment does evaluate breast cancer as a potential endpoint, and agency toxicologists anticipate that even more information on breast cancer risk will be included in the final assessment document. Lymphoid cancer, however, is the primary endpoint in TCEQ’s assessment and the primary contributor to ethylene oxide risk in EPA’s assessment. Based on the dose-response analyses for lymphoid cancer, females appear to have less risk from ethylene oxide exposure than males. Therefore, using risk results based on males results in even greater protection for females.

Yes, risks are greatest for people who are exposed to EtO for their entire lifetime, and lifetime risks are greater when exposure begins in childhood. To ensure that children are also protected, the EPA and TCEQ both use age adjustment factors.

Please refer you to TCEQ’s response to the fifth question, as well as to the FDA: https://www.fda.gov/news-events/press-announcements/statement-fda-commissioner-scott-gottlieb-md-steps-agency-taking-prevent-potential-medical-device

The current long-term value for ethylene oxide is 1 ppb, which is an interim value intended to be conservative until a fully scientifically rigorous assessment with extensive dose-response analyses could be conducted. By TCEQ having done so, such an assessment fully supports the proposed value of 4 ppb.

TCEQ works with EPA daily on a wide range of issues. TCEQ and EPA typically agree on issues, but sometimes do not. It’s not unusual for states to disagree with EPA. However, in this case, TCEQ demonstrated that its assessment more accurately predicts the cancer potency of ethylene oxide than the EPA’s assessment (see answers given above).

TCEQ does not regulate occupational exposures. You may try contacting the Occupational Safety and Health Administration.

TCEQ began considering its currently proposed ethylene oxide assessment in early 2017 and put out a public request for information on Aug. 16, 2017. It is the agency’s understanding that EPA collected the initial ethylene oxide air samples around the Willowbrook, Ill., facility in May 2018. This May 2018 EPA sampling event was the beginning of Illinois’ (and other state’s) ethylene oxide activities.