The U.S. Environmental Protection Agency’s (EPA) research on the chemical ethylene oxide offers a perfect example of the adverse impacts of government-generated junk science and the importance of correcting it when it happens.

In this case, a study created by a research arm of the EPA known as the Integrated Risk Assessment or IRIS vastly overstated risks associated with this chemical, which is used to sterilize about half of the nation’s medical supplies. Because of that flawed study, state and local governments shut down several medical sterilization plants just before the COVID-19 crisis hit, which contributed to medical supply shortages. All but one reopened, but the hype related to this junk-science study may lead to future closures and supply shortages, which is why EPA needs to correct the IRIS study.

So how did IRIS produce such a misleading assessment? One key reason relates to the exposure estimates in the underlying three studies (see herehere, and here) on which IRIS relied for its assessment. All three were conducted by researchers from the National Institute for Occupational Safety and Health (NIOSH) and all focused on the same group of workers exposed to ethylene oxide between 1938 and 1985. IRIS disregarded other studies and data that could have proved valuable in its assessment (for details see this paper).

The EPA’s assessment explains that the agency relied on these studies alone because the group of workers it covered was large, tracking the cancer rates among 18,254 workers at 14 plants that used ethylene oxide to sterilize medical equipment or treating spices over a period of 16 years. Supposedly, data underlying these studies also had fewer confounding factors than did other ethylene oxide worker studies, according to the EPA. 

Overall, these three studies showed one thing: that cancer rates among workers exposed to ethylene oxide over several decades were lower than cancer rates within the general population. These studies also reported weak associations for a few rare cancers, but the researchers had to parse through the data to tease out those associations.

While it is plausible that ethylene oxide could cause cancer in the cases of relatively high, long-term exposures, the NIOSH studies did not have actual exposure measurements for most of the timeframe workers were exposed. Instead, the first study from 1991 developed average exposure estimates to cover all years between 1938 and 1985. Since they had no data on worker exposures prior to 1976, the researchers developed estimates based on air samples collected in the later years–between 1975 and 1984. Using this data, the final estimated exposures for the entire time frame starting in 1938, was quite low, at an average of 4.3 ppm for workers operating sterilization machines and a low of 2 ppm for other workers in the plants. 

Yet common exposure levels in the industry for all years prior to the late 1970s, as found in other studies at that time, were many magnitudes higher than estimated in the NIOSH studies. A meta-analysis that reviewed all the major studies found that during the 1970s ethylene concentrations inside these facilities were much higher and periodically even exceeded the odor threshold, which is 400 parts per million, indicating exposures vastly higher than the NIOSH studies developed. Such high levels during the early years are not surprising since facilities began to implement working exposure measures during the late 1970s and early 1980s.

Moreover, a large percent of the group was likely exposed to these much higher levels, which magnifies the impact of relying on underestimates. “About 86% of the workers had exposures before 1978, when the exposure levels were believed to have been higher,” the meta-analysis explained. Hence, the NIOSH studies likely underestimated exposuresfor a significant number of the workers in the study, and potentially by a large amount.

Accordingly, by relying on these studies, which vastly underestimated exposures, without attempting to correct the data, IRIS staff assured that their assessment would vastly overstate ethylene oxide risks. Using these exposure estimates, IRIS established a “reference concentration” for ethylene oxide of 0.1 parts per trillion. The reference concentration is supposed to represent the level “of a continuous inhalation exposure” at which the chemical is expected to pose no “appreciable risk of deleterious effects during a lifetime.”

But when you compare the IRIS reference concentration to the actual human exposures that form naturally inside the human body and the environment, it’s easy to see that IRIS’s figure has no basis in reality. The human body produces ethylene oxide at a level that is 19,000 times higher than IRIS’ reference concentration. IRIS’ reference concentration is also 1,000 to 2,000 times lower than the background levels reportedly found in urban air around the nation, which EPA data indicate is about 0.1-0.2 ppb.

Clearly, no one should accept the IRIS assessment as useful for regulatory purposes, but until EPA corrects it, it’s going to continue generated unfounded fears and fuel dangerous activist-led campaigns.

Note: There were lots of other problems with the EPA assessment that are too much to detail in a blog post. You can find more details in my recently released paper for the Competitive Enterprise Institute.