EPA distorts the scientific evidence and fails to protect kids’ brains in its proposed limit for perchlorate in drinking water

Tom Neltner, J.D., is Chemicals Policy Director and Maricel Maffini, Ph.D., Consultant

Today, the Environmental Protection Agency (EPA) proposed a Maximum Contaminant Level (MCL) of 56 parts per billion (ppb) for perchlorate in drinking water – more than three times less protective than an interim health advisory level set in 2008. To justify this increase, EPA turned its back on scientific evidence showing that this potent neurotoxin undermines childrens’ motor development and control and can increase their anxiety and depression. The agency’s reasoning is inconsistent with its own analysis published in a draft report in late 2017 and the findings of a peer review panel it convened last year to review that report.

If the agency had used the most protective scientific study and the most sensitive endpoint evaluated in the proposed rule, the MCL would likely be 4 ppb – more than three times more protective than the current health advisory. As a result, the agency fails to adequately protect children from a lifetime of harm. With this MCL, EPA is allowing pregnant women to be exposed to perchlorate in the first trimester of pregnancy at levels that pose much greater risk of impaired neurodevelopment in their children.

The proposed MCL – and how the agency reached it – was both a disappointment and a surprise to us. In late 2017, we applauded the agency’s scientists for developing an innovative model connecting a mother’s perchlorate exposure in the first trimester to fetal harm. We were not alone – in early 2018, EPA’s peer review panel congratulated the agency’s scientists on their analysis. We also complimented EPA’s population-based approach to developing an MCL by estimating the percent of pregnant women, and their children, with borderline thyroid dysfunction due to low iodine intake.

So how did EPA abruptly change course and estimate an MCL less protective than the current health advisory? By altering its analysis in three subtle but significant ways:

1. Rejecting five epidemiology studies showing harm at even lower exposure levels in favor of one IQ study by Korevaar et al. in 2016.
2. Choosing an MCL that allows an IQ loss of 2 points even though the study showed a 1 point loss was statistically significant.
3. Dismissing an alternative, population-based method that EPA proposed in 2017 that reinforces the need for a more protective standard.

As a result of these distortions, EPA’s proposed MCL of 56 ppb falls far short of the mandate in the Safe Drinking Water Act that EPA propose a limit that is “at the level at which no known or anticipated adverse effects on the health of persons occur and which allows an adequate margin of safety.”[1] The only good news is that the agency has a chance to redeem itself since this is only a proposed rule and it has 18 months to review the public comments before issuing a final rule.

To help understand how flawed EPA’s analysis is, we describe the subtle but significant ways in which EPA distorted the science below.

Rejecting five epidemiology studies showing harm at even lower exposure levels in favor of one IQ study by Korevaar et al. in 2016.

In the 2017 draft report, EPA’s scientists identified a set of five robust epidemiology studies that showed a statistically significant quantifiable relationship between a pregnant woman’s levels of a key thyroid hormone and several endpoints relevant to neurodevelopment. Two of those studies identified IQ loss as the endpoint. Three others showed loss in measures of motor development and control. The peer review panel agreed with EPA concluding that, “[t]hese studies provide data from human subjects in a form suitable for determining quantitative relationships between maternal fT4 [free thyroid hormone T4] and the extent of neurodevelopmental impairment.”

In its proposed rule, EPA dropped all but one study – the IQ study by Korevaar et al. in 2016 and reanalyzed that study’s data. This revised study had the least sensitive endpoint, essentially enabling the agency to propose the highest MCL.

Even though EPA favored IQ because “intelligence quotient (IQ) – is more straightforward to interpret because there is more national and cross-national data available”, the agency completely ignores one of the IQ studies, a 2004 study by Vermiglio et al.. It is not even mentioned in the preamble to the proposed rule despite its being included in the set of robust studies in the 2017 draft report. This study showed IQ loss associated with reduced fT4 at levels three times lower than the IQ study EPA picked to develop the MCL.

In addition, EPA dismisses the three studies showing statistically significant evidence of harm to motor development and control (Pop et al. 1999, Pop et al. 2003, and Finken et al. 2013). Two of these studies would have resulted in an MCL about 2.5 times more protective than the modified Korevaar et al. 2016 study that EPA picked to develop the MCL. The agency justified its decision because they have fewer individuals and were less straight forward to interpret than IQ points. While these considerations are not unreasonable, they are hardly sufficient to reject these robust studies showing adverse effects at lower levels than the re-analyzed Korevaar study.

In its proposed rule, EPA scientists identified a new study by Endendijk et al. published in 2017 that they considered sufficient to determine a quantitative relationship between maternal fT4 levels and children exhibiting greater anxiety or depression at levels six times lower than the IQ study EPA used to develop the proposed MCL. As with the motor control studies, it dismissed this study using a similar – and inadequate – rationale.

Choosing an MCL that allows an IQ loss of 2 points even though the study showed a 1 point loss was statistically significant.

The Korevaar et al. 2016 study found a statistically significant correlation between a drop in maternal fT4 during the first trimester and a loss of one IQ point in the child. In its 2017 report, the agency listed a one IQ point loss as sufficient to calculate an MCL. This approach is consistent with past agency practice and its guidance.

However in the proposed rule, EPA uses a loss of two IQ points instead, while misleadingly claiming that it is sufficient to assure “no known or anticipated adverse effects on the health of the most sensitive population and allows for an adequate margin of safety.” This is the safety standard for an MCL in by the Safe Drinking Water Act.

Dismissing an alternative, population-based method that EPA proposed in 2017 that reinforces the need for a more protective standard.

Despite positive feedback by the peer review panel on its proposed population-based method in the 2017 draft report, the agency chose not to use it to develop an MCL. It dropped the approach, in part, because of the potential “difficulties communicating the population at risk for developing this precursor effect as a result of perchlorate exposure.” In other words, EPA had the opportunity to use the mother’s decreased fT4 level (the so-called precursor effect) as an early marker of exposure to perchlorate to protect a larger number of children. However, because it found it difficult to define these women, it chose the more straight-forward, easier to define IQ point loss, thus increasing the risk of impaired brain development to potentially thousands of children.

Summary

To protect children from anticipated harm to their brain development from perchlorate exposure during the first trimester, we calculated that the MCL should be around 4 ppb. Instead, the agency proposed 56 ppb, 14 times less protective than it should be and more than 3 times what it has had in place since 2008 as a health advisory. The agency only got to 56 ppb by distorting the science and essentially throwing out much of the important and innovative work it had done that showed the need for a more protective standard. Plain and simple, this decision puts children’s brain development at risk.

[1] 42 U.S.C §300g-1(b)(4)(A).