“Epic fail” in West Virginia chemical spill: Poor information, poor communications, poor decisions

Richard Denison, Ph.D., is a Senior Scientist.

[UPDATE 1/28/14:  See updates at several places in this post regarding a 2011 Eastman safety data sheet on crude MCHM – which, though more recent than the 2005 version initally circulated, still does not reference the additional oral toxicity studies conducted by Eastman in the 1990s.]

Little more than two weeks after the January 9, 2014, spill of multiple chemicals into West Virginia’s Elk River, it’s becoming increasingly clear that the private and public sectors at all levels failed miserably with regard to protecting the public’s health.  There is plenty of blame to go around.

Our focus in the blogging we’ve done about this terrible incident has been and remains on the lack of reliable information available on the chemicals involved in the spill, the failure to promptly share what was available with the public, and the shaky science upon which decisions and public communications as to the critical safety questions were based.  In this post, I revisit several aspects of the initial and ongoing information gaps to add some additional perspective. 

I discuss in some detail below two major problems that I believe demand close examination in the Chemical Safety Board’s and others’ investigations into the causes and consequences of the spill:

1. State and federal officials appear to have initially relied on Eastman Chemical Company’s incomplete and out-of-date Material Safety Data Sheet (MSDS) on “crude MCHM,” and as a result sowed confusion from the outset that has led to widespread public mistrust.
2. Those same officials appear to have accepted without scrutiny the adequacy, accuracy and relevance of Eastman’s additional toxicity studies of MCHM, based only on summaries of those studies when they were finally provided by Eastman.

I end by briefly describing some of the implications of this tragic incident that need to be addressed going forward.

One caveat:  Because information on this incident has emerged in a piecemeal and haphazard manner, I cannot vouch for the accuracy of every detail provided in this post.  I have strived to the best of my ability to accurately describe the sequence and nature of events based on the available information. 

An incomplete Material Safety Data Sheet set in motion much of the ensuing mess

In situations like this, MSDSs are supposed to provide the information needed to inform appropriate responses.  Because quick decisions need to be made, it is essential that such documents are complete and current.  In this case, nothing could have been further from the truth.

Eastman Chemical Company’s 2005 MSDS for “crude MCHM,” which surfaced in the first days after the spill, provided a single acute toxicity value for oral ingestion:  a median lethal dose, or LD50, of 825 milligrams per kilogram of body weight.  A press report in the Charleston Gazette on January 11 indicated that Eastman regarded the unpublished 1998 study that was the source of that LD50 value to be proprietary and hence would not be made public.

[UPDATE 1/28/14:  Eastman’s recently posted Q&A on the West Virginia spill provides a link to a 2011 version of its “safety data sheet” for crude MCHM.  This 2011 document refers to the same LD50 value as did its 2005 MSDS and notably still does not provide any reference to the “no-observed-effect” level discussed further below.]

[UPDATE 1/31/14:  This morning, Eastman posted an updated version of its Q&A document on its website (linked to in the above paragraph), and took down the earlier version.  Here is the original version, the updated version dated 1/31/14, and a redline comparison of the two versions.]

It is now quite clear, from that January 11 press report citing West Virginia health officials as well as from the Centers for Disease Control’s (CDC) own January 20 “Summary Report of Short-term Screening Level Calculation,” that CDC used this acute lethality value as the sole basis for initially setting the one part per million (1 ppm) “safe” level for MCHM in drinking water.  CDC’s methodology for doing so has never been disclosed.  I blogged previously about how this reliance on a median lethal dose to set a safe level for nonlethal effects was fraught with problems.

The first indication that additional studies on MCHM had been located came six days after the spill, in a January 15 letter from the Director of CDC to the West Virginia Secretary of Health.  That same evening, West Virginia health officials issued their “Water Advisory for Pregnant Women” advising pregnant women not to drink the water at all until none of the chemical could be detected in the water.

It was not until a full week after the spill – late on January 16 – that Eastman decided to make its studies public, and even then, it only provided summaries of them – more on this below.  That same day, in a press call, CDC for the first time described in more detail how it had derived the 1 ppm “safe” drinking water level for MCHM. 

But what CDC did not say was that it had pulled a switch:  Gone was any mention of the first and at that time only available study – the acute lethality study referred to in Eastman’s MSDS – on which it had first relied.  Instead, CDC said it set the value based on one of the newly found studies, a 1990 Eastman study that claimed to have identified a “no-observed-effect” level of 100 milligrams per kilogram of body weight per day (100 mg/kg/day).  Miraculously, this second study CDC said supported the same “safe” level of 1 ppm it had set previously using the LD50 study.

I’ll have more to say about CDC’s reliance on the second study in a moment.  But let’s go back to why all this confusion and midstream switching happened in the first place:  All indications are that state and federal officials relied on that incomplete and out-of-date MSDS from Eastman.  As I noted, that 2005 MSDS provides only a single oral toxicity value, the LD50 derived from Eastman’s 1998 study.  Despite the fact that Eastman’s additional study that provided the “no-observed-effect” level CDC later used was conducted in 1990, no mention of it is made in its MSDS issued 15 years later.

[UPDATE 1/28/14:  Eastman’s recently posted Q&A on the West Virginia spill provides a link to a 2011 version of its “safety data sheet” for crude MCHM.  In the initial days after the spill, government officials appear to have been using the 2005 MSDS and only later was the 2011 version being circulated.  However, this discrepancy is largely or entirely moot, as neither of the versions would have alerted officials to the additional data:  the only acute oral toxicity data for crude MCHM provided in the 2011 version is the same LD50 value provided in the 2005 MSDS; the 2011 version still does not provide any reference to the “no-observed-effect” level.]

Eastman has now made available summaries of or references to several other unpublished oral and dermal toxicity tests, as well as a mutagenicity test – all conducted by Eastman long before 2005.  Yet none of these is referenced in its MSDS.  [UPDATE 1/28/14:  These omissions apply to the 2011 as well as 2005 versions of the MSDS.]

[Believe it or not, this spill involves yet another example of an incomplete MSDS leading to confusion:  As we now all know, a second chemical (actually apparently a mixture of chemicals) – called “PPH, stripped” – was in the leaking tank.  That information was first revealed by the tank owner, Freedom Industries on January 21, 12 days after the spill.  At that time, Freedom Industries made available its MSDS on the substance, which is dated October 2013.  That MSDS references very little in the way of toxicity data.  Yet the next day Freedom Industries revealed that “PPH, stripped” was actually a mixture of two products made by The Dow Chemical Company.  It then provided Dow’s MSDSs for those products, both of which predate Freedom Industries’ MSDS and describe a considerable amount of toxicity data – virtually none of which is included in Freedom Industries MSDS.]

Under the Occupational Safety and Health Administration’s (OSHA) Hazard Communication Standard (HCS), it is the obligation of chemical manufacturers and employers to maintain and provide accurate and current MSDSs.  Obviously these requirements have been badly breached in this case.

CDC’s apparent reliance on only a summary of Eastman’s “no-effect-level” study and its unquestioning acceptance of Eastman’s interpretation of the data were unwarranted

As noted earlier, Eastman posted summaries of its studies to its website a week after the spill, on January 16.  This allowed the public for the first time to scrutinize the data CDC had relied on.  None of these studies has been published or peer reviewed, and up until their posting Eastman had indicated that all of the studies were proprietary.

Upon examining the key “no-observed-effect” study CDC is now indicating it used as the basis for the 1 ppm level, two things were quickly apparent:

First, the study was conducted on a different test substance than what actually spilled.  While the LD50 study had tested “crude MCHM” – the material that actually spilled into the Elk River on January 9 – the “no-observed-effect” study had been conducted on “pure” (97.3%) MCHM – a rather different test substance.  Some experts argue that distinction is far from trivial.  At the very least, CDC should have prominently flagged this difference between what was tested and what was spilled, and discussed its implications.

Second, the study summary indicates that effects from exposure to MCHM were actually observed at the “no-observed-effect” level asserted by Eastman, as well as at the lowest dose tested in the study.  Nonetheless, Eastman had dismissed their significance and still used the study to set the no-effect-level at 100 mg/kg/day.  This is common practice in industry toxicity studies:  Present the data and then argue why an observed effect isn’t real.

NRDC toxicologist Dr. Jennifer Sass pointed all this out in an excellent post to her blog dated January 19.  She notes that the study showed changes in both body and organ weight even at the lowest dose tested, but that Eastman had rejected them by arguing that the effects either were not statistically significant or did not change uniformly with dose.  She points out that very small numbers of rats – only two males and two females – were exposed at each dose level, which calls into question the statistical grounds on which Eastman dismissed rather than included the effects seen even at the lowest dose.

The implications are far from trivial:  In her blog post, Dr. Sass notes that use of the lowest dose tested as a lowest-observed-effect level would have reduced CDC’s “safe” level by a factor of 40.

While experts may disagree in interpreting these kinds of data, here’s the problem: CDC seems to have accepted Eastman’s conclusions without any question.  And that’s especially problematic because, as Sass notes, the Eastman document is only a summary of the full study. 

Indeed, Tables 1 and 2, which present the results, are titled “Summary of Repeated Exposure Study.”  They only provide qualitative indicators of the actual results – all assigned by Eastman.  So, for example, rather than providing actual measurements, the authors merely indicated that various measured parameters were “normal” or used up or down arrows and a numeric 1-3 scale to indicate whether there were “slight,” “moderate” or “great” increases or decreases.  None of these terms is defined, nor is how the authors determined whether the results were statistically significant or not.

In the absence of access to the actual study data, there is simply no way for independent observers – whether from CDC or the public – to draw their own conclusions about what effects the study did or did not find and at what doses.

There is no indication that Eastman has provided the full study to anyone; if CDC has it and has analyzed it, then it needs to state so publicly and provide its own assessment of the data.

It should be noted that Eastman provided summaries of multiple studies, and this same concern about lack of scrutiny by CDC and lack of access to full study results applies to all of them.

Simply accepting at face value the chemical manufacturer’s own interpretation and conclusions from the data, and doing so without access to and examination of the underlying data, is a huge breach of the public’s trust.

Implications

Unfortunately, the damage to public trust has been done and will not be easily undone.  I do empathize to some extent with government officials, who were caught by the spill with little to go on and had to make decisions on the fly.

But in some ways, that’s the whole point:  Flaws in the system like the ones I have described often don’t become apparent until there is a crisis and the need for information and utmost transparency becomes paramount.  And there is nothing to indicate that the possibly of a spill like the one we have just witnessed could not have reasonably been anticipated.

All of this means we need a system that prepares in advance for such circumstances as we’ve seen play out in West Virginia this month:  That would be a system that:

• requires chemical safety information to be developed and available to the maximum extent practicable,
• requires assessment – independent of chemical manufacturers and users – of the quality and interpretation of the available information before a crisis is unfolding,
• requires available information and the basis for decisions being made to be quickly disseminated to the public by all parties (private entities as well as government), and
• requires that gaps or uncertainties in that information (which will always exist) be forthrightly acknowledged and disclosed, and that steps be ongoing to narrow those gaps and uncertainties.