EPA’s New Chemicals Program: TSCA dealt EPA a very poor hand

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

[The first post in this series can be found here.]

Some in the chemical industry point to EPA’s New Chemicals Program as a robust program, one that could serve as a model for reform of the Toxic Substances Control Act (TSCA).  Most recently, the National Petrochemical & Refiners Association (NPRA) did so in its testimony at a recent House of Representatives subcommittee’s TSCA oversight hearing.  So just how robust is EPA’s program on new chemicals? 

When it was enacted in 1976, TSCA drew a clear line between “new” and “existing” chemicals.  Existing chemicals – those in commerce in the mid-late 1970s when the TSCA Inventory was established – did not have to be reviewed by EPA in order to remain on the market.  In contrast, new chemicals – defined as those not on the TSCA Inventory – cannot enter commerce without such an EPA review.  Once that review is completed and manufacture commences, a new chemical is placed on the Inventory and essentially becomes an existing chemical.

Of the roughly 84,000 chemicals now listed on the Inventory, 62,000 were in commerce when TSCA was enacted, and about 22,000 new chemicals have since entered commerce.  EPA reviews about 1,500 new chemicals every year, about half of which go on to enter commerce and be listed on the Inventory.

So right off the bat, it’s important to recognize that the great majority of chemicals in commerce are not and will not be addressed by EPA’s New Chemicals Program (NCP).  Rather, advocates for NCP as a model chemicals policy argue that TSCA reform could consist of simply applying the program’s basic approach and review processes to chemicals already on the market.

Unfortunately, while it makes a lot of sense to mandate that new chemicals be reviewed before their commercialization, Congress dealt EPA an exceedingly poor hand to use in conducting such assessments.  Below I’ll explore the following limitations, each of which illustrate why the NCP is not a good model for TSCA reform:

  • No data, no problem: No up-front testing requirement or minimum data set applies to new chemicals.
  • Guessing game: EPA is forced to heavily rely on limited models and methods to predict the toxicity or behavior of a new chemical.
  • Catch-22: While EPA can require testing of a new chemical on a case-by-case basis, it must first show the chemical may pose a risk – not an easy task without any data in the first place!
  • One bite at the apple: EPA typically gets only a single opportunity to review a new chemical.
  • Crystal-ball gazing: EPA has to try to anticipate a new chemical’s for-all-time future production and use.
  • Black box: New chemical reviews lack transparency.
  • Anti-precaution: In deciding whether to require testing or controls for a new chemical, EPA equates lack of evidence of harm with evidence of no harm.

The bottom line is that – precisely because it suffers from TSCA’s major structural flaws – the NCP falls far short of serving as a model for the statute’s reform.

No data, no problem:  No up-front testing requirement or minimum data set applies to new chemicals.

TSCA requires any company planning to produce or import a new chemical to notify EPA at least 90 days before commencing manufacture.  The notification typically takes the form of a Premanufacture Notification (PMN), unless the company seeks any of several exemptions EPA provides (for more on the available exemptions, see here).

PMNs must include basic information on anticipated use, production volume, exposure and release – to the extent such information is known or reasonably foreseeable by the submitter at the premanufacture stage.

However, TSCA does not require companies to submit a minimum base set of data on a chemical’s toxicity or ecotoxicity or its environmental fate and behavior.  This oddity of TSCA stands in contrast to the policies of virtually every other developed country in the world.  Although EPA encourages such data to be included in the PMN, the great majority of PMNs do not:  67% of PMNs contain no test data, and 85% of PMNs contain no health data; and more than 95% of PMNs contain no ecotoxicity data.

TSCA actually precludes EPA from requiring up-front development and submission of a minimum set of data on a new chemical’s hazards, by limiting the data required to be submitted in a PMN to that which is either already “in the possession and control” or is already known to, or “reasonably ascertainable” by, the notifier.  See TSCA Section 5(d).

Despite this lack of even basic data on a chemical’s hazards, TSCA typically requires that EPA complete its review of a new chemical within 90 days and decide whether a chemical needs more data or any restrictions placed on its manufacture or use; if EPA fails to act, manufacture can commence.

What’s an agency to do?

Guessing game:  EPA is forced to heavily rely on limited models and methods to predict the toxicity or behavior of a new chemical.

To make the best of the bum hand it was dealt, over the last several decades, EPA’s New Chemicals Program has pioneered the development of – and relies heavily on the use of – predictive methods and models, most notably structure-activity relationships (SARs).  SARs attempt to predict the behavior or toxicity of an untested chemical based on how similar its chemical structure is to those of other tested chemicals.  SARs include both mathematical models and algorithms that yield quantitative estimates, and more qualitative approaches that group chemicals into categories based on structural similarities and “read across” from tested to untested chemicals within the category.

EPA’s efforts to develop ways to predict the hazards of new chemicals in the face of the constraints under which it must operate under TSCA have been admirable.  But they hardly represent an ideal or model approach to data development.

Among other things, SARs themselves have severe limitations:

  • SARs are only available and reasonably robust for a subset of even basic hazard endpoints of concern, and are best for predicting properties like persistence and bioaccumulation potential. In particular, with the notable exception of genotoxic carcinogenicity, validated SARs are lacking for most human health-relevant endpoints, especially for chronic concerns such as reproductive and developmental effects.
    A good illustration of this limitation is the fact that EPA’s otherwise-useful online SAR tool, the PBT Profiler, includes only one toxicity endpoint – a value for chronic toxicity to fish.
  • SARs don’t work or haven’t been developed for many major classes of chemicals, including highly fluorinated compounds, inorganic chemicals, salts of organic chemicals, even moderately reactive chemicals, surfactants, and any multi-component process stream or mixture.
  • Even for covered endpoints and chemicals, SARs can provide estimates that differ dramatically from measured data. For example, EPA has found that its SARs predicted that the acute aquatic toxicity of dichloroacetyl chloride would be three orders of magnitude lower than found when directly measured in aquatic organisms.

Catch-22:  While EPA can require testing of a new chemical on a case-by-case basis, it must first show the chemical may pose a risk – not an easy task without any data in the first place!

TSCA does give EPA authority to require, on a case-by-case basis, testing or data development for new (as well as existing) chemicals.  And, for a small fraction of new chemicals, EPA has done so.  These are cases where EPA can meet the statutory burdens to require testing:

  • In a classic Catch-22, EPA must already have substantial information about a chemical – enough to demonstrate that it “may present an unreasonable risk” or that it will be produced in large quantities and result in significant environmental releases or human exposures.
  • EPA must also demonstrate that insufficient information exists to determine the effects of the chemical on health or the environment, and that testing is necessary to develop such information.

Arguably EPA has somewhat more leverage to compel testing of a new chemical relative to a chemical already in commerce.  Depending on the case, EPA can suspend its review of the PMN pending development of the data, or complete its review but include conditions that require testing prior to manufacture or impose controls pending completion of testing.  It can also negotiate a Consent Order (more on this below) or a Voluntary Testing Action in cases where PMN submitters agree on the need for testing.

Where EPA imposes a prohibition or limitation on a new chemical while testing is done, TSCA requires that any such controls be temporary and no longer apply upon submission of the specified information (see TSCA Section 5(e)).  Any permanent regulation of a new chemical would still require EPA to find that it “presents or will present an unreasonable risk” – the same near-impossible burden that applies to existing chemicals under Section 6 of TSCA (see TSCA Section 5(f)).

Just how often EPA has compelled or negotiated testing is unclear, as EPA has never publicly reported such data without lumping them together with other actions it has taken on new chemicals.  What is known is that EPA has taken action of some sort on fewer than 10% of the PMNs it has reviewed, only a fraction of which entail testing obligations.

One bite at the apple:  EPA typically gets only a single opportunity to review a new chemical.

TSCA grants EPA typically only one bite at the apple for new chemicals – a one-time, 90-day review opportunity at the premanufacture stage that takes place well before the full picture of the actual production, use and exposure, and lifecycle impacts of a chemical has emerged.

Once that review is completed and manufacture commences, the new chemical is placed on the TSCA Inventory, becomes an “existing” chemical, and any company can manufacture and use it without even having to notify EPA it is doing so.

In the relatively rare cases that EPA imposes any conditions or testing requirements on a new chemical, they apply only to the original notifier – unless EPA decides to go through a wholly separate rulemaking process to promulgate what is called a Significant New Use Rule (SNUR) specific to that chemical.

EPA has issued SNURs for about 7% of new chemicals.  They typically specify the same conditions imposed on the original notifier, and extend them to any other manufacturer that wants to begin producing the chemical.  If – and only if – such a company wants to produce or use the chemical in a manner that falls outside of the conditions specified in the SNUR, it must first notify EPA.

SNURs themselves only require notification so that EPA can review the new use, and do not themselves impose new regulatory controls.  To do that, EPA would have to either promulgate yet another regulation or negotiate a Consent Order with the new company.

All that just to ensure proper use of one new chemical.

Here again, TSCA is out of step with the laws of other developed countries.  Both Canada and the European Union have tiered notification and assessment processes for new chemicals.  As a chemical enters commerce and its production volume increases, these countries require additional notifications, impose additional data requirements and provide for renewed scrutiny by government.

Crystal-ball gazing:  EPA has to try to anticipate a new chemical’s for-all-time future production and use.

The lack of up-front hazard data requirements for new chemicals under TSCA reflects in part the fact that notification takes place at a relatively early point in the course of developing, manufacturing, and marketing a new chemical, when it may not be realistic to expect a company to have conducted much testing.  Government intervention at this stage has the advantage of flagging overt potential concerns before manufacturing has commenced and before significant financial investment has been made by the producer.  It also has the potential to allow redesign of the manufacturing process or the chemical itself to eliminate or reduce the concern in advance of commercialization.

However, the lack of data on a chemical’s hazards and other properties, combined with the rather speculative nature of information a company provides in its PMN on the chemical’s potential uses, releases, and exposures, severely limits the robustness of any risk evaluation conducted at this stage.  And for the vast majority of new chemicals, no matter how significant a change in production or use occurs subsequently, a company is not even required to let EPA know it’s happened.

Essentially, EPA has a single, highly time- and data-constrained opportunity to gaze into a crystal ball to predict the for-all-time trajectory of a chemical’s production, use and disposal – and if it guesses wrong, EPA’s only remaining regulatory recourse is under Section 6 of TSCA.  And recall that EPA has almost never succeeded in using its Section 6 authority to control dangerous substances, failing even to ban asbestos.

Black box:  New chemical reviews lack transparency.

I often hear from EPA staff that its new chemical reviews really are well-done and protective – they just can’t reveal enough of the details to convince me of that fact.

Three major factors render EPA’s review of new chemicals largely opaque to the public.

First, and perhaps most understandable, TSCA highly constrains how EPA can handle any information it receives that the submitter claims to be confidential business information (CBI).  CBI designations for new chemicals are the absolute norm: about 95% of PMNs contain information, including chemical identity, designated by the submitter as CBI.  As I have discussed elsewhere, both the latitude with which CBI claims can be asserted under TSCA and the constraints that apply to EPA in handling such information (including risking criminal charges and imprisonment for its wrongful disclosure; see TSCA Section 14(d)) have yielded a culture and practice at EPA that bends over backwards to prevent disclosure at all costs.

Second, because TSCA’s requirements for developing test rules or regulatory controls are so onerous with regard to time, resources and evidentiary burdens, EPA instead resorts to using Section 5(e) Consent Orders to impose such requirements on new chemicals.  Over the course of TSCA’s life, EPA has issued well over 1,000 such Consent Orders, compared to 4 – that’s right, 4 – regulatory restrictions on new chemicals under its Section 5(f) authority.

However, although it is not prohibited from doing so, EPA rarely makes its Consent Orders public, even in redacted form to remove CBI.  This is because, according to EPA staff, these documents need not be made public because they are considered products of adjudication.  (In contrast, if EPA follows up a Consent Order with a SNUR, the latter is a regulation subject to full notice and comment, and hence, disclosure.)

Finally, it is exceedingly difficult to assemble even a rough, let alone accurate, picture of the number and nature of EPA decisions on new chemicals.  EPA does not publicly track such data even in the aggregate.  And on the few occasions it has published statistics on how often it has used various authorities or instruments to review or control new chemicals, the data lump together or obscure important facts – such as for how many new chemicals EPA has imposed a testing requirement.

Even within TSCA’s constraints, EPA could certainly be far more transparent about – and engender more public confidence in – its review of new chemicals.  The fact that it has failed to do so says much about the need for statutory reform to shift the prevailing culture at EPA from one that errs on the side of keeping chemical information within its four walls, to one that affirmatively drives more and better information into the public domain.

Anti-precaution:  In deciding whether to require testing or controls for a new chemical, EPA equates absence of evidence of harm with evidence of no harm.

One of the best peeks into the black box I’ve had of late is when EPA took the unusual step of leaking a “sanitized” or CBI-redacted copy of a TSCA Section 5(e) Consent Order it negotiated with the manufacturer of a carbon nanotube.  I have already blogged at some length about this, here and here.  My angle here is to highlight again the decision logic EPA used, which staff tell me is not at all unusual in new chemical reviews at EPA.

In order to take action to control a new chemical – whether to impose conditions on use, restrict workplace exposures or disposal, anything – EPA must first find as a matter of law that the chemical “may present an unreasonable risk.”  (To require testing only, EPA can alternatively find that a chemical will be produced in substantial quantities and may be result in significant release or exposure.)

EPA can and typically does make separate findings for human health and for the environment.  In the case of the carbon nanotubes, its basis for these different findings is telling:

  • Human health: The Consent Order notes that this PMN was like most others in that no test data were submitted. EPA’s actual finding, it would seem, logically follows: “EPA is unable to determine the potential for human health effects” from exposure to this nanomaterial, and hence it “may present an unreasonable risk of injury to human health.” But the basis for this finding is not actually the absence of any test data for the nanotubes that are the subject of the PMN review. Rather, it is based on the existence of other affirmative evidence that certain carbon nanotubes act like asbestos. As a result, EPA is requiring one test to be performed, and imposing some modest restrictions on use and workplace exposures.
  • Environment: One might think the wholesale absence of test data would be enough for EPA to conclude that it is also “unable to determine the potential for environmental effects as well.” One would be wrong: The Consent Order actually affirmatively states that EPA has determined that “no significant environmental effects are expected.” As a result, EPA is not requiring any testing for environmental effects and is not imposing any controls on environmental releases.

Given how little environmental data exist on nanomaterials in general, let alone this particular carbon nanotube, how could EPA have reached this conclusion?

I’ll repeat here what I said in my earlier post, because it bears repeating:  In reviewing a new chemical, EPA’s decision logic is as follows:  Only if EPA already has evidence of a potential effect can it conclude that it is unable to determine whether there is an effect and call for testing.  If EPA doesn’t have evidence of a potential effect – even if it has no data at all – it’s ready to conclude that no significant effects are expected.

The precautionary principle, which the U.S. chemical industry so loves to malign, calls for initiation of proportionate action on a chemical where there is evidence of potentially dangerous impacts, even where that evidence is still incomplete or uncertain.

That principle also underpins the European Union’s REACH Regulation, which calls for the development of sufficient data on both new chemicals and those already in commerce, so as to be able to prioritize among them and act to control those that pose significant risks on an intelligent and well-informed basis

Some maintain that the New Chemicals Program is TSCA’s version of precaution, arguing that it seeks to prevent harm before it happens.  But the example I’ve just provided suggests that, if anything, TSCA forces the New Chemicals Program to employ an anti-precautionary principle.

Next up:  I’ll dissect the argument that EPA’s Chemical Assessment and Management Program (ChAMP) is a model for TSCA reform.

This entry was posted in Health policy, Regulation and tagged , , , , . Bookmark the permalink. Both comments and trackbacks are currently closed.