In an opinion piece titled "Chemical regulators have overreached" in the August 27, 2009 issue of Nature, two prominent animal welfare advocates claim that vastly larger numbers of chemicals will have to be tested under the European Union's REACH regulation than previously estimated, and hence that 20 times more laboratory animals will be sacrificed.  They call for a moratorium on some animal tests.  Well, a closer look reveals that it's the opiners themselves that have greatly overreached.

[Update 8/28:  The European Chemicals Agency (ECHA) has just issued this press release also disputing the findings of this new study.]

The authors of the Nature opinion piece are Thomas Hartung and Costanza Rovida.  Hartung is the director of the Center for Alternatives to Animal Testing (CAAT), while Rovida is identified as a private consultant, but was formerly affiliated with the European Centre for the Validation of Alternative Methods (ECVAM), as was Hartung. 

The Nature piece cites a longer, 22-page report by the same authors released by the Trans-Atlantic Think Tank for Toxicology (t4).  t4 is a creation of CAAT.

The report is laid out to look like a peer-reviewed journal article but is self-published (more later on what the authors claim to be the expert review conducted of the report).  [Note added 8/27: The report is to be published in a journal called ALTEX.  According to its website, ALTEX is "the official journal of CAAT ... and t4, the transatlantic think tank of toxicology."  According to an article appearing in today's BNA Daily Environment Report (p. A-4):  "The study was prepared with funding from the Transatlantic Think Tank for Toxicology, which works with [CAAT]."  Hence my characterization of the report as "self-published" is quite appropriate.]

This study has used numerous demonstrably false or highly questionable assumptions, one piled on another, to grossly inflate the number of chemicals requiring testing under REACH, and the number of animals involved.

Both the opinion piece and the accompanying report reflect a fundamental misunderstanding of the basics of REACH and an apparent willingness to inflate every number in long chains of calculations to yield the largest possible estimates for the number of animals to be sacrificed under REACH. 

In this post, I will address in detail some of the more egregious claims.  They include:

• Vastly overstating the number of chemicals in commerce, to be registered and required to tested under REACH.
• Vastly overstating the number of high-production-volume chemicals in the EU.
• Overstating the number of animals required for at least certain tests.
• Claiming expert review of its report, when 7 of the 8 reviewers are either close colleagues of the authors or representatives of the chemical industry.  Not a single representative of the European Commission or the European Chemicals Agency reviewed the report.

Prepare for a fairly deep dive, with lots of numbers, because that's what the authors have based their claims on.

Some context

But first, some context.  During the nearly decade-long debate over the final text of REACH, animal welfare advocates extracted major concessions from the EU.  In addition to peppering REACH with statements to the effect that animal testing would be done only as a "last resort," the changes forced by animal welfare advocates included elimination of all animal testing for existing chemicals produced below 10 tons per year per manufacturer, and a requirement that only testing proposals, not test data, be submitted at the time of registration for any tests involving laboratory animals. 

Most notably, an entire Title of REACH is devoted to "Data Sharing and Avoidance of Unnecessary Testing," setting in motion the mandatory formation of so-called Substance Information Exchange Forums (SIEFs) among makers and users of a chemical that have become the latest poster child for the chemical industry's ongoing gripes about REACH.

Let me be clear:  I personally, and EDF organizationally, strongly support taking all possible measures consistent with good science and sound chemicals safety policy to reduce unnecessary animal testing.  That includes unearthing and utilizing all available data, allowing and facilitating the appropriate use of alternatives to animal testing, including in vitro methods, read-across within chemical categories, and estimation models based on structure-activity relationships (SARs).  It also means aggressively developing more alternatives, including high-throughput screening methods and computational toxicology – approaches that form the core of the long-term vision embodied in the National Academy of Sciences' seminal report Toxicity Testing in the 21^st Century.

But we also need to address the fact that tens of thousands of chemicals are in active use today that have never been sufficiently tested or assessed for safety, due to policies put in place decades ago that simply presumed them to be safe.  That is a very deep hole to dig ourselves out of.

But it's not nearly as deep as Hartung and Rovida would have us believe.  Let's examine some of their claims:

Claim #1:  "More than 100,000 synthetic chemicals are used in consumer products."

That's the very first sentence in the Nature opinion piece, and it's flat wrong.  This number is derived from the number of chemicals listed in the EU's inventory of all chemicals that were in commerce in the EU at the time the inventory was developed in 1981.  It is not an accurate count of chemicals currently in commerce.

In the US, about 84,000 chemicals are listed on the cumulative TSCA Inventory, first set in 1979, but again not all of those are currently in commerce.  EPA's latest count of those manufactured or imported above 25,000 pounds/year is less than 7,000 chemicals.  While that is clearly an underestimate as there are many chemicals below this threshold, and the reporting system has a number of exemptions, nowhere near 84,000 chemicals are in active commerce in the U.S.  Given the global nature of the chemicals market, it seems highly unlikely that the situation is radically different in the EU.

Claim #2.  "Our report … is based on the pre-registration of chemicals [under REACH]."

The authors' primary analysis is based on the gross number of substances that were pre-registered under REACH last year.  However, as the European Chemicals Agency (ECHA), which administers and oversees REACH, has made clear, pre-registration is not an accurate representation of the number of chemicals to be registered under REACH. 

ECHA's press release from March of this year states:

• "ECHA does not expect all of these [preregistered] substances to be registered."
• "In ECHA’s opinion the list contains many preparations and substances that did not require registration."

ECHA has already found that the list of pre-registered substances contains many substances (as well as items such as articles) that are duplicates or are entirely exempt from or inapplicable under REACH and will not need to be registered at all.  Pre-registrations were filed not only by chemical makers and importers, but by downstream users, as well as contract testing labs, consultants and others, mining for business opportunities.

Bizarrely, Hartung and Rovida acknowledge "a large abuse of preregistration" as well as significant duplicative entries.  Yet they proceed unfazed to base much of their analysis on the inflated pre-registration numbers.

Claim #3.  "The latest published list of REACH chemicals contains 143,835 substances that are supposed to be fully registered, each requiring a chemical safety report."

     AND

There are a total of "140,008 substances that may require extensive testing for registration."

These sentences contain several significant errors.  First, they reflect the gross number of pre-registered substances.  It is true that ECHA's pre-registration list contains more than 140,000 entries.  But as noted above, that number is highly inflated and the number of substances to be registered under REACH is expected by ECHA to be far lower. 

In a statement sent to Nature by ECHA in response to Hartung and Rovida's study (referred to in NatureNews here), ECHA reiterates that, based on its review of the pre-registration lists, it still believes its original estimates for the number of unique substances to be registered under REACH (about 30,000) is quite close to accurate.

Second, only those registered substances above 10 tonnes/year are required to have chemical safety reports (CSRs).  The EU estimates that the large majority (about two-thirds) of all registered substances will fall under this threshold and not require CSRs.  For these chemicals, no animal testing is to be required under REACH.

Claim #4.  We estimate "68,000 chemicals falling under REACH, and this is the lower (optimistic) estimate in our study."

The authors characterize the estimate they derived from pre-registration lists as "worst-case," yet they use it as the primary basis for their analysis.

But even their "best case" number of 68,000 chemicals is also highly inflated.  Its derivation is frankly, laughable:

• They start with the EU's own estimate that about 30,000 chemicals will be registered under REACH.  That number was derived by data collected by the EU in the mid-1990s, compelling the authors to seek to "update" it.
• First they note that chemical production as measured by sales volume has increased substantially in the EU, nearly doubling between the mid-1990s and today.  I have no reason to doubt this.
• Second, they point out that the EU itself has grown by accepting into its ranks a number of new countries.  They put that growth at about 20%.  Again, all fine.
• But then, astoundingly, they assume that the number of chemicals produced in the EU has increased in direct proportion to these growth factors.  That leads them to multiply the 30,000 EU estimate by about 2 and then again by about 1.2, to yield the 68,000.

The notion that recent growth in the sales and volumes of chemicals in the EU was derived entirely by introduction of new chemicals, and not primarily by increases in production of existing chemicals, is contradicted by all empirical evidence – including the statistics cited by the authors themselves in the very first paragraph of the Nature opinion piece. 

They point out that "existing 'old' chemicals represent about 97% of those in use today and 99% of the production volume."  I'll let you do the math to conclude that there is simply no way that 38,000 new REACH-eligible chemicals have been introduced in the EU since the mid-1990s.  OK, I'll do the math:  That would mean, among other things, that the "old" chemicals would account for well under half of those in use today, not 97%!

Indeed, the actual number of new chemicals registered in the EU since 1981 (which is cited by the authors elsewhere but ignored here!) is about 4,400.

Claim #5.  After going through more arcane calculations, the authors finally arrive at the following numbers of chemicals that they claim will require extensive animal testing:

• 47,858 chemicals marketed above 1000 tonnes/year, to which a 2010 registration deadline applies
• 53,040 chemicals marketed above 100 tonnes/year, to which a 2013 registration deadline applies

The former of these numbers represents what the EU calls high-production-volume (HPV) chemicals.  The authors claim there are nearly 48,000 such HPV chemicals.  The EU estimates there are only a few thousand.  Who's right?

The Organization for Economic Cooperation and Development (OECD) maintains a list of HPV chemicals produced in its 33 member countries.  OECD includes not only all of the EU, but also the U.S., Japan, Australia, Canada, Korea and all of the rest of the developed world.

How many HPV chemicals does the OECD list?  About 5,000.

So yet again, Hartung and Rovida grossly overstate reality:  They are off by at least an order of magnitude.

Claim #6.  "The two-generation study for reproductive toxicity … consumes an average of 3,200 rats per chemical."

The authors zero in on this particular test as a primary culprit, calling for a moratorium on such testing under REACH.  Let's look at the claim.

The authors claim this "average" number was calculated in a paper by Höfer et al (2004).  That paper, however, merely asserts the number and provides no calculation.  It does, however, characterize the number as a "maximum" number, and includes it in a table of "theoretical extrapolation of a maximum number of animals to be used."

The authors allude to a second paper by Cooper et al. (2006) that estimates only 2,600 rats per test, but doggedly stick with the higher number for all of their calculations.  Even that number seems high to experts we have contacted.  The Cooper et al. estimate assumed an average of 15 offspring per mated pair of rats; Hartung and Rovida themselves cited data that the average litter size for rats is only 8.2 offspring, while others put it at around 10.  Yet the authors appear unaware of and certainly never flag this major discrepancy.

There are, of course, many reasons why understanding a chemical's effects on reproduction is critical, and there is a large number of chemicals for which we are already finding such effects.  ECHA's statement summarizes the need for this test as follows:

     "The two generation study is the only study where functional fertility (including mating, fertility, number of implantations and litter size) is investigated in parental animals exposed during vulnerable life stages from conception, in utero up to puberty. Such an exposure design may be of special importance, e.g., for endocrine disrupting chemicals. This is not covered by any other reproductive study, including one-generation study protocols, as long as mating of the F1 generation [offspring of the exposed parents] is not performed."

Claim #7.  "The plausibility of our assumptions and calculations was checked by eight experts from industry, academia and regulatory authorities."

This paper has not been peer-reviewed in any normal sense of the term. 

A footnote on the first page identifies two reviewers.  One is the current Chair of the Board and former director of CAAT, the organization Hartung now directs.  The other is a colleague of Hartung's at the University of Konstanz in Germany, where Hartung has a joint appointment.

Six other expert reviewers are cited in the Acknowledgement section of the paper.  Five of the six work for the chemical industry or its trade associations:  ECETOC (a trade association "financed by its membership, which comprises 50 of the leading companies with interests in the manufacture and use of chemicals"), Dupont, Shell, Exxon-Mobil and BASF.  CAAT's advisory board is also well-stocked with industry representatives.

This is no accident:  There is, shall we say, a strongly shared interest between the chemical industry and animal welfare advocates in undercutting chemical testing programs.  This isn't the first instance of such close cooperation, and I very much doubt it will be the last.

A single reviewer was drawn from government (a German federal agency). 

The paper received no review whatsoever from anyone from the European Commission or ECHA.  Perhaps had that occurred, some of the huge errors might have been caught before publication.

Conclusion

As noted at the start, this study has used numerous demonstrably false or highly questionable assumptions, one piled on another, to grossly inflate the number of chemicals requiring testing under REACH, and the number of animals involved.

Why?  One need only look at the last concluding sentence of the author's study for what I think is at least part of the answer:

     "It is be­yond dispute that the primary aim of REACH is protecting hu­man health and the environment from unwanted consequences of exposure to chemicals.  The challenge will be to do it sensibly within the context of REACH while using all the information and experience we have and recognizing that most chemicals have been produced and used safely for many years without ex­tensive testing on animals.  (emphasis added)

That naïve assumption – that what we haven't tested can't hurt us – is what got us into this mess in the first place.  I cited many sources of information that demolish that argument  in the Introduction to my 2007 report, Not That Innocent.

There is a near-total absence in either the Nature piece or the accompanying study of mention of concern for the need to protect human health from the effects of toxic chemicals.  More striking, given the animal welfare orientation of the authors, is their utter failure to recognize or acknowledge that gaining a better understanding of chemical hazards is essential to protecting animals in the wild from toxic chemicals. 

Our knowledge of the endocrine-disrupting effects of chemicals originated with studies of animals in the wild.  DDT's devastating effects first came to light through witnessing the dramatic declines in reproductive success of ospreys and eagles in the wild.  Growing evidence indicates that the widespread and increasing deformations and gender-bending effects seen in wild fish and amphibians are the result of chemical exposures.  We now know that wildlife in the remotest parts of the Earth carry dangerous levels of persistent substances in their bodies.

All of these impacts of untested and under-assessed chemicals affect untold billions or trillions of animals in the wild.

Doesn't that matter?

[My EDF colleague and toxicologist, Dr. Cal Baier-Anderson, helped with some aspects of the content of this post.]

Many of the screening-level hazard data being collected and analyzed under ChAMP that pertain to human health are derived from traditional laboratory animal studies.  The National Academy of Sciences (NAS) recently offered a "new paradigm for toxicity testing" in its 2008 report Toxicity Testing in the 21^st Century: a Vision and a Strategy.  Can ChAMP hazard data be used to facilitate the development of new testing strategies? 

First, more about the new toxicity testing paradigm envisioned by NAS:  Instead of exposing whole animals to chemicals to examine their effects, which takes a lot of time, costs a lot of money and may not always be a good model for how humans can be affected, might we use human cells or tissues (e.g., liver, kidney) to determine how chemicals act biologically?  Where chemicals interact with the cells and alter (e.g., inhibit or over-stimulate) critical biochemical pathways, such perturbations may well represent early events or signals that can eventually lead to an adverse effect.

ToxCast

There is an EPA initiative called ToxCast that is exploring the use of high-throughput screening (HTS) test methods to predict hazard, characterize so-called "toxicity pathways," and prioritize among large numbers of chemicals.  According to the ToxCast website:  "In its first phase, ToxCast^TM is profiling over 300 well-characterized chemicals (primarily pesticides) in over 400 HTS endpoints."  The endpoints include tests that measure DNA and protein synthesis activities, multi-cell interactions and developmental assays in zebrafish.  These methods are among those being developed to realize the vision described in NAS' report. 

The results of the first set of ToxCast assays are still being validated – meaning that they are being analyzed to determine how similar the outcomes are to those of standard laboratory animal tests.  This is one important step in determining whether the ToxCast testing array can actually substitute for some standard laboratory animal testing.

But we would suggest there are already two timely and useful applications of the data that ToxCast can generate in aiding EPA's effort to evaluate hazard information under ChAMP.

Supporting alternatives analysis

The first idea would be to compare the relative capacity for chemicals in the same functional use class (e.g., solvent, chelator, fragrance) or chemical category (e.g., fatty nitrogen derived cationics) to effect biological changes detectable by the ToxCast HTS methods.  In doing so, we would be testing the hypothesis that chemicals that demonstrate less capacity for effecting changes in the HTS tests are more likely to have lower hazard profiles. 

As we described in a previous post, alternatives analysis is a tool that compares hazard assessments of chemicals that are used to perform similar functions, in order to identify those that have a lower hazard profile.  Using ChAMP hazard data and characterizations for this purpose would start by comparing the results of laboratory animal study data.  By running some of the ChAMP chemicals through the ToxCast HTS system and comparing the results of the HTS tests with the ChAMP hazard data, EPA could both support the identification of safer chemicals and test how well the HTS methods predict in vivo effects – a true win-win.

Validating (or refuting) chemical categories

About 80% of the chemicals sponsored under EPA's HPV Challenge Program are members of proposed chemical categories.  Within these categories, sponsors and EPA propose that hazard data for tested category members can be "read across" to untested members, as an alternative to direct testing of each chemical.

Grouping chemicals into a category starts with an hypothesis that distinct chemicals that show similarity or regularity in their physical-chemical properties and chemical structures actually possess similar or predictably regular patterns of biological activity.  Establishing a valid category requires that the hypothesis actually be demonstrated to be true, once the available data on physical-chemical properties, environmental fate and toxicity/ecotoxicity for the proposed category members are assembled.  Then, some degree of read-across among category members can be justified.

However, as our comments on the several ChAMP assessments we've analyzed in recent posts make clear (see here, here and here), both industry sponsors under the Challenge and EPA under ChAMP appear frequently to be over-relying on category approaches, lumping together chemicals that are insufficiently related or similar to warrant read-across.

HTS data could play a very useful role, therefore, in helping either to demonstrate – or to negate – an hypothesis that members of a proposed category of chemicals actually exhibit similar biological activity.  If category members in fact exhibit similar or regular patterns of activity across an array of different cellular and subcellular assays, then the case for grouping them for purposes of read-across would be much stronger than a case based solely on similarity or regularity in their physical-chemical properties and chemical structures.  Alternatively, if HTS data do not show such patterns, the category should not be utilized and either it should be broken up into smaller groupings or its putative members should be tested individually.

Both of these uses of ChAMP hazard data and characterizations side-by-side with HTS data developed under ToxCast – in supporting alternatives analysis and in category validation – would offer the added benefit of helping to advance the transition to the "new paradigm for toxicity testing."

John Balbus, M.D., M.P.H., is Chief Health Scientist.

A new paper by Shaw et al., published in May in the Proceedings of the National Academy of Sciences, “suggests a generalizable and scalable method for the systematic characterization and comparison of novel nanomaterials” using high throughput in vitro tests.  Does this mean that the National Academy of Sciences’ vision for toxicity testing in the 21st century – proposed for conventional chemicals – is already here for nanomaterials?  Not quite. 

The National Academy of Sciences (NAS), the Environmental Protection Agency, industry, environmental NGOs, and animal rights groups all look to high throughput in vitro testing to reduce animal testing while increasing the efficiency and lowering the cost of assessing chemicals and other materials for safety concerns.  But, as the NAS report acknowledges, achieving this shared vision will require a tremendous input of new research (and monetary resources) in order to refine and validate the next generation of toxicological test methods.  I discussed this in a previous blog post.

The Shaw et al. study makes a meaningful advance in the path towards in vitro screening.  The investigators used a variety of assays, cell types, and doses to test 50 different nanoparticles, thereby developing a multidimensional data profile for each.  They then applied statistical clustering techniques to examine whether differences in the structures of the nanoparticles, subtle or otherwise, changed their toxicity profiles.  For three nanoparticles that showed interesting differences in their profiles, the authors conducted short-term in vivo assays to seek to validate what they had seen in vitro.

The authors were able to demonstrate that using four different assays and a range of doses and cell types provided a robust set of data with which to categorize and distinguish the different nanoparticles.  The authors did not try to use the assays to predict specific health effects, but rather to create a classification system for previously untested nanoparticles and then use comparisons with nanoparticles that had already been thoroughly assessed by in vivo tests to make inferences about potential toxicity. 

The three nanoparticles chosen for in vivo validation in this study are all intended for use as medical diagnostic agents, and hence have been widely tested in animals.  The study demonstrated that the nanoparticle that appeared to have the highest toxicity of the three in the in vitro assays also caused the greatest response in vivo.

While the Shaw et al. study has considerable strengths, a couple of critical limitations highlight the large gap that still needs to be bridged if in vitro screening is to be effective in protecting human and ecosystem health from risks arising from the wide array of nanomaterial applications under development.

First, all of the 50 nanomaterials tested were designed for intravenous injection as medical diagnostic agents.  Because of this, the nanoparticles were water-soluble, making them easier to use in in vitro systems.  Most nanoparticles intended for non-medical purposes are poorly soluble and therefore difficult to apply and dose accurately in the liquid medium found in cellular assays.  And while in vitro assays using cells grown and exposed to a nanoparticle in liquid media reasonably simulate an injection exposure pathway, making inferences about health effects from exposure via other routes of exposure that are more relevant to non-medical uses – inhalation, dermal contact and even ingestion – is more difficult.

Second, both the nanomaterials and the endpoints assessed were of limited relevance to evaluating the kinds of effects that could arise from consumer, environmental, or occupational exposures to nanoparticles.  Most were variants of iron oxides, which are not commonly found nanomaterials in consumer products.  In addition, the investigators were most concerned with acute reactions to the injected materials, and therefore used acute assays and cell types relevant to intravenous injection:  vascular cells, monocytes (a type of white blood cell), and hepatocytes. 

The supporting in vivo test was also a short-term assay designed to look for changes in monocyte counts.  This test is most relevant to identifying immediate systemic inflammation, but would be very unlikely to detect perturbational changes that occur over longer time periods and could lead to chronic effects such cancer, neurodevelopmental toxicity, endocrine disruption, etc.  The latter types of effects are of greater concern for occupational, consumer, and environmental exposures. 

The authors note that this approach is most valuable as an initial screen in the relatively rigorous evaluation process required by the Food and Drug Administration (FDA) for approval of pharmaceuticals.  They point to the ability to identify the most promising nanomaterials by comparing their profiles to those of materials that have already gained FDA approval – and therefore were already extensively tested.  Of course, any nanomaterials selected through this screening process would then themselves be put through an extensive set of safety tests. 

This is a very different application of high throughput screening than, for example, would be needed to help evaluate a novel material for which a Premanufacturing Notice is submitted under the Toxic Substances Control Act, where there is no backup of rigorous safety testing to assure protection.  Such an approach may have value, however, in screening alternative nanoparticles as companies develop new products.

So while this study is a step in the right direction, there is still a lot more good science needed if high throughput in vitro screening tests are to become a reliable means to assess the safety of occupational, consumer, and environmental exposures to nanomaterials.

John Balbus, M.D., M.P.H., is Chief Health Scientist.

The history of health and environmental impacts of fuel additives is not a pretty one.  From tetra-ethyl lead to methyl tert-butyl ether (MTBE), we’ve learned the hard way that what goes in the tank ends up in our bodies and the environment sooner or later.  Getting a thorough understanding of the potential risks of a new fuel additive at an early stage is essential to avoid a lot of harm, suffering, and economic costs down the line. 

A new study by Park et al. has assessed the potential respiratory risks of a fuel additive called Envirox (nanoparticulate cerium oxide), giving it a clean bill of health based only on in vitro tests.  Is this the vision of the future of risk assessment?  Should we feel safe?

Nanoparticulate cerium oxide is touted as a solution to both global warming and particulate air pollution.  Added to diesel fuel as a combustion catalyst, it has been shown to reduce both fuel consumption and fine particle concentration in diesel exhaust.  But what happens when these tiny particles of cerium oxide blow out of the tail pipe? 

The study by Park et al. uses short-term in vitro assays and exposure data to conclude negligible risk of oxidative stress and pulmonary inflammation from chronic exposure to Envirox-augmented diesel exhaust.  The authors note that they have only examined oxidative stress and pulmonary inflammation and do not generalize more broadly about other potential health risks.

But does this really show Envirox is safe?  The authors note, “this assessment assumes that the in vitro exposure data can be accurately projected to the in vivo situation.”  What they don’t say is that it also assumes that short-term in vitro tests accurately predict effects from chronic exposure.  Neither of these assumptions is seriously examined in the paper’s discussion, and I’ve questioned whether current in vitro tests can be relied upon to predict actual toxicity in a previous blog.  But let’s assume for the moment that these assumptions are true.

Part of the concern with nanoparticles is the potential for translocation around the body, including to places where larger particles cannot go.  Could cerium oxide nanoparticles reach and build up in the bone, kidneys, or spleen (areas where non-nano cerium oxide particles accumulate)?  What about the developing brain (which other nanoparticles have been shown to access)?  Can they harm those organs over time?  Unfortunately, exposing lung slices in a petri dish can’t tell us about translocation or harm to these other organs. 

In the 1920s, tetra-ethyl lead got the green light as a gasoline additive after a short-term test of effects in adults showed no harm.  Of course, the worst effects of its use were chronic effects in children.  More than twenty years after lead was taken out of gasoline, children are still affected by residual lead contamination in urban soils.  We should be much smarter now. 

Even if the present study can be extrapolated to suggest it is unlikely that Envirox will cause pulmonary oxidative stress and related harm in real people, we need to know much more than that before concluding that its widespread use as a fuel additive is safe.