Unsafe at any Dose? Diagnosing Chemical Safety Failures, from DDT to BPA

[Cornucopia has re-posted this article for its sole merit, which does not imply endorsement of the author’s opinions expressed elsewhere.]

Independent Science News
by Jonathan Latham, PhD

Piecemeal, and at long last, chemical manufacturers have begun removing the endocrine-disrupting plastic bisphenol-A (BPA) from products they sell. Sunoco no longer sells BPA for products that might be used by children under three. France has a national ban on BPA food packaging. The EU has banned BPA from baby bottles. These bans and associated product withdrawals are the result of epic scientific research and some intensive environmental campaigning. But in truth these restrictions are not victories for human health. Nor are they even losses for the chemical industry.

For one thing, the chemical industry now profits from selling premium-priced BPA-free products. These are usually made with the chemical substitute BPS, which current research suggests is even more of a health hazard than BPA. But since BPS is far less studied, it will likely take many years to build a sufficient case for a new ban.

But the true scandal of BPA is that such sagas have been repeated many times. Time and again, synthetic chemicals have been banned or withdrawn only to be replaced by others that are equally harmful, and sometimes are worse. Neonicotinoids, which the International Union for the Conservation of Nature (IUCN) credits with creating a global ecological catastrophe, are modern replacements for long-targeted organophosphate pesticides. Organophosphates had previously supplanted DDT and the other organochlorine pesticides from whose effects many bird species are only now recovering.

So the big and urgent question is this: if chemical bans are ineffective (or worse), what should anyone who wants to protect themselves and everyone else from flame retardants, pesticides, herbicides, endocrine disruptors, plastics and so on—but who doesn’t expect much help from their government or the polluters themselves—do?

What would an effective grassroots strategy for the protection of people and ecosystems from toxic exposures look like?  Ought its overarching goal be a reduction in total population exposures and/or fewer chemical sales? Or should it aim for sweeping bans, such as of entire chemical classes? Or bans on specific usages (e.g. in all food or in all of agriculture)? Or on chemical use in particular geographic locations (e.g in/around all schools)? Or perhaps a better demand would be the dismantling (with or without replacement) of existing regulatory agencies, such as the culpable EPA? Or should chemical homicide be made a statutory crime? Or all of these together? And last, but not least, how can such goals be achieved given the finances and politics of our age?

The first task of chemical campaigning is to strip away the mythologies which currently surround the science of toxicology and the practice of chemical risk assessment. When we do this we find that chemical regulations don’t work. The chief reason, which is easy to demonstrate, is that the elementary experiments performed by toxicologists are incapable of generating predictions of safety that can usefully be applied to other species, or even to the same species when it exists in other environments or if it eats other diets. Numerous scientific experiments have shown this deficiency, and consequently that the most basic element of chemical risk assessment is scientifically invalid. For this reason, and many others too, the protection chemical risk assessments claim to offer is a pretense. As I will show, risk assessment is not a reality, it is a complex illusion.

This diagnosis may seem improbable and also depressing, but instead it reveals promising new political opportunities to end pollution and create a sustainable world. Because even in the world of chemical pollution, the truth can set you free.

The ensuing discussion, it should be noted, makes no significant effort to distinguish human health effects from effects on ecological systems. While these are often treated under separate regulatory jurisdictions, in practice, risks to people and ecosystems are difficult if not impossible to separate.

The story of the toxicological alarms surrounding BPA, which are diverse and scientifically extremely well substantiated, make an excellent starting point for this task.

Ignoring the full toxicity of BPA

According to the scientific literature, exposure to BPA in adulthood has numerous effects. It leads to stem cell and sperm cell defects (humans), prostate cancer (humans), risk of breast cancer (human and rats), blood pressure rises (humans), liver tumours and obesity (humans and mice) (Grun and Blumberg 2009; Bhan et al., 2014; Prins 2014). However, foetuses exposed to BPA suffer from a significantly different spectrum of harms. These range from altered organ development (in monkeys) to food intolerance (in humans) (Ayyanan et al., 2011; Menard 2014; vom Saal et al., 2014). Also in humans, early BPA exposures can lead to effects that are nevertheless delayed until much later in life, including psychiatric, social and behavioural abnormalities indicative of altered brain functions (Braun et al., 2011; Perera et al., 2012; Evans et al., 2014).

The above examples are just a representative handful. They are drawn from a much larger body of at least 200 publications (some have estimated a thousand publications) finding harms from BPA. The sheer quantity of results, the diversity of species tested, of consequences found, and of scientific methodologies used, represent a massive accumulation of scientific evidence that BPA is harmful (reviewed in Vandenberg et al., 2012). The evidence against BPA being safe, in short, is as close to unimpeachable as science can manage.

Nevertheless, such a large evidence base indicates that anti-BPA campaigning has been only partially successful. All the bans and the commercial withdrawals still ignore the implications of some of the most alarming scientific findings of all. For example, bans on baby bottles will not prevent foetal exposure. Nor will they prevent harms that result even from very low doses of BPA.

Ignoring the toxicity of BPS

The chemical most frequently used to make BPA-free products is called BPS. As its name implies, BPS is very similar in chemical structure to BPA (see Fig. 1). However, BPS appears to be absorbed by the human body significantly more readily than BPA and is already detectable in 81% of Americans (Liao et al., 2012).

Research into the toxicology of BPS is still at an early stage, but BPS is now looking likely to be even more toxic than BPA (Rochester and Bolden 2015). Like BPA, BPS has been found to interfere with mammalian hormonal activity. To a greater extent than BPA, BPS alters nerve cell creation in the zebrafish hypothalamus and causes behavioral hyperactivity in exposed zebrafish larvae (Molina-Molina et al., 2013; Kinch et al., 2015). These latter results were observed at the extremely low chemical concentrations of 0.0068uM. This is 1,000-fold lower than the official U.S. levels of acceptable human exposure. The dose was chosen by the researchers since it is the concentration of BPA in the river that passes their laboratory.

Chemical substitutions are business as usual

The substitution of one synthetic chemical for another, wherein the substitute later turns out to be hazardous, is not a new story. Indeed, a great many of the chemicals that environmental campaigners nowadays oppose (such as Monsanto’s best-selling herbicide Roundup) are still considered by many in their industries to be “newer” and “safer” substitutes for chemicals (such as 2,4,5-T) that are no longer widely used.

Thus, when the EU banned the herbicide atrazine, Syngenta replaced it with terbuthylazine. Terbuthylazine is chemically very similar and, according to University of California researcher Tyrone Hayes, it appears to have similar ecological and health effects.

The chemical diacetyl was forced off the market for causing “popcorn lung“. However, it has been largely replaced by dimers and trimers of the same chemical. Unfortunately, the safety of these multimers is highly dubious since it is believed that, in use, they break down into diacetyl.

The Bt pesticides produced inside GMO crops are considered (by farmers and agribusiness) to be safer substitutes for organochlorine, carbamate, and organophosphate insecticides. These chemicals replaced DDT, which was banned in agriculture following Rachel Carson’s Silent Spring. DDT was itself the replacement for lead-arsenate. Many other examples of what are sometimes called regrettable substitutions can be found.

Chemical bans (or often manufacturer withdrawals) that precede such substitutions are nevertheless normally celebrated as campaigning victories. But the chemical manufacturers know that substitution is an ordinary part of business. Because weeds and pests become resistant and patents run out, they are usually looking for substitutes irrespective of any environmental campaigning.

Manufacturers also know that, since approvals and permits initially rely primarily on data supplied by the applicant (and which is often anyway incomplete), problems with safety typically manifest only later, as independent data and practical experience accumulate. Given this current system it is almost inevitable that older (or more widely used) chemicals typically have a dubious safety record while newer ones are considered “safer”.

“Bad actors”: the rotten apple defence in toxicology

In these cycles, of substituting one toxin for another, BPA is likely to become a classic.

Environmental health non-profits become active participants in this toxic treadmill when they implicitly treat certain chemicals as rotten apples. Some even explicitly refer to particular chemicals as “bad actors“. The chemical “bad actor” framing strongly implies that the methods and institutions of chemical regulation are not at fault.

But we can ask the question, in what chemical or biological sense can BPA be termed a bad actor? Is there, for example, a specific explanation for how it slipped through the safety net?

The very short answer to this question is to recall the results noted above: BPA impairs mammalian hormonal and reproductive systems; it disrupts brain function; it impacts stem cell development; it causes obesity and probably cancer; it causes erectile dysfunction. Many hundreds of research papers attest that BPA’s harmful effects are numerous, diverse, prolonged, reproducible and found in many species. In short, they are easy to detect (e.g. vom Saal et al., 2014).

So while hundreds of scientists outside the regulatory loop have found problems, the formal chemical regulatory system has never flagged BPA, even though astonishingly, long before it was thought of as a plastic, BPA first came to the attention of science in specific searches for estrogen-mimicking (i.e. hormone-disrupting) compounds. And despite the overwhelming nature of the published evidence regulators still resist concluding that BPA is a health hazard. And so the clear answer to the “bad actor” question is that there is no special reason why BPA should have slipped through the regulatory process; instead, the case of BPA strongly suggests a dysfunctional regulatory system.

Framing the problem of pollution as being caused a few “bad actor” chemicals is equally inconsistent with the facts in other cases too. Chemical regulatory systems initially approved but have sometimes later banned or restricted (and always under public pressure): atrazine, endosulfan, Roundup (glyphosate), lindane, methyl bromide, methyl iodide, 2,4,5-T, chlorpyrifos, DDT and others. Many other chemicals are strongly implicated as harmful by extensive and compelling independent scientific evidence that has so far not been acted on. And of course, chemical regulators have graduated whole classes of “bad actors”: the organophosphate pesticides, PCBs, organochlorine pesticides, chlorofluorocarbons, neonicotinoids, phthalates, flame retardants, perfluorinated compounds, and so on.

How many bad actors ought it to take before we instead indict the whole show?

Chemical regulation in theory and practice: the limits of toxicology

An alternative approach to judging regulatory systems by their results, is to analyse them directly and assess their internal logic and rigour. Thus one can ask what is known about the technical limitations of toxicology and the overall scientific rigour of chemical risk assessment? And, secondly, one can direct attention to the social and institutional practices of chemical regulation. Are chemical risk assessments, for example, being applied by competent and well-intentioned institutions?

The technical limitations of chemical risk assessment are rarely discussed in detail (but see Buonsante et al., 2014). A full discussion would be lengthy, but some of the most important limitations are outlined in the paragraphs below.

The standard assays of toxicology involve the administration (usually oral feeding) of chemicals in short term tests of up to 90 days to defined strains of organisms (most often rats or mice). These test organisms are of a specified age and are fed standardised diets.

The results are then extrapolated to other doses, other age groups and other environments. Such experiments are used to create estimates of harm. Together with estimates of exposure they form the essence of chemical risk assessment. When specific chemicals are flagged as being worthy of further interest, other techniques may be brought to bear. These may include epidemiology, cell culture experiments, and biological modeling, but the basis of risk assessment is always the estimation of exposure and the estimation of harm. To say that both estimates are prone to error, however, is an understatement.

Part I: limits to estimating chemical exposures

Fifty years ago no one knew that many synthetic chemicals would evaporate at the equator and condense at the poles, from where they would enter polar ecosystems. Neither did scientists appreciate that all synthetic fat-soluble compounds that were sufficiently long-lived would bio-accumulate as they rose up the food chain and thus reach concentrations inside organisms sometimes many millions of times above background levels. Nor until recently was it understood that sea creatures such as fish and corals would become major consumers of the plastic particles flushed into rivers. These misunderstandings are all examples of historic errors in estimating real world exposures to toxic substances.

A general and broad limitation of these estimates is that real world exposures are very complex. For instance, commercial chemicals are often impure or not well defined. Thus PVC plastics are a complex mixture of polymers and may be further mixed with Cadmium or Lead (in varied concentrations). One implication of this is that it is impossible for experiments contributing to risk assessment to be “realistic”. The reason is that actual exposures are always unique to individual organisms and vary enormously in their magnitude, duration, variability, and speed of onset, all of which influence the harm they cause. Whose specific reality would realism mimic?

Additionally, many regulatory decisions do not recognise that exposures to individual chemicals typically come from multiple sources. This failing is often revealed following major accidents or contamination events. Regulatory agencies will assert that actual accident-related doses do not exceed safe limits. However, such statements usually ignore that, because regulations function in effect as permits to pollute, many affected people may already be receiving significant exposures for that chemical prior to the accident.

Returning to the specific case of BPA, no one appreciated until 2013 that the main route of exposure to BPA in mammals is absorption through the mouth and not the gut. The mouth is an exposure route whose veinous blood supply bypasses the liver, and this allows BPA to circulate unmetabolised in the bloodstream (Gayrard et al. 2013). Before this was known, many toxicologists explicitly denied the plausibility of measurements showing high BPA concentrations in human blood. They had assumed that BPA was absorbed via the gut and rapidly degraded in the liver.

Part II: limits to estimating harms

Similarly significant obstacles are faced in estimating harm. Many of these obstacles originate from the obvious fact that organisms and ecosystems are enormously biologically diverse.

The solution adopted by chemical risk assessment is to extrapolate. Extrapolation allows the results of one or a few experiments to “cover” other species and other environmental conditions.

Most of the assumptions required for such extrapolations, however, have never been scientifically validated. Lack of validation is most obvious for species not yet discovered or those that are endangered. But in other cases they are actively known to be invalid (e.g. Seok et al., 2013).

For example, in their responses to specific chemicals, rats often do not extrapolate to humans. Indeed, they often do not extrapolate even to other rats. Thus individual strains of rats respond differently (which of course is why they get used); but also young and old rats give different responses. So do male and female rats (vom Saal et al., 2014). So too do rats fed non-standard diets (Mainigi and Campbell, 1981)

Even more extreme extrapolations are employed in ecological toxicology. For example, data on adult honey bees is typically extrapolated to every stage of the bee life cycle, to all other bee species, and sometimes to all pollinators, without the experimenters citing any supporting evidence. Such extrapolations may seem absurd but they are the primary basis of the claim that chemical risk assessment is comprehensive.

There are many other limits to estimating harm. Until it was too late, scientists were not aware that a human with an eighty-year lifespan could have a window of vulnerability to a specific chemical as short as four days. Neither was it known that the effects of chemicals could be strongly influenced by the time of day they are ingested.

Another crucially important limitation is that, for budgetary and practical reasons, toxicologists necessarily focus on a limited number of specific “endpoints”. An endpoint is whatever characteristic the experimenter chooses to measure. Typical endpoints are death (mortality), cancers, organism weight, and organ weights; but endpoints can even be more subtle measures like neurotoxicity. There is a whole politics associated with the choice of endpoints, which reflects their importance in toxicology, including allegations that endpoints are sometimes chosen for their insensitivity rather than their sensitivity; but the inescapable point is that no matter what endpoints are chosen, there is a much vaster universe of unmeasured endpoints. These typically include: learning defects, immune dysfunction, reproductive dysfunction, multigenerational effects, and so on. Ultimately, most potential harms don’t get measured by toxicologists and so are missing from risk assessments.

Another example of the difficulty of estimating real life harms is that organisms are exposed to mixtures of toxins (Goodson et al., 2015). The question of toxin mixtures is extremely important (Kortenkamp, 2014). All real life chemical exposures occur in combinations, either because of previous exposure to pollutants or because of the presence of natural toxins. Many commercial products moreover, such as pesticides, are only available as formulations (i.e. mixtures) whose principal purpose is to enhance the potency of the product. Risk assessments, however, just test the “active ingredient” alone (Richard et al., 2005).

Consider too that all estimates of harm depend fundamentally on the assumption of a linear (or at least simple) dose-response relationship for the effect of each chemical. This is necessary to estimate harms of doses that are higher, lower, or even in between tested doses. The assumption of a linear response is rarely tested, yet for numerous toxins (notably endocrine disrupting chemicals) a linear dose-response relationship has been disproven. Thus the question for any risk assessment is whether the assumption is reliable for the novel compound under review (reviewed in Vandenberg et al., 2012).

Replacing doubts with false certainty

To summarise, the process of chemical risk assessment relies on estimating real world exposures and their potential to cause harm by extrapolating from one or a few simple laboratory experiments. The resulting estimates come with enormous uncertainty. In many cases the results have been extensively critiqued and shown to be either dubious or actively improbable (Chandrasekera and Pippin, 2013). Yet extrapolation continues—even though we know that the various errors must multiply—because the alternative is to actually measure these different species, using different mixtures and under different circumstances. Given the challenges this would entail, the continued reliance on simplistic assumptions is understandable.

Nevertheless, one might  have thought that such important limitations and assumptions would be frequently noted as caveats to risk assessments. They should be, but they are not. Following the UK’s traumatically disastrous outbreak of BSE (mad cow disease) in the 1980s, during which most of the UK population was exposed to infectious prions following highly questionable scientific advice, this exact recommendation was made in the Phillips report. Lord Phillips proposed that such caveats should be specifically explained to non-scientific recipients of scientific advice. In practice however, Phillips changed nothing.

When an unusual scientific document does discuss the limitations of chemical risk assessment (such as this description of the failure of interactions between pesticides to extrapolate between closely related species), it rapidly becomes obvious just how much the knowledge and understanding available to us are dwarfed by actual biological and system complexities. As any biologist ought to expect, the errors multiplied and the standard assumptions of risk assessment were overwhelmed even by ordinary life situations.

For good reason many scientific experts are therefore concerned about the number and quantity of man-made chemicals in our bodies. Recently, the International Federation of Gynecology and Obstetrics linked chemical exposure to the emergence of new diseases and disorders. They specifically mentioned obesity, diabetes, hypospadias and reproductive dysfunction and noted: “The global health and economic burden related to toxic environmental chemicals is in excess of millions of deaths” (Di Renzo et al., 2015). The Federation acknowledged this to be an underestimate. Nor does it count disabilities.

Conflicts of interest in chemical risk assessment

In addition to the technical difficulties, there is also the problem that the scientists who produce scientific knowledge often have financial (and other) conflicts of interest. Conflicts, we know, lead to biases that impact on science well before it is incorporated into risk assessment (e.g. Lesser et al., 2007).

A fascinating example of apparent unconscious bias comes from a recent survey of scientific publications on the non-target effects of pesticidal GMO (Bt) crops in outdoor experiments. It was commissioned by the Dutch government (COGEM 2014). The report observed that researchers who found negative consequences of GMO Bt crops were disregarding their own findings, even when these were statistically significant. Even more interesting to the Dutch authors, was that the rationales offered for doing so were oftentimes illogical. Typically, researchers were using experimental methods specialised for detecting ecotoxicological effects that were “transient or local”, but when such effects were found the researchers were dismissing the significance of their own results for being either transient or local. The COGEM report represented prima facie evidence that researchers within a whole academic discipline were avoiding conclusions that would throw doubt on the wisdom of using GMO Bt crops. Apparently the Bt researchers had a prior ideological commitment to finding no harm of the kind that scientists are supposed to not have.

Corporate capture and institutional dysfunctionality

Chemical regulation occurs primarily within a relatively small number of governmental or “independent” regulatory institutions.

Of these, the United States Environmental Protection Agency (EPA) is the most prominent and widely imitated example. The EPA has a variety of institutional and procedural defects that prevent it being an effective regulator. Perhaps the best known of these is to allow self-interested chemical corporations to conduct the experiments and provide the data for risk assessment. This lets them summarise (or even lie about) the results. As was once pointed out by Melvin Reuber, former EPA consultant, it is