The assumption of a threshold is false in risk assessments for endocrine disrupting chemicals (EDCs) when endogenous hormones being disrupted are already above any possible threshold: a policy failure by the US FDA

Abstract

The US FDA continues to assure the public that if they are exposed to an endocrine disrupting chemical (EDC), there is a threshold daily exposure level below which everyone is safe throughout the lifespan (the Acceptable or Tolerable Daily Dose). This assurance is false when the endogenous hormone systems being disrupted by EDCs are already above a threshold for producing adverse effects or there are cumulative effects of mixtures of similar chemicals. Decades of published experimental research, and multiple mathematical analyses, demonstrate the absence of a threshold (safe) dose for hormones, hormonal drugs or EDCs. However, this entire literature has been rejected in decision making by the US FDA food safety division. We review experiments directly falsifying the threshold hypothesis for hormones, drugs and EDCs. Our analysis includes application to hormone and EDC experimental data sets of the Michaelis–Menten (MM) equation, which explains mathematically hormone (or EDC)-receptor binding that mediates normal as well as adverse effects at extremely low exposures. The MM equation regresses to zero at zero hormone or EDC dose, i.e., there is no threshold dose. We derive the MM equation in Supplemental Materials for those interested in the math. We also identify that hormone and EDC potency is the consequence of a number of cell-specific mechanisms not included in the MM equation. We begin by reviewing the history of the now falsified assumption that a threshold exists for EDCs that are generally classified as systemic toxicants. In particular, the threshold hypothesis is problematic during fetus/infant/child development during which homeostatic defense systems have not yet fully developed, rendering them largely defenseless against the disruptive effects of EDCs. Published findings implicate EDC exposures at very low environmentally relevant doses with every non-communicable disease, which are all increasing in incidence on a global scale. We identify here the empirical and endocrinological evidence falsifying the determination by chemical regulatory agencies that there are safe daily exposure levels for EDCs below an estimated threshold that is not actually experimentally examined. These findings indicate that current human exposures to EDCs are harming the general population as well as causing widespread environmental harm.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12940-026-01265-z.

Background

Overview of the review

In risk assessments for non-carcinogenic chemicals (systemic toxicants) conducted by regulatory agency toxicologists in the USA, EU, UK, Asia, and countries in other regions, a core assumption is that there is a threshold dose below which there are no adverse effects [1–5]. The “safe” dose that is calculated is based on the threshold assumption and is presented to the public as a daily exposure that is safe throughout the lifespan for the general population. This estimated safe threshold dose is referred to in the European Union (EU) as the Tolerable Daily Intake (TDI) and as the Acceptable Daily Intake (ADI) by the US Food and Drug Administration (FDA), or the Reference Dose (RfD) by the US Environmental Protection Agency (EPA).

However, the assumption that there are thesholds for a specific class of environmental chemicals, endocrine disrupting chemicals (EDCs), has been demonstrated to be false by decades of research in endocrinology and toxicology showing that there is no threshold or safe daily exposure below which this special category of systemic toxicants, EDCs, do not cause adverse effects. Importantly, the prediction of a safe threshold dose is not accepted to be an assumption but rather as fact by the toxicologists who work for chemical corporations or US regulatory agencies. The existence of thresholds for EDCs continues to be an accepted assumption by US regulatory agency administrators, although the threshold issue has been the subject of debate for decades [6–8], even within the FDA [9, 10].

EDCs represent a unique category of toxic environmental chemicals, since exogenous EDCs interfere with endogenous hormone action, and the hormones being disrupted are critical for ongoing normal development and then coordination of normal function in adults [11, 12]. The US EPA defined an EDC as “an exogenous agent that interferes with synthesis, secretion, transport, metabolism, binding action, or elimination of natural blood-borne hormones that are present in the body and are responsible for homeostasis, reproduction, and developmental processes” [11].

We review here evidence that falsifies the threshold assumption for EDCs that bind to cellular receptors for endogenous hormones that are already above threshold levels of activity. However, even if there is not an endogenous hormone providing an additive effect to a chemical, exposure to multiple chemicals (referred to as mixture or cumulative effects) will result in the same flaw in the FDA’s chemical regulation paradigm, namely, that a focus on the activity of a single chemical and not the totality of factors (both endogenous and exogenous) impacting a response will lead to false estimates of safety. Current risk-based approaches in the US do not address the critical issue that for EDCs, the assumption that a level of risk can be calculated that is protective of the general public or for those exposed occupationally ignores the data presented in our review. We describe that for EDCs in household products, such as baby food and packaging, estimating risk [13, 14] should not be necessary, and any risk is unacceptable.

In this review we focus on the US chemical regulatory system and specifically on the FDA food safety division, since the US EPA is currently being dismembered (e.g., there has been a complete elimination of the EPA Office of Research and Development; ORD). Also, in the EU the food and food packaging regulatory system generally uses approaches that are different from the approaches used by the Human Foods Program (formerly the Center for Food Safety and Applied Nutrition; CFSAN) at the FDA. For example, EU regulatory agencies apply the Precautionary Principle and assume that chemicals need to be proven to be safe, as opposed to FDA regulators that consider chemicals used in food, food packaging and household products to be safe until proven to be harmful (inexplicably, this is opposite to the FDA’s drug division approach to regulating hormonally active drugs). However, the regulatory agency testing system for chemicals in food packaging in the EU is still primarily focused on genotoxicity, not endocrine disruption [15].

We thus do not attempt here to compare chemical regulatory guidelines in the EU vs. the US, which is quite complicated and would require a separate review. What is critical for the present review is that all regulatory agencies in the US and EU (as well as other countries in the Organization for Economic Co-operation and Development; OECD) currently accept that the assumption of a threshold is valid for EDCs [8], although this does not rule out that some EDCs might also be mutagens [16, 17].

We will focus here on estrogenic EDCs and use the manmade chemical bisphenol A (BPA) and the estrogenic drug diethylstilbestrol (DES), as well as the endogenous hormone estradiol, as our primary examples. Our focus on estrogens is based on the extensive published knowledge regarding estrogen action over more than six decades and that estrogenic EDCs have continued to be a major focus of EDC research over the last 30 years [8, 12, 18, 19].

Our focus on BPA is based on the fact that: 1) BPA is one of the highest volume chemicals in global production and is used as an additive in a myriad of products in addition to being the monomer used to produce polycarbonate plastic, 2) there are thousands of published studies on the adverse effects of BPA on the environment and animal and human health, and 3) there was a massive error in the initial assessment by the US EPA and US FDA of a safe threshold dose for BPA relative to the most recent BPA risk assessment conducted in the EU [8]. Critically, the presence of a background endogenous hormone (estradiol) binding to the same receptors as an EDC (such as BPA), as well as the numerous other ubiquitous estrogenic chemicals to which humans are being exposed simultaneously, has not been addressed in risk assessments conducted by US regulatory agencies. US regulatory agencies conduct risk assessments for one chemical at a time, as opposed to classes of structurally and functionally related chemicals that interact with endogenous hormones [8, 10, 20–22].

Our view is that methods used in risk assessments for systemic poisons (such as cyanide or some heavy metals), which have traditionally been the basis for FDA assumptions used in chemical risk assessments, are inappropriate for EDCs, particularly regarding the issues of estimating safety, dose–response relationships and the importance of life stage [11, 19, 23]. Here we review data from studies of EDCs that have been the focus of scientists interested in disruption of endocrine homeostasis [8, 12, 16, 24, 25].

What is an assumption?

An assumption is a guess rather than an observation based on actual experimental data. This is referred to as inductive logic, where you take a specific issue and generate a general theory that “true believers” refuse to abandon, a process described by Kuhn in his analysis of resistance to paradigm shifts by those heavily invested in the old paradigm being challenged [26]. Experimental biologists take the opposite approach, called deductive logic, where a hypothesis is posed and then subjected to experiments to determine whether the hypothesis is or is not falsified. This process requires experiments that include both positive as well as the negative controls [27], which has been a controversial issue in toxicological studies conducted for regulatory purposes [28]. Deductive reasoning is the modern scientific method that involves proposing a hypothesis that is testable by appropriate experimentation.

When a hypothesis is falsified by experimental evidence, the result for a scientist would be that the hypothesis is rejected and a new hypothesis incorporating the new experimental data is proposed. New hypotheses that generate new experiments can thus be supported (based on what is specifically subjected to experimental analysis) or falsified (not supported by the data). In other words, a hypothesis can never be proven to be correct beyond what is narrowly experimentally determined, but it can be clearly falsified, requiring revision of the hypothesis and incorporation of the new data. The threshold hypothesis for EDCs has been falsified based on both mathematical [6] and empirical evidence [9, 10, 13, 16], but regulatory agency toxicologists at the FDA in the US have not considered these findings in their decision making regarding chemical safety. A recent example of this was the FDA decision that current exposures to BPA are safe [29–31].

Industry abuse of the generally regarded as safe (GRAS) loophole

Importantly, only a very small percent of the chemicals in products have actually been subjected to toxicological testing. For example, most chemicals present in processed food have not been tested for health effects because they have been allowed to be declared “generally regarded as safe (GRAS)” by the chemical manufacturers without the FDA engaging in regulatory oversight [32]. This is still the regulatory standard used by the FDA, although the 1958 Food Additive Amendment Law was intended to require FDA approval of safety before a new food additive could be used. The GRAS exemption was thus initially meant to apply to items such as table salt, not manmade chemicals that end up in food that chemical companies are allowed to state are safe with essentially no regulatory oversight [33].

In reality, beginning after passage of the Food Additive Amendment Act in 1958, the FDA allowed tens of thousands of chemicals to continue being used because they were already in products, although without any safety testing having been conducted. The GRAS exemption has thus allowed thousands of previously untested chemicals to be added to food and food packaging in the US. Subsequently, the US FDA gave industry the authority to self-determine chemical safety [34]. While the EU is still struggling with this issue, it is far ahead of the US FDA [15, 35]. A critical related issue is the adverse consequence for the public health of the FDA accepting that there are arbitrary safe exposure levels for chemicals that they declare safe without any experimental data regarding whether these assumed “safe levels” are, in fact, safe.

History of the threshold assumption

Paraselsus and “the dose makes the poison” fallacy

The core risk assessment assumptions regarding thresholds and dose–response relationships used by regulatory toxicologists are based on a 500-year old hypothesis proposed by Paracelsus (a pseudonym) in the early 1500s to describe the behavior of poisons. What Paracelsus actually wrote in about 1537 was: “All things are poison, and nothing is without poison; the dosage alone makes it so a thing is not a poison”. This has been revised and has become toxicological dogma as not meaning that all chemicals are poisons, but instead that only “the dose makes the poison” [5], which is a complete inversion of the original Paracelsus quote. A core risk assessment assumption is that this means that for poisons, adverse effects always increase monotonically with dose, and that there is always a threshold dose below which there is no adverse (poisonous) effect.

Belief in a threshold dose is coupled with rejection of the published adverse effects of EDCs at very low doses [23] because the same effects that occur at low doses were not seen to increase monotonically over a wide (e.g., 10,000-fold) dose range in a regulatory guideline study conducted by the FDA [30]; this exceptionally wide range of doses in a collaborative project was dictated by the National Institute of Environmental Health Sciences (NIEHS) [36]. There have been some industry-backed ideas that have been published proposing the bizarre notion that consuming a little bit of extremely toxic chemicals, such as dioxins (TCDD), is actually good for you [37]. Needless to say, we, and others, do not think that this is a good idea and certainly should not be applied to estrogenic or other EDCs [38, 39].

In fact, Homburg and Vaupel [40], whose expertise is on the history of toxicology, have argued that it was not until the middle of the 1900 s that industrial toxicologists rediscovered the writings concerning poisons by Paracelsus. This led to misinterpreting his use of the term “dose” and the rephrasing of his writing to read “the dose makes the poison”. Homburg and Vaupel argue that this was actually not consistent with Paracelsus’ original statement that all things are poison. This has to be viewed from the perspective that Paracelsus was writing in the early 1500s during the time that chemists were alchemists who had a very different set of assumptions about chemicals, namely, that all matter was composed of four elements (earth, air, fire, and water) as well as in the belief in transmutation (e.g., lead could be transmutated into gold).

Paracelsus's writings about doses of poisons, when viewed from the perspective that he was an alchemist, meant something completely different from the way they are being interpreted today by twenty-first century toxicologists. Dose in 1537 was a purely qualitative concept, in which terms such as larger or smaller played no role. The ‘right dose’, for instance, referred to a harmonious equilibrium of forces in nature. The now famous Paracelsus quote was rarely cited before 1900, and it would be another fifty years until it started to be used frequently by twentieth century industrial toxicologists and then regulators [40].

Importantly, Paracelsus lived prior to the ability of chemists to accurately quantify chemicals and prior to the scientific advances that have revealed the exquisite sensitivity of humans, animals and plants to hormonal signals and EDCs (the concept of hormones did not exist in the sixteenth century). What is strange is that “the dose makes the poison” concept in the mid 1900 s, that is still used by twenty-first century regulatory toxicologists, occurred at the same time that endocrinologists were developing the tools to identify and measure hormones [41] and determine the mechanisms of hormone action via binding to hormone receptors [42, 43]. This misinterpretation by industrial toxicologists of the publication by Paracelsus almost 500 years ago resulted in the threshold hypothesis that has become the basis for the assurance by regulatory agencies worldwide that there are exposure levels for chemicals below which risk is negligible for the general population [44].

During the 1900 s industrial toxicologists were thus instrumental in creating a regulatory system based on the minimum lethal dose and the maximum exposure limits on chemicals. For a while toxicologists used a lethal dose 50 (LD₅₀) that resulted in 50% of animals being killed in studies for risk assessments. A problem was that the public found the lethal dose (animal death) criterion to be objectional, and in the 1980 s the maximum tolerated dose (MTD) was used as the starting point in dose selection. After testing a few high doses, either the No Adverse effect Level (NOAEL) or Lowest Adverse Effect Level (LOAEL) was calculated. Use of the LOAEL meant that the lowest dose tested caused an adverse effect. The NOAEL or LOAEL were used as points of departure for calculating a threshold dose. For BPA the LOAEL of 50 mg/kg/day was based on a decrease in body weight, since only high doses were tested and 50 mg/kg/day was the lowest dose tested. This LOAEL was thus chosen as the point of departure for estimating via linear extrapolation a threshold (reference or acceptable daily) dose of 50 µg/kg/day by dividing the LOAEL by 1000; i.e., three tenfold “uncertainty” factors: tenfold for using a short-term study to estimate effects of chronic exposure in people, tenfold for extrapolating from animals to humans, and tenfold to account for variability among people [45].

Throughout the latter part of the twentieth century, industries promoted the idea that industrial work was not dangerous if exposure to chemicals was below an amount that was actually set at an arbitrary, but to chemical regulators, what seemed like a low level. Environmental estrogenic chemicals such as BPA have been shown cause effects in pituitary cells at extremely low doses (10^–14 M or 0.0023 pg/mL BPA) [46]. The exquisite sensitivity of hormone receptors to vanishingly small concentrations of EDCs has not been appreciated by regulatory toxicologists.

The importance to industrial toxicologists of setting an arbitrary exposure level that was labeled as being too low to be of concern, was to eliminate the need to subject chemicals to experimental safety testing. The intent was to avoid the possibility of any regulation of their manufacturing procedures. The idea of safe limits (thresholds) rested on the assumption that organisms (including fetuses and babies) as well as ecosystems, such as rivers, had “self-cleaning capacity” that would make the discharged toxic chemicals harmless. This was presented to the public concerned with dumping pollutants into rivers as “dilution is the solution” that was used to justify unsafe toxic chemicals being discharged into rivers [40].

Systemic toxicants vs. carcinogenic chemicals: the Delaney clause

Prior to the twentieth century a common view regarding poisons was that they could kill people at very low doses, exactly the opposite of the position of regulatory agencies today. The exception to the current rejection of evidence that EDCs can act at very “low doses” concerned carcinogenic chemicals due to the “Delaney Clause” in the Food Additive Amendment Act in 1958. The Delaney clause prohibited the use of any food additive found to cause cancer in animals or humans, based on the assumption that a single mutagenic event could potentially end up leading to development of cancer. Due to the Delaney Clause banning any carcinogenic chemical for use in food or food additives, a big distinction in regulatory toxicology was the assumption of an absence of thresholds for carcinogens, but the presence of thresholds for all systemic (non-carcinogenic) chemicals [16].

EDCs are generally classified as systemic toxicants for regulatory purposes. However, if an EDC was found to be carcinogenic in a study that was recognized by the FDA as following regulatory guidelines, the assumption of a threshold would presumably not have to be applied. After passage by the US Congress of the Food Quality Protection Act (FQPA) in 1996, the FDA has applied a risk-based (risk–benefit) standard to regulating carcinogenic chemicals (e.g., acceptable levels of risk could be determined), and thus the FDA abandoned the zero-risk standard for carcinogenic chemicals set by the Delaney Clause in the 1958 Food Additive Amendment Act by the US Congress [44].

An interesting contrast to the prior regulatory process is that the FDA recently announced that it was “repealing the color additive regulations that permit the use of Red Dye 3 [erythrosine] in foods (including dietary supplements) and in ingested drugs” because of evidence that “Red Dye 3 can induce thyroid tumors in male rats”. Thyroid tumors in male rats were assessed to result from the perturbation of the hypothalamic-pituitary-thyroid hormonal axis. The FDA concluded that the perturbations were initiated by an increase in circulating thyroid stimulating hormone (TSH), causing thyroid follicular hyperplasia, which has the potential to develop into thyroid tumors [47]. However, the experimental data show that there is no threshold predicted for this EDC adverse effect [10]. This is potentially a paradigm-changing decision by the FDA. In the US prior to 1996, carcinogenic chemicals were supposed to be banned from food or food additives at any dose, while thousands of presumed non-carcinogenic chemicals ended up being considered acceptable for use in products based on the “generally regarded as safe” loophole [32, 33, 40], although this was not the intention of the 1958 Food Additives Amendment law.

The “Reasonable certainty” standard

The 1958 Food Additives Ammendment Law required the FDA to define "safe" for food additives as a "reasonable certainty in the minds of competent scientists that the substance is not harmful under the intended conditions of use", which was intended to encompass a variety of harm beyond acute toxicity (21 C.F.R. § 170.3(i)). Instead, the FDA turned over control of food additive safety assessments to panels consisting of people being paid by the corporations making the products, in direct violation of the 1958 law [32, 33]. For example, the National Institute of Environmental Health Sciences (NIEHS) invited 38 internationally recognized experts to a meeting to assess the hazards posed by bisphenol A (BPA). In 2007 these scientific experts published a consensus statement: “Chapel Hill bisphenol A expert panel consensus statement: integration of mechanisms, effects in animals and potential to impact human health at current levels of exposure” [48].

The consensus conclusion of all experts divided into four separate panels at this NIH-sponsored meeting on BPA was that “The published scientific literature on human and animal exposure to low doses of BPA in relation to in vitro mechanistic studies reveals that human exposure to BPA is within the range that is predicted to be biologically active in over 95% of people sampled. The wide range of adverse effects of low doses of BPA in laboratory animals exposed both during development and in adulthood is a great cause for concern with regard to the potential for similar adverse effects in humans.” [48]. Importantly, this subsequently led to a number of epidemiological studies showing that low dose effects of BPA in wildlife and laboratory animals were also related to BPA exposure in human populations [25]. This consensus statement was accompanied by 5 peer-reviewed publications by conference participants reviewing the published evidence for human exposure, evidence from in vitro, wildlife and laboratory animal studies, as well as evidence for carcinogenesic effects [49–53]. The next year the FDA simply ignored all of this information and declared BPA to be safe, and did not change the ADI (safe dose) from the 50 µg/kg/day dose established in 1988 by the EPA [45]. However, it was required under the 1958 law based on the “reasonable of certainty of no harm” standard for the FDA to consider the published opinion of scientific experts in their 2008 risk assessment for BPA [54].

More than 15-years later, the presumed safe daily dose for BPA was reduced by the European Food Safety Authority (EFSA) to 0.2 ng/kg bw/day [8]. This was a huge (250,000-fold) decrease compared to the presumed safe daily exposure dose of 50 µg/kg/day (50,000 ng/kg bw/day) still predicted by the EPA and FDA in the US based only on very high dose studies conducted in the 1980s [45]. The basis for this huge discrepancy is explained by US FDA rejecting all research conducted by independent academic scientists, in contrast to the EU approach of conducting a systematic review of all recent BPA publications in its BPA risk assessment [8]. However, we note that both EU and US estimates are based on the premise that there is a threshold dose for BPA below which exposure is safe. Because the EU TDI for BPA is now 0.2 ng/kg/day, BPA is now effectively banned from food and beverage packaging in the EU because the threshold dose is set so low that it cannot be measured.

Sadly, in the US in spite of overwhelming evidence that BPA is implicated in causing an incredible number of adverse effects, the FDA has refused to take any regulatory action [8, 25, 35]. For the tens-of-thousands of chemicals in food for which there are not the thousands of BPA publications showing adverse effects, there is little hope for regulatory action by the FDA.

The 1958 Food Additive Amendment law stipulated that regulators should also take into consideration “the cumulative effect of additives in the diet of man or animals, taking into account any chemically or pharmacologically related substance or substances in such diet.” The importance of cumulative effects has been extensively addressed [55] but has not been incorporated into safety assessments by in the US by the FDA. There have been proposals to harmonize the FDA’s drug and food division regulatory guidelines [34]. The drug division at the FDA (The Center for Drug Evaluation and Research; CDER) requires drugs to be tested for efficacy and a determination of the level of risk vs. benefits of the drug along with reporting of adverse effects, interactions with other drugs and potential cumulative effects. In sharp contrast, the food safety division at the FDA considers all chemicals in food to be safe until they are proven to cause harm, although as indicated above, most chemicals in food have never undergone any testing for safety.

The position of the FDA is based on the assumption “that reasonable assurance of safety can be given even in the absence of chemical-specific toxicity data, providing that the intake is sufficiently low, i.e. that an exposure level can be defined below which the FDA has determined that there is no significant risk to human health” [56]. It is worth reiterating that the estimated “negligible risk” dose is arbitrary [44] and is not subjected to testing for safety because it is accepted as fact rather than as only an assumption [21]. The FDA does not inform the US public that the threshold levels that the FDA declares safe for food additives are in most cases not based on any data but on the misunderstanding of a prediction about the toxicity of poisons posed about 500-years ago [8, 23, 44, 56]. This has huge health implications because the US, that spends the most amount of money on health care in the world, has the highest rates of chronic diseases of any developed nation [57].

The most critical issue is that there should have been a realization by FDA regulatory toxicologists that endogenous estrogen-mediated responses occur at doses between a thousand to over a million-fold lower than their threshold “low”, parts-per-billion (ppb), estimated safe dose levels for BPA. This type of regulatory failure should not have happened. The lack of understanding of the principles of hormone action was not just limited to the FDA, but was also expressed at the Health and Safety Executive, Chemicals Regulation Division in the UK [3]; however, regulatory decisions in the UK are no longer coupled to those in the EU after Brexit.

In the UK the use of this arbitrary “safe” exposure was referred to as a “threshold of adversity” [3]. This is highly problematic in that the issue of what is “adverse” depends on what apical outcome is examined, the sensitivity of the assay being used, the timing of exposure, age of animals at investigation, and dose levels. These factors determine the estimation of the potency and thus risk of an adverse effect due to exposure to a chemical. Figure 1 shows that determining whether a single chemical causes adverse effects can vary markedly based on the apical outcome examined. A huge problem is that individual chemicals are tested as if there are no other chemicals with similar activity that a person might also be exposed to.

Fig. 1.

Potency is the measure of a chemical’s ability to have a measurable effect that is deemed adverse, which can vary dramatically depending on specific cells, tissues, or organs chosen for examination [25]. Risk assessment decisions regarding the potency of EDCs such as bisphenol A (BPA) not only depend on which organs are measured but the sensitivity of the assays being used. From: [58] with permission

It is reasonable to expect that regulatory agencies will use the most sensitive outcome in the most relevant and sensitive model system to regulate chemicals to which people are exposed, rather than use model animals based on “convenience and familiarity” just because it is the animal model that the testing facility has previously used [59]. This proposal by members of the US National Toxicology Program (NTP), unfortunately, has not been followed [60–62].

Rejection by the FDA of academic research on EDCs

For the relative few chemicals that are subjected to any hazard testing, regulatory guideline studies are often used exclusively by regulatory agency toxicologists in chemical risk assessments conducted by the FDA [30, 54]. These guideline studies, conducted using Good Laboratory Practice (GLP) protocols, are based on insensitive methods and expensive, detailed record keeping developed in the last century [61, 63]. Guideline GLP studies involve using very high doses, traditional histopathology (hematoxylin and eosin staining of sections), and organ weights. For example, Stump et al. [64] conducted a neurotoxicological study of BPA funded by BPA manufacturers that followed outdated US regulatory guidelines and used brain weight as a primary outcome for determining effects of BPA on the brain in rats. Stump et al. concluded that the No Adverse Effect Level (NOAEL) for effects of BPA on the brain was a dose of 164,000 µg/kg/day during gestation. This NOAEL dose was over 65,000-fold higher than the Lowest Adverse Effect Level (LOAEL) in a FDA and NIEHS funded collaborative study with the same rat strain that involved assays using twenty-first century approaches to examine brain function, and adverse effects on the brain were reported at 2.5 µg/kg/day, which was the lowest dose tested in the study [31]; there has not been a NOAEL established for BPA if the totality of the published findings is examined.

Guideline approaches are thus inappropriate for determining adverse effects due to exposure to low doses of EDCs to which people throughout the world are exposed on a daily basis. In fact, the FDA’s draft 2008 risk assessment for BPA (that rejected over 1000 academic published studies and was based entirely on two plastic industry funded studies) was criticized by the FDA Science Review Subcommittee that stated: “The draft FDA report does not articulate reasonable and appropriate scientific support for the criteria applied to select data for use in the assessment. Specifically, the Subcommittee does not agree that the large number of non-GLP [Good Laboratory Practice] studies should be excluded from use in the safety assessment.” [65]. The FDA subsequently discarded its draft 2008 risk assessment for BPA and in 2010 released a statement about BPA that concluded that: “FDA shares the perspective of the National Toxicology Program that recent studies provide reason for some concern about the potential effects” [66, 67]. However, the FDA still would not state that exposure to BPA was unsafe at current exposure levels.

It was thus stunning that 10 years after the rebuke in 2008 by their science advisory review panel regarding having rejected without explanation all research findings by independent academic scientists, the FDA once again rejected all findings about BPA from academic scientists [29]. Specifically, the FDA participated in a collaborative program referred to as “The Consortium Linking Academic and Regulatory Insights on BPA Toxicity; CLARITY-BPA” [36]. This research program involved an estimated 50 million dollar collaborative study of BPA by scientists at the FDA, the NTP, and the National Institute of Environmental Health Sciences (NIEHS). NIEHS funded academic scientists who submitted competitive grant proposals to compare their state-of-the-art research approaches [31] with the traditional guideline study results obtained by FDA scientists [30]. In violation of the contract all collaborators signed, the FDA issued a statement, prior to peer review, that based only on their guideline studies, they could assure the public that the use of BPA in food and beverage packaging was safe [29]. However, the results from dozens of endpoints in the studies conducted by independent academic collaborators using tissues from the same rats used by the FDA, as well as the FDA’s own data, contradicted this conclusion by FDA administrators [31].

Given the global nature of commerce, it is likely that even without any regulatory action by the FDA, the use of BPA in food packaging in the US will decline due to the recent dramatic reduction in the threshold dose in the EU [8]. There has also been regulatory action by state agencies (e.g., BPA is listed under Proposition 65 as a developmental and reproductive toxicant in California).

Low doses of EDCs and chronic diseases

This issue is of immense importance because the threshold hypothesis has resulted in US regulatory agencies rejecting the overwhelming evidence of adverse effects due to exposure to the FDA’s estimated threshold doses of EDCs. This is particularly problematic during critical periods in fetal/neonatal organ and physiological system development when an EDC can permanently redirect gene function through epigenetic processes, with life-altering disease consequences [68]. Adverse effects of developmental exposure to EDCs occur at the generally low levels reported to be relevant to exposures in the USA, EU, Asia [52, 62, 69] and elsewhere. Effects of these exposures include a predisposition to chronic endocrine-related diseases: neuro-behavioral; cardiovascular and other metabolic diseases, such as obesity, diabetes, liver and kidney disease; reproductive disorders, such a subfertility/infertility, as well as reproductive system cancers; immune system diseases and other disorders [11, 12, 70–72].

Critical issues relevant to the threshold assumption

Nonmonotonic dose–response curves

The threshold assumption is intimately coupled to a second assumption, namely, that all dose-responses are monotonic; this assumption has also been falsified by experimental data [23, 31]. The belief by the FDA that dose–response relationships for all chemicals are monotonic has involved disregarding their own findings of nonmonotonicity based on the belief that the findings are not “biologically plausible” [30]; however, this decision by the FDA was rebutted with extensive additional data [31]. As discussed previously, the reason the rejection of nonmonotonic dose–response assumptions are important is that chemical risk assessments involve applying a linear correction procedure based on the assumption that the dose–response curve is monotonic [19, 23].

The mechanisms mediating entirely different effects due to hormone receptor saturation at high doses vs. occupancy of a small percent of receptors at very low doses has been examined by endocrinologists. It has been established that very low and very high doses of hormones, drugs and EDCs cause very different effects. A fascinating set of findings have involved experiments in which oscillating dose–response curves have been reported, but these studies involved examining a very wide range of doses spanning at least 7 orders of magnitude [46, 73]. It is well known that BPA along with other endogenous and exogenous estrogens bind to multiple different receptors as dose increases (referred to receptor cross-talk), which is likely a factor in these findings [13, 25, 74]. Research conducted by Angel Nadal and colleagues has revealed that in mouse pancreatic cells in cell culture, a low (100 pM – 1 nM) dose of BPA decreased membrane calcium flux, mediated through ERß, while higher doses of BPA (100 nM – 1 µM) increased calcium flux back to control levels, with this effect mediated by ERα and different pathways. Calcium flux is a crtical regulator involved in cell signaling and multiple other physiological processes. These findings could have resulted in the illusion that BPA was having no effect on membrane calcium flux if only very high doses had been used [75].

EDCs can cause adverse effects at very low doses

That EDCs can alter cell function at extremely low doses has been described, but due to nonmonotonic dose-responses, these low-dose effects are not apparent if only high doses are tested [23, 46, 73]. However, there are different definitions of what is a “low dose”, which can cause confusion. First, “low dose” effects of environmental endocrine disrupting chemicals generally refer to effects being reported at doses lower than those used in traditional toxicological studies for risk assessment purposes. Second, “low dose” is also commonly used to refer to environmentally relevant doses, such as identified in biomonitoring studies [22, 62]. A third definition (e.g., for an estrogenic chemical) would be a dose of an estrogenic EDC that matches the activity of estradiol and results in biological effects in animals and humans [59, 76].

Hormone receptors and EDCs: implications for the threshold assumption

Here we will use one of the most studied steroid hormones (estradiol) and the first steroid receptor that was identified, estrogen receptors (ERs), that bind with high affinity estrogenic hormones and estrogenic EDCs, such as BPA [19, 77]. However, what we discuss applies to all hormones and other EDCs that bind to other hormone receptors, since they all follow the basic principles of hormone action, which is a subject that has not traditionally been part of the toxicology curriculum.

Hormone receptor specificity and sensitivity

One of the hallmarks of hormone receptors is their exquisite sensitivity and specificity at low doses for a specific class of hormone and the EDCs that mimic those hormones. The exquisite sensitivity of hormone receptors to low doses of hormones or EDCs has been most clearly demonstrated for G-protein coupled receptors (GPCR), which stimulate massive amplifying intracellular rapid signaling systems following binding of a hormone or xenohormone to even a single receptor [78]. Importantly, environmental estrogenic chemicals such as BPA have been shown to be equipotent to estradiol for effects mediated by a G-protein-coupled estrogen receptor (GPER) [25, 46, 79]. However, regulatory agencies seeking to minimize the findings that BPA causes adverse effects at doses below those declared to be safe continue to refer to BPA as “weakly estrogenic” [80].

What factors determine hormone and EDC potency?

The number of receptors that respond to a hormone (or an EDC) is a critical factor in determining the cell’s sensitivity to the hormone or other chemicals that bind to the receptor, and thus the resulting cellular response [81]. Additionally, the number of receptors that respond to a hormone can change as a result of exposure to a hormone or an EDC, resulting in increased or decreased sensitivity and magnitude of response by cells that express the receptors that bind the hormone or EDC. In up-regulation, the number of receptors increases in response to rising hormone or EDC levels, but only if the increase in the hormone or EDC is in a physiological dose range [82, 83].

Receptor up-regulation due to low dose exposure makes the cell more sensitive to the hormone, resulting in a greater response at the same dose. For example, a 0.1 pg/ml increase during fetal life in serum free (bioactive) estradiol, a low dose of the drug diethylstilbestrol or the estrogenic EDC BPA, permanently increased prostatic androgen receptors in male mice [82–84]. Subsequently, other studies showed that developmental or adult exposure to estradiol, DES or BPA resulted in diseases of the prostate and urogenital system in mice and rats [85–89].

When the number of receptors decreases in response to high hormone levels, called receptor down-regulation or desensitization (due to receptor internalization or chemical modification such as phosphorylation), cellular response is reduced in proportion to the reduction in receptors [81, 90]. For example, desensitization was used [81] as a mechanism to demonstrate a reduction in the cellular response to a hormonal drug in proportion to the reduction in the number of cell membrane receptors (Fig. 2).

Fig. 2.

The effect in human airway cells of a decrease in receptors mediating the response to a high efficacy bronchodilator agonist, formoterol, as receptors (ß₂-adrenoceptors) are systematically deleted from the cells through the process of desensitization. Total ß₂-adrenoceptors receptors were reduced from 100% to 0.2%, resulting in a proportional decrease in the maximum response (ED_max). Additionally, the sensitivity of the response decreased, as the dose required to result in 50% of the maximum response (ED₅₀) was increased (shown as red circles). These findings are true for both transmembrane receptor systems (G-protein coupled receptors; GPCR) as well as nuclear receptor systems (e.g., estrogen receptors alpha and beta) [91]. Thus, when receptor number (which is “concentration” for cellular kinetics) declines, the ED₅₀ moves to the right or weakens, in addition to the decrease in the maximal response, which indicates reduced potency of the ligand. Modified from: [81], with permission

A problem that has not been appreciated by regulatory toxicologists is that, in addition to the number of receptors present in a cell, potency of EDCs is not determined by just the affinity of a chemical for a receptor. There are factors that occur after binding of an EDC to a receptor that are not included in the Michaelis–Menten equation describing hormone-receptor binding and the establishment of a binding (affinity) constant (Supplemental Materials; MM Section 1; Equation 1). Thus, potency cannot be determined by just examining the number of receptors expressed in cells [81]. Instead, potency is controlled by additional factors such as co-regulatory proteins that become part of the transcriptional apparatus that regulates gene expression.

For example, Kuman and McEwan noted that: “The coupling of the functional and structural dynamics of steroid hormone receptors with differences in the local concentrations of potential co-regulatory proteins is likely to result in receptors bringing differing sets of binding partners together in response to agonist or antagonist ligands, such that an agonist in one cell type can be an antagonist in another cell type” [92]. This complexity of hormonal and EDC action has been completely absent in discussions of EDC safety thresholds by the regulatory toxicology community. We present next published evidence that falsifies the assumption that all EDCs can be assumed to have thresholds below which exposure is safe, and that this is not a harmless academic argument.

Experimental and mathematical evidence falsifying the threshold hypothesis for hormones and EDCs

The traditional view of regulatory toxicologists is that “There must be a threshold below which things are not toxic. If that weren’t the case, we would all be dead” according to James W. Bridges, Emeritus Professor of Toxicology, University of Surrey, UK [93]. This myopic focus on short-term acute toxicity was demonstrated to not apply to EDCs by findings that human fetal exposure to the prototypical EDC, diethylstilbestrol (DES), which is a drug that when prescribed to pregnant women prior to the third trimester, caused lifelong harm in offspring. For some female offspring (i.e., DES daughters) fetal DES exposure resulted in vaginal adenocarcinoma in young adulthood, about 20 years after their fetal exposure to DES, and by age 50 DES surviving daughters showed a threefold increase in risk for developing breast cancer [94, 95]. The evidence is that DES is not a mutagen and that mechanisms altering uterine development in females include disruption of the HOX and other gene signaling pathways in the developing fetal Mullerian ducts [96, 97].

Fetuses are not little adults

A critical issue that continues to be ignored by chemical regulatory agencies is that exposure to EDCs during pregnancy poses a unique problem, since fetuses are highly vulnerable to permanent damage caused by maternal EDC exposure. Fetuses do not have the developed homeostatic systems that can counteract the disruption of development by EDCs. A central tenet in pediatric medicine is that: fetuses, infants and children are not little adults, and thus they are more susceptible to environmental stressors, such as EDCs. The problem is that when physiologic regulatory systems are in the process of developing, they can be permanently deranged by exposures during “critical periods” in development [98].

In a study that we conducted with pregnant mice [99], the adult pregnant females showed the expected linear decrease in endogenous estradiol in response to low vs. high doses of the potent estrogenic drug DES (Fig. 3-A). However, at a low dose of DES fed orally to the pregnant dam (0.1 µg/kg maternal body weight/day), the fetuses showed an effect opposite to the effect in the dam: an increase in endogenous estradiol, suggesting the absence of the homeostatic (compensatory) estrogen feedback system present in the pregnant mother (Fig. 3-B). These data suggest that fetuses exposed to low doses of DES were not only harmed by the DES, but by the stimulation by DES of excess estradiol production, a “positive feedback” double whammy effect, which is the opposite of compensatory homeostasis. This was particularly evident in female mouse fetuses. For “DES daughters” there were life-long consequences of fetal DES exposure not only for them, but also for their offspring [94, 95]

Fig. 3.

Panel A The effect of an oral dose (administered via feeding diethylstilbestrol; DES) dissolved in corn oil 0 (vehicle control), 0.1 µg/kg body weight (low dose) or 50 µg/kg body weight (high dose) to pregnant CD-1 mice from gestation day (GD) 11–17 (parturition in CD-1 mice occurs on GD 19). Maternal and fetal male (M) and female (F) serum was collected on GD 18 and analyzed for estradiol using a highly sensitive and specific radioimmunoassay [100]. There was an expected suppression of endogenous estradiol in maternal serum as a function of DES dose. Panel B Male and female GD 18 fetal serum estradiol showed that the low dose of DES (0.1 µg/kg maternal body weight) resulted in an increase in fetal serum estradiol. In contrast the high dose of DES (50 µg/kg maternal body weight) had the expected effect of decreasing endogenous estradiol in fetuses. There were thus opposite responses to maternal and fetal exposure to a low dose of DES (a nonmonotonic dose–response curve in fetuses). These findings suggest that in fetuses there is an absence of the same homeostatic systems that are present in adults. The figure is based on data presented in [99] with permission

In contrast to the low dose DES findings in fetal mice, at the much higher DES dose, within the acute toxic range of 50 µg/kg/day, the mother and both male and female fetuses showed a similar significant decrease in endogenous estradiol [99]. Importantly, the opposite effects in the pregnant female and her fetuses in response to a low dose of DES would not have been discovered if only high toxic doses of DES had been administered, which would be the standard protocol in a regulatory guideline chemical study focused only on the effects of a few very high doses [13].

Taking into account the endogenous dose of hormones in EDC research: the turtle experiment

The issue of the need to account for the endogenous, background dose of estradiol (or any other hormone) when there is exposure to chemicals that act through the same receptor mechanism as the endogenous hormone, such as BPA binding to estrogen receptors, has been the subject of scientific inquiry for many decades [101]. Sheehan et al. [9] conducted the clearest high-powered experimental demonstration regarding the issue of whether or not a threshold exists when there is a background level of hormonal activity already impacting the outcome under normal conditions of hormone action, and an exogenous hormonally active chemical is administered that acts through the same mechanism as the endogenous hormone.

Red eared slider turtle embryos exhibit temperature-dependent sex determination, as opposed to mammals that have XY chromosome (genetically) based sex determination. Temperature-dependent sex determination in these turtle embryos is mediated by estradiol (as estradiol increases, more embryos become females) due to a temperature-dependent increase in the enzyme that produces estrogen, aromatase (also referred to estrogen synthetase; CYP19A1). In the Sheehan et al. study an incubation temperature of 28.6 °C temperature was used so that 23.5% of the control group embryos (with zero additional dose of estradiol) developed as females, indicating that this temperature was high enough to significantly increase aromatase activity and thus endogenous estradiol to a level that impacted sex determination from male to female in the most sensitive embryos.

The study by Sheehan et al. thus focused on whether the dose–response curve intersected the X (dose) axis above zero on the X (dose) axis, which is a requirement if the prediction is that there is a threshold dose (positive dose value on the X axis} below which there is no response to the administered chemical. Sheehan et al. demonstrated experimentally that there was not a threshold, because any dose above zero external dose increased the percent of embryos that became females. The basis for this result was that the background level of endogenous estradiol (the chemical being experimentally administered) was already impacting the outcome (increasing the percent of female embryos) at zero external dose. Thus, at zero external dose of estradiol, the dose–response curve intersected the Y axis above zero, revealing a background estrogen effect had to be taken into account in determining the total exposure dose, and that fitting the data to the Michaelis-Menden equation resulted in intersection of the X axis at a negative value, revealing the background value (Fig. 4).

Fig. 4.

Results from a large experiment using 2,400 red-eared slider turtle fertilized eggs administered 7 doses of estradiol (E₂), plus controls (n = 300 per group), to the turtle embryos to alter sex determination (red-eared slider turtles show temperature dependent sex determination). The data show that the administered estradiol caused effects as a result of acting additively with the endogenous estradiol of the embryo. The results demonstrated at P < 0.05 that there was no threshold dose for administered estradiol below which a response did not occur. An embryo incubation temperature of 26.6^○C resulted in 100% of embryos becoming males, while as temperature increased to 31^○C, 100% of embryos became females. As temperature increases above 26.6^○C, the enzyme aromatase (CYP19A1) is progressively increased, resulting in an increase in endogenous estradiol synthesis that is the mechanism that increases the proportion of embryos that become females. However, there is variability in the sensitivity of embryos to the temperature increase and resulting estradiol increase. At the temperature of 28.6^○C, used in this experiment, the endogenous production of estradiol resulted in 23.5% of embryos becoming females. The solid line, calculated from the Michaelis–Menten equation, fit the data with a correlation coefficient of r = 0.90. Dashed lines indicate upper and lower 95% confidence intervals. The solid line strikes the x axis (total dose) determined to be equivalent to an external dose of estradiol of 1.7 ng/embryo; this was thus the endogenous dose of estradiol (D_o) in units of applied dose (D; ng/embryo). The actual median effective dose (ED₅₀) of 5.0 ng is the sum of the endogenous and applied doses (ED₅₀ = 3.3 + 1.7 = 5.0 ng/embryo). The Michaelis–Menten equation was adjusted to include the endogenous estradiol that contributed to the total dose (Percent Response = ED_max * [D + D₀]/ (ED₅₀ + [D + D₀]). The adjusted MM equation for percent female = 100 * [D + 1.7]/(5 + [D + 1.7]) (Supplemental Materials; Section 1; MM Equation 3-D). Importantly, in this experiment there were no background non-hormone factors impacting the results, since at a temperature of 26.6^○C without endogenous estradiol synthesis being stimulated at this temperature, 0% of offspring were females. ED_max = maximum effective dose; D = administered dose; D_o = endogenous concentration of estradiol converted to administered dose units of ng/embryo; ED₅₀ = median effective dose (50% females). Modified from: [9] with permission

The Sheehan et al. experiment involved 8 doses (control and 7 doses of estradiol) being administered to a total of 2,400 red eared slider turtle (Trachemys scripta elegans) fertilized eggs (embryos) equally randomly divided among the dose groups (300 embryos/group; Fig. 4). The high number of doses and the very large number of embryos examined in this study gave the study sufficient power to exclude a threshold effect. The absence of a threshold is apparent by examining the 95% confidence interval (that does not exceed zero on the X axis). This finding statistically precludes the possibility of a threshold at P < 0.05 probability.

When administered dose of a hormone or EDC is zero and there is a positive response value above zero on the Y axis (as shown in Fig. 4), there is already a background level of activity in the system that is being impacted by the administered chemical. Hoel [6] showed mathematically that when there is additivity of an endogenous hormone and an EDC, there can be no threshold for the applied EDC dose when the dose–response curve intersects the Y axis above zero and instead reaches the X (dose) axis below zero externally administered dose (Fig. 4; Supplemental Materials; Section 1; MM Equation 3-D). Hoel [6] as well as Blair [10] and Sheehan [21] thus all demonstrated mathematically that administration of a chemical that acts through the same hormone receptor mechanism as an endogenous hormone cannot have a threshold beneath which there is no adverse effect.

Regarding the concept of “threshold of adversity” [3], irreversibly eliminating males from the population due to embryos being exposed to estrogenic chemicals that act additively with estradiol [19, 20] would clearly be an adverse effect (extinction) on that turtle population (Fig. 4). This adverse effect of an exogenous estrogen would be exacerbated by the turtles being exposed to gradually increasing global temperature due to climate change, leading to a further increase in aromatase in turtle embryos and fewer males in the population. This is a clear example of the potential for an interaction of climate change and EDC exposure.

In summary, the turtle embryos have been shown to be a very sensitive system for testing the hypothesis that an external dose of estradiol would act additively with the background endogenous dose of estradiol. Interestingly, something similar has also been shown in an in vitro study to determine whether EDCs with a lower affinity for estrogen receptors relative to estradiol had significant cumulative effects at very low doses. When administered in the presence of estradiol there was a cumulative effect of very low doses of a number of estrogenic EDCs, but there was no effect of the same low doses of estrogenic EDCs when they were administered individually [20]. Clearly, examining only one chemical at a time rather than considering cumulative effects of multiple exposes, as well as the interaction with endogenous hormones, can lead to false conclusions about safety thresholds.

Validation of the Sheehan et al. findings for multiple EDCs

To determine whether prior experiments concerning the effects of EDCs agreed with the Sheehan et al. turtle findings, Blait et al. [10] analyzed data from 26 separate experiments that included 24 different endpoints and 9 different EDC treatments, with a total of 178 data points. The objective was to determine whether the dose–response data from each study statistically fit the Michaelis-Menetn equation that describes hormone-receptor binding (with the caveat that the unmodified equation applied to receptors such as those for estradiol with one hormone binding site).

Since the data analyzed by Blair et al. were from separate experiments with different units and scales, five parameters were considered: 1) The dose that resulted in the maximum response in each experiment (ED_max), 2) The dose that resulted in 50% of the maximum response (ED₅₀), 3) The administered dose of EDC in the experiment (D). 4) The endogenous hormonally controlled background dose (D₀), and 5) Any other background non-hormonal (Bnh) factor impacting the response (typically through an unknown mechanism). These are the parameters required to fit the data to the Michaelis–Menten equation (Supplemental Materials Section 1; MM Equation 3). To accomplish this both X and Y variables were normalized. This involved normalizing the X axis dose values to 0 to 10 arbitrary dose units, and the Y axis values to 0 to 1.5 arbitrary response units (Fig. 5).

Fig. 5.

Blair et al. [10] plotted 178 data points from 26 separate experiments that examined 24 different endpoints from 9 different EDC treatments. To accomplish this both X and Y variables were normalized. This involved normalizing the X axis dose values to 0 to 10 arbitrary dose units, and the Y axis values to 0 to 1.5 arbitrary response units. The fully normalized data were presented as R’/R_max), where R’ = Response = the effect of any background effect not due to endogenous hormones. Blair et al. identified that the data from these 26 studies of EDCs fit the Michaelis-Menten equation with a correlation coefficient of r = 0.962, demonstrating a close to perfect correlation of these data to the Michaelis-Menten equation. Importantly, Blair modified the Michaelis-Menten equation to take into account background endogenous hormonal activity and, in addition, other non-hormonal factors in 26 studies examining different chemicals and different dose ranges. D₀ = endogenous dose contributing to the total doses, ED₅₀ = the dose that resulted in 50% of the maximum response, R’ = response – any non-hormonal background contributor to the response, R_max = dose resulting in the maximum response. From [10] with permission

Blair et al. identified that the data from these 26 studies of EDCs fit the Michaelis–Menten equation with an average correlation coefficient of r = 0.962, demonstrating a close to perfect correlation of these data to the Michaelis–Menten equation modified to take into account background hormonal and other factors in studies examining different chemicals and different dose ranges. In each of these experiments the administered EDC was found to have had an additive effect with the endogenous background hormone. In addition, other non-hormonal background effects on the response were assessed. None of the data from these experiments are consistent with the presence of a threshold dose below which no response occurs. See Suplemental Materials (Section 2) for more details of the Blair et al. study.

The mathematical description of hormone-receptor binding: the Michaelis–Menten equation

The Michaelis–Menten equation was initially developed to describe the rate of enzyme-catalyzed reactions involving binding of a reactant to an enzyme. The Michaelis-Menton equation has also been applied by endocrinologists to describe the reversible binding of hormones to receptors [102], which follow the same mathematical kinetics (Supplemental Materials; Section 1; MM Equation 1; MM Equation 2). An essential part of hormone signaling is the capacity to rapidly respond to both an increase or a decrease in circulating hormone and thus an increase or decrease in receptor occupancy and response. The consequence is that the non-covalently bound hormone-receptor complex can quickly respond to changes in circulating hormone concentrations leading to a change in the cellular responses that occur as a result of a change in hormone binding to a receptor.

Since the Michaelis–Menten equation is the central equation used to describe hormone-receptor binding in the field of endocrinology, it is incomprehensible that the mathematical principles used to describe hormone-receptor interactions in endocrinology are rejected by toxicologists in regulatory agencies responsible for regulating EDCs that act by binding to the same receptors as endogenous hormones. But it is the rejection of the decades of research on hormone action conducted by endocrinologists that allows regulatory toxicologists to continue to adhere to principles that do not apply to hormones, hormonally active drugs or EDCs. Regulatory toxicologists continue to insist that there are thresholds below which EDCs and hormonally active drugs are safe, which experimental and mathematical analyses described above show to be an assumption that has been clearly falsified.

Note that the Michaelis–Menten equation plots to zero response only at zero dose when using a linear scale for dose (Fig. 6-A). When the Y axis (hormone-receptor binding) is zero, intersection with the X (dose) axis at a higher concentration than zero would be required for there to be a threshold dose above zero at which no response occurs. However, this is not what the mathematical or experimental findings show to be the case [10, 21, 103]. An important issue shown in Fig. 6-A (insert) is that it has been determined that if the hormone concentration is below 30% of maximum receptor binding level (less than K_d/2), the relationship between dose and response is highly proportional and essentially linear to zero on the X and Y axis.

Fig. 6.

Panel A. This is a linear plot of the relative percent of receptors occupied by a ligand as a function of dose resulting in a typical saturation curve for a receptor–ligand interaction. Without log transformation the shape of the dose–response curve would resemble this curve plotted relative to a linear increase in dose. As dose approaches zero the dose–response curve gets closer and closer to being a straight line, as shown in the insert. The curve was generated from data with B_max = 100; K_d = 10; dose range from 0.10—990. The insert shows that the binding observed at very low concentrations of ligand (dose < K_d/2) appears linear and therefore might erroneously be interpreted as nonspecific. Panel B. The Michaelis–Menten semi-log sigmoid plot of estradiol binding to ERα in human breast cancer (MCF-7) cells. The X axis is a log scale covering a 100,000-fold range of doses. The Y axis is presented as the percent of cell proliferation relative to maximum proliferation of MCF-7 cells administered estradiol. The increase in estradiol dose from 0—10 pM leads to a highly proportional increase in receptor binding. Half maximal response for estradiol binding to ERα is at 100 pM, 91% response at 1000 pM and 99% response at 10,000 pM. To go from 1% receptor occupancy to 99% receptor occupancy spans 10,000-fold concentration range. The top grey line is the ECmax. Within the physiological range of free estradiol in blood (below the K_d), there is a relatively proportional change in receptor binding, while above 10,000 pM estradiol, receptor saturation has occurred and no further increase in receptor occupancy or estradiol-stimulated response via this receptor can occur with any further increase in dose. The MM equation plots to 0 on the X and Y axis within the physiological range of estradiol in animals, including humans. Panel A: From: [103] with permission. Panel B (W. Welshons, unpublished). This figure, as well as the insert in Panel A, show the absence of a threshold

Hormone-binding data are often presented as a semi-log plot. Figure 6-B shows estradiol binding to nuclear estrogen receptors using a log₁₀ scale for dose. The semi-log plot covers a wider dose range (creating the familiar sigmoidal plot) than is possible using a linear plot, but does not provide the precise information provided in a linear plot about what occurs as you approach zero dose. However, the closer the data get to zero dose on the X axis (e.g., below 10% receptor occupancy, even with the log scale) the greater the proportionality of dose and response, and the dose–response curve never actually reaches zero on the Y axis at zero dose.

Regarding high doses, using ERα as an example, the concentration of estradiol that results in virtual saturation (total saturable binding) of ERα in human MCF-7 breast cancer cells is 100-fold higher than the hormone concentration that occupies 50% of available receptors, based on application of the Michaelis–Menten equation (Fig. 6-B).

The Michaelis–Menten equation describes the relationship of the number of occupied hormone receptors that lead to the response, which is driven by the binding of a hormone to its receptor when the receptor has one binding site (e.g., the binding of estradiol to an estrogen receptor), forming a hormone-receptor complex (Supplemental Materials; Section 1; MM Equation 2). We recognize that the equations get more complex when there are multiple binding sites and more complex outcomes (e.g., cooperativity) occur. However, the underlying math of receptor saturation still applies [102].

The dissociation constant

The dissociation constant (K_d) is a physical–chemical constant identifying the hormone or EDC concentration at which 50% of receptors available in the system being studied are occupied at approximately steady state. The K_d is determined based on the concentrations of the hormone and the concentration of its receptor in a temperature-dependent assay system and is thus typically described as the apparent K_d.

Hormone-receptor binding is based on the law of mass action that asserts that the rate of the reaction is proportional to the product of the concentrations of both the hormone and receptor, which is described by application of the Michaelis–Menten equation for hormone-receptor binding as it approaches equilibrium and accurately describes the kinetics of hormone-receptor binding (Supplemental Materials; Section 1; MM Equation 2).

Importance of percent receptor occupancy

Once receptor saturation has occurred, no further increase in hormone occupancy or hormone-stimulated response is possible, and in many cases, response inhibition occurs (such as desensitization; Fig. 2), which can occur due to a number of factors [8, 19, 23]. Normal hormone action requires occupancy of only a very low percent of available receptors to elicit large effects [19, 77] (Fig. 7). These extra receptors are referred to as “spare receptors” that can impact physiological processes [81] (Fig. 2). A high affinity ligand such as estradiol typically only binds a small percent of a cell’s receptors to produce a maximal response (Fig. 7). However, if the number of receptors is greatly reduced due to endocrine disruption or other factors such as desensitization (Fig. 2), the critical importance of the excess (spare) receptors is eliminated, and the sensitivity of cells to normal hormonal systems or hormonal drugs is greatly reduced [81]. None of these basic principles of hormone action are taken into account by regulatory agencies in the determination of the potency and safety of EDCs (Fig. 1).

Fig. 7.

The effect of very low (femtomolar) to high (nanomolar) concentrations of estradiol on proliferation of human breast cancer MCF-7 cells (presented as a percent of maximum response in relation to percent receptor occupancy. The data show the physiological range for proliferative response to estradiol is within the sub-picomolar to low picomolar range. Above the nanomolar range, the nuclear receptors are saturated and no additional estradiol can cause an increase in response, as no further increase in receptor occupancy is possible. In this cel culture system, 50% of the maximum response to estradiol occurred at ~ 1 percent receptor occupancy (~ 1 pM; ~ 0.3 pg/ml). At ~ 10% receptor occupancy ~ 90% of the maximum response occurred at a tenfold higher dose of ~ 10 pM; 3 pg/ml). At 50% receptor occupancy (the K_d) the proliferation of MCF-7 cells was maximal (at ~ 100 pM; ~ 30 pg/m). At doses higher than the K_d it is not uncommon to see response begin to decrease, resulting in a non-monotonic dose–response curve. These findings demonstrate that normal hormonal responses occur at very low doses and at very low receptor occupancy, and that studies using only high doses are of no value for assessing effects of EDCs. This figure was based on data published in [19] with permission

At very low hormone concentrations, first order linear kinetics apply (the rate of binding to available receptors is proportional to the concentration of the hormone being studied). Above the K_d, the reaction kinetics become more complicated (referred to as zero-order kinetics), where hormone-receptor interactions are not just determined by the hormone concentration. As the occupancy of all hormone receptors (receptor saturation) is reached, eventually there can be no further increase in receptor occupancy, which is required for a change in a hormone-stimulated response with increasing dose [102].

Also, up to about 10 percent receptor occupancy there is high proportionality (close to linearity) between the increase in dose and increase in receptor occupancy, and thus the system is very sensitive to small changes in dose in the very low-dose range (Fig. 6-A; Fig. 7). However, above the K_d dose, there has to be a big change in dose to achieve a small increase in receptor occupancy, and so the system being studied in this dose range is less sensitive to hormones or EDCs that act via these receptors. As receptor saturation is approached, the system being studied becomes insensitive to effects mediated by the hormone’s primary receptor (Fig. 6-B), and effects could be mediated by the hormone or EDC interacting with other receptors via receptor crosstalk [104].

The effective dose vs. the dissociation constant

It is important to distinguish hormone-receptor binding (and calculation of a K_d) from studies of the response that is initiated (Supplemental Materials; MM Equation 3). The cellular response to an EDC is identified as the effective dose (ED), which is expressed as a percent of the maximum potential response for the system being measured (ED_max). The ED is a function of both the hormone dose and receptor concentrations, as well as co-regulatory proteins in the cell, which are essential to determining the response to a hormone or EDC in a specific tissue (Fig. 1); this is why estrogenic EDC are referred to as “selective estrogen receptor modulators” [105, 106]. Determining the role of co-regulatory proteins in triggering cellular responses is not considered in the process of calculating binding by a hormone to a receptor using the Michaelis–Menten equation (Supplemental Materials; Section 1; MM Equation 2; MM Equation 3).

In Fig. 7 data from human MCF-7 breast cancer cell proliferation in response to estradiol are shown. This reveals the exquisite sensitivity of these human breast cancer cells to estradiol, with 50% of maximum proliferation (ED₅₀) occurring at a dose of 0.3 pg/ml (1 pM). However, at concentrations above 10⁻⁶ M (~ 0.3 µg/ml; parts per million; ppm), estradiol is cytotoxic and results in cell death regardless of whether MCF-7 cells, which express estrogen receptors and are estrogen responsive, or estrogen receptor–negative C4-12–5 cells are examined (estrogen-non-responsive C4-12–5 cells were derived from MCF-7 cells). This demonstrates that cell death identified in this experiment is not a function of a hormone response mediated by estrogen receptors, but is due to acute cytotoxicity of a very high dose of a lipophilic chemical [19].

Need to differentiate between thresholds for endogenous hormones vs exogenous EDCs

Even though the Michaelis–Menten equation describing hormone-receptor binding has no threshold term (no threshold is the default assumption), in reality the question regarding whether an endogenous hormone has a threshold above zero dose can never actually be experimentally verified, because it would take an infinite number of samples to achieve the power to answer this question as the plot approaches zero for dose on the X axis as well as a detection system with infinite sensitivity to determine some infinitesimally low threshold.

However, whether or not the endogenous hormonal system being disrupted by an EDC does or does not have a threshold is irrelevant. This is because the endogenous hormonal system is already clearly operating above any putative threshold, and at a level at which adverse effects can occur, which include cancers of female reproductive organs. It is not possible for a threshold to exist for any estrogenic EDC dose that is added to endogenous estrogenic hormones that are already bioactive [77] (Fig. 4; Supplemental Materials; Section 1; MM Equation 2; MM Equation 3).

Perchlorate: a thyroid disrupting EDC not mediated by receptor binding

It is important to note that not all effects of EDCs are mediated by disruption of hormone-receptor binding, but can still result in endocrine disruption. An interesting example of an EDC that disrupts thyroid hormone synthesis is perchlorate, used as an explosive enhancer in rocket fuel and bombs. In addition to describing the binding of hormones such as estradiol to estrogen receptors, the Michaelis–Menten equation also describes the interaction of EDCs such as perchlorate (Cl-O4) with cell membrane transport proteins that are not classified as hormone receptors. The active transport of iodide from the blood into the thyroid gland by a cell membrane protein (the sodium-iodide symporter) occurs at concentrations of iodide in blood far below the K_d of iodide for the symporter. This is similar to the action of estradiol at concentrations in blood far below the K_d of estradiol for nuclear estrogen receptors (Fig. 7), although unlike iodide, no active transport is involved for estrogens to enter cells. It has been shown that perchlorate (an EDC oxyanion) interferes with iodide transport by the NIS by interfering with sodium binding to the symporter, resulting in a decrease in thyroid hormone in the thyroid gland (endocrine disruption) [107]. Perchlorate also interferes with the transport of iodide across the placenta as well as into breast milk, thus additionally impacting fetuses and nursing infants during critical periods when neural development and IQ is altered by even a small decrease in thyroid hormone levels [108].

Discussion and conclusions

We reviewed here that, based on the core principles of endocrinology, there can be no threshold (safe) level of exposure for EDCs that cause effects via binding to receptors when there are endogenous hormones that are already at concentrations in blood above a threshold for causing adverse effects. However, identifying whether a chemical is an EDC will require a major paradigm shift by regulatory agencies. This will include abandoning their current focus on outdated guideline GLP study protocols developed in the last century to examine effects of only a few high doses of a chemical.

An approach to identifying the “Key Characteristics” of EDCs has been developed [109]. In addition, a set of protocols to determine experimentally whether any chemical, including EDCs, can be identified as safe and sustainable for use in consumer products: “The Tiered Protocol for Endocrine Disruption (TiPED)”, has been developed. However, TiPED is intended to be an evolving set of protocols based on the current state of the science [110]. New approaches to chemical testing will have to incorporate data showing the need to consider cumulative effects instead of exposure to one chemical at a time [20]. The need to consider cumulative exposures was mandated in the Food Additive Act in 1958 in the US and has been the subject of numerous reviews [22].

Mechanisms mediating non-monotonic dose–response relationships are crucial to understand the profound flaw in studies of EDC risks when the study protocols are only directed at determining effects caused by very high doses of an EDC (starting at the maximum tolerated dose) and then calculating a benchmark dose as a point of departure for applying linear safety factors to estimate a threshold dose [111]. This approach results in an error in calculating whether the intersection of the zero response point on the Y axis is above zero on the X (dose) axis, which is required for there to be a threshold dose below which no effect determined to be adverse occurs [19, 21].

In the Supplemental Materials (Section 1) we derive the core Michaelis–Menten equation that describes the binding of hormones or EDCs to receptors that evolved to activate cellular processes in response to extremely low hormone concentrations. Large effects also occur in response to a very small percent of receptors being occupied (Fig. 7) [19, 77].

The Michaelis–Menten equation requires a large enough number of doses in order for the shape of the dose–response curve to be determined. A problem with regulatory toxicology studies, as opposed to the design of the Sheehan et al. [9] study to experimentally test the threshold hypothesis, is that administration of only 3 or 4 very high doses (often beginning at the maximum tolerated dose and decreasing as little as 50 fold) has been common in traditional guideline GLP studies conducted for chemical risk assessments [21, 61].

To calculate the shape of a dose–response curve, typically 5 or more doses covering a wide dose range is required to conduct a statistical analysis to determine whether the dose–response curve fits a monotonic or nonmonotonic function; an example would be the method described in Angle et al. in which a control and 5 doses of BPA were fed to pregnant mice covering 5–50,000 µg/kg/day [112]. The typical regulatory toxicology study in which 3 high doses are administered is thus unable to be used to assess the actual shape of the dose–response curve, which is why studies with only 3 doses were rejected for use in the examination of the threshold hypothesis by Blair et al. [10].

Another issue is the basis for the selection of only high doses for analysis in guideline studies conducted for risk assessments. Dose selection should be based, when known, on an EDC’s physiological range of activity [76, 113]. A second issue in dose selection is environmental relevance [62]. For follow-up health effect studies due to occupational exposures [114, 115] or exposures from medical equipment used in hospitals [116], this would involve examining higher doses relative to those found in the general population based on biomonitoring studies [22, 62]. For studies showing very high exposures in occupational and medical settings, or after evidence of high exposures due to accidents, such findings should rapidly lead to follow-up examination of the health consequences of the relevant high exposures. This would require coordination between the FDA, EPA, CDC, NIOSH and other federal and state agencies, which has not often occurred.

A core risk assessment assumption used to justify the current misguided approach to dose selection is that all dose–response curves are monotonic (i.e., the toxicological dogma that “the dose makes the poison”). In sharp contrast, for EDCs, non-monotonic dose–response curves (e.g., U-shaped or inverted U-shaped) are common [59]. Non-monotonicity of hormone and EDC dose–response relationships results from numerous mechanisms that have been discussed in detail in other reviews [8, 23], but these findings have been ignored by toxicologists at the FDA [30]. Critically, using safety factors to calculate a threshold dose after examining only a few high doses of a EDC that exhibits a non-monotonic dose response curve has been shown to result in a large error in estimating a presumed safe (threshold) dose [19].

Specifically, for hormones, hormonally active drugs, and EDCs, high doses can cause the opposite effect seen at lower doses [23, 75, 82]. This is the basis for tight control of the therapeutic dose of hormonal anticancer drugs for breast (tamoxifen) and prostate (Lupron) cancers, by scientists and physicians in The Center for Drug Evaluation and Research (CDER) at the FDA. These hormonally active drugs stimulate cancer cells at low doses (called “flare”) but have the opposite effect of inhibiting cancer cell proliferation at much higher therapeutic doses; this was only discovered for tamoxifen after the drug was in clinical use [23].

If physicians followed the principles of toxicology when administering these anticancer drugs, they could kill their patients if the exposure dose got into the low-dose “flare” range that a regulatory toxicologist would declare safe based on ignoring that these drugs show non-monotonic dose–response relationships. There are numerous other examples of non-monotonic dose–response relationships for hormones, vitamins, drugs and EDCs [23]. That there could be opposing positions about non-monotonic dose responses within the CDER and Human Foods Program within the FDA demonstrates a significant lack of coherent leadership at the FDA.

The basic assumption within the FDA regarding chemicals in food is that chemicals are safe until proven to cause harm, which is the opposite of the approach concerning drugs, which need to be shown both safe (relative to risk) and effective prior to use. Everyone accepts that there is some risk associated with drugs taken to cure illnesses. What is not accepted is that there should be any risk associated with eating food wrapped in packaging leaching toxic EDCs into the food.

There is no easy solution regarding replacing guideline study protocols and improving food safety on the horizon, because the criticisms of the outdated, insensitive and extremely expensive guideline GLP studies that use massive numbers of animals and result in false conclusions [61] are still being ignored by the FDA [5, 29]. The abuse of the “Generally Regarded as Safe” standard for chemicals in food, where the FDA simply turned over to industry the determination of chemical safety, also has to end [33, 117].

FDA administrators involved in food safety recently stated, regarding the 500-year-old misinterpreted threshold assumption that the dose males the poison: “The presence of a chemical alone isn’t what determines whether a food is safe to eat. To assess the safety of chemicals in food, scientists at the U.S. Food and Drug Administration and others worldwide look at information about the chemical’s safety, as well as how much of a chemical is in the food and how much a person eats or drinks. It’s the amount that counts” [5]. What is implied here is that that “more is worse”, which we have shown here is false for EDCs [23]. The issue is not whether low or high doses of EDCs are worse, but that they lead to very different outcomes.

Another problem is that exposure assessment is the weakest, least accurate aspect of chemical risk assessments. Exposure studies are expensive, and regulatory decisions are often based on exposure models that are based on inappropriate assumptions [8, 35]. It is thus important that the dose range studied for identifying the hazards of EDCs for the general population be within the range identified in biomonitoring studies when these data are available instead of relying just on toxicokinetic models [8, 62, 118]. It has thus fallen to academic investigators, not federal regulators, to show that estimated safe doses for numerous EDCs are, in fact, not safe [11, 12, 19, 25, 31, 72].

A paradox is that regulatory agencies are under pressure from chemical corporations and animal rights groups to reduce laboratory animal testing of chemicals and adopt New Approach Methods (NAMs). In fact, scientists at the FDA noted in a study in which they identified a rapid assay with an inexpensive non-traditional invertebrate model animal that “Due to the high cost and long duration of traditional testing methods… only a small fraction of chemicals that humans are exposed to have been assessed” [119]. Our view is consistent with this concern in that changes in the traditional approaches used to identify chemical hazards are needed. For example, we propose that if a chemical can bind to hormone receptors in an in vitro assay (which, by definition, will result in endocrine disruption), this represents an intrinsic hazardous property of the chemical, and the result should become the basis for classifying the chemical as an EDC and regulating its use. Further, if a chemical can interfere with normal hormone signaling through any mechanism, it represents an intrinsic hazardous property, and is therefore a potential risk, particularly when exposure occurs during developmental critical periods. Because very low doses of EDCs are hazardous [8, 19], the risk can be dramatically underestimated using traditional guideline approaches [109]. We thus agree with the concern expressed by Hunt et al. [119] and the need to expand approaches for assessing chemical hazards while not completely eliminating animal testing that is especially critical for assessing long-latency adverse health effects [11, 12].

Because regulatory toxicologists do not accept that hormone- or EDC-receptor binding occurs at exceedingly low doses, they do not accept that biomonitoring studies identifying low EDC levels in the general population [62, 120] pose a risk (Risk = hazard x dose). The “elephant in the room” is that no in vitro assay exists that could have predicted the decades-long latency for development of cancers that was revealed by the DES iatrogenic tragedy involving the offspring of the millions of pregnant women administered this drug over decades [95].

There was hope that the CLARITY-BPA model of creating a consortium of academic experts to share tissues using state-of-the-art methods could replace outdated guideline protocols and bring regulatory toxicology into the twenty-first century at significantly reduced cost [31]. However, the FDA has rejected collaborating in risk assessments with academic scientists [29, 30], which is described in detail by Heindel et al. [31]. Instead, the FDA continues to assure the public that chemicals added to food or food packaging in the US are safe, even though studies have reported the FDA’s lack of regulation of thousands of chemicals in food [15, 121–123].

We have reviewed that for hormonally active EDCs that cause adverse effects by binding to receptors for endogenous hormones, the threshold assumption is falsified by decades of research in endocrinology. We note that the threshold hypothesis has been falsified by not only a large number of studies by academic scientists, but by the FDA’s own scientists [9, 10, 21]. Importantly, the public health costs associated with the myriad non-communicable diseases shown to be related to exposure to different EDCs have been identified as more than tenfold higher than the industry profits that are being protected by regulatory agencies in the US [124]. We conclude with a list of flaws in current chemical risk assessments (Fig. 8) and suggestions to improve the chemical regulatory system in the US (Fig. 9).

Fig. 8.

Flaws in chemical risk assessments in the US by the FDA

Fig. 9.

Suggestions that the FDA should follow in revising their approach to regulating toxic chemicals in food and food contact materials

Abbreviations

ADI

Acceptable daily intake set by the US FDA

B_max

Maximum occupancy of receptors

BPA

Bisphenol A

CDC

Centers for Disease Control and Prevention

CDER

Center for Drug Evaluation and Research at the FDA

CLARITY-BPA

The Consortium Linking Academic and Regulatory Insights on BPA Toxicity

D

Administered dose

D₀

Endogenous background dose

DES

Diethylstilbestrol

ED

Effective dose resulting in a response

ED₅₀

Dose resulting in 50%% of maximum response

EDC

Endocrine disrupting chemical

EPA

US Environmental Protection Agency

EU

European Union

FDA

US Food and Drug Administration

GD

Gestation day

GLP

Good laboratory practices

GPCR

G-protein-coupled receptors

GPER

G-protein-coupled estrogen receptors

GRAS

Generally regarded as safe

IQ

Intelligence quotient

K_d

Dissociation constant, dose resulting in 50% occupancy of available receptors

LD₅₀

Lethal dose for 50% of animals administered a chemical

MTD

Maximum tolerated dose with an adverse effect but without death

NAM

New approach methods

NIOSH

National Institute for Occupational Safety and Health

NIEHS

National Institute of Environmental Health Sciences

NIH

National Institutes of Health

NIS

Sodium-iodide symporter

NTP

US National Toxicology Program

pM

Picomolar

ppb

Parts per billion

RfD

Reference dose (safe daily intake dose) set by the US EPA

R_max

Maximum response

TCDD

2,3,7,8-Tetrachlorodibenzo-p-dioxin

TDI

Tolerable daily intake set by the EU

TiPED

The tiered protocol for endocrine disruption

TSH

Thyroid stimulating hormone

UK

United Kingdom

Authors’ contributions

FvS and WVW contributed equally to the conceptualization and writing of the manuscript. FvS primarily revised the manuscript. WVW primarily revised the figures and legends. FvS and WVW prepared the Supplemental Materials.

Funding

No funding was provided for this review.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.