There are things we know we know

While we have made considerable progress in understanding the complex interactions of EDC exposures, the environment and human health, this has been achieved by investigating a very narrow set of chemicals. Here, the ‘dirty dozen’ includes diethylstilbestrol, ethinylestradiol (both pharmaceuticals), bisphenol A (BPA, a plastic monomer), phthalates (a group of plasticisers), atrazine (a widely used herbicide), polychlorinated biphenyls and polybrominated diphenyl ethers (industrial chemicals), dichlorodiphenyltrichloroethane (DDT; an insecticide) and its metabolite DDE.1 Among these, BPA is by far the best-researched compound: out of the ∼17 500 publications on EDCs listed by ISI Web of Science in June 2016, almost one-third (5600) contain the term ‘bisphenol A’ (figure 1). According to current knowledge, there are over a thousand suspected EDCs.3 However, 49.6% of the pertinent studies deal with eight chemicals.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Research on Endocrine Disruptors is dominated by a few well-researched compounds. Number of publications containing the compound's name as a subset of all publications containing the term ‘endocrine disrupt*’ (ISI Web of Science, retrieved June 17, 2016). EE2, ethinylestradiol; PCBs, polychlorinated biphenyls (360 publications); DES, diethylstilbestrol (296); DDT, dichlorodiphenyltrichloroethane (197); atrazine (119).

This preoccupation with a small set of chemicals has historical ground: the identification of EDCs was based on chance findings rather than a structured approach. For example, oestrogenic nonylphenols were inadvertently found to leach from plastic labware. Basically, the process of selecting chemicals to study was driven by what compounds analytical chemists were able to detect or toxicologists tested positively in their bioassays. This has led to a situation in which the compounds producing the most spectacular effects in terms of exposure or hazard received the most scientific attention and were studied in even more detail. Although nothing is wrong with that per se, one can argue that the selection bias inherent in the process (establishing ‘relevance’ by phenomenological excitement and later by the sheer number of studies) puts too much emphasis on the known EDCs.

While it appears obvious to point out that human populations (not to speak of wildlife) are exposed to a much larger number of chemicals, among them an unknown number of EDCs, the community appears to have a lasting obsession with the known EDCs. That is to say, 25 years after the discovery of the classical EDCs, more than 60% of the platform presentations at this year's Gordon Research Conference on ‘Environmental Endocrine Disruptors’ dealt with the ‘dirty dozen’. Hansen and Gee4 have attributed this phenomenon, which is not specific to EDCs but also observed for other environmental contaminants, to ‘scientific inertia’5 caused by the desire for high levels of proof, reliance on existing intellectual and technological resources, publication pressure and a conservative, risk-avoiding behaviour by researchers, reviewers and funders.

There is another mechanism reinforcing this preoccupation: reports on adverse effects of economically important EDCs are often contested either as part of the scientific discourse or the product defence strategies of chemical manufacturers. This produces an insatiable need to increase the weight of evidence by further research, which, however, does not necessarily result in greater certainty in establishing causality. The reasons for that are inherent in the scientific process. Research tackling complex interaction often uncovers even more complexity and thus increases uncertainty. In addition, the causes of endocrine-related diseases (as well as most other non-communicable diseases) are inevitably multifactorial so that establishing a high weight of evidence for a certain EDC inducing a condition is almost impossible. In that sense, we may promote deadlock when trying to ‘ultimately prove’ the adversity of a single compound by producing more data. One way to resolve this would be the more stringent application of the precautionary principle. By regulating compounds based on a lower weight of evidence, researchers would be encouraged to move from the ‘dirty dozen’ to other unstudied/understudied EDCs.

There are some things we do not know

For whatever reasons, toxicologists, exposure scientists, epidemiologists and regulators continue to investigate and assess a handful of EDCs in great detail. Admittedly, this approach has been very successful in generating funding and knowledge as well as societal awareness. This in turn helped to reduce exposures to some compounds (eg, phthalates and BPA) in some applications (eg, baby toys and water bottles) and some countries. However, elsewhere in this journal, it has been criticised that the scientific community places too much emphasis on the problem instead of on the solution.6 While this is certainly true and much more effort needs to be devoted to developing safer chemicals, we also need to critically revisit the underlying assumption of our current strategy. By spending the majority of resources on the known EDCs, we hope to improve public health without knowing their actual contribution to disease caused by chemical exposures at large.

Although data on this question is absent, it appears implausible that a handful of chemicals will realistically represent the complex and changing chemical mixtures that human populations are exposed to. Thus, we are at risk of missing relevant drivers of disease with regard to mixture toxicity7 and unknown EDCs. Recent studies from environmental toxicology shed more light on this issue: it is now widely acknowledged that the pollutants regularly analysed in environmental samples make up only the tip of the iceberg and rarely explain the observed toxicity. For instance, Tang et al8 demonstrated that 299 known pollutants contributed to <3% to the cytotoxicity and oxidative stress induced by water samples. Along the same line, Kortenkamp et al9 proposed that the reason why epidemiological studies often fail in establishing chemical–disease relationships is the focus on few prominent compounds. This should provide a strong incentive to identify the chemicals that are actually relevant.

Approaching the unknown

Our current focus on a few EDCs is too reductionist and does not allow for a holistic understanding of the exposome. Instead of picking out selected chemicals (by whatsoever criteria) and working our way upwards by describing sources and exposures as well as mechanisms and trying to establish causation with disease (figure 2, bottom-up), we need to develop alternative strategies. Currently, there are two approaches: The chemical-based approach involves the high throughput screening of thousands of chemicals for selected mechanisms of toxicity in vitro. This has been successfully implemented within the US EPA's Toxicity Forecaster (ToxCast) program, which after screening 1859 compounds in 27 assays established that the range of potential EDCs is much broader than was previously thought.10 Broadening the range of compounds and the suite of toxicological end points, however, remains nothing but a reiteration of the current approach with improved efficiency.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Two approaches to investigate endocrine disrupting chemicals (EDCs): the current approach (left) starts with selected compounds and investigates their exposures, effects and links to disease (bottom-up). An alternative approach (right) would determine the biological effects in human samples (induction of molecular initiating events of endocrine disruption) and apply non-target chemical analysis to identify the causative EDCs (top-down).

An effect-based approach, in contrast, inverts the direction of the current scientific workflow (figure 2, top-down). Within the framework of effect-directed analysis (EDA, sometimes also referred to as toxicity identification evaluation), complex samples are screened for in vitro toxicity integrating the biological effects of mixtures as well as unknown chemicals. A subsequent non-target chemical analysis can then be used to identify the causative compounds.11 EDA can be applied to screen for the molecular initiating events of endocrine disruption (eg, receptor activation) in biological samples (eg, blood or urine) and ultimately enable the discovery of novel EDCs. The major advantage of EDA is that by combining exposure (biological samples) and hazard (biological effects) it generates information on EDCs that are relevant in terms of both.

There are also theoretical and practical challenges involved: First and foremost, extrapolating from in vitro to human health effects in vivo is not straightforward because of the different levels of complexity. This also links to the open question as to which in vitro end points are best suited to predict disease, both in terms of specificity and sensitivity. The answer in turn depends on a better understanding of mechanisms and ultimately causality. Here, the concepts of quantitative in vitro to in vivo extrapolations12 and adverse outcome pathways13 will help to frame these issues and develop a better suite of bioassays. Until now, ecotoxicological studies have provided cause for optimism as they have shown an excellent agreement between wastewater-induced oestrogen receptor activation in vitro and markers of fish feminisation in vivo.14

The major practical limitation of the EDA approach is the identification of truly unknown chemicals, which is challenging when using mass spectrometry because it relies on the availability of reference spectra. Accordingly, these can only be generated from analysing analytical standards, that is, one has to know in advance which chemicals to search for. Notwithstanding recent advances in non-target mass spectrometry, the fast and unambiguous identification of a large number of chemicals in a complex sample remains the bottleneck of EDA and other applications (eg, metabolomics). Here, nuclear magnetic resonance is better suited to identify unknown compounds but depends on a high amount and purity of the analyte, which is rarely obtainable from complex samples.

These limitations notwithstanding, environmental toxicology demonstrates the feasibility and usefulness of EDA. For instance, Rostkowski et al15 screened bile from rainbow trout exposed to wastewater for antiandrogenic activity in vitro. They found in the fish several unexpected EDCs, including chlorophene and triclosan, which contributed with about 50% to the observed effects. Compared to that, the well-known BPA and nonylphenol only accounted for <1%. In epidemiological studies, the endocrine activity has been determined in vitro as a biomarker integrating the total burden of EDCs. For instance, an increased breast cancer risk in leaner women was related to the oestrogenicity detected in adipose tissue.16 More recently, serum oestrogenicity was strongly associated with breast cancer risk in a case-control study.17 Along the same line, the antiandrogenic activity of placenta extracts was associated with the risk of urogenital malformations.18 Fully implementing the effect-based approach in epidemiology will enable the identification of EDCs inducing the in vitro effects and potentially contribute to the health outcomes. Finally, this will create a more holistic picture of the human exposure to EDCs.

Old and new acquaintances in unexpected places

EDA can also be applied to search for EDCs in unexpected places by screening different sources of exposure. For example, we discovered recently that plastic baby toys used to soothe teething ache contain and leach EDCs. We detected in vitro oestrogenic and antiandrogenic activities leaching from 2 out of 10 baby teethers. Using non-target gas chromatography coupled to mass spectrometry, we identified methylparaben and propylparaben as compounds causing the endocrine activity in one product.19 The parabens readily leached to water, simulating the baby's saliva, implying that an exposure is likely. Interestingly, the compounds inducing the antiandrogenicity detected in the second product remain unidentified. This again highlights that we are potentially exposed to more EDCs than we are currently aware of. While the EC has recently banned parabens from certain baby cosmetics, our work shows that known EDCs from unexpected sources escape the regulatory as well as the parental view.

Another case is the widespread contamination of bottled water with EDCs, whose toxicity can be detected in bioassays.20–23 As in the case of environmental samples, the usual chemical suspects cannot explain the observed toxicity. Although searching for truly unknown compounds is a formidable challenge, it can produce valuable insight: For instance, we used non-target chemical analysis and detected between 2000 and 3000 organic chemicals in bottled water. We also identified a novel antioestrogen, which bears some structural resemblance to phthalates but has never been analysed before.24 Given the popularity of drinking bottled water (260 million m³ are consumed p.a. globally), an extensive exposure of human populations is plausible, yet no toxicological data are available. This case highlights that characterising chemical exposures based on few selected EDCs can be too short-sighted: It provides a false impression of safety in case these compounds are not detected but other so far unidentified EDCs are present.

We need to widen our view

Given the complexity of the exposome, investing the majority of our material and intellectual resources in the study of a handful of EDCs that are already well investigated is too simplistic. By continuing this business-as-usual strategy, we are missing the chance of more comprehensively understanding the exposome, establishing links between complex exposures and disease, and ultimately protecting public health. Accordingly, we need to widen our view to so far unknown EDCs. With the availability of high-throughput bioassays and non-target chemical analysis, the tools to discover EDCs that are relevant in terms of exposure and hazard are at hand. Now, the scientific community, including reviewers and funders, needs to overcome the inertia and approach the unknown.