Non-monotonic dose-response (NMDR) relationships are more frequently reported today in experimental studies than they were 10 years ago [1–3]. The endocrine disrupting chemicals (EDCs) are regularly associated with NMDR relationships. Until recently, NMDR relationships were not considered plausible, and thus they were not published, reported, or interpreted as relevant biological phenomena. An increasing number of scientists think that NMDR relationships represent a toxicological reality, but supplementary effort is required to revisit the Paracelsus principle of “the dose makes the poison”. Consequently, it is important to avoid rejecting studies merely because they do not match the classic toxicological concept of monotonicity. Before affirming that such phenomenon exists in a study, a screening analysis of individual NMDR profile is required to reveal better understanding of the underlying biological mechanism (s) involved and support plausibility of the observed NMDR.

The term “NMDR” describes a dose-response relationship characterized by a curve whose slope changes direction within the range of tested doses. Non-monotonicity represents a challenge to fundamental concepts in toxicology and risk assessment. Indeed, environmental risk assessment approaches used by regulatory agencies around the world were developed on the basis of a methodology published by the National Academy of Sciences [4]. For the hazard characterization step, it is generally accepted that once detectable, a response of an organism to a toxicant increases proportionally with the level of exposure until reaching an upper-limit or maximal-effect level (E_max) beyond which increasing toxicant dose will not increase the response (known as a monotonic dose-response). To assess the dose-response relationship of a chemical, several doses are typically tested to define the no observed adverse effect level (NOAEL) and/or the lowest observed adverse effect level (LOAEL). The NOAEL is considered a conservative default threshold below which a chemical is not expected to induce adverse effects, irrespective of the dose [5].

In contrast to this well-accepted hyperbolic or curvilinear dose-response relationship, which when plotted as toxicant effect (ordinate) and the logarithm (log10) of dose (abscissa) results in a sigmoidal shaped curve, a NMDR relationship may present as a bell-shaped profile. This profile, also called inverted-U shape, is characterized by responses at intermediate dose (s) and a decreased response or no response observed at low- and high-exposure levels. Also observed in the literature are U-shaped profiles with the highest responses at low- and high-exposure levels. In such cases, the standard notion of threshold, as defined by the NOAEL concept, is not inclusive of all potentially harmful effects. As clearly stated by Vandenberg et al. [3], “if a non-monotonic relationship occurs between the doses tested in traditional toxicology studies (i.e. the NOAEL or NOEL) and the calculated “safe” or reference dose, this would still have serious implications for risk assessment”. Thus, the standard approaches used for setting safe human and environmental exposure levels by extrapolation from high dose testing might not be applicable in case of NMDR profiles. There is a class of toxicants (i.e., EDCs) for which NMDR relationships, have been experimentally described with relative high frequency compared to other chemicals [6–11]. The EDCs act through several modes of action to cause effects on sensitive tissues. As a result, EDCs can commonly act on many physiological systems in addition to endocrine tissues. Effects observed in sensitive target tissues can also be caused by exposure-related changes in mechanisms governing normal negative feedback regulation of endocrine tissues and hormone secretion; such effects may also influence the dose-response relationship depending on exposure duration. With increasing awareness, more studies are being specifically designed to address whether the dose-response relationships for EDCs and other chemicals are described with an NMDR curve. As a result, NMDR relationships are being reported with an increasing frequency, which may have consequences for risk assessment. As reported by Vandenberg et al. [3], regardless of mechanistic underpinnings of each NMDR, their existence alone challenged traditional means of risk assessment.

The work described in this manuscript aimed to (1) perform a focused literature review, (2) extract a representative set of observed NMDR relationships for some EDCs and (3) use that information to develop a methodology to assess whether such a reported dose-response relationship is sufficiently reliable in order to be used in risk assessments. The screening analysis method evaluating the likelihood and the plausibility of the biological mechanisms involved was applied for each identified NMDR. Then, a stepwise decision tree was developed as a tool to standardize analysis for data reliability of observed in vivo NMDR relationships for risk assessment.

The purposes of this work were to review the literature to identify NMDR relationships observed for some EDCs, to develop a methodology to assess whether those dose-response relationships were sufficiently reliable for use in risk assessments. In this qualitative approach, a judgment on the quality of the reviewed studies was not introduced, except for the case study on BPA. Rather, the aim was to derive tools that allow consideration of NMDR relationships in risk assessments. In recent years, NMDR profiles have been reported in the literature with an increasing frequency from a variety of in vitro and in vivo toxicological models involving substances that affect hormonal systems [10, 74, 75]. In an extensive review by Vandenberg et al. [10], they reported hundreds of examples of possible NMDR relationships for more than 20 natural hormones and more than 70 putative EDCs. Those examples were from studies performed in cultured cells, on whole animals or on human. It was not a surprise that such relationships were observed because of the complexity resulting from the many modes of action through which EDCs may influence the actions of hormones. Along with the potential for complex pharmaco/toxicodynamic influences (e.g., expression of varying levels of multiple receptors and possible interactions with native ligands), the critical feedback mechanisms involved with regulation of hormonal systems creates a level of increased complexity. Dose-response relationship for EDCs would reflect this complexity and would likely result in a non-monotonic dose-response. Studies using several hormone-sensitive cell lines have shown that NMDR relationships can result from a variety of mechanisms [9, 76]. However, that type of relationship is not exclusive to EDCs and is also observed for chemicals substances that do not act on the endocrine system, and can be elicited by non-chemicals stressors, such as ionizing radiations [7, 77].

This work intentionally considers the in vivo and the in vitro studies in which the authors of identified publications claimed that the dose-response relationships observed during their experiments had an NMDR profile. Here, the aim was not to compare monotonic dose-response relationships with NMDR profiles, but to evaluate the plausibility of non-monotonicity when assessing the dose-response relationship from all available data. Indeed, NMDR profiles are still a controversial issue, especially regarding their usefulness for risk assessments [3, 78].

In this current study, 51 experimental studies describing 170 NMDR profiles concerning EDCs and/or natural hormones were selected. Out of these 51 investigations, the two epidemiological studies [15, 16] were excluded since the scoring criteria procedure was not applied on them.

For the risk assessment point of view, the reproducibility of the effects displaying NMDR profiles should be taken into account because it supports the validity of an observed relationship, and improves the confidence in the biological observations reported. Considering reproducibility of dose-response relationships, most of the in vivo studies reporting NMDR profiles did not lead to further experiments by other scientific teams which prevent to assess their reproducibility. When studies appear to have been replicated, there were often variations in the experimental designs. This can be illustrated by the in vitro studies of Kochukov et al. [28] and Wozniak et al. [34] that both show U-shaped relationships for the effects of BPA on prolactin release from rat pituitary cells. However, due to differences in experimental systems, differences in the shape of the resulting concentration-response curves are observed that result in major differences in the apparent potency of BPA.

To some degree, the observation of a NMDR for a compound at different endpoints can increase confidence in the validity of NMDR relationship. In the Marmugi et al. study [50], 18 NMDR profiles related to different metabolic endpoints were reported: regulation of plasma insulin (1 NMDR relationship) and changes in gene expression related to lipid biosynthesis (17 NMDR relationships). In a weight of evidence approach, these observations are considered with lower score compared to NMDR relationships reported from 2 independent studies. It was not the aim of the current study to collect and compare all of the available measurements concerning one compound for the same endpoints. The literature search, as part of the current study, focused only on studies claiming NMDR relationships, which did not afford the opportunity to retrieve studies that do not report NMDR relationships. Therefore, it is currently unfeasible to establish with confidence the degree to which specific NMDRs are reproducible.

The scoring assessments used for this current study have been adapted from those previously used by Calabrese and Baldwin and Calabrese ans Blain [12, 14] for the quantitative identification of hormesis. We consider these criteria to be a useful starting point to develop a more general approach for analysis of NMDR relationships. As a result of the current study, we have identified some possible limitations of the approach in its current form. For example, the zero equivalent point (ZEP) criteria (Scores A and B) is an arbitrary definition applicable to hormetic responses, which requires reversal of the response and a complete return to the control value. However, this value is not always observed or even expected for EDCs. Future efforts should be made to elaborate more specific criteria that are independent, but inclusive of hormesis, that would result in a generalizable approach. Additionally, it will be necessary to develop a scoring criterion that considers dose range and the number of dose levels.

The specific nature (curve shape) of an NMDR relationship observed from a toxicological assessment has important ramifications for the risk assessment. In different cases, NMDR may be observed in concentration ranges for which the effect is still adverse. Thus, U-shaped dose-effect relationships could be observed with maximal effects at the extreme doses of the U-shape. In this case, it may be expected that at a dose far lower than the lowest tested dose levels, the effect will decrease again, because infinite toxicity cannot be observed at dose levels close to the zero dose. Conversely, bell-shaped dose-effect relationships may be observed with minimal effects induced by the two extreme doses that are not equal to the control response. In this case, at very high doses, unspecific effects will, for example, appear because the organism would be completely overwhelmed by the substance. Thus, it appears credible that U-shaped and bell-shaped dose-response relationships might correspond to two parts of a same dose-response relationship. At high doses, the substances could impair biological processes by non-specific mechanisms (e.g., choatropic effect, membrane disruption, unspecific binding to proteins). However, because a complete profile is rarely available, we are proposing two ways to use NMDR in risk assessments. For bell-shaped profiles, a NOAEL or a benchmark dose could be extrapolated by modeling the ascending part of the dose-effect relationship. For a U-shaped profile, it could be recommended to experimentally verify, or to model, whether the exposure levels are in the same range of doses for which an NMDR relationship is observed. The main improvement in the testing strategy would be to conduct a better investigation of the dose-response relationship by testing more doses, especially in the lower dose range, as it was proposed during the workshop on “The low dose effects and non-monotonic dose responses for endocrine active chemicals: Science to practice”, held in Berlin, Germany, on September 12–14^th, 2012 [79]. These proposals could be included in the revision of OECD’s guidelines. If more dose groups are used, then the benchmark dose approach might be preferred because of the possible reduction in the number of animals per dose group. This approach was also proposed at the European Food Safety Authority’s 17^th Scientific Colloquium on Low Dose Response in Toxicology and Risk Assessment, held in Parma, Italy, on June 14 and 15, 2012 [2].

The relevant EDC modes of action that explain these NMDR relationships are still poorly investigated. Of the BPA studies fully considered, only 1 study explored the hypothesized mechanisms of action [44]. In a recent review from Vandenberg et al. [10], several mechanisms producing NMDR relationships were discussed, but for most of them only hypotheses were proposed which could be considered purely speculative, and their biological basis could not be fully assessed. Mechanisms involved in a non-monotonic profile are often described in vitro, but defining the in vivo mode of action could be much more complicated.

A significant biological plausibility of NMDR does not systematically imply the causality of adverse effects. For the purpose of risk assessments, the endpoints and the potential impact on human health must be considered when defining adversity. Determining whether an effect is adverse can be difficult for subtle or small-magnitude effects. In some situations, as previously mentioned, it would be reasonable to think that some of the effects characterized as presenting an NMDR might have a linkage to compensatory effects and not to endocrine disruption, leading to an adverse effect [7]. However, some of the endpoints identified in this study (Table 3) might be considered to be adverse. Moreover, adversity should also be considered regarding the exposure period. Indeed, there are windows of sensitivity in which specific subpopulations may be more vulnerable [80, 81]. In animals, EDCs have the potential to cause reproductive or developmental toxicity or several types of cancers in hormonally responsive organs such as the uterus or the mammary gland in females and the prostate in males [82, 83]. Thus, biological effects can be reversible or weakly severe in adults, but can be serious and damaging if exposures occur during specific critical life period (often prenatal or perinatal). For some substances, the NMDR profiles are observed exclusively when exposure occurs during a sensitive period of development (i.e., fetuses) when hormone-sensitive tissues such as uterine are most susceptible to the effects of EDCs. Consequently, during the fetal windows, low doses may impair physiological mechanisms responsible for NMDR relationships that do not exist during adulthood [9, 76, 81, 82].

For this current work, we developed a methodology to consider individual NMDR relationships in a risk assessment context. We expanded a stepwise decision tree to assess the likelihood of NMDR relationships reported. This analysis can be completed with a weight of evidence approach to evaluate all the data available for the substance of interest. This approach enables assessing the consistency of the dose-response relationship. The steps in this approach involve assembling the relevant data (either positive or negative), evaluating these data for quality and relevance, and integrating the different points of evidence to support conclusions concerning specific properties of the substance. Thus, the relevance and the impact of NMDR relationships reported on EDCs can be better understood and incorporated into the risk assessment method to determine the potential impacts of such substances on human health. Another important point to determine is the extent of experimental exposure (the range of exposure level in which an NMDR relationship occurs). The literature review performed for this current study enabled an analysis of the NMDR relationship profiles reported by the authors.