The “Low Dose” Hypothesis: Validity and Implications for Human Risk

Abstract

In the late 1990s, a “low dose” hypothesis was proposed based on studies that purported to show that hormonally active environmental agents were causing a variety of effects, mainly reproductive and developmental, at “low doses.” The supporters of this hypothesis claim that traditional “high-dose” toxicity studies are not adequate to assess adverse effects from these hormonally active agents in that they do not detect effects that are occurring at “low doses.” In addition, it is claimed that these “low dose” effects are occurring at levels comparable to those to which humans are being exposed. These claims have been controversial and expert panels evaluated the evidence behind them in the early 2000s. Although these panels generally concluded that such “low dose” effects were not conclusively established, proponents of the “low dose” hypothesis assert that a large number of more recent studies now provide clear support for their hypothesis. This review carefully examines both recent and older studies that have been cited to support the “low dose” hypothesis, including their relevance for the human population. These include in vivo and in vitro laboratory studies as well as a very limited number of epidemiological investigations. Based on the evidence, it is concluded that these “low dose” effects have yet to be established, that the studies purported to support these cannot be validly extrapolated to humans, and the doses at which the studies have been performed are significantly higher than the levels to which humans are exposed.

Background

In the early 1990s, as a result of publications linking reproductive and developmental problems in wildlife to hormonally active synthetic compounds in the environment, concerns were raised about the possibility of similar effects in humans. It was suggested that male and female reproductive and developmental problems in human populations reflected the effects of exposures to the same types of compounds, often called endocrine disruptors. Examples of such purported effects included decreased male sperm counts, altered male-female sex ratios at birth, and increased incidence of breast and testicular cancers. However, closer examination of the data revealed that assertions about the occurrence of such reproductive effects, e.g., increases in male reproductive tract problems and altered sex ratios, did not withstand close scrutiny. Similarly, the purported linkages between environmental agents and other effects, such as breast cancer, were not borne out in large-scale epidemiological studies (Safe 2005).

In the late 1990s, the earlier claims were replaced by assertions that these hormonally active compounds were causing a variety of effects on experimental animals at “low doses” and that these effects had not been detected earlier because the standard toxicology studies had been performed at high doses. The effects observed in this “low dose” research were often not the same ones claimed earlier, e.g., impaired reproductive function, but rather more subtle effects that some scientists claimed were harbingers of more serious ones. On the basis of these observations and the claim that the “low doses” used in the studies were similar to the doses to which humans are exposed in their daily lives, as well as in utero, these researchers concluded that humans are currently at risk. These assertions are parts of what has become known as “the low dose hypothesis,” the subject of this report (vom Saal and Sheehan 1998). Scientists who support this hypothesis do not believe that traditional high dose toxicology studies provide accurate representations of the toxicity of agents, particularly their toxicity at “low doses.”

Although a variety of hormonally active compounds have been claimed to have “low dose” effects, public attention has been focused on just a few, especially bisphenol A (BPA). One reason for this is that BPA exposure may occur in both adults and children through food and drink. This compound has also been the subject of the most intensive study by the “low dose” researchers and the studies on BPA have been used as the bases for the critical conclusions of these researchers. A recent review article, for example, postulates that the extensive literature on the “low dose” effects of BPA affirms the need for a new risk assessment paradigm (vom Saal and Hughes 2005). In light of this research focus, and the similarity of claimed effects among BPA and other “endocrine disruptors,” this article will utilize the literature on the effects of BPA in a critical analysis of the “low dose” hypothesis. Other chemicals that are claimed to produce “low dose” effects will be discussed briefly in the Toxicity Assessment section.

Definition of “Low Dose”

Although, on the surface, the term “low dose” suggests a specific part of the dose spectrum, namely, very small doses, this is not the case. As it is defined in the “low dose” hypothesis, it can refer to a very wide range of doses spanning many orders of magnitude because it is framed in terms of the relationship between the experimental dose and the lowest dose found to cause effects in higher dose experiments; i.e., the lowest observable adverse effect level (LOAEL). Thus, for example, “low dose” for a compound with a LOAEL of 10 mg/kg/day would be 100 times higher than “low dose” for a compound with a LOAEL of 0.1 mg/kg/day. Supporters of the “low dose” hypothesis generally define “low dose” as any dose below the LOAEL and this is the definition that will be adopted for the purposes of this report.

Hormesis and “Low Dose”

The idea that agents can have effects at doses much lower than those shown to cause adverse toxicological effects is not new. For example, pharmaceuticals are well known to have beneficial effects at “low doses” and adverse effects at high doses; often these high doses are orders of magnitude greater than those that produce benefit. In fact, the drug discovery process emphasizes identifying pharmaceuticals that have as large a margin of safety as possible; i.e., the highest possible ratio of the dose that causes adverse effects to the dose that provides benefit.

In recent years, an extensive review of the literature has shown that a wide variety of chemicals cause different effects at doses below those traditionally used for risk assessment purposes, e.g., the LOAEL, than they do at much higher ones. The term “hormesis” has been used to describe this phenomenon. The literature suggests that the hormetic effects occurring below the LOAEL are characterized by nonlinear dose-response curves; the curves are either J-shaped or inverted U-shaped. Because of the complex nature of the curves, studies of hormesis may require more data than other toxicological investigations (Calabrese 2005).

Hormetic effects can be either stimulatory or inhibitory. They are generally thought to represent transient overcompensation of the organism to exposure to an agent. This response is believed to help assure that the organism can successfully deal with any changes induced by the exposure and that the organism will be better prepared for subsequent exposures to this agent. Thus, hormesis is a normal process by which organisms deal with the multitude of both natural and synthetic agents to which they are exposed on a daily basis. In most cases this overcompensation will result in benefits to the organism but it is also possible that the outcome will be neutral or even detrimental in some cases (Calabrese 2005).

Although the results of a few of the “low dose” studies cited in support of the “low dose” hypothesis show characteristics of hormesis, e.g., J-shaped or inverted U-shaped dose-response curves, it is often not clear whether a hormetic response has been demonstrated because of the variability in response from study to study and/or the lack of appropriate data. However, in light of the hormesis literature, it is not surprising that changes are observed in organisms exposed at “low doses.”

Previous Evaluations of “Low Dose” Studies

From 2000 to 2002, scientific panels were convened in the U.S. and Europe to carefully evaluate the evidence for “low dose” effects, because of failures to replicate the results of some “low dose” studies, especially those on BPA. One group, the U.S. National Toxicology Program (NTP) Endocrine Disruptors Low Dose Peer Review Panel, concluded that it was not “persuaded that a low dose effect of BPA has been conclusively established as a general or reproducible finding” (Melnick et al. 2002; NTP 2001). Another group, a Scientific Committee of the European Commission, concluded that “studies on the reproductive and developmental effects of bisphenol A … do not support low-dose effects” (European Commission 2002a). A more recent evaluation of “low dose” effects of BPA by an expert panel concluded on the basis of detailed scrutiny of 19 studies that “the results … do not suggest that these small positive or negative changes in biological endpoints result in neoplastic or negative reproductive outcomes” (Gray et al. 2004). A very recent update of this study critically evaluated more than 50 additional studies and concluded that “the weight of evidence does not support the hypothesis that low oral doses of BPA adversely affect human reproductive and developmental health” (Goodman et al. 2006).

The Current Evaluation

It has been claimed by the “low dose” advocates that new research findings published since the governmental reports of 2001 and 2002 provide convincing evidence of a variety of estrogenic effects at doses below the LOAEL, thus supporting the “low dose” hypothesis (vom Saal and Hughes 2005). To evaluate this claim and thus the “low dose” hypothesis, basic precepts of scientific validity will be applied to examination of the data. The most important of these is reproducibility, i.e., the same results are seen from the same causes each time a study is conducted. In toxicological studies, this means that not only are the same effects seen but also that the dose response for these effects is the same from study to study. Related to this precept is consistency, i.e., the results all fit a pattern. This means that results from studies across species and under a variety of conditions fit in with a common explanation. A third criterion is that the studies are conducted properly, i.e., they include proper controls and are performed under appropriate experimental conditions. In the studies under consideration here, this latter criterion means that they must include more than one dose so that a dose response can be established and also that they must be performed for a long enough period of time to establish whether the effects are transient or permanent. The evaluation of the degree to which these criteria are met will include a discussion of the assertions that deviations from these criteria can be explained by reference to known factors that introduce variability into the results.

Because the claims under consideration here include not only that effects have been established in experimental animals but also that the results in these animal studies are applicable to humans, additional criteria must be applied. These criteria include whether or not the conditions under which the experiments have been performed are relevant to humans. Addressing this includes consideration of whether the route of exposure is relevant to human routes of exposure and whether the doses administered in the experiments are similar to those to which humans are exposed. Another consideration is whether or not the disposition of the chemicals in humans is similar to that in the experimental species, and also whether the likely sites of action are similar in the two species.

The degree to which these criteria are met will be the focus of the critical assessment of the claims of “low dose” effects and their applicability to humans. As will be demonstrated in this report, this assessment will show that the general conclusions arrived at in earlier evaluations remain valid. It is still the case that the “low dose” studies of BPA lack consistency and reproducibility and also that there is little support for the contention that the “low dose” results suggest humans are at risk from very low exposures to BPA or other hormonally active substances.

In addition, the assertion that the “low dose” results imply that the basic principles of risk assessment must be reexamined will be evaluated. Related to this, the claim that such a reevaluation will lead to lower acceptable exposure values, more stringent regulation, and increased public health will also be addressed. It will be shown that the assertions are not substantiated and that the implications for regulation and public health have not been demonstrated.

TOXICOLOGICAL ASSESSMENT OF “LOW DOSE” EFFECTS

Types of Chemicals Studied by “Low Dose” Researchers

Consistent with the background described earlier for the “low dose” hypothesis, the focus of its proponents has been on synthetic chemicals that have been labeled as “endocrine disruptors.” They include a variety of halogenated organic compounds including dioxin, polychlorinated biphenyls (PCBs), polybrominated diphenyl ethers (PBDEs) and DDT, phthalates, nonyl- and octylphenol, and BPA. The range of effects associated with these chemicals is quite broad so that it is not possible to consider them as a toxicologically coherent set of compounds. Even though it might be possible to assign them the same label, i.e., “endocrine disruptors,” this does not imply a common mechanism of action. Indeed, the traditional risk assessments of these chemicals reflect significant chemical-to-chemical differences in the modes of action deemed to be of most toxicological significance. For some chemicals, such as dioxin, the current risk assessment approach is based on the assumption of no threshold whereas for others, such as phthalates and BPA, a threshold is assumed. This poses a problem for applying the “low dose” hypothesis to the former class of chemicals since it is not possible to define a “low dose” if there is no threshold (LOAEL).

Because of the large number of chemicals identified as of concern, it is not surprising that there has not been equal scrutiny of all of the above chemicals, and classes of chemicals, by the “low dose” researchers. The most intensive investigation has been applied to those that are said to have estrogen-like activity and the one chemical in this group that has been the subject of most research by the “low dose” hypothesis proponents is BPA. Because the data on BPA have been used as the strongest arguments in favor of the “low dose” hypothesis, this article will focus on an evaluation of the validity of the BPA studies.

Review of Studies of “Low Dose” Effects

Studies of “low dose” effects of BPA have almost exclusively been performed on laboratory animals, especially rats and mice, or on in vitro systems utilizing cells in culture (vom Saal and Hughes 2005). The laboratory experiments have been performed on a range of animal species and strains and at a wide array of doses, using a variety of exposure routes and techniques and have investigated a large assortment of endpoints. Many of these end points are biochemical or morphological in nature rather than focusing on function. For example, the studies may examine DNA synthesis in reproductive cells or prostate weight in animals exposed in utero rather than fertility or reproductive success. Thus, these effects do not necessarily have an adverse impact on function, even in the animal subjects under investigation. Although there have been a very few epidemiological investigations claiming to show similar effects in humans, there are serious questions about the interpretation of the results and none of these studies have been replicated. In sum, as can be seen from the following detailed discussion, the data do not support the existence of consistent, reproducible, adverse effects from “low dose” exposures to “endocrine disruptors,” including BPA.

Reproductive and Developmental Effects

By far, the majority of studies of “low dose” effects have focused on reproductive and developmental end points. In part, this is because high doses, e.g., greater than 50 mg/kg/day of BPA, have been shown to affect these end points, although generally only at doses high enough to cause systemic toxicity. From a variety of studies it appears clear that at least some of these agents, e.g., BPA and some phthalates, are capable of causing estrogenic effects at such high doses (e.g., Tyl et al. 2002).

The reproductive and developmental research performed by “low dose” scientists has examined possible effects on both males and females in both animal and in vitro systems and over a wide range of doses (vom Saal and Hughes 2005). In BPA studies, doses administered to whole animals have varied from 0.002 μg/kg/day (Peknicova et al. 2002) to 40,000 μg/kg/day (Aloisi et al. 2001) (compared to the LOAEL of 50,000 μg/kg/day). Animals have been dosed with BPA in food, by gavage, using an implanted pump or an implanted capsule, by subcutaneous injection, and by intraperitoneal injection, as well as by other routes. BPA has been dissolved in a variety of vehicles including corn oil, olive oil, peanut oil, sesame oil, ethanol, and DMSO. Animals have been exposed prenatally, perinatally, as well as postnatally, for varying durations, in different experiments (see references in vom Saal and Hughes 2005; Gray et al. 2004; Goodman et al. 2006).

Whole-Animal (In Vivo) Studies

End points that have been assessed in “low dose” animal studies of the effects of BPA exposure during gestation include body weight, prostate weight, uterine weight, sperm density, sperm production, testis weight, anogenital distance, testosterone levels, uterine epithelial cell height, mammary gland morphology, time of estrus, day of vaginal opening, number of live pups/litter, and sex ratio of offspring (vom Saal and Hughes 2005). Many of these end points, e.g., body or organ weight, are not measures of reproductive or developmental functioning, but rather changes that may or may not be indicators of subsequent effects on functioning. Although an end point by end point discussion of the results is not possible within the framework of this paper, a discussion of some representative results will illustrate the strengths and limitations of the conclusions that have been drawn from this database.

One good example is the claim that prostate weight in male offspring is increased by in utero BPA treatment (Nagel et al. 1997). Although careful attempts have been made at reproducing these studies, the results have not borne out the initial findings (Ashby, Tinwell, and Haseman 1999; Cagen et al. 1999). Another example is the claim that sperm numbers and production are decreased in adult males exposed to BPA (Sakaue et al. 2001). Again, these effects are not seen consistently in all studies (Ashby et al. 2003). In addition, neither sperm characteristics nor prostate weight in male offspring were affected in multigenerational studies of BPA (Ema et al. 2001; Tyl et al. 2002). Similar disparities have been found among studies of effects on females; e.g., some studies show that uterine weight is increased by BPA treatment (Al-Hiyasat, Darmani, and Elbetieha 2004) and others indicate no change (MacLuskey, Thajszan, and Leranth 2005). Thus, although effects have been detected in some experimental animals under some conditions, it appears that there is no clear pattern of reproductive and developmental changes.

Even if it is accepted that effects are occurring, there is no consistency in the dose at which they are seen to occur or in the response of the experimental animals to changes in dose. Indeed, in a number of experiments only one dose has been administered, making it impossible to determine if there is a dose response, much less what it is (Farabollini et al. 2002; Howdeshell et al. 1999). In some studies, the lower dose causes an effect and the higher one does not (Akingbemi et al. 2004); in others, the opposite is true and effects only occur at the higher dose (Al-Hiyasat, Darmani, and Elbetieha 2004). In yet other studies, hormesis appears to be occurring, i.e., different effects are seen at different doses within the “low dose” range (Chitra, Latchoumycandane, and Mathur 2003). Further, some studies indicate that effects are transient and occur at only a particular time, e.g., a certain number of days postnatally (Nikaido et al. 2004). Because the time frame of many studies is limited, it is not possible to determine whether the effects observed in these studies are transient, and thus of limited toxicological significance.

Last, a number of BPA studies by “low dose” researchers and others that measured reproductive functioning have found no effect. These includes studies at a wide range of doses from 0.2 to 3000 μg/kg, using a variety of routes of exposure and performed on a number of different species and strains of experimental animals including DBA/1J, CF-1, and ICR/Jcl mice as well as Sprague-Dawley, Wistar, and F344 rats (Cagen et al. 1999; Honma et al. 2002; Kwon et al. 2000; Negishi et al. 2003; Ramos et al. 2003; Yoshino et al. 2004). These results have shown no effect on a number of measures of reproductive function, including percent of animals pregnant, number of offspring, litter size, and gender ratio.

In Vitro Studies

A number of studies have been performed using cultures of various types of cells, including yeast cells, breast cancer cells, prostatic cancer cells, and testicular cancer cells. In addition, some experiments have been performed using embryonic cells (Takai et al. 2001). In studies on BPA, the range of concentrations administered to these cell or embryo systems is quite large, from about 10⁻¹² to 10⁻⁴ M, representing a 100,000,000-fold difference from highest to lowest dose (Wozniak, Bulayeva, and Watson 2005; Takai et al. 2000). The end points investigated in cell cultures include measures of DNA synthesis and gene expression as well as binding to estrogen and androgen receptors. Studies on embryonic cells have focused on measures of the development of these cells in vitro and also after implantation into female mice.

One very significant limitation of these in vitro studies is they examine the effects of an agent that has not undergone the normal processes that occur in an organism after exposure, e.g., absorption, metabolism, storage, and excretion, processes that can alter both the nature and concentration of that agent. Further, the cells being exposed are not in their normal biological milieus and so are not likely to react as they would as part of a living organism. Thus, it is impossible to apply the in vitro results directly to whole animals, much less humans, either qualitatively or quantitatively. It is especially inappropriate to extrapolate concentrations used in in vitro studies to doses administered to whole animals or those to which humans are exposed. Making valid qualitative and quantitative extrapolations would require much more information about the effects of internal biochemical processes and cell milieus than is available. In addition to the limitations discussed above, the usefulness of the cell culture studies has also been hampered by the lack of consistent results from study to study and any clear dose response of the effects observed. As a result, the in vitro studies are mainly useful in attempting to understand the mechanism of action of the agent under study—recognizing that the caveats above suggest that there are serious limitations in achieving even that goal.

Epidemiological Studies

There was a great deal of publicity surrounding a recent study suggesting that maternal exposure to phthalates (compounds mainly used as plasticizers) affects anogenital distance (AGD) in male offspring (Swan et al. 2005). However, care must be taken in assessing the significance of this one study showing a correlation between an environmental agent and an effect. First, there have been other cases in which highly publicized epidemiological studies suggested that an environmental agent was associated with an effect but subsequent larger and more careful studies revealed that this association did not hold. Good examples are the studies claiming that polychlorinated biphenyls (PCBs) were related to breast cancer, i.e., Falck et al. 1992; Wolff et al. 1993, and subsequent reports showing no association, e.g., Laden et al. 2001 and Gammon et al. 2002. Second, the end point mentioned in the title of this study, anogenital distance (distance between the anus and the genitalia) in male humans, has not been well characterized, e.g., the extent of variability has not been established, and the biological significance, if any, of this end point is not known. The author of the only article examining the significance of measures of AGD concluded “Whether this particular measure, or other measures of AGD in humans, has any utility as markers of exposure in utero to hormonally active agents remains to be seen” (Salazar-Martinez et al. 2004).

Two other epidemiological reports claiming evidence of reproductive effects in humans include one describing a correlation between BPA levels and ovarian dysfunction (Takeuchi et al. 2004) and another describing an association between BPA levels and miscarriage (Sugiura-Ogasawara et al. 2005). Interestingly, in the first study, the authors suggested the difference in BPA levels was likely a result of the effects of hormones on BPA and not that the study provided evidence that BPA caused the ovarian dysfunction.

A number of questions also arise about the second study. For one, it compares the serum BPA levels in women who had a history of three or more first-trimester miscarriages with healthy nonpregnant women who had not experienced live births, infertility, and miscarriages, so it is not clear whether hormonal changes in pregnancy itself could be associated with BPA levels. In addition, it does not address the issue of whether the BPA serum level differences resulted from changes in the pregnant female or caused them. Thus, neither of these studies cited by “low dose” proponents provides significant support for the contention that BPA effects in humans have been demonstrated.

Other Effects

Although there are a number of studies of BPA that focus on other end points, including neurobehavioral (e.g., Negishi et al. 2003) and immunological (e.g., Yoshino et al. 2004) ones, they are much fewer in number than those directed towards reproductive and developmental changes. Although researchers have investigated a wide range of changes, there has been little effort at replicating individual studies and thus it is more difficult to assess the consistency of the studies from one species or strain to another or even from one laboratory to another. Also, as was the case for the reproductive and developmental studies, there are serious issues that limit the strength of the conclusions that are being drawn. These include the use of only a single dose for some experiments, the lack of a clear dose response, limits on the duration of studies—(even though other studies of this same class of effects shows that they may be transient), and use of nonenvironmentally relevant routes of exposure. Interestingly, a number of the studies of other types of effects also examine reproductive parameters and reflect the same types of results seen in the reproductive studies, including a lack of effect on a variety of reproductive function indices.

Study Variability

Overview

As has been noted in the above discussions of the “low dose” studies that have been performed, there has been a great deal of variability from experiment to experiment and, in a number of cases, effects cannot be replicated. The most parsimonious explanation of the lack of consistency is that the end points that have been chosen are inherently variable, i.e., sensitive to a variety of factors, and this variability may be even greater at lower doses. A study that is consistent with this interpretation is one showing that the results of “low dose” studies on the effects of BPA and other putative “endocrine disruptors” on reproductive organ weights are dependent on the weights of these organs in the control animals; in particular the effects are seen only when control weights are at the extremes (Ashby et al. 2004). This problem of lack of reproducibility is recognized by the “low dose” researchers, who have attempted to explain these inconsistencies in experimental results, in particular, to explain the lack of effects in some studies. Each of the major sources of variability proposed by these researchers will be examined in detail.

Estrogenic Content of Diet as a Source of Variability

One explanation provided for the variability in response to specific agents is that the presence or absence of natural estrogenic chemicals in the animal feed can have a large impact on whether or not changes are found (vom Saal and Hughes 2005). In particular, it is claimed the presence of such estrogenic compounds in feed can counteract the effects of the test compounds and so lead to data erroneously indicating that there are no effects.

There are, however, questions about the validity of this claim. For one, many of the studies of “low dose” effects do not specify the estrogenic content of the feed provided to the animals so it is not possible to determine if the correlation between effects and estrogenic content of diet is a general phenomenon (see studies referenced in vom Saal and Hughes 2005). In addition, a recent study clearly shows even diets that are claimed to be estrogen-free may produce estrogenic effects (Ciana et al. 2005). Further, feeds differ in other ways in addition to the estrogenic content so it would require more detailed studies to demonstrate whether the natural estrogens are the source of any of the differences (Odom et al. 2001). In the absence of consistent inclusion of data on the estrogenic content of feeds, questions as to estrogenic potency of feeds, and lack of data on the possible effects of other feed components, it is impossible to support the contention that variability in the estrogen content of feed is the source of response variability seen in conflicting studies.

Even if the claim that estrogenic compounds in feed can counteract “low dose” effects is valid, it suggests that studies of animals fed estrogen-free diets cannot be extrapolated to humans who consume a diet that includes of significant amounts of phytoestrogens. It also suggests that extrapolating results from animals whose diet does not have comparable estrogenic content to the human diet is also suspect.

Strain and Species Differences as Sources of Variability

Another explanation for the lack of consistency in results from one study to another, particularly the absence of effects in some studies, is that different species and strains react differently to estrogenic chemical exposures. For example, Sprague-Dawley rats are said to be refractory to such exposures, i.e., they are resistant to estrogenic effects, whereas other rat strains are responsive to estrogenic chemicals. There is evidence, however, that suggests that this explanation may not be valid. For example, the authors of a study comparing estrogenic effects in three different rat strains including the Sprague-Dawley rat conclude that there are not pronounced differences among rat strains in their estrogenic responses to chemical exposures (Diel et al. 2004). In addition, as discussed earlier, an absence of effects on reproductive functioning was observed in studies of a wide range of experimental animals, including the “responsive” F344 rat as well as the “refractory” Sprague-Dawley rat. Thus, the claim that inconsistencies in results can be accounted for by the differential sensitivity of particular strains of animals to estrogenic chemical exposures is not supported by the evidence.

Presence/Absence of Control as a Source of Variability

It has been claimed that studies that do not show “low dose” effects may be defective because they do not include an appropriate control (vom Saal and Hughes 2005). In this context, a control is a substance or condition that has been determined to result in (or not result in) a particular effect; in most of the cases under consideration in this report, some estrogenic effect. In particular, diethylstilbestrol (DES) is posited by the authors as the most appropriate control for BPA because it has been seen to produce some of the same effects as BPA, although at lower doses. The appropriateness of such a control is dependent on whether or not DES will consistently produce the effect in question at the appropriate dose. It is clear from the literature cited by the “low dose” proponents that DES does not mirror all of the “low dose” effects claimed for BPA or other compounds thought to be of concern nor does it produce effects consistently from study to study (Gray et al. 2004). In addition, the dose of DES required to produce the desired effect in the species/strain under study often has not been established. Thus, DES cannot serve as a universal control for BPA and studies that do not use DES cannot be claimed to be deficient on this basis alone. Indeed, the variability in results of exposures to the same compound from study to study suggests that more research needs to be done to identify appropriate controls. In the absence of this research, it is not possible to support the argument that the wide variety of results seen in “low dose” studies are due to the absence of an appropriate control.

Toxicity Assessment Summary

It is clear that “low dose” studies of BPA lack consistency and reproducibility and that most of the critical sources of the variability have not been identified. Further, those sources that have been identified have not been quantified. In light of this, it becomes very difficult to provide a scientifically valid answer to the question of what “low dose” effects have been established and for those that have, at what doses and at what times the effects occur—the basic questions that need to be answered if toxicity is to be adequately characterized. This characterization, in turn, is critical if the results of the toxicology studies are to be used for human health risk assessment. In a subsequent section, the implications of this conclusion on the usefulness of the “low dose” study results will be addressed.

HUMAN EXPOSURES TO “LOW DOSE” CHEMICALS

Overview

It is very difficult to determine daily exposures of human populations to chemicals in the environment, especially those chemicals present at very low levels. First, each individual leads a unique life so there is likely to be significant interindividual variability in exposure sources as well as the magnitudes, frequencies, and durations of exposure. The difficulty of making reasonable estimates of exposure is actually even greater than suggested by interindividual variability because such estimates depend on knowledge of the relative importance of various routes of exposure, e.g., food, water, air, knowledge that is often quite limited. Without such information, it is not clear exactly what data are needed in making the estimate or what data need to be collected to improve the estimates of exposure.

Because addressing the possible risks from environmental contaminants requires quantitative exposure values, estimates of exposure are made using whatever information is available, despite the uncertainties. Often these estimates are based on calculations of concentrations of the agent present in particular sources, e.g., drinking water, combined with measures of human behavior, e.g., amount of water ingested. In other cases, more direct measures of exposure are made by monitoring levels of chemicals in body fluids such as urine or blood. This type of monitoring, which provides direct measures of body levels, might seem to eliminate the need to know the sources and the contributions of each to the overall exposure. However, utilizing the monitoring data for estimating risk requires translation of body levels into exposure doses—a process that depends on knowledge of the time course of exposure and the quantitative relationship between fluid levels and exposures amounts and frequencies. Because such information is often lacking or incomplete, doses estimated by this method may not accurately reflect actual exposure values.

The case of BPA illustrates the difficulties of exposure assessment and some ways that these can be addressed. Originally, estimates of human exposure to BPA were based on studies of how much BPA could possibly leach, mainly from plastic bottles and coated cans, into food and drink, the two routes of exposure thought to be most significant. This led to estimates of daily exposure of approximately 1 μg/kg/day (European Commission 2002b). Subsequently, using values from urine monitoring and information about the disposition of BPA in the body, a range of daily exposure from 0.002 to 0.3 μg/kg/day was calculated. Even better information was collected in a study where actual measurements of BPA levels in food, liquid, and the environment were made for a small population. These values were combined with ingestion and inhalation amounts and led to a median intake of about 0.04 μg/kg/day with a high value of 0.07 μg/kg/day. Although the actual range of exposure is most likely broader than this because only one population was studied, it suggests that the 1 μg/kg/day originally estimated is most certainly a significant overestimate and a better estimate is between 0.01 and 0.1 μg/kg/day (Kamrin 2004).

“Environmentally Relevant” Exposures

It is quite common for investigators of “low dose” effects to include in their publications a statement to the effect that the doses being investigated are “environmentally relevant,” meaning they are comparable to the doses to which humans are exposed (e.g., Akinbemi et al. 2004; Takai et al. 2001; Adriani et al. 2003; Kawai et al. 2003). Often, this statement is made without further discussion or justification or is based on dated estimates of exposure. Because this assertion of relevance is so important to the claims of the “low dose” advocates, it is important to critically examine it in light of the human exposure analysis presented above.

To perform this critical assessment, a number of important criteria need to be applied. The first, and simplest, criterion is that the size of the dose administered in the “low dose” study is within the range of doses to which humans are estimated to be exposed. Comparison of the human exposure estimates presented in the last section to the experimental doses administered in “low dose” studies shows that this criterion is definitely not satisfied for the chemical of most interest, BPA. Although the broadest estimates of daily exposure are in the range of 0.002 to 0.4 μg/kg/day, values as high as 400 μg/kg/day are claimed to be environmentally relevant (Dessi-Fulgheri, Porrini, and Farabollini 2002). Indeed, almost all of the reported “low dose” effects of BPA resulted from exposures well above the estimates of human exposures.

The second criterion is that the route of exposure is environmentally relevant. Again, the BPA case provides evidence that this criterion is often not satisfied. Various whole-animal BPA studies are performed using routes of exposure that are clearly not environmentally relevant, e.g., doses administered using pumps or capsules implanted under the skin, subcutaneous injections, and intraperitoneal injections—yet authors of studies using nonenvironmental routes of exposure often claim they are relevant (e.g., Markey et al. 2005). In addition to the lack of relevance of such clearly nonenvironmental exposures, even exposures by more relevant routes may be difficult to interpret because it has been clearly shown that the route of exposure has a large impact on the levels of the chemical in the organism and, thus, on the amount of chemical reaching the site of action (Pottenger et al. 2000). Further, it is very clear from basic toxicological principles that in vitro exposures do not have direct environmental relevance because the exposures do not simulate those by which human cells are exposed.

In summary, the claim that doses administered in “low dose” experiments are “environmentally relevant” does not withstand even limited scrutiny. Careful examination of the “low dose” BPA studies clearly demonstrates that most of these do not satisfy even the most basic criteria by which such a claim can be judged to be toxicologically sound. Thus, the implication that such studies are applicable to humans because of the doses employed is not warranted.

RISK ASSESSMENT OF “LOW DOSE” EXPOSURE TO HUMANS

A very significant problem in evaluating the significance of “low dose” results for humans is that the data do not form a coherent whole, i.e., the results vary substantially from study to study as well as from species to species and strain to strain. As discussed in the previous section, attempts to explain these inconsistencies are not convincing. This, in turn leads to questions about the validity of extrapolating these results, even qualitatively, from one animal species or strain to another, and thus even more serious questions about attempts to extrapolate the data from rodents to humans.

Even if one were to accept that effects did exist and some were relevant to humans, it would be very difficult to make any scientifically valid quantitative extrapolations from the rodent studies to humans. This results from a number of factors, especially (1) the lack of a clear dose response in many studies, including those that involve administration of a single dose; (2) data showing that opposite effects occur at the extremes of the dose range or that show effects only at a middle dose and not at the high and low doses; (3) the lack of comparability of doses that cause effects in different species and strains—so it is impossible to know which dose, if any, is the one from which to extrapolate; (4) the utilization of routes of exposure and test systems that have no direct relevance to human environmental exposures; and (5) data showing the toxicokinetics and dynamics of BPA are different in humans as compared to the test animals (Volkel et al. 2002; Volkel, Bittner, and Dekant 2005).

In addition, as pointed out above, despite assertions that doses used in “low dose” studies are comparable to doses to which humans are exposed, it is clear that vast majority of studies of BPA have been performed at doses much higher than those to which humans are exposed environmentally. Thus, even if one ignored the serious flaws in making either qualitative and quantitative conclusions about humans from the “low dose” rodent studies, it is clear that many of these effects would not be seen in humans, since their exposures are much lower than those of the experimental animals.

Further, it is important to recognize that even if effects are established in laboratory studies, this does not imply that these effects are adverse. In addition to the studies on synthetic compounds, some “low dose” researchers have looked at naturally occurring phytoestrogens, such as resveratrol and genistein. The results indicate that these compounds can show many of the same effects seen in studies of synthetic chemicals. This is not surprising because these natural agents can interact with the organism in the same way as the synthetic ones, such as BPA. However, although it has been shown that the phytoestrogen resveratrol can cause changes similar to those of BPA, and has similar potency (Henry and Witt 2002), it also appears that this agent has beneficial properties, particularly in the prevention of cancer and heart disease (Upham et al. 2004; Wu et al. 2001). Thus, the overall toxicological impact of this compound appears to be beneficial although it can interact with the body in similar ways to BPA, a compound that is claimed to pose a risk to humans. This clearly illustrates that the changes seen in the “low dose” experiments are not necessarily adverse and a comprehensive toxicological assessment of these studies requires a more sophisticated analysis of the short- and long-term effects as well as effects on a larger variety of organ systems than is characteristic of “low dose” studies.

Thus, from a variety of perspectives, the existing data are not likely to provide sound bases for assessing the doses at which effects are expected to occur in humans, predicting whether or not these effects might occur from current environmental exposure levels, and predicting whether the overall impacts of these exposures will be beneficial or detrimental.

IMPLICATIONS OF “LOW DOSE” STUDIES FOR RISK ASSESSMENT AND RISK MANAGEMENT

Noncarcinogenic Risk

It has been suggested by “low dose” researchers that the current risk assessment paradigm needs to be modified to take into account effects occurring at much lower doses than those normally administered in standard toxicology tests. In particular, they propose that toxicological testing for noncarcinogens should routinely include studies performed at “low doses” and these studies should form the bases for newly calculated acceptable daily intake values (vom Saal and Hughes 2005). The exact way that these animal studies are to be used to make quantitative judgments about human risks is not specified but if current approaches are used, this means that the “low doses” would serve as the points of departure for the application of safety factors that are traditionally employed in the estimation of acceptable daily intake values. Thus, it is claimed that this new approach will lead to lower acceptable daily intake values and correspondingly to more stringent regulation. It is further asserted that these lower values will provide greater protection of public health.

First, as pointed out above, the various problems with the current “low dose” data sets, including inconsistencies and lack of clear dose response, make it very difficult to use such data for risk assessment. To establish a valid point of departure for a noncarcinogen, there must be a clearly defined effect that is agreed upon to be adverse and a clearly defined dose at which such an effect occurs. Neither of these conditions are currently met for the chemicals subject to “low dose” study, including BPA, which has been most intensely studied.

Second, although supporters of the hormesis dose-response model have also suggested that the risk-assessment paradigm takes lower dose effects into account, their approach is different (Calabrese 2005). They do not believe that you can take a point of departure at which an effect occurs and then add safety factors to this to arrive at an acceptable daily intake because, as is the case in many of the lower-dose studies, the dose response is J-shaped or inverted-U shaped. It does not seem appropriate toxicologically to incorporate safety factors based on the assumption that the response changes monotonically with dose when it does not. Thus, changing the paradigm to include “low dose” data in noncarcinogen risk assessment is not a straightforward process and likely to require significant reexamination of the basic risk assessment assumptions. The impact of such a reevaluation on acceptable intake values is not clear.

Third, those who have investigated the hormesis data from studies on a wide range of agents have shown that in some cases the lower-dose effects are beneficial or neutral (Calabrese 2005). This is not surprising because it is well known that a number of naturally occurring chemicals, such as trace minerals, are essential at lower doses but toxic at higher ones. Thus, the conclusion that “low dose” studies will necessarily lead to lower acceptable daily intakes of noncarcinogens does not necessarily follow from changes in risk assessment practice to take into account “low dose” effects.

Carcinogenic Risk

Carcinogenesis has not been an end point in the studies of the “low dose” researchers because cancer is an extremely low-frequency event at these “low doses” and therefore difficult to detect using any reasonable number of experimental animals. However, examination of the one lower-dose study that was performed using a very large number of animals to determine the carcinogenesis dose response at very low doses might shed some light on the implications of including “low dose” data in carcinogen risk assessments. This large-scale study showed that although the agent (acetylaminofluorene) in question clearly caused cancer at high doses, it was protective against cancer at very low doses, i.e., the cancer incidence was decreased compared to the control animals (SOT Task Force 1981). This suggests consideration of “low dose” carcinogenic effects may call into question the assumption of a linear “low dose” response that has been used by the U.S. EPA in estimating the risk of carcinogenesis. A reexamination of this assumption could well lead to an increase in acceptable intake levels.

CONCLUSIONS

Validity of “Low Dose” Effects

It is clear from the data summarized by scientists studying hormesis that a variety of chemicals, including natural compounds, can cause biological changes at what can be defined as “low doses.” This includes some estrogenic effects. However, the data presented by the advocates of the “low dose” hypothesis do not provide clear and consistent evidence for adverse estrogenic effects caused by the compounds they have investigated, including the one that has been most heavily researched, BPA. The results they present are not reproducible with regards to the species and strains that show each change, the doses at which these changes occur, and the dose response for each change. In addition, the explanations provided to explain the lack of consistency and reproducibility do not stand up under close scrutiny. Thus, it is not possible to predict with a reasonable degree of certainty the changes that will be brought about by a given dose of chemical in a specific species or strain, nor to extrapolate results from one species or strain to another; critical toxicological indicators of whether or not these changes have been adequately characterized.

In addition, the toxicological significance of each change has not been established. Biochemical or structural changes are not equivalent to adverse effects. Indeed, based on the data collected from thousands of studies that show lower-dose effects, including those studies investigating estrogenic effects, the changes observed may be beneficial, neutral, or adverse. Hormesis researchers have suggested that many of these changes represent adaptations of the organism to exposure and are not predictive of adverse effects (Calabrese 2005). This is borne out by the large number of “low dose” studies that show no changes in reproductive function (e.g., fertility or sex-ratio), whereas at the same time report a variety of estrogenic changes in the experimental animals. In addition, the possibility such “low dose” effects may be beneficial is supported by the findings that some natural estrogenic compounds may behave as inhibitors of carcinogenesis and may protect against heart disease (Upham et al. 2004; Wu et al. 2001).

It is also clear from the data that the wide variety of chemicals included under the “low dose’ hypothesis do not represent a class of compounds that cause identical or similar effects by a common mechanism or group of mechanisms, although there are some common effects shown by subgroups of these chemicals. Thus, generalizations about effects from one compound to another are not warranted.

Applicability of “Low Dose” Studies to Humans

As discussed in the previous section, the variability and lack of reproducibility in the data make it impossible to make either qualitative or quantitative predictions of the effects that occur in experimental animals exposed to a specific dose of a particular chemical. As a result, it is not possible to extrapolate the data from one animal species or strain to another. Extrapolation from experimental animals to humans is thus even less likely to be scientifically valid.

In addition, even if consistent results were achieved, the fact that many of the routes of exposure employed in the experimental animal studies have no relevance to human environmental exposures would preclude simple extrapolation of results from these studies to the human situation. Further, although it is often asserted that the dose levels used in the animal studies are comparable to those to which humans are exposed, careful examination of the data for BPA reveals the vast majority of these studies are performed at doses significantly higher, and often many orders of magnitude higher, than those to which humans are exposed.

More recently, the “low dose” researchers have cited a few epidemiological studies that are claimed to show a relationship between “endocrine disruptors” and indicators of adverse effects on human reproduction. However, questions have been raised about the significance of the changes that have been found and the associations posited. Even if one were to accept these at face value, none of them have been repeated so that their validity has yet to be established. In sum, there is no clear epidemiological evidence linking chemicals claimed to have “low dose” effects with such effects in humans.

Implications of “Low Dose” Results for Risk Assessment and Risk Management

The demonstration of effects at doses below those ordinarily administered in toxicology studies designed to be used for risk assessment can pose questions about the generality of the current approach for both carcinogens and noncarcinogens. However, as indicated earlier, these effects first need to be much better characterized before they can serve as the bases for risk assessments. If this can be accomplished, then a reexamination of the basic assumptions and approaches of both noncarcinogen and carcinogen risk assessment would be appropriate. This is the conclusion reached by researchers who have examined the phenomenon of “hormesis.” It is not clear what the result of such a reevaluation would be; however, it is very possible it would not support the assertion of the “low dose” proponents that this reassessment will lead to lower acceptable risk values. Because a number of the lower-dose effects are beneficial, not adverse, it could very well be that this reassessment could lead to higher acceptable risk values.

Overall Weight of Evidence with Regard to Risks from “Low Dose” Exposures

A critical examination of the literature cited by proponents to support the “low dose” hypothesis, particularly regarding the representative chemical BPA, and data on human exposures to both synthetic and natural estrogens, reveals that there is no compelling evidence that humans are at risk from current exposures to chemicals purported to exhibit “low dose” effects. This includes both carcinogenic risks and risks of adverse impacts on reproduction and development.