Evaluate the Evidence to Examine Non-Cancer Effects

This step for examining non-cancer health effects involves comparing site-specific doses and concentrations to observed-effect levels in available human and animal studies. You will perform this step when assessing non-cancer effects for contaminants that have health guidelines (looking at critical studies used to derive non-cancer health guidelines such as MRLs, RfDs, and RfCs). Health assessors can also perform this step for contaminants that do not have available health guidelines (see text box on the right for another approach ATSDR has used in site-specific cases when there is no health guideline). This step will help you to determine where your site-specific doses or concentrations lie in relation to the observed-effect levels. You will also see if differences between the study data and the exposure scenario you are evaluating make health effects more or less likely.

Two Key Components of This Step

1. Compare site exposure doses or concentrations with effect levels observed in the critical study.
2. Carefully consider study parameters in the context of site exposures.

Health assessors will examine the available evidence—in light of uncertainties—and give an overview of whether non-cancer health effects are likely under site-specific exposure conditions. In the non-cancer effects evaluation, you will consider various factors:

• Relevance of the critical study to the site-specific scenario. Examples:
  • Factoring in the duration of exposure reported in the critical study and the duration of exposure identified in the site-specific scenario.
  • Some reproductive or developmental effects result from in utero exposure, and as such, may not apply to children or adults. For example, if the critical study showed that in utero exposure resulted in birth defects, these effects typically do not apply to children or adults. (Some developmental effects identified from in utero exposure, however, may also apply to very young children who are also rapidly developing.)
  • In experimental studies, a high bolus dose of a chemical administered to an animal could have a different effect than low-dose chronic or intermittent exposures in humans.
• Publication date for different studies used to derive different health guidelines (e.g., a more recently published study might serve as a better basis for a health guideline by one agency compared to an older study that was used by another agency).

Using the information pulled and examined from the original study that served as the basis for the health guideline (see that process outlined in this section on identifying data studies) and from available studies for contaminants without health guidelines, consider numerically comparing the site-specific exposure doses or concentrations to the study’s health effect doses or concentrations. In addition to directly comparing the values, health assessors also can determine how close the site-specific doses or concentrations are to doses or concentrations associated with an effect. Make this comparison by dividing the critical study’s LOAEL, BMDL, or HED (as reported in Appendix A in the Toxicological Profile or IRIS) by the site-specific exposure doses or concentrations. Thus, example equations would include

• LOAEL/site-specific exposure dose,
• BMDL/site-specific exposure dose, and
• HED/site-specific exposure dose.

The resulting value of this equation is often referred to as a margin of exposure, or MOE. This evaluation should be done, but calling it an MOE in your document is optional. An MOE below 1, like in this example, indicates that your site-specific dose is above the health effect study level and will need further examination. A value above 1 means your site-specific dose is below the health effect study level. But even in cases where your site-specific dose is below the observed-effect level, you will still need to review the tox information to determine whether the site-specific dose is close enough to the observed-effect level. Use professional judgment to determine risk of harmful effects.

ATSDR does not use a NOAEL in the MOE calculation, but a robust NOAEL may be factored into other components of the in-depth evaluation. For instance, if an MOE is above but approaching 1 (i.e., an MOE above 1 indicates that the site-specific dose or concentration is lower than the health effect study level, but approaching 1 means it is getting close to that effect level) and requires closer evaluation, health assessors can visually examine their site dose or concentration with doses and concentrations from the LSE figure and table. You can see how the site dose or concentration compares to the NOAEL, in addition to the LOAEL and other effect levels presented (as discussed in this section on reviewing tox data for other health effects). During this step, you can examine the NOAEL to determine whether it is robust (e.g., based on a study with a large sample size). If the NOAEL is robust, factor it into your decision about whether health effects are possible. For instance, the NOAEL for arsenic is robust because it is based on a large human population. When comparing your site-specific dose to the arsenic NOAEL, since it is a strong NOAEL, a dose at or approaching the NOAEL may not be considered an issue, especially if the exposure duration was short.

When based on a LOAEL, BMDL, or HED, these comparisons provide added insight into how close site-specific doses and concentrations are to effect levels. When deciding the potential for harmful effects, an HED is better to use than a BMDL, which is better than a LOAEL. In general, a LOAEL is better to use than a NOAEL because it tells you how close the site-specific doses or concentrations are to harmful doses or concentrations. Keep in mind that NOAELs may not represent a true no-effect level because of limitations in study design.

When you have only a NOAEL and a LOAEL, if the site-specific exposures are well below a NOAEL based on a robust human study, the likelihood for harmful health effects in the exposed population would be low, if any. If, however, the NOAEL is based on a weak study, exposure doses near the NOAEL could be of concern because of several uncertainty conditions:

• Small number of animals tested
• Short duration of exposure
• Insufficient monitoring of endpoints
• Applicability of the effect in animals to humans
• Seriousness of the toxic response
• Frequency of toxic responses at the LOAEL

Quite often, a health guideline is derived from a LOAEL, BMDL, or HED, rather than from a NOAEL or an HED_NOAEL. The likelihood of adverse health effects increases as site-specific doses or concentrations approach or exceed a LOAEL, BMDL, or HED derived from either a human or animal study. In such cases, the text should describe the harmful effects that might be expected and include a statement about uncertainty in deciding the potential for harmful effects. This description should form the basis for selecting ATSDR’s conclusion categories.

In the absence of contrary information, assume that effects in animals will also occur in humans and that humans are as sensitive as or more sensitive to the chemical than are animals. Consider consulting a toxicologist with questions, such as which effect level is appropriate to use, and also refer to the section on examining the Study Applicability to Site-Specific Exposures. See an example in the box below for performing a numeric comparison of site-specific doses to observed-effect levels (also called the MOE).

The health assessor should also consider the relevance of the health guideline’s critical study to the site-specific exposure conditions and the exposed population. If the health guideline was based on a NOAEL in adults and the site population includes a sensitive population (such as children), the NOAEL might not apply to all population segments. The assessor should also consider the exposure scenario of the health guideline study. To examine how this might affect whether the studies relate to your site-specific exposures, health assessors should refer to the Toxicological Profile MRL worksheet for the discussion about the critical study to verify if there may be significant differences between that study and the site-specific exposure that is being assessed. Also, the assessor should consider the confidence in the health guideline study; if similar findings have been reported in other studies, your confidence is enhanced.

ATSDR does not have a specific scheme to determine the confidence (e.g., low, medium, high) in the MRL, the study, or the endpoints. Health assessors should use information from the Toxicological Profile or IRIS, their evaluation of evidence, and their professional judgment when deciding whether non-cancer health effects might be possible. Health assessors should always include a brief statement about the degree of uncertainty. The following are a few examples of areas of uncertainty:

• Pathways of exposure
• Human contact with the contaminant
• Estimating site-specific human doses
• Using animal data to predict effects in humans
• Robustness of human data

Study Applicability to Site-Specific Exposures

When deciding the possibility of harmful effects from site-specific exposure to a contaminant, it might be useful to review certain aspects of the study design used to derive the health guideline. Remember, you are not deciding whether the study is of good quality—that decision has already been made. With an MRL, the decision about study quality has been made by ATSDR, the agency profile manager, and the MRL workgroup. As for the study design, the study’s exposure conditions (e.g., duration, exposure route) might be important in deciding whether or not harmful effects are possible. For example, if the critical study used to derive the MRL was based on birth defects caused by in utero exposure, you know that children and adults exposed to this contaminant are not at risk of birth defects. Rather, the harmful effects identified by the study apply only to newborns. In this case, you should use the estimated doses for pregnant women to decide the risk of birth defects rather than the estimated doses for young children. Another way to illustrate this is to imagine you are looking at exposure to a chemical that has an MRL derived using a critical study based on over 20 years of exposure. Your hazard quotients (HQs) were above 1 for the youngest age group, birth to < 1 year, but not for the other age groups. You closely examine the study applicability related to the HQ above 1 for the youngest age group, and determine that health effects are not likely because their exposure duration (i.e., 1 year) above the MRL was much shorter than the duration from the critical study (in this case, > 20 years) used to derive the MRL.

You should also evaluate other factors, such as the number of animals per dose. For example, some animal studies might have identified a NOAEL based on 10 animals per dose. Keep in mind that a NOAEL based on 10 animals is a weak NOAEL and should not be considered a no-effect level. In this case, you should determine whether the study identified a LOAEL or used benchmark dose modeling to identify a BMDL or used dosimetric adjustment to identify an HED. In general, as explained previously in this section, you should use an HED over a BMDL and a BMDL over a LOAEL when deciding the possibility of harmful effects.

Determining Epidemiologic Study Suitability

Understanding the strengths and weaknesses of different epidemiologic studies (e.g., occupational studies, community-based studies) will help you determine their suitability in supporting and drawing conclusions about possible health effects. For example, you will sometimes find a human study where many individuals (e.g., several thousand) were used to identify a NOAEL. You could consider this as a robust NOAEL. These points are often highlighted in Toxicological Profiles, particularly in the MRL worksheet (Appendix A), if the human study is used to derive the MRL. Also, when applying the study results to your site scenario, keep in mind the criteria that epidemiologists use to evaluate the strength of human data:

• Were study objectives clearly presented?
• Was the study design appropriate for the research questions that are being asked?
• Was the methodology for data collection and analysis well-documented and able to be replicated?
• Were exposed groups and control groups (if applicable) properly selected and characterized?
• Were exposures adequately characterized?
• Was there sufficient length of follow-up (e.g., integration of a latency period) to allow for observance of toxicity or disease?
• Were the causes of disease and death confirmed?
• Were confounding factors and potential bias adequately considered?
• Was the sample size adequate to identify an effect?
• Were the appropriate statistical methods used?
• Were methods adequate for addressing missing data?
• Were study results clearly documented and easy to interpret (e.g., comparison populations, reference values), including study limitations?
• Was the study ecologic (i.e., focused on effects on groups rather than effects on individuals) so that a true causal relationship cannot be inferred?

Epidemiologists also use criteria to judge the causal significance of associations revealed in studies. You should indicate the intent of a study because epidemiologists cannot determine a causal relationship from incidence/prevalence studies and ecologic type analyses. Individual criteria, if met, can support a causal relationship but cannot prove it. The more criteria that are met, the more likely an observed health effect is causally related to the exposure under study. Use the following criteria [Fedak et al. 2015; Hill 1965] to evaluate and explain the strength of the evidence linking a contaminant with a health outcome of interest:

• Time sequence. Exposure must precede the onset of the disease. A logical sequence of events must be demonstrated.
• Strength of association. The magnitude of the relative risk in well-designed analytic studies (comparison of disease incidence in those exposed to incidence in those who are not) can be a valuable measure of the strength of the association.
• Dose-response relationship. The probability or severity of the effect should increase with increasing exposure intensity and duration.
• Specificity of association. If the effect is unusual and is specific to the studied exposure, a causal relationship is more easily demonstrated.
• Consistency. A relationship should be reproducible (i.e., observed in other studies or analyses).
• Biologic plausibility (or coherent explanation). The link between the cause and the effect should make sense biologically by what is known about the disease and the exposure under study. The findings should be validated by what is known about animal models.
• Experiment. Methods (e.g., laboratory tests, epidemiologic studies) used to test whether a decrease in disease occurs from stopping or intervening with a particular exposure may provide the most robust evidence for concluding if there is a cause-and-effect relationship.
• Analogy. The premise that when robust evidence supports a causal association between exposure to a specific chemical and a certain disease, scientists can use less robust evidence for a chemical known to be similar.

Questions and Examples to Consider When Evaluating Studies

Many factors affect the relevance of animal and human study data to site-specific exposures, but going over all of them in detail is beyond the scope of this manual. Consider the following types of questions when evaluating how study features might make the identification of harmful effects more or less plausible.

• Is the critical study based on human or animal data? Clearly, a study based on human data holds the greatest weight in describing relationships between an exposure and a human health effect. Fewer uncertainties exist about potential outcomes documented in well-designed epidemiologic studies. Even if the MRL is exceeded, estimated site-specific doses below a robust NOAEL reported in a human study could support a conclusion that adverse effects are unlikely. Remember, a weak NOAEL based on a small number of people or animals should not be considered a no-effect level. Before deciding whether harmful effects might be possible, consider the site-specific applicability and size of the exposed group. Also consider similarities and differences between available study data and your site-specific exposure conditions (e.g., exposure route, chemical form).
• For animal studies, how relevant is the dosing method to site exposures? The method by which the test animal received its dose (e.g., gavage/water, gavage/oil, water, food, vapor) influences the relevance of an experimental study to environmental exposures. Often, the exposure route in experimental studies is different from the route by which people living near a site could be exposed. Consider this example:
  • A laboratory study in which animals were administered a substance via gavage or drinking water might not directly apply to a soil-exposure scenario. This is because solubility often influences how much and how quickly substances are absorbed, which can affect the dose-response for the toxic effect. The chemical form of the substance tested in water and gavage can differ considerably from the form in soil. For similar reasons, a dietary animal study might not adequately represent exposures from drinking water. When scientific information is not available on the bioavailability of contaminants in a media (e.g., soil) compared to the media used in the animal or human study (e.g., water), ATSDR assumes that the contaminant is 100% bioavailable. These differences based on dosing methods, if known, can be covered in the uncertainty discussion in your document.
• For animal studies, how might dosing regimens influence the interpretation of the study data? Dosing regimens can also influence the absorption and ultimately the effects observed in experimental studies. Examine how closely, in relative terms, the study conditions match site-specific exposure conditions. Ask the following questions:
  • Were animals dosed continuously or intermittently?
  • Were animals dosed over the short or long term? For example, the same dose administered in the shorter term (e.g., 28 days) might produce different effects than those produced after a longer-term dose administration (e.g., 90 or 180 days). Because different dosing regimens can produce different effects or affect the severity of the observed effect, you can be more confident the more closely study data match site-specific conditions. If only intermediate dose data are available, state the uncertainties should you apply such data to chronic exposures. Do not apply acute exposure data to chronic exposure scenarios.
• Is the form of the contaminant in the selected study the same or different from the form detected at the site? The chemical form, including the valence state or salt complex of a substance can affect its bioavailability, its distribution within the body, and ultimately its toxicity. If study data are not available for the form of the substance present at your site, determine and explain in your document whether the chemical form at your site could be more or less bioavailable, or more or less toxic, than the form used in the study. Also include this point in the uncertainty discussion of your document. Below are two examples:
  • Per ATSDR’s Toxicological Profile for Uranium, the oral intermediate MRL for uranium (soluble salts) is derived from a drinking water study. This is an important consideration when estimating doses for the soil ingestion pathway. A data review indicates that the fractional absorption of soluble uranium compounds is significantly greater than insoluble uranium compounds. In weathered soils, insoluble uranium compounds will predominate. Therefore, using the MRL based on absorption from water to assess exposure to uranium in soil would lead to an overestimate of the absorbed dose, because of the reduced bioavailability of uranium in soil compared to water. In this situation, you will still use the MRL but will discuss this issue of different absorption rates across the gut as part of the document’s uncertainty discussion.
  • Based on ATSDR’s Toxicological Profile for Arsenic, most arsenic in fish is in an organic form, known as arsenobetaine (commonly called “fish arsenic”), which has little to no toxicity. Inorganic arsenic, which is considerably more toxic, makes up only a small amount of total arsenic in fish. Ideally, therefore, unless we have analytic information to the contrary, we assume that arsenic in fish is the non-toxic arsenobetaine. However, some shellfish have, in addition to arsenobetaine, small amounts of inorganic arsenic in their tissue. When evaluating arsenic in shellfish, health assessors need to review the literature to determine the percentage of inorganic arsenic in the type of shellfish sampled or request speciated arsenic data. You may need to estimate the dose of inorganic arsenic from eating certain shellfish based on the estimated percent of inorganic arsenic from a literature search or from the Toxicological Profile for Arsenic.
• Are the effects observed in animals expected in humans? If you are using dose levels from animal toxicity studies (e.g., in mice, rats, monkeys) to evaluate site exposures, determine whether any human or in vitro studies suggest similar toxic effects can occur in humans. Metabolism or mechanistic data can also help you assess whether observed effects are unique to, or different in, the study animal compared to humans. If these data do not exist, assume that similar effects would occur in humans. For instance, when an MRL is based on a critical toxic effect from an animal study, ATSDR considers the relevancy of that effect for human exposure during the development of that MRL value. The following are possible scenarios:
  • The metabolism of a chemical in animals could produce more or less toxic intermediates than in humans.
  • The metabolism in humans could occur by another pathway and produce more or less toxic intermediates.
  • Toxic intermediates could be produced at high levels of exposures that may be administered in the animal studies, but not at lower exposure levels that may occur in your assessment scenario. (See also discussion on toxicokinetics.)
• How relevant are other health endpoints? Health-based guidelines are typically designed to be protective of the most sensitive adverse health effect. However, you need to understand the range of other possible health effects that could occur at the exposure dose or concentration you have calculated for the chemicals in the site assessment. This additional comparison might also help address concerns about other health effects that community members have reported observing in their community.
  • For example, if an MRL is based on increased kidney weight in rodents, you should also evaluate the possibility of other harmful effects. This evaluation would require reviewing the LSE figures and tables in the Toxicological Profile to determine whether other organs and systems are also affected at the calculated site-specific doses.
  • This additional evaluation might also be useful in addressing concerns expressed by community members about harmful effects they believe are occurring in their community. Your evaluation might support their concerns or might be useful in explaining that those effects are not likely at the estimated site-specific doses (see section on Reviewing Tox Data for Other Health Effects).
• Does the bioavailability of the contaminant differ in the study matrix versus the environmental matrix being evaluated? The bioavailability of a contaminant influences how much is absorbed by the human body and ultimately the potential for harmful effects. The bioavailability of a contaminant is discussed in the toxicokinetics section of the Toxicological Profile.
  • Contaminants in solid matrices (e.g., soil) might be less absorbed in the digestive tract than the same contaminants in water because of the differences in the contaminant solubility within the soil matrix. There are in vitro assays that can measure the relative bioavailability (RBA) of specific metals (e.g., arsenic), which can be used to refine the estimation of the exposure dose. Remember that you must know the RBA to account for differences in absorption between two media (e.g., soil, water). Site-specific doses should not be adjusted using only the bioavailability value.
  • Some forms of a salt can bind tightly to soil, thereby reducing its bioavailability. For instance, some forms of arsenic bind tightly to soil and are therefore not readily absorbed in the digestive system. The form of arsenic in drinking water from the critical study used to derive the chronic oral MRL is more soluble and thus more readily absorbed (see EPA Recommendations for Default Value for Relative Bioavailability of Arsenic in Soil).
  • In general, ATSDR assumes that chemicals ingested in water, soil, beverages, and food are 100% absorbed or have the same absorption rate as studies used to develop health guidelines. So far, only arsenic in soil has been assigned an RBA less than 100% (i.e., 60%) in PHAST. Therefore, the estimated dose from soil ingestion is reduced by 40%. Health assessors may change this default RBA in PHAST only if they have site-specific information about the relative bioavailability for arsenic at their site. Note that EPA’s Integrated Exposure Uptake Biokinetic (IEUBK) Model for lead does not use an RBA value but rather uses the bioavailability measured usually from a bioaccessibility (i.e., solubility) test using site-specific soils. If using EPA’s IEUBK lead model, health assessors may change the default bioavailability of lead (i.e., 30%) if they have site-specific bioavailability data.

Describing Uncertainty

After describing possible health effects, health assessors will need to describe the uncertainty that exists in making decisions about health effects. Sometimes this uncertainty deals with animal or human study limitations, but uncertainty can stem from factors related to various other components in the PHA process, such as in the exposure pathway, the exposed population, and estimations of site-specific exposure doses and air concentrations.

When deriving health guidelines (MRLs, RfDs, RfCs) from human and animal studies, ATSDR and EPA use uncertainty factors (UFs) to account for the following:

• Variation in sensitivity among the members of the human population.
• Uncertainty in extrapolating animal data to humans.
• Uncertainty in using LOAEL data rather than NOAEL data.

A default for each individual UF is usually 10. However, if benchmark dose modeling or dosimetric adjustments are used, this UF of 10 is often reduced to 3. A modifying factor can also be used sometimes to account for limitations in the overall database of available toxicity data.

The MRL worksheet (Appendix A) in the Toxicological Profile will describe each critical study and the uncertainty factors used in deriving the MRL. The worksheet will also provide a summary of other studies considered for the MRL derivation. The MRL worksheet may also provide useful information about other sensitive health effects that might be expected. IRIS summaries also discuss uncertainties and EPA’s confidence in the critical study or studies used to derive RfDs and RfCs. Some of this information might be useful as you discuss uncertainty in your public health documents.