An official website of the United States government.

We've made some changes to If the information you are looking for is not here, you may be able to find it on the EPA Web Archive or the January 19, 2017 Web Snapshot.


EPA's Approach for Assessing the Risks Associated with Chronic Exposure to Carcinogens

Background Document 2
January 17, 1992


The purpose of this Background Document is to provide a framework to interpret the information presented in the Carcinogenicity Assessment Section of the IRIS chemical files.

In the analysis of data regarding the potential human carcinogenicity of chemical agents, the U.S. EPA uses the approach described in its Guidelines for Carcinogen Risk Assessment (51 FR 33992-34003, Sept. 24, 1986). Detailed examples of how the Guidelines can be applied are found in two documents: (1) Health Assessment Document for Epichlorohydrin (EPA-600/8-83-032F, Dec., 1984, p. 7-32 to 7-48); and (2) OTS Assessment of Health Risk of Garment Workers and Certain Home Residents from Exposure to Formaldehyde, Appendix 4 (Apr., 1987).

The U.S. EPA approach to evaluating carcinogenicity data follows the general format set forth by the National Academy of Sciences (NAS) in its description of the risk assessment process (see Risk Assessment in the Federal Government: Managing the Process, Washington, DC: NAS Press, 1983). The risk assessment process is an activity that is independent of the process of formulating regulatory control options and is independent of the economic and political factors influencing the regulatory process. The U.S. EPA recognizes the distinction between these regulatory concerns (referred to as "risk management considerations" in the 1983 NAS report) and the risk assessment process.

In the NAS report, the four elements of the risk assessment process are defined as follows:

  1. Hazard identification, in which a determination is made as to whether human exposure to the agent in question has the potential to increase the incidence of cancer;
  2. Dose-response assessment, in which a quantitative relationship is derived between the dose, or more generally the human exposure, and the probability of induction of a carcinogenic effect;
  3. Exposure assessment, in which an evaluation is made of the human exposure to the agent. Exposure assessments identify the exposed population, describe its composition and size, and present the type, magnitude, frequency, and duration of exposure.
  4. Risk characterization, in which the exposure and dose-response assessments are combined to produce a quantitative risk estimate and in which the strengths and weaknesses, major assumptions, judgments, and estimates of uncertainties are discussed.

The Carcinogenicity Assessment Section of the IRIS chemical files is designed to supply information resulting from the hazard identification and dose-response assessment steps. To complete the risk assessment, one needs to develop estimates of exposure and combine these estimates with dose-response characteristics to develop estimates of risks. One then needs to characterize the risks by considering the strengths, weaknesses, assumptions, policies, and uncertainties associated with the case.



The purpose of the hazard identification step is to determine whether the agent in question poses a carcinogenic hazard in exposed humans. The major types of evidence bearing on this question are these: (1) human studies of the association between cancer incidence and exposure to the agent; and (2) long-term animal studies under controlled laboratory conditions. Also considered is supporting evidence such as short-term tests for genotoxicity, metabolic and pharmacokinetic properties, toxicological effects other than cancer, structure-activity relationships, and physical/chemical properties of the agent.

The agent's potential for human carcinogenicity is inferred from the available information relevant to the potential carcinogenicity of the chemical and from judgments as to the quality of the available studies. A weight-of-evidence approach is used by the U.S. EPA to classify the likelihood the agent in question is a human carcinogen. A three-stage procedure is followed. In the first stage, the evidence is characterized separately for human studies and for animal studies. Secondly, the human and animal evidence are combined into a presumptive overall classification. In the third stage, the provisional classification is adjusted upwards or downwards, based on analysis of the supporting evidence. The result is that each chemical is placed into one of the following five categories.

Group Category
A Human carcinogen
B Probable human carcinogen:
B1 indicates limited human evidence;
B2 indicates sufficient evidence in animals and inadequate or no evidence in humans
C Possible human carcinogen
D Not classifiable as to human carcinogenicity
E Evidence of noncarcinogenicity for humans


The dose-response assessment step in carcinogen risk assessment is to define the relationship between the dose of an agent and the likelihood of a carcinogenic effect on the assumption that the agent in question is a human carcinogen. The data obtained in the dose-response assessment is combined with that obtained in the exposure assessment to yield a numerical estimate of risk. Numerical estimates of risk can be presented in one or more of the following four ways: (1) unit risk; (2) the concentration corresponding to a given level of risk; (3) individual risk; and (4) population risk. The IRIS chemical files include only unit risks and risk-related air and water concentrations. The activity of numerical risk estimation is not dependent upon the likelihood that the agent is a human carcinogen, as characterized in the hazard identification process. Instead, it is an independent piece of information, to be used in combination with the hazard identification information in making regulatory decisions.

Dose-response assessment usually entails an extrapolation from the generally high doses administered to experimental animals or exposures noted in epidemiologic studies to the exposure levels expected from human contact with the agent in the environment. It also includes considerations of the validity of these extrapolations. Extrapolation is ordinarily carried out first by fitting a mathematical model to the observed data and then by extending the model (or a bound on the risks it predicts) from the observed range down toward risks expected at low exposure.

Dose-response assessment includes (1) selection of the appropriate data sets to use; (2) derivation of estimates at low doses from experimental data at high doses, using an extrapolation model; and (3) choice of an equivalent human dose when animal data sets are used. CHOICE OF DATA SETS

In addition to data quality, the choice of data sets to use for quantification included the following considerations:

  1. Human data are preferable to animal data;
  2. In the absence of appropriate human data, information from an animal species whose biological responses are most like those of humans (e.g., similar metabolism) is preferable;
  3. In the absence of the ability to identify such a species or to select such data, data from the most sensitive animal species/strain/sex combination are given the greatest emphasis;
  4. The route of administration which most resembles the route of human exposure is used. Where this is not possible, the differences in route are noted as a source of uncertainty;
  5. When the incidence of tumors is significantly elevated at more than one anatomical site by the agent, estimates of overall risk are made by determining the number of animals with tumors at one or more of these sites;
  6. Benign tumors are generally combined with malignant tumors, unless the benign tumors are not considered to have potential to progress to the associated malignancies of the same historgenic origin [see McConnell et al. (1986) for guidance]. CHOICE OF EXTRAPOLATION MODEL

Since risk at low exposure levels cannot be measured directly either by animal experiments or by epidemiologic studies, a number of mathematical models and procedures have been developed for use in extrapolating from high to low doses. Different extrapolation models or procedures, while they may reasonably fit the observed data, may lead to large differences in the projected risk at low doses. The choice of a low-dose extrapolation method in EPA assessments is dependent upon chemically specific information bearing on the mechanism of carcinogenesis and other relevant biological information, and not solely on goodness-of-fit to the observed tumor data. When data are limited, however, and when uncertainty exists regarding the mechanisms of carcinogenic action, models or procedures which incorporate low-dose linearity are preferred when compatible with the information available. EPA usually employs the linearized multistage procedure in the absence of adequate information to the contrary.

The first step of the linearized multistage procedure calls for the fitting of a multistage model to the data. Multistage models are exponential models approaching 100% risk at high doses, with a shape at low doses described by a polynomial function. When the polynomial is of first degree, the model is equivalent to a one-hit model, which produces an approximately linear relationship between dose and cancer risk at low doses.

In the second step of the linearized multistage procedure, an upper bound for the risk is estimated by incorporating an appropriate linear term into the statistical bound for the polynomial. At sufficiently small exposures, any higher-order terms in the polynomial will contribute negligibly, and the graph of the upper bound will appear to be a straight line. The slope of this line (formerly called the potency) is called the slope factor in the IRIS chemical files. Its units are (proportion of individuals with tumors) /mg/kg/day. Since the slope at higher exposures may differ from that at lower exposures, IRIS chemical files identify exposures associated with a risk greater than or equal to 1 in 100, as above the range where the slope factor in the file can be applied.

Other models that may be used for dose-response assessment include the Weibull, probit, logit, one-hit, and gamma multi-hit models. These models are defined in the Glossary of Terms. Except for the one-hit model, they all tend to give characteristic S-shape dose-response curves of many biological experiments, with varying curvature and tail lengths. Their upper bounds tend to parallel the curvature of the models themselves, unless a procedure has been devised to provide otherwise, as is the case with the linearized multistage procedure. The slope factor designated in the IRIS chemical files for these models is the slope of the straight line from the upper bound risk at zero dose to the dose producing an upper bound risk of 1%.

With regard to the spontaneous background rate of tumor occurrence, two alternative approaches have been used. Both are summarized by slope factors. The first approach defines "added risk" (AR) as the difference between the total response rate under a given exposure condition (dose d), and the background incidence in the absence of exposure (dose zero). The corresponding equation is AR = P(d) - P(0). The second approach, called "extra risk," (ER) can be described as the "added risk" applied to the portion of the population which did not show background tumors. The corresponding equation is ER = [P(d) - P(0)]/[1-P(0)]. "Extra risk" is the most commonly used approach, but the alternative approach, that of "added risk," is being explored by the U.S. EPA for its utility in certain circumstances, and has been used in several cases. When the background response is sizable, "extra risk" is larger than "added risk," and when the background is small, both types of risk are about equal. DETERMINATION OF HUMAN EQUIVALENT DOSES

When animal data are used as a basis for extrapolation, the human dose that is equivalent to the dose in the animal study is calculated using the assumption that different species are equally sensitive to the effects of a toxin if they absorb the same dose per unit of body surface area. This assumption is made only in the absence of specific information about equivalent doses for the chemical in question. Since surface area is approximately proportional to the 2/3 power of body weight, the equivalent dose is expressed in milligrams per (body weight raised to the 2/3 power) per day. In the calculation of human equivalent doses, the actual animal weight in the bio-assay is used whenever that information is available; otherwise, standard species weights are used. It follows that if the animal dose is expressed in units of mg/kg/day, the equivalent human dose (assuming a body weight of 70 kg), in the same units, is smaller than the animal dose by a factor of 13 for mice (weight of 30 grams) and 5.8 for rats (weight of 350 grams).

In using animal inhalation experiments to estimate lifetime human risks for partially-soluble vapors or gases, the air concentration (ppm) is generally considered to be the equivalent dose between species based on equivalent exposure times (e.g., a lifetime exposure to a 1 ppm). With regard to the inhalation of particulates or completely-absorbed gases, the amount absorbed per unit of body surface area is considered to be the equivalent dose between species. ROUTE-TO-ROUTE EXTRAPOLATION

In order to evaluate human risks for both air and water contamination when only one route has been tested in animals or evaluated in humans, additional assumptions, with corresponding additional uncertainties, may be introduced for route-to-route extrapolation. For this reason, the Summary Tables in the Carcinogenicity Assessment Section of the IRIS chemical files specify the routes of exposure that were used for the calculation of air and drinking water unit risks.


Quantitative risk estimates have several uses, and the expression employed should be tailored to each use. To apply these estimates exposures must be provided as daily averages over a lifetime.

The slope factor is the cancer risk (proportion affected) per unit of dose. In the IRIS chemical files the slope factor is expressed on the basis of chemical weight [milligrams of substance per kilogram body weight per day (mg/kg/day)]. The slope factor can be used to compare the relative potency of different chemical substances on the basis either of chemical weight (as above) or moles of chemical (m moles/kg/day).

To estimate risks from exposures in food, one multiplies the slope factor (risk per mg/kg/day), the concentration of the chemical in the food (ppm) and the daily intake (in mg) of that food. The total dietary risk is found by summing risks across all foods.

For evaluating risks from substances found in certain other environmental sources, dose-response measures are expressed as risk per concentration unit. These measures are called the unit risk for air (inhalation) and the unit risk for drinking water (oral). The continuous lifetime exposure concentration units for air and drinking water are usually micrograms per cubic meter (ug/cu.m) and micrograms per liter (ug/L), respectively. If the fraction of the agent is absorbed from the diet for humans and animals differs, the U.S. EPA applies a correction when extrapolating the animal-derived value to humans.

For determining the concentrations of air or water at certain designated levels of lifetime risk (risk-specific concentrations), the U.S. EPA calculates the ratio of that level of risk to the unit risk for water or air. For example, one may want to know the water concentration corresponding to an upper bound risk of 1 in 100,000 (E-5) given a water risk of 4.0E-5/ug/L. This would be 2.5E-1 ug/L.

In summary, the quantities appropriate for calculating upper-bound risks for air, drinking water, and food are, respectively, the air unit risk (risk per ug/cu.m of air), the drinking water unit risk (risk per ug/L of drinking water) and the oral slope factor (risk per mg/kg body weight/day), corresponding to dietary intake risk.


The discussions of confidence presented in the IRIS chemical files reflect only the EPA Carcinogen Risk Assessment Verification Endeavor (CRAVE) Work Group's judgments about the ability of the risk measures derived from dose-response assessment on the agent to estimate the risks of that agent to humans. This judgment is based on consideration of factors that increase or decrease confidence in the numerical risk estimate. These factors fall into the following five dimensions:

  1. Appropriateness of data to estimation of human carcinogenic risk;
  2. Quality of study design;
  3. Strength of study results;
  4. Appropriateness of model application to the data; and
  5. Support of risk estimate by data from collateral studies.


McConnell, E.E., H.A. Solleveld, J.A. Swenberg and G.A. Boorman. 1986. Guidelines for Combining Neoplasms for Evaluation of Rodent Carcinogenesis Studies. JNCI. 76(2): 283-289.