EPA Carcinogen Classification: Descriptors, Evidence, and IRIS

The EPA classifies chemicals for cancer risk under two systems that both remain in force: the 1986 letter groups (A through E) and the 2005 narrative descriptors that run from “Carcinogenic to Humans” down to “Not Likely to Be Carcinogenic to Humans.” Which label a chemical carries depends on how strong the human evidence is, what animal studies show, and what biological data explains how the substance behaves inside the body. That label then drives drinking water limits, air pollution rules, cleanup thresholds, and workplace hazard communication across the federal government.

The Current Narrative Descriptors

In 2005, the EPA moved away from single-letter labels and adopted a narrative approach in its updated Guidelines for Carcinogen Risk Assessment (EPA/630/P-03/001F).¹ Assessors now write a weight-of-evidence narrative that lays out the studies, how they fit together, and what uncertainties remain. Every narrative closes with one of five standard descriptors.

Carcinogenic to Humans is reserved for substances where convincing epidemiological evidence shows a causal connection between human exposure and cancer. Trichloroethylene, a common industrial solvent, carries this descriptor.²

Likely to Be Carcinogenic to Humans applies when the weight of evidence demonstrates carcinogenic potential but stops short of the certainty needed for the top descriptor.¹

Suggestive Evidence of Carcinogenic Potential signals that the data raises concern but is too thin to support a “likely” conclusion.¹

Inadequate Information to Assess Carcinogenic Potential means the available studies are too few, too limited, or too contradictory to support any conclusion.¹

Not Likely to Be Carcinogenic to Humans applies when the data is robust enough to conclude the substance does not pose a cancer hazard under the conditions studied.¹

The narrative format lets scientists explain how mode of action, study quality, and data gaps factor into the conclusion, context that a single letter could never convey.¹

Why the Old Letter Groups Still Show Up

The 1986 Guidelines for Carcinogen Risk Assessment, published at 51 FR 33992, sorted chemicals into five groups. Group A (Human Carcinogen) required enough human epidemiological evidence to establish a causal link. Group B (Probable Human Carcinogen) split into B1, where limited human evidence sat alongside animal data, and B2, where animal studies showed sufficient evidence but human data was inadequate or nonexistent. Group C (Possible Human Carcinogen) applied when animal evidence was limited and no human data existed. Group D (Not Classifiable) captured cases where both human and animal evidence were inadequate or conflicting. Group E (Evidence of Non-Carcinogenicity) applied when at least two well-designed animal studies in different species showed no increase in tumors, or when adequate human and animal studies agreed on that point.³

People often read these letters as a potency ranking, like school grades, but they were never that. A Group B chemical is not necessarily less dangerous than a Group A chemical; the difference is the type and amount of evidence, not how potent the substance is. That misreading was one of the reasons the EPA moved to narrative descriptors.

The agency did not go back and reclassify every chemical already assessed under the 1986 system.⁴ Older EPA assessments still carry Group A through E designations. Both systems are considered valid within the agency.

What Evidence Drives the Classification

Three broad categories of data feed every cancer classification, and the EPA weighs them together rather than treating any single category as decisive.⁵

Human Epidemiological Studies

Epidemiological data is the most directly relevant evidence because it tracks cancer outcomes in people who were actually exposed. Cohort studies follow an exposed group over time and compare their cancer rates to an unexposed group. Case-control studies work in reverse, starting with people who already have a diagnosis and looking back at their exposure history.¹ Consistent elevated cancer rates across studies produce strong evidence. Small samples, confounding variables, or inconsistent results weaken it.

Animal Bioassays

Long-term animal studies, usually in rodents, expose test subjects to a chemical at controlled doses over most of their lifetimes to see whether tumors develop.¹ These studies matter most where human data is scarce, which describes most industrial chemicals. Positive results across multiple species or tumor sites strengthen the case; a single study with marginal tumor increases in one species carries far less weight.

Mechanistic and Supporting Data

Beyond tumor counts, scientists examine how a chemical behaves at the cellular level. That includes genotoxicity tests for DNA damage, studies of how the body absorbs and processes the chemical, and comparisons with structurally similar compounds already known to cause cancer.¹ A chemical that produces tumors in animals and also directly mutates DNA will land in a higher category than one where the tumor mechanism is unclear.

Mode of Action and Dose-Response Modeling

Knowing that a chemical causes cancer at high doses in the lab doesn’t automatically say what happens at the low doses people encounter in daily life. The EPA uses mode of action analysis to bridge that gap, figuring out the biological steps through which a chemical produces tumors and then modeling risk at real-world exposure levels.¹

Two main approaches apply. When a substance directly damages DNA or triggers mutations, the EPA assumes any amount of exposure carries some degree of risk. This is linear extrapolation: the dose-response curve is treated as a straight line all the way down to zero. The same default applies whenever there isn’t enough mode-of-action data to justify an alternative.¹

When the evidence shows a non-mutagenic mode of action, say a chemical that causes cancer only by chronically irritating tissue, the EPA may use nonlinear extrapolation. That approach assumes a threshold below which the biological process leading to cancer doesn’t get triggered.¹ The distinction matters for regulation. Linear modeling produces lower permissible exposure limits because even tiny doses contribute some theoretical risk. Nonlinear modeling allows somewhat higher thresholds.

The quantitative outputs of these models are what regulators actually use. The oral slope factor expresses cancer risk per unit of daily dose, and regulators apply it to exposures through food and drinking water. The inhalation unit risk does the same for airborne exposure, expressed as risk per concentration in air. Both are upper-bound estimates set conservatively so that regulatory limits built on them err toward safety.⁶

Extra Protection for Early-Life Exposure

Children aren’t just smaller adults when it comes to cancer risk. The EPA’s Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens applies age-dependent adjustment factors when a chemical causes cancer through a mutagenic mode of action:⁷

Birth through age 2: cancer risk estimates are multiplied by 10.
Age 2 through 15: cancer risk estimates are multiplied by 3.
Age 16 and older: no adjustment.

The tenfold factor for infants reflects animal data showing substantially higher tumor rates from exposure during periods of rapid cell division and organ development. The threefold factor for older children is a midpoint estimate for the transitional years when growth is slowing but tissues remain more vulnerable than in adults.⁷ These factors apply only to mutagenic carcinogens. If a chemical causes cancer through a different mechanism, or chemical-specific early-life data exists, the standard adjustments don’t apply.

Where the Classifications Live: IRIS

Finished cancer classifications and their toxicity values are published in the Integrated Risk Information System (IRIS), the EPA database maintained by the Office of Research and Development. IRIS assessments feed regulatory decisions across the federal government.⁸ Each assessment moves through draft development, internal EPA review, interagency science consultation, public comment and external peer review, revision, a final round of agency and interagency review, and publication. The EPA targets roughly 26 months for a standard assessment and 39 months for a complex one.⁹

How a Classification Turns Into a Rule

A classification sitting in a database doesn’t protect anyone by itself. Its regulatory force comes from how other EPA programs and federal agencies use it. Program and regional offices combine IRIS toxicity data with their own exposure assessments to characterize risks for specific scenarios, and those characterizations drive enforceable standards.⁸

IRIS assessments inform the Maximum Contaminant Levels set under the Safe Drinking Water Act, which cap contaminant concentrations in public water supplies.¹⁰ They feed into the National Emission Standards for Hazardous Air Pollutants under the Clean Air Act, which limit industrial air releases. Under CERCLA (the Superfund law), a chemical’s cancer classification can influence its reportable quantity, the threshold at which a spill triggers mandatory federal notification. Benzene’s reportable quantity was reduced to 10 pounds specifically because of its carcinogenicity.¹¹

Workplace rules respond as well. OSHA’s Hazard Communication Standard requires chemical manufacturers, importers, and employers to classify chemicals for health hazards including carcinogenicity, and to communicate those hazards through labels and safety data sheets.¹² A change in EPA classification can ripple across all of these programs at once.

EPA vs. IARC vs. NTP

The EPA is not the only body classifying carcinogens, and this is where confusion often starts. Two other major systems operate in parallel: the International Agency for Research on Cancer (IARC), part of the World Health Organization, and the National Toxicology Program (NTP), run by the U.S. Department of Health and Human Services. All three may evaluate the same chemical and reach different conclusions because their criteria and evidentiary standards differ.

IARC uses a numbered group system. Group 1 means carcinogenic to humans, based on sufficient human evidence or a combination of strong mechanistic evidence and sufficient animal evidence. Group 2A (probably carcinogenic) requires at least two positive lines of evidence; Group 2B (possibly carcinogenic) requires only one.⁴ IARC focuses on whether a substance can cause cancer under any circumstances, meaning it identifies hazard, not the level of risk at typical human exposures. That distinction explains why IARC sometimes classifies substances as carcinogenic that don’t trigger regulatory action under other systems.

The NTP uses two categories: “known to be a human carcinogen,” which requires sufficient human evidence of a causal relationship, and “reasonably anticipated to be a human carcinogen,” which covers chemicals with limited human evidence, sufficient animal evidence, or convincing mechanistic data.¹³ The system is simpler but less granular; there is no NTP equivalent of the EPA’s “suggestive evidence” or “inadequate information” categories.

A rough alignment across the three systems:⁴

Confirmed carcinogen: EPA “Carcinogenic to Humans” / IARC Group 1 / NTP “Known”
Probable carcinogen: EPA “Likely to Be Carcinogenic” / IARC Group 2A / NTP “Reasonably Anticipated”
Possible concern: EPA “Suggestive Evidence” / IARC Group 2B / no direct NTP equivalent
Insufficient data: EPA “Inadequate Information” / IARC Group 3 / not listed by NTP
Probably not carcinogenic: EPA “Not Likely” / IARC Group 4 / not listed by NTP

These are approximate parallels, not exact equivalents. Each agency applies its own standards, reviews its own set of studies, and updates on its own timeline. A chemical labeled “Carcinogenic to Humans” by the EPA will not necessarily hold IARC Group 1 status, and vice versa. Checking all three databases gives the fullest picture of what the scientific community has concluded about any specific substance.

1
Environmental Protection Agency. Guidelines for Carcinogen Risk Assessment
2
U.S. Environmental Protection Agency. Trichloroethylene IRIS Assessment
3
U.S. Environmental Protection Agency. Guidelines for Carcinogen Risk Assessment – Section: IV. EPA Classification System
4
Agency for Toxic Substances and Disease Registry. Cancer Classification Systems
5
U.S. Environmental Protection Agency. Risk Assessment for Carcinogenic Effects
6
U.S. Environmental Protection Agency. EPA’s Approach for Assessing the Risks Associated with Chronic Exposure to Carcinogens
7
U.S. Environmental Protection Agency. Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens
8
U.S. Environmental Protection Agency. Basic Information about the Integrated Risk Information System
9
U.S. Environmental Protection Agency. Assessment Development Timelines
10
U.S. Congress. U.S. Environmental Protection Agency’s Integrated Risk Information System
11
eCFR. 40 CFR 302.4 – Designation of Hazardous Substances
12
Occupational Safety and Health Administration. Hazard Communication Standard 1910.1200
13
National Toxicology Program. Report on Carcinogens Process and Listing Criteria