The short answer, if that is all you came for
| Group | Meaning | Agents |
|---|---|---|
| 1 | Carcinogenic to humans | 135 |
| 2A | Probably carcinogenic | 98 |
| 2B | Possibly carcinogenic | 329 |
| 3 | Not classifiable | 498 |
There is no group meaning "does not cause cancer." Group 4 held one substance in its history and was abolished in 2019, so a Group 3 classification is not a clearance.
"Limited evidence" does not mean weak. It is a defined term: a causal interpretation is credible, but chance, bias or confounding could not be ruled out.
And the group tells you hazard, not risk. Tobacco smoking is Group 1. Salted fish is Group 1. Both correct, not comparable dangers. The category describes the strength of the evidence that something can cause cancer, never the size of the effect.
A Category Is Not a Verdict
"IARC Group 2B" appears on this site and across the internet as though everyone knows what it means. Almost nobody does, including people who quote it confidently in both directions. It is used to frighten and it is used to reassure, and the same four words can do either job depending on who is holding them.
So before any classification on this site is worth anything to you, here is the machinery behind it: who decides, out of what, what the words mean, and the four things a classification deliberately does not tell you.
This takes about ten minutes to understand once, and then every carcinogen headline you ever read again is legible.
What IARC Is, and What It Is Not
The International Agency for Research on Cancer is the cancer research arm of the World Health Organization, based in Lyon. Its Monographs programme convenes working groups of scientists who review the published evidence on a single agent and reach a classification. The results are published free.
Four things it does not do
- ●It does not set exposure limits. No safe dose, no threshold, no parts per million.
- ●It does not perform risk assessment, meaning it does not estimate how many cancers an agent causes at real-world exposures.
- ●It does not regulate anything, ban anything, or approve anything.
- ●It does not rank agents against each other. There is no league table inside a group.
All four of those are other bodies' jobs. IARC answers one question only: can this cause cancer in humans, under some conditions? That question is called hazard identification, and keeping it separate from everything above is the single most useful thing on this page.
The Groups, and How Many Are In Them
Current as of 3 July 2026, covering Monograph Volumes 1 to 142:
| Group | Meaning | Agents |
|---|---|---|
| 1 | Carcinogenic to humans | 135 |
| 2A | Probably carcinogenic to humans | 98 |
| 2B | Possibly carcinogenic to humans | 329 |
| 3 | Not classifiable as to carcinogenicity | 498 |
1,060 agents in total. The list changes at every working group meeting, so the live version is the authority.
Look at the shape of that table before anything else. The largest category, by a wide margin and holding nearly half of everything ever evaluated, is Group 3: we cannot say. Only 135 agents in sixty years of work have reached Group 1.
That is the answer to "IARC says everything causes cancer." It does not, and its own numbers say so.
Group 4 No Longer Exists
There used to be a fifth category, Group 4, "probably not carcinogenic to humans." In its entire history exactly one agent was ever placed there: caprolactam, an industrial chemical used in making nylon.
The 2019 revision of the Preamble, which is the rulebook the working groups follow, removed Group 4 altogether and moved caprolactam into Group 3.
This matters for how you read everything else. There is now no category that means "does not cause cancer." The system can conclude that an agent does cause cancer, probably does, possibly does, or that the evidence does not permit a conclusion. It has no way to issue a clearance, and after 2019 it no longer pretends to.
So anyone citing a Group 3 classification as evidence of safety is not reading the system slightly wrong. They are using a category that was specifically designed not to mean that.
Where the Evidence Comes From
A working group weighs three separate streams, and the group an agent lands in depends on how they combine.
- ●Human evidence. Epidemiology: studies of actual people with actual exposures.
- ●Animal evidence. Experimental studies, where dose is controlled and the result is not confounded by everything else a human life contains.
- ●Mechanistic evidence. Whether a biological pathway by which the agent could cause cancer has been demonstrated, and whether that pathway operates in humans.
The three grades, which are defined terms
Human and animal evidence are each graded. These are not casual adjectives; they have specific meanings that are worth knowing precisely, because the whole system turns on them.
Sufficient. A causal relationship has been established.
Limited. A causal interpretation is credible, but chance, bias or confounding could not be ruled out with reasonable confidence.
Inadequate. The available studies do not permit a conclusion regarding the presence or absence of a causal association.
Read "limited" again. It does not mean weak or doubtful. It means the finding is believable and the study design could not exclude the alternatives. That is a long way from nothing, and it is routinely reported as nothing.
How the Groups Are Actually Built
Simplified, but this is the logic:
| Group | Typical basis |
|---|---|
| 1 | Sufficient evidence in humans. |
| 2A | Limited in humans plus sufficient in animals. Or inadequate in humans, sufficient in animals, and strong evidence that the mechanism also operates in humans. |
| 2B | Limited in humans and less than sufficient in animals. Can also rest solely on strong mechanistic evidence. |
| 3 | Inadequate in humans, with inadequate or limited evidence in animals. |
The 2019 change worth knowing: mechanistic evidence was given more power. An agent can now be moved up a group on the strength of a demonstrated biological pathway, and Group 2B can be reached on mechanistic grounds alone.
That cuts both ways, honestly. It means the system can act on a credible mechanism before the human epidemiology has accumulated, which is a precautionary strength. It also means some 2B classifications rest on laboratory biology rather than on anybody actually getting cancer, which is worth knowing when you see one quoted.
The Mistake Everyone Makes: Hazard Is Not Risk
This is the one that matters most, and the one that makes nonsense of most reporting.
Hazard is whether something can cause cancer at all, under some conditions. Risk is how much cancer it causes at the amount you actually meet. IARC answers the first and is explicit that it is not answering the second.
Tobacco smoking is Group 1. Salted fish is Group 1. Both classifications are correct, and anyone who concludes those are comparable dangers has misread what the category claims. A group tells you the strength of the evidence that an agent can cause cancer. It tells you nothing about the size of the effect.
Which means a classification on its own cannot answer the question you actually have. For that you need the effect size, the dose, and the route, and those live in the studies rather than in the category. It is why every entry on this site gives a relative risk or an odds ratio where one exists, rather than stopping at the group.
The reverse error is just as common and more consequential. "Only Group 2B" is used to wave something away, when 2B can mean a credible human signal that no study has yet been able to clean up. Neither direction is served by the category alone.
Where the Criticism Is Fair
A page that only defended this system would be doing the same thing it accuses the headlines of.
- ●Agencies disagree. The best-known case is glyphosate, which IARC placed in Group 2A in 2015 while other regulators reached different conclusions. Part of the divergence is genuine scientific disagreement and part is that the bodies are answering different questions: IARC asks whether an agent can cause cancer, while a regulator asks whether it does so at approved use levels. Both answers can be correct and they will not match.
- ●The literature IARC reviews is the literature that exists. If an exposure has been poorly studied, or studied mostly by the industry that profits from it, that shows up as "inadequate evidence" and lands in Group 3, which then gets quoted as reassurance.
- ●Agents are selected for review because they are already suspected. This is the honest answer to "everything is a carcinogen." The 1,060 evaluated agents are not a random sample of the world; they are the ones someone had reason to question. The proportion landing in Group 1 says nothing about the proportion of all substances that cause cancer.
- ●Classifications lag. Evaluation happens when a working group is convened, which can be decades after an exposure becomes widespread, and re-evaluation happens when evidence accumulates. Coffee was reclassified in 2016. Saccharin left the list in 1999. The list is a snapshot of an argument, not a settled fact.
None of that makes the Monographs unreliable. It makes them what they are: the most careful public summary available of what is known, at the moment it was written, about agents someone thought worth examining.
How to Read One in Practice
Six questions, and they take less than a minute once you know them.
- ●Which group, and in which direction is it being used? Someone quoting it to alarm you and someone quoting it to reassure you are both leaving something out.
- ●Which evidence stream carried it? Human epidemiology, animal studies, or mechanism alone. A 2B built on mechanism is a different claim from a 2B built on human data.
- ●Which cancer? Classifications are frequently site-specific, and the site is usually dropped in the retelling.
- ●What is the effect size? Not in the group. Go to the studies.
- ●What exposure, by what route, at what dose? Also not in the group.
- ●When was it last evaluated? And has the evidence moved since.
A classification is a starting point that someone has done an enormous amount of careful work to hand you for free. It was never meant to be the end of the question, and the people who wrote it say so in the Preamble.
What was withheld from you was never the finding. It was the key to reading it.
Sources
All of it is free to read. The Preamble is the rulebook, and it is the document that answers most arguments about what a classification means.