In June 2016, the International Agency for Research on Cancer did something it rarely does: it removed a substance from its list of possible carcinogens. Coffee had been classified as Group 2B, possibly carcinogenic to humans, since 1991. A working group reconvened, reviewed the accumulated evidence, and reclassified it. In the same evaluation, it added a new Group 2A listing for something almost nobody had been worrying about.
What the classifications actually mean
The IARC groups are the most misunderstood thing in cancer communication, so this needs establishing first. IARC classifies hazard, not risk. It answers the question of whether something is capable of causing cancer under some circumstances, not how likely it is to do so at real-world exposure levels. This is why processed meat and tobacco smoking sit in the same group despite being vastly different in actual danger.
The distinction is worth an example. Sunlight and processed meat both appear on IARC's lists, and so does plutonium. The classification tells you how confident the agency is that a thing can cause cancer under some conditions. It says nothing about how much exposure it takes, how common the resulting cancer is, or whether the exposure you actually have carries any meaningful risk. A great deal of alarming coverage comes from treating a hazard classification as though it were a risk estimate.
Group 1 means carcinogenic to humans. Group 2A means probably carcinogenic. Group 2B means possibly carcinogenic. Group 3 means not classifiable as to carcinogenicity, which is where a substance lands when the evidence is inadequate to reach any conclusion, or when it consistently fails to show an effect.
The 2016 evaluation
The findings were summarised by Loomis and colleagues in The Lancet Oncology[1], with the full assessment published as Volume 116 of the IARC Monographs[2]. The working group reviewed more than a thousand studies covering coffee, maté and very hot beverages.
Coffee was moved from Group 2B to Group 3. The specific reasoning was that the evidence for an association with any cancer site was inadequate. The 1991 classification had rested largely on suggestive findings for bladder cancer, and by 2016 the accumulated evidence no longer supported that, with the earlier signal attributed substantially to inadequate control for smoking, the same confounder that distorted the early mortality research covered in our article on coffee and lifespan.
The working group went further than a neutral verdict. It reported that coffee consumption was associated with reduced risk of cancer of the liver and of the endometrium, findings consistent with the liver research covered in our article on coffee and the liver, where Kennedy and colleagues[3] found an extra two cups a day associated with 35% lower hepatocellular carcinoma risk.
What changed between 1991 and 2016
It is worth asking what actually shifted, since coffee did not change in those 25 years. Three things did.
The first was the volume and quality of evidence. By 2016 the working group had access to well over a thousand studies, including very large prospective cohorts that did not exist in 1991, which allowed far better separation of coffee's effects from those of the behaviours accompanying it.
The second was the handling of smoking. The 1991 bladder cancer signal is now generally attributed to inadequate adjustment for tobacco, the same confounder that made coffee look deadly in early mortality research. Coffee and cigarettes travelled together for most of the twentieth century, and statistical methods for separating them improved considerably.
The third was the framing. The 2016 group evaluated drinking coffee as a behaviour, and separately evaluated the temperature of hot beverages. Separating those two exposures is what allowed one to be cleared while the other was flagged, and it is a good example of a classification improving because the question was posed better rather than because the world changed.
The Californian labelling case
The most visible real-world consequence of the acrylamide question played out in California, where litigation under a state right-to-know statute sought cancer warning labels on coffee. A court initially ruled that coffee sellers had not demonstrated the absence of significant risk, and warning labels briefly appeared.
The state's own environmental hazard assessment agency then took the unusual step of proposing a regulation stating that cancer risk from drinking coffee does not require a warning, citing the weight of the epidemiological evidence, including the IARC evaluation. The regulation was adopted and the labels were not required.
The episode is a useful illustration of a general point about acrylamide claims: the compound is present, it is genuinely a probable carcinogen in the abstract, and the studies of people who actually drink coffee, which necessarily include whatever acrylamide they consumed, do not show elevated cancer risk. Where a whole-food epidemiology and a single-compound toxicology disagree, the epidemiology is usually the better guide to what the food does.
The finding that replaced it
The same evaluation classified drinking beverages at temperatures above 65°C as Group 2A, probably carcinogenic to the human oesophagus. This applies to any beverage: coffee, tea, maté, or hot water. The mechanism is thermal injury to the oesophageal lining and the chronic inflammation and cell turnover that repeated scalding produces.
The evidence came substantially from populations where drinking beverages extremely hot is customary, including studies of maté drinking in South America and of very hot tea consumption in parts of Asia and the Middle East, where associations with oesophageal cancer were observed independent of the drink involved. That independence is what pointed at temperature rather than at any beverage's chemistry.
The 65 degree rule, in practice
This is the most immediately useful thing in the whole evaluation, so it is worth making concrete. Coffee is typically brewed at around 90 to 96°C and served somewhere between 70 and 85°C. A cup left in ordinary room conditions falls below 65°C within roughly five to ten minutes depending on the vessel, the volume and whether milk was added.
Adding cold milk drops the temperature immediately, which is one reason a latte or a cup with milk poses less of an issue than black coffee taken straight from the pot. Insulated travel mugs work against you here, since they are designed to hold temperature and can keep a drink above the threshold for a long time.
The practical test needs no thermometer: if you cannot hold a comfortable sip in your mouth, the drink is too hot. That is roughly where the threshold sits for most people. Waiting a few minutes also improves what you taste, since very hot liquid suppresses the perception of sweetness and aroma, which is why professional cupping protocols evaluate coffee as it cools rather than immediately after brewing.
It is worth keeping the magnitude in proportion. This is a risk documented in populations habitually drinking beverages far hotter than most people find tolerable, and oesophageal cancer is not common in most countries. The point is not to be anxious about hot drinks, it is that letting your coffee cool for five minutes is free and there is no reason not to.
This is a genuinely useful, actionable finding buried under a headline about coffee. Sixty-five degrees is hotter than most people drink comfortably; freshly brewed coffee is typically served between 70 and 85°C and generally needs a few minutes to fall below the threshold. Waiting for a drink to become comfortable to sip is, on this evidence, the sensible practice, and it is also better for tasting the coffee, since very hot liquid suppresses flavour perception.
The acrylamide question
Acrylamide is a compound formed when foods containing certain sugars and the amino acid asparagine are heated to high temperatures. It appears in fried and baked starchy foods, and it forms during coffee roasting. It is classified by IARC as a probable human carcinogen based on animal studies, which has periodically generated alarm about coffee, most visibly in a Californian legal case that briefly required cancer warning labels on coffee.
The 2016 IARC evaluation is the relevant response, because the working group assessed coffee as consumed, acrylamide included, and still concluded the evidence was inadequate to classify it as a carcinogen. In other words, the presence of acrylamide in coffee was already accounted for in the evaluation that cleared it. The exposure from coffee is small relative to dietary acrylamide overall, and the epidemiological studies of coffee drinkers, which capture whatever acrylamide they consumed, do not show elevated cancer risk.
What the broader evidence shows
The umbrella review by Poole and colleagues[4] found high versus low coffee consumption associated with an 18% lower risk of incident cancer overall, a relative risk of 0.82, alongside lower risks of several specific cancers. The large cohort studies point the same way with one instructive exception: Freedman and colleagues[5], in 402,260 American adults, found inverse associations between coffee and death from heart disease, respiratory disease, stroke, injuries, diabetes and infections, but specifically not for death from cancer.
Where coffee and cancer research still has open questions
Saying the evidence is inadequate to classify coffee as a carcinogen is not the same as saying every question is closed. A few remain genuinely open.
Bladder cancer, the original basis for the 1991 classification, is the one most argued about. Some analyses continue to report a weak association in smokers, and separating coffee from tobacco in a population where they co-occur is difficult even with good methods. The working group's conclusion was that the evidence was inadequate, which is a statement about uncertainty rather than a clean exoneration.
Very hot coffee remains a live issue precisely because the 2A classification applies to it. In populations that habitually drink beverages scalding hot, coffee is the vehicle for a real risk even though coffee itself is not the cause.
And the protective associations, for liver and endometrial cancer, rest on observational data with all the usual limitations, covered in our article on coffee and the liver where the definitive meta-analysis rated its own evidence very low quality under GRADE. Coffee looking neutral-to-favourable on cancer is well supported; coffee actively protecting against specific cancers is not established.
That distinction is worth holding onto. Coffee's favourable associations across the literature are concentrated in cardiovascular, metabolic, liver and neurological outcomes. Cancer is the area where coffee looks most neutral, and the appropriate summary is an absence of harm rather than a demonstrated benefit, with liver cancer the clearest exception.
The summary
IARC removed coffee from its possible-carcinogen list in 2016 after finding the evidence for an association with any cancer site inadequate, and noted associations with reduced liver and endometrial cancer risk. The 1991 classification had largely reflected uncontrolled confounding by smoking. In the same evaluation, IARC classified drinking any beverage above 65°C as probably carcinogenic to the oesophagus, which is the actionable finding most coverage missed.
The general lesson is worth keeping for the next food scare. A hazard classification answers whether something can cause cancer under some conditions; it does not answer whether your exposure does. The two questions get conflated constantly, and the conflation is where most of the anxiety comes from.
Coffee happens to be a case where the evidence eventually got good enough to resolve the question, which is not true of most items that reach a hazard list.
Most stay unresolved, which is why a reclassification is worth noticing when it happens.
The 65-degree finding is the part of this evaluation worth carrying into daily life.
Everything else here is a reason to stop worrying rather than to start.
Practically: coffee is not a cancer concern on current evidence, acrylamide in coffee was already accounted for in the evaluation that cleared it, and the thing worth changing is letting your drink cool before you start on it. Our guide to coffee and health covers the everyday version.
Letting a drink cool for a few minutes is a small enough habit that there is no reason to weigh it up at all.
It is a rare case of a health scare being replaced by a smaller, more actionable finding that almost nobody heard about.