As I wrote three weeks ago, the state of California is deliberating whether to list coffee as a carcinogen under its Prop 65 statute. This statute requires the posting of warnings alerting the public to the presence of carcinogenic and toxic substances in food, beverages, consumer products, and the environment. The warning being considered stems from the presence of trace levels of acrylamide, a chemical which is produced during roasting and which causes tumors in animals when administered at very high doses. As I was informed after posting my article, in fact, there are already warnings posted in coffee shops in California.
Since there is widespread concern about the presence of chemicals in our food and our environment generally, usually at very low levels, the move to list coffee under Prop 65 provides an opportunity to examine a case in which we actually have much more high-quality information to go on than is usually the case.
For most potential causes of cancer that arise we do not have good data on actual human exposure and have to resort to animal experimental data. This is true of most industrial and agricultural contaminants. It is also true of other exposures that are difficult to measure – a prime example being diet.
In the case of coffee we have an abundance of high-quality human data. Over the past two decades the association of coffee drinking with cancer risk has been examined in a large number of cohort studies in various countries. In this type of study, a defined population is enrolled at one point in time and information is obtained on health status, lifestyle behaviors, and other factors. The cohort is then followed for a number of years, and the occurrence of disease is recorded. This enables researchers to compare the incidence of disease in individuals with the exposure of interest (in this case, coffee) with the incidence in individuals who have no or minimal exposure (those who do not drink coffee).
The information on coffee intake collected in these studies is of high quality because coffee drinking is a relatively stable habit, and people can tell researchers with reasonable accuracy whether they drink coffee at all and, if so, how many cups, on average, they drink per day. Furthermore, there are people who don’t drink coffee at all. This means that researchers can examine a wide range of coffee intake and compare heavy drinkers, moderate drinkers, and light drinkers to non-drinkers of coffee.
In recent years, a substantial number of meta-analyses of these studies examining coffee drinking in relation to cancer at different sites have been published. Meta-analysis is a procedure which combines data from individual studies in order to obtain a “summary estimate” of an association, which is more stable than the estimates derived from smaller, individual studies. It is a process of statistical averaging. However, a meta-analysis is only as good as the data that are being summarized. A well-conducted meta-analysis involves examining the quality of the studies being compiled, the possibility of publication bias (i.e., whether studies that failed to find an association never got into the published literature), heterogeneity between studies, and, where heterogeneity is found, attempting to identify its source.
Of the dozen or so cancer sites which have been subjected to meta-analysis for their association with coffee, several have shown an inverse association with coffee-drinking. An “inverse association” means that coffee-drinking is associated with decreased risk and, hence, might suggest a possible protective effect of coffee drinking. Cancer of the liver, endometrium, melanoma, and oral and pharyngeal cancer all showed evidence of an inverse association. Postmenopausal breast cancer, colorectal, and prostate cancer also showed inverse associations in some meta-analyses; however, the results were less consistent for these cancers.
No association – that is, no increase or decrease in risk with coffee consumption — was seen with cancers of the ovary, pancreas, stomach, bladder, kidney, or thyroid.
The strongest inverse association was seen for liver cancer – roughly a 50% reduction in risk for the highest coffee consumers compared to non-drinkers/occasional drinkers. For endometrial cancer the reduction was close to 30% in one meta-analysis and 20% in another. For oral and pharyngeal cancer the reduction was 30%.
These results suggesting an inverse association of coffee drinking with certain cancers have led some people to conclude that coffee drinking may be beneficial to health and, specifically, that it may have chemo-preventive effects.
But we have to be careful. These are observational studies, and we can’t jump from findings of an association to the assumption of causality.
The apparent reduction in risk of certain cancers associated with coffee-drinking could be distorted by other factors. Historically coffee drinking has been correlated with smoking habits. That is, particularly in the past, when smoking was more prevalent, those who were heavier smokers also tended to consume coffee more heavily. Since smoking is a strong risk factor for a number of cancers, failure to correct adequately for how much a person smoked could make it appear that coffee drinking was associated with cancer when, in fact, it was not. (It is interesting that a meta-analysis of coffee-drinking and lung cancer – a disease which is strongly associated with smoking habits – showed a doubling of risk associated with coffee drinking). Particularly for liver cancer and oral and pharyngeal cancer, the inverse association has to be interpreted cautiously due to the strong association of these cancers with smoking and alcohol intake. Also, we should note that the largest reduction in risk seen for liver cancer is about a 50 percent, and this is still modest in magnitude compared to the positive association seen for smoking and lung cancer, which is on the order of 2,000 percent.
Because of the limitations of observational studies, some have proposed carrying out randomized controlled clinical trials in which participants are assigned to a coffee intervention or to a control group. However, it is difficult to see how such a clinical trial could work. First, such a study would be unrealistic because of its required size and cost. Second, coffee drinkers who are assigned to the “no coffee” control group would be very likely to backslide and resume drinking coffee.
So what is the bottom line? The bottom line is that – with the possible exception of lung cancer – coffee drinking does not appear to increase the risk of any cancer, and for certain cancers, it may be associated with decreased risk. These results are based on a large number of cohort studies carried out in different countries. Again, in these studies respondents were able to give reasonably accurate information about their coffee consumption over their lifetime. This information is roughly comparable to that regarding smoking, and probably superior to that regarding alcohol consumption, because, unlike with alcohol, there is no stigma concerning coffee drinking. Both smoking and alcohol consumption show strong and robust positive associations with particular cancers, whereas coffee drinking does not.
Concerning the move under Prop 65 to require coffee sellers to post warning labels about the possible carcinogenic effects of trace amounts of acrylamide in coffee, the human/ epidemiologic evidence shows that coffee, as it is actually consumed by hundreds of thousands of people all over the world, does not result in any increase in the incidence of cancer.
Second, in 1991 the International Agency for Research on Cancer (IARC), whose assessments of potential carcinogens are heavily relied on by California’s Office of Environmental Health Hazards Assessment (OEHHA), classified coffee as Group 2B “possible carcinogen.” More recently, in 2016, the agency downgraded coffee to Group 3 – “not classifiable as to carcinogenicity.”
IARC is unlike any other risk assessment agency in that it focuses on the potential, no matter how theoretical or far-fetched, of an agent to cause cancer, and ignores the degree of exposure, which, in the real world is, crucial in determining risk. Because of this, IARC’s classification scheme is disposed to emphasize positive findings. And this penchant is reflected in the astounding fact that, of the roughly one thousand agents that the agency has evaluated, exactly ONE has been placed in Group 4 – “probably not carcinogenic to humans.” The chemical that has this distinction is caprolactam, which is used in the manufacture of nylon.
IARC recognizes that where solid human/epidemiologic data are available, such data provide the most relevant evidence regarding carcinogenicity. Since we have an abundance of high-quality human data indicating that coffee does NOT cause cancer in people, if any agent merits being placed in Group 4, coffee does.