The International Agency for Research on Cancer assigned in 1991 coffee a 2B rating which classified the beverage as a "possible human carcinogen." The agency has now changed its opinion. Specifically, drinking coffee was supposedly associated with a higher risk of bladder cancer. The designation of coffee as a possible carcinogen basically rested on a series of weak case-control studies. As I explain in The End of Doom, in case-control studies epidemiologists identify a population that suffers from a disease or other condition and then attempt to match them with a similar population free of the disease or condition. They then look for differences in lifestyle, diet, or the environment that might account for the disease. More than 50 different types of bias have been identified in epidemiological studies.
Case-control studies may have some use for generating hypotheses, but should almost never be used as guides for regulation and policy. Why not? The problem is that epidemiologists generally find vastly more false positives than they do true positives—that is, they identify far more associations between phenomena than eventually are found to be the case. How do we know that there are far more false positives than true positives? Because the vast majority of epidemiological studies are not replicated. In other words, other, later researchers do not find that the risk factor identified in the initial observational study is in fact associated with a disease. S. Stanley Young of the US National Institute of Statistical Sciences estimates that only 5 to 10 percent of observational studies can be replicated.
Even with the best of scientific intentions, it is not easy to sort actual risk factors from the statistical background noise of confounders and researcher biases. "With epidemiology you can tell a little thing from a big thing. What's very hard to do is to tell a little thing from nothing at all," said Michael Thun, an American Cancer Society epidemiologist, in 1995. Former Boston University epidemiologist Samuel Shapiro agrees: "Epidemiologists have only primitive tools, which for small relative risks are too crude to enable us to distinguish between bias, confounding, and causation."
Well, such cautions about the manifold defects of case-control epidemiology did not forestall the IARC from classifying coffee as a possible human carcinogen twenty-five years ago. The IARC evaluation process essentially enacts the precautionary principle which in which the absence of evidence is never sufficient evidence of absence. As a result of this highly precautionary bias only 1 substance out of 989 agents evaluated by the IARC has ever been declared probably not carcinogenic to humans. The one compound placed in IARC Group 4—probably not carcinogenic to humans—is caprolactam, a precursor the manufacture of nylon fibers.
Well, the IARC has announced today that it now moving coffee from Group 2B to Group 3, that is, your morning cuppa is "not classifiable as to its carcinogenicity to humans." But the IARC is still anxious that you not drink very hot beverages since they are still classified as possible human carcinogens.
Last year, Harvard researchers reported in three prospective cohort studies that moderate coffee consumption "was associated with reduced risk of death from cardiovascular disease, diabetes, neurological diseases such as Parkinson's disease, and suicide. Coffee consumption was not associated with cancer deaths. The analyses took into consideration potential confounding factors such as smoking, body mass index, physical activity, alcohol consumption, and other dietary factors."
Let's just say that it's possible that the IARC's implementation of the precautionary principle may have caused premature deaths by convincing some people to forego coffee.