The results are interesting, but there's a methodological limitation the article isn't emphasizing as strongly as it should. An illustrative analogy: Suppose y

The results are interesting, but there's a methodological limitation the article isn't emphasizing as strongly as it should. An illustrative analogy: Suppose you organize a group of people who often take painkillers for headaches, and a group that takes painkillers only rarely. You find that headaches of the regular painkiller-takers are no better with their painkillers than the non-painkiller-takers are without painkillers. Can you conclude from this that painkillers have no long-term effects on headaches? You can't, because the reason the first group was taking pain killers in the first place was probably because they have bad headaches, while the second group doesn't. The painkillers just bring them into parity. The problem is the two groups the coffee study compares were not assigned their subjects randomly. They're made up of people who choose freely to drink or not drink coffee. The fact that regular coffee drinkers are no better with coffee than non-drinkers are without coffee doesn't necessarily mean that coffee has no long-term effect. It could well be that non-drinkers already have endogenously whatever benefits caffeine provides, and drinkers do not, and coffee brings them up to parity. In other words, the current study has nothing to say on what happens to people when their long-term coffee drinking habits are changed from what they would normally prefer them to be, and so you can't infer much about the benefits or drawbacks of such changes. You could control for this by assigning people randomly two two groups, have one group drink coffee regularly for a few months while the other abstains, and then repeat the experiment performed here. The results of this modified experiment would determine whether current ad-lib drinkers should stop drinking coffee and whether current non-drinkers should start.