A newspaper headline says 'Coffee drinkers more likely to get lung cancer'. What questions would you ask about the research behind it, and how would you design a study to find out the truth?

My first questions would be about what kind of study it was. Was it an observational study that just found an association, or a trial? How many people were included, and how were coffee drinking and cancer measured — did people remember how much coffee they drank years ago? And how big was the increase — is it a large relative risk of a rare outcome¹, which might mean a very small absolute difference?

The biggest issue I'd suspect is confounding. Historically, people who drank a lot of coffee were more likely to smoke — the two often went together. Smoking causes lung cancer, so if the study didn't account for smoking properly, coffee could look like a cause when it's really just associated² with the real cause.

There could also be other problems, like selection bias in who took part, or chance if the researchers tested lots of foods and one came up positive.

To investigate properly, a randomised trial — making some people drink coffee for decades — wouldn't be practical or ethical. So I'd design a large prospective cohort study: recruit people without cancer, record their coffee intake and smoking carefully, and follow them for many years. The key analysis would be looking at coffee and lung cancer only in people who have never smoked³. If the association disappears in never-smokers, that suggests smoking was the explanation.

I'd also want to know whether there's a plausible biological mechanism, and whether other studies found the same thing — one headline is rarely the full story.