How was smoking shown to cause lung cancer without a trial?
Nobody randomised smoking. Doll and Hill found heavier smokers died of lung cancer at many times the rate of non-smokers, with risk rising with dose. When Fisher argued an inherited trait might cause both, Cornfield showed in 1959 that such a trait would need to be at least nine times as common in smokers. No plausible trait came close.
Last time's quick check: each point is where supply met demand that year, and both curves were moving, so the points trace neither. After the Second World War, lung cancer deaths in Britain were rising fast. In 1950, Richard Doll and Austin Bradford Hill compared patients with lung cancer in London hospitals with similar patients who had other illnesses. Almost all the lung cancer patients smoked, and heavy smoking was far more common among them.
Then they did something bolder. In 1951 they wrote to the doctors of Britain and asked about their smoking. Over the following years they tracked who died, and of what. Heavier smokers died of lung cancer at many times the rate of non-smokers, and the risk climbed with every step up in how much they smoked. That's a dose–response pattern.
The fiercest critic was Ronald Fisher, the father of the randomised experiment. Nobody had randomised smoking, he pointed out. What if some inherited constitution made people both more likely to smoke and more likely to get cancer? Then smoking might not cause cancer at all. As a point of logic, Fisher was right: an unmeasured common cause could, in principle, produce the pattern.
He was also a smoker himself, and a consultant to the tobacco industry. But the question deserved a better answer than "trust us", and it got one. In 1959, Jerome Cornfield and colleagues turned Fisher's objection into arithmetic. Smokers had roughly nine times the lung cancer risk of non-smokers.
For a hidden factor to explain all of that, it would have to be at least nine times as common among smokers as among non-smokers. No plausible inherited trait came close. It was the first sensitivity analysis. Not "is there a hidden cause?" (there always might be), but "how strong would it have to be to change our conclusion?"
In 1965 Bradford Hill listed nine things to weigh when deciding whether an association is causal: how strong it is, whether it's consistent, whether the cause comes first, whether more dose brings more effect, and others. He insisted they were aids to judgement, not a checklist.
The same arguments run through food and farming today. Most questions about diet and health, or soil and nutrients, can't be randomised for decades. So we do what Cornfield did: measure carefully, and ask how strong a hidden cause would need to be. That's where the ideas came from. Next, in Module 2: drawing causal maps, and the roles every variable can play.
In 1965 Austin Bradford Hill listed nine things to weigh when deciding whether an association is causal, including how strong it is, whether it is consistent, whether the cause comes first, and whether more dose brings more effect. He insisted they were aids to judgement, not a checklist.
In 1959 Jerome Cornfield and colleagues turned Fisher’s objection into arithmetic. Smokers had roughly nine times the lung cancer risk of non-smokers, so a hidden factor explaining all of it would have to be at least nine times as common among smokers as among non-smokers. No plausible inherited trait came close.
Causal evidence is changing how farming is measured and rewarded. Leave your email and we will tell you when new films, courses and findings are published.
Prefer to read online? Vitagri also writes on Substack. See our privacy policy for how we use your email.