Richard Doll

Chapter 1 · Why Had Lung Cancer Increased Fifteenfold?

1947, and a question nobody could answer

1912 – 1947 · Hampton · St Thomas's · North Africa · London

Richard Doll was born in Hampton, west London, in October 1912, the son of a doctor. He read mathematics at Cambridge for a term, drank too much the night before an examination, failed it, and went to St Thomas's to study medicine instead — his own account of one of the more consequential hangovers in medical history.

He qualified in 1937, served as an army medical officer through the Second World War in North Africa and the Middle East, and came home a committed socialist and an enthusiast for the new National Health Service.

In 1947 the Medical Research Council asked him, with the statistician Austin Bradford Hill, to investigate something alarming. Deaths from lung cancer in Britain had risen roughly fifteenfold in twenty-five years. Before the First World War it was a rare disease that a physician might see a handful of times in a career. By the 1940s it was common and climbing steeply.

Something had changed. The question was what.

Doll's own expectation — and he was explicit about it afterwards — was that the cause would be *tar from newly surfaced roads*, or exhaust fumes from motor cars. Both had increased enormously over the same period, both were plausible respiratory irritants, and both fitted the pattern of a disease more common in towns than in the country.

He was a smoker himself, as most men of his generation were, and it did not occur to him as a candidate.

“I gave up smoking two-thirds of the way through the study.”

— Richard Doll, recalling the 1948–49 case-control study

Why this matters

Doll began with a clear hypothesis that was wrong, and the value of the study lies in the fact that his own data changed his mind.

You have the fifteenfold rise and the wrong expectation. What would you ask?

Ask Doll

• “Why did you expect road tar to be the answer?”
• “How rare was lung cancer before the First World War?”
• “Why did smoking not occur to you as a cause?”
• “What did the war do to how you thought about medicine?”
• “How does a mathematician end up in medicine?”

Chapter 2 · Interviewing Patients in London Hospitals

The case-control study, 1948 to 1950

1948 – 1950 · London hospitals

The design was what is now called a *case–control study*. Take people who have the disease — the cases. Take comparable people admitted to the same hospitals for other reasons — the controls. Ask both groups the same questions about their lives, and see what differs.

Doll and his interviewers spoke to around 1,700 patients in twenty London hospitals. They asked about occupation, housing, heating, exposure to fumes, where they lived — and, among many other things, about smoking.

The design has a critical feature: the interviewers asked patients before their diagnosis was necessarily known to the interviewer, so that expectation could not shape the questioning.

The pattern came out clearly and it was not the roads. Patients with lung cancer were overwhelmingly more likely to be heavy smokers than patients admitted for other reasons. There were almost no lifelong non-smokers among the lung-cancer cases at all.

Doll's own reaction is the part worth teaching. He was convinced by the data before publication, and he stopped smoking two-thirds of the way through the study. He said afterwards that the results were so much stronger than he had expected that he changed his own behaviour before he published.

They published in the *British Medical Journal* in September 1950. An American study by Wynder and Graham appeared in the same year with the same conclusion.

The response was largely scepticism. And some of the objections were serious.

Why this matters

A case-control study compares people with a disease against comparable people without it, and the design choices that keep expectation out of the answers are what make it trustworthy.

You have the interviews and the answer you did not expect. What is your question?

Ask Doll

• “How do you choose the right people to compare against?”
• “Why does it matter what the interviewer knows?”
• “What convinced you before you had published?”
• “What were the strongest objections to your result?”
• “Did it feel strange to be a smoker finding this?”

Chapter 3 · The Best Statistician in the World Says You Are Wrong

Fisher's objection, and why it had to be answered

1950 – 1958 · London · Cambridge

Ronald Fisher was the most important statistician of the twentieth century — the man who essentially invented experimental design and the analysis of variance. And he attacked the smoking–cancer conclusion, repeatedly and in print.

His objection was not stupid. It was this: an association between two things does not tell you which causes which, or whether a third thing causes both. Perhaps there is a genetic predisposition that makes a person both more likely to take up smoking *and* more susceptible to lung cancer. On that account smoking and cancer would go together without smoking causing anything. He pointed to twin studies suggesting smoking habits have a heritable component.

He also objected that the causal direction might be reversed: perhaps a pre-cancerous lung produces mild irritation that makes a person want to smoke.

These are proper objections, and the case-control design cannot answer them. Fisher was also, it emerged, a paid consultant to the tobacco industry and a heavy pipe smoker himself — both worth knowing, and neither of which makes the logical objection invalid.

Answering it required a different kind of study, and this is what Doll and Bradford Hill did next. In October 1951 they wrote to every doctor on the British medical register — around 60,000 people — asking about their smoking habits. About 34,000 replied.

And then they *waited*, following those doctors, recording who died and of what, for the rest of their lives.

Why this matters

Fisher's objection — that a common cause could produce the association — is a real logical problem that a case-control study cannot resolve, and answering it required fifty years.

You have the objection and the fifty-year answer to it. What would you ask?

Ask Doll

• “What exactly was Fisher's objection?”
• “Why can a case-control study not settle it?”
• “Does Fisher's tobacco consultancy invalidate his argument?”
• “Why write to doctors rather than the general public?”
• “What does following people forward prove that looking back cannot?”

Chapter 4 · Fifty Years

The British Doctors Study, 1951 to 2001

1951 – 2004 · Britain · Oxford

The British Doctors Study ran for fifty years, and the results arrived in stages.

By 1954, after just over two years, there was already a clear excess of lung-cancer deaths among smokers. By 1956 the *dose–response* relationship was established: the more a person smoked, the greater the risk, in a graded way. That matters enormously, because a dose relationship is very hard to explain by a hidden common cause.

The study also answered Fisher directly, in the most decisive way available. Doctors *stopped smoking* in large numbers during the study — they were, after all, reading the results. And in those who stopped, the risk *fell*, progressively with the years since quitting. A genetic predisposition cannot explain that. Nothing about the person changed except the exposure.

The fifty-year report, published in 2004, gave the full account, and the numbers are worth stating precisely. About *half* of all persistent cigarette smokers are killed by the habit. Smokers lose on average about *ten years* of life. And those who stop — even at fifty, even at sixty — regain much of what they would have lost; stopping at thirty recovers almost all of it.

Doll and Bradford Hill also helped set out the criteria by which an observed association is judged causal: strength, consistency across studies, specificity, temporal order, dose–response, plausibility, coherence, experiment, and analogy. These are the *Bradford Hill criteria*, and they are how every claim that an exposure damages health is now argued.

Doll also demonstrated the cancer risk of asbestos, the danger of X-raying pregnant women, and the protective effect of aspirin against vascular disease.

Why this matters

The falling risk in doctors who quit is the observation that defeats the genetic-confounding objection, because nothing changed except the exposure.

You have half a century of following the same people. What is your question?

Ask Doll

• “Why is a dose-response relationship such strong evidence?”
• “What happens to risk when somebody stops smoking?”
• “How do you keep a study running for fifty years?”
• “When is an association strong enough to call a cause?”
• “What did you find about asbestos and about X-rays in pregnancy?”

Chapter 5 · The Payments Disclosed After His Death

A reputation, and a complication

1969 – 2005 and after · Oxford

Doll became Regius Professor of Medicine at Oxford in 1969, founded the cancer epidemiology unit there, was knighted, and was showered with honours. He worked into his nineties and remained mentally formidable to the end.

He died in July 2005, aged ninety-two, having been a non-smoker for more than half a century.

After his death, documents came to light showing that from the 1970s he had received consultancy payments from chemical companies including Monsanto and from the Chemical Manufacturers Association — reportedly on the order of a substantial daily fee over a period of years — while producing assessments on the carcinogenicity of substances including Agent Orange components and vinyl chloride, in which he concluded the risks were smaller than others claimed. These arrangements had not been disclosed in the relevant publications.

Doll's defenders argue that non-disclosure was normal practice at the time, that he directed some payments to Green College at Oxford rather than taking them personally, and that no analysis of his has been shown to be wrong because of the money.

His critics argue that the appearance is bad regardless, that the modern norm of disclosure exists precisely because readers cannot assess influence they do not know about, and that a man who spent his life on the effects of industrial exposures should have seen the difficulty.

Both positions have force. The tobacco work stands — it has been replicated worldwide many times over and does not depend on trusting Doll personally. But the episode is a real part of the record, and it is a useful case in why disclosure rules exist.

Why this matters

Doll's undisclosed industry consultancies are a genuine complication in a distinguished record, and they illustrate exactly why disclosure norms were later made compulsory.

You have the honours and the payments. What would you ask him?

Ask Doll

• “Why did you not disclose the consultancies?”
• “Does taking money from industry change what you can conclude?”
• “Why should a reader be told about a payment at all?”
• “Does the tobacco work depend on trusting you?”
• “What would you have done differently about the money?”

What Doll changed

Doll turned a statistical association into an accepted cause and created the methods modern epidemiology runs on. The British Doctors Study established that about half of all persistent smokers are killed by the habit, that smokers lose around ten years of life, and that stopping — even in middle age — recovers most of it. Every claim that a habit or exposure damages health is now argued the way he argued this one: by strength, by dose, by consistency, and by what happens when the exposure stops.

A debate that continues

Consultancy payments from chemical companies, disclosed only after Doll's death and not declared in the relevant publications, remain a genuine criticism of his independence, though no analysis of his has been shown to be wrong because of them.

Keep exploring — ask Doll

• “Which exposure should be investigated the same way now?”
• “How do you argue with an industry that funds the doubt?”
• “What is the hardest kind of cause to prove?”

Related lives

• James Lind — The Man Who Ran the First Fair Trial
• Florence Nightingale — The Lady with the Lamp
• Edward Jenner — Father of Vaccination
• Ernest Starling — The Man Who Named the Hormone

Related themes

Correlation and causation · Smoking and health · Evidence and study design