Doll, Hill, Wynder and Smoking: Who Was "First" to Show the Link with Lung Cancer, and by Which Criterion? (1950)
In this article
Doll, Hill, Wynder and Smoking: Who Was "First" to Show the Link with Lung Cancer, and by Which Criterion? (1950)
Yesterday we looked at which question a drug's approval answered: is it safe? Today's question comes in through another door: what causes this disease? In 1950 several answers pointing the same way were published within a few months of each other. Which one was "first"? There is no short answer, because that year the word "first" measures at least four different things. Let us begin with the questions asked at the patient's bedside.
London, 1948: questions and matched controls
Doll and Hill's study was planned in 1947. From April 1948 to October 1949, London hospitals sent in notifications of suspected lung cancer. The interviews were carried out with the same questionnaire by four full-time almoners (hospital social workers) employed on the study. The patient was asked what he or she had smoked over a lifetime, at what age smoking began and stopped, how many cigarettes or how much pipe tobacco a day, and whether the smoke was inhaled. The definition of "smoker" was explicit: at least one cigarette a day for at least one year.
The co-author A. Bradford Hill was Professor of Medical Statistics at the London School of Hygiene and Tropical Medicine and honorary director of the MRC Statistical Research Unit; Richard Doll was in the same unit. According to the LSHTM's own page, both initially expected to blame the rise on new road surfacing (tarmac) and motor-vehicle exhaust. The same page says Doll stopped smoking as a result of the findings; it gives no date, and I give none.
Two columns of 709
Doll and Hill's paper "Smoking and carcinoma of the lung: preliminary report" appeared in the BMJ of 30 September 1950, on pages 739-748. In twenty London hospitals, 709 patients with lung carcinoma (649 men, 60 women) were studied; each was matched to a patient without cancer of the same sex and five-year age group (709 controls). Of 2,370 notifications, 2,140 were eligible for interview; 408 could not be interviewed, and no patient refused.
The most striking figure concerned men who had never smoked: of 649 men with lung cancer only 2 (0.3 percent), against 27 of the 649 control men (4.2 percent), had never smoked. In women the difference was smaller: 19 of 60 patients (31.7 percent) against 32 of 60 controls (53.3 percent). Those smoking 25 or more cigarettes a day made up 26.0 percent of the male patients and 13.5 percent of the male controls. The authors concluded that smoking is an important factor in the cause of lung cancer, but also wrote that this was not necessarily to say that smoking causes lung cancer (conceptual summary).
The same year, the same question: JAMA in May
Four months before Doll and Hill, on 27 May 1950, volume 143 of JAMA carried Ernst L. Wynder and Evarts A. Graham's paper "Tobacco smoking as a possible etiologic factor in bronchiogenic carcinoma"; the number in its title was 684 "proved" cases (pages 329-336). Wynder had been born in 1922 in Herford, Germany, and emigrated to the United States with his family in 1938; he was a medical student at Washington University in St. Louis. According to Thun, in 1948 he began interviewing lung cancer patients without any training in epidemiology. Evarts Graham, a thoracic surgeon, was sceptical of the link, and himself a heavy smoker. Even so, he allowed the interviews and supplied letters of introduction that helped extend the study across the country.
In the same issue, immediately afterwards (pages 336-338), came M. L. Levin and colleagues' "Cancer and tobacco: a preliminary report." According to Thun, Levin's smaller study had been submitted to JAMA shortly before Wynder and Graham's; the journal hesitated because of the unfamiliar design and preliminary data, and when it received the larger study with a similar design and a respected co-author, it printed both in the same issue. So even in 1950, "first submitted" and "first published" do not point to a single name.
Two papers from Germany
There were studies asking this question before 1950. Doll and Hill's paper reports that Müller (1939) found 3 of 86 male lung cancer patients, and 14 of 86 healthy men, had never smoked; the paper had appeared in volume 49 of the Zeitschrift für Krebsforschung (pages 57-85, according to Davey Smith and Egger). The second paper comes from Jena: in 1942, at the anti-tobacco institute there, Schairer and Schöniger (1943) sent a questionnaire on smoking habits to the relatives of 195 patients diagnosed with lung cancer in Jena between 1930 and 1941 (Morabia). The study appeared in volume 54 of the same journal, pages 261-269; an English translation was reprinted in the International Journal of Epidemiology in 2001. Davey Smith and Egger also note that reviews on the subject existed some twenty years before 1950. I have not read these papers in full; I relay their descriptions from secondary sources.
"First" is at least four different questions
The first is the earliest publication to record the association: the sources cite Müller (1939) and Schairer-Schöniger (1943), and there are earlier reviews too. The second is the design that sets patients beside a comparison group: both German studies had done this; what was new in 1950 was less the design than its scale. The third is the study that offers a large sample, matched controls, a uniform interview and rising rates with rising amounts smoked together: Doll and Hill's 709 + 709 patients carry this combination; but I did not verify that no earlier study had it. The fourth is the study that convinced the medical community: that is spread over years rather than fixed in one publication; according to Thun, Doll was surprised at the "great reluctance" of many researchers and physicians to accept the results.
So each of the sentences "Doll and Hill showed it first," "Wynder and Graham showed it first" and "Müller showed it first" is incomplete when said without a criterion.
Cost and limits: association, not cause
First, the design. These were retrospective case-control studies: smoking history was asked after the disease had appeared, so recall and control-selection bias were possible. Doll and Hill were aware of this: they re-interviewed 50 controls six months or more later (reliability) and compared true cases with patients first thought to have lung cancer and later given another diagnosis (interviewer bias).
Second, the numbers. The paper's risk estimates (6, 19, 26, 49 and 65 times, by daily cigarette amount; about 50 times for 25 or more a day) were, in the authors' own words, not measures of the actual risks but "admittedly speculative." The finding on inhaling was called "anomalous" and went unexplained; there were only 60 women.
Third, the human cost and the shadow of history. Evarts Graham stopped smoking in 1951 (Thun) or 1952 (ASCO Post) and lost his life to lung cancer in 1957. The Jena institute was founded in 1942; according to Davey Smith it received 100,000 Reichsmark from Hitler's personal funds, while Davey Smith and Egger's letter says it was "partly" supported. The same letter rejects the idea that acknowledging this history would cast a shadow over present-day tobacco control. This piece does not settle that debate; it notes its link with the ethical question we discussed at Nuremberg (day 121).
What remains
- The place of the case-control design in the search for causes
- The dose-response idea: more cigarettes, a higher rate
- The knowledge that the question "who was first" depends on a criterion
What fell or was limited
- The unconditional title of "first" given to a single study
- The assumption that hospital-selected controls pose no problem
- Treating "association" and "cause" as the same thing
Four studies pointing the same way, four different denominators. The height is the percentage of men who had never smoked in each study; the orange bar shows the lung cancer cases and the green bar the comparison group; the numbers in brackets are the sizes of the case and comparison groups. The values rest on what Doll and Hill's paper reports (for Müller, 3.5 and 16.3, which I calculated from 3/86 and 14/86). The comparison groups and the definition of "non-smoker" changed from study to study, so the bars should not be compared one to one; what they share is that the case bar is shorter everywhere.
Seeing the same link in your own hospital
I found no verified source on how these studies were received in Türkiye; I do not invent a local link. I can offer a general inference: a case-control study requires asking a patient group about smoking history systematically, choosing a suitable comparison group and putting the same questions to everyone in the same way; seeing this association in a country's own data depends on the order of its records and the training of those who conduct the interviews.
From 1939 to today
What was confirmed, what conflicts
- Doll-Hill: BMJ 30 September 1950;2:739-748. 709 patients (649 men, 60 women), 709 matched controls; twenty London hospitals; notifications April 1948 - October 1949; 2,370 notifications, 2,140 eligible, 408 not interviewed, no refusals; four almoners. The text I read covered pages 740-748; the first page was not visible.
- Hill's role: In the paper he appears only as co-author; the details of his contribution are not in the text, and this piece contains no sentence such as "Hill designed the study."
- Doll's expectation and his stopping smoking: According to the LSHTM page both expected road surfacing and exhaust; Doll stopped as a result of the findings, no date given. Thun writes that both were smokers. I have not read Doll's own account; the reminiscence in PMC does not contain these details.
- Wynder-Graham: JAMA 1950;143(4):329-336, 27 May 1950 (I could not reach the PubMed page; the citation is from ASCO Post and Wikipedia). The 684 in the title and the 605 men (1.3 percent never smoked) and 882 comparison men (14.6 percent) reported by Doll and Hill may describe different sets; I did not resolve this. The Wikipedia page gives internally inconsistent figures, so I took no numbers from it.
- Year Graham stopped smoking: Thun 1951, ASCO Post 1952; the conflict is unresolved. His loss of life in 1957 (two sources).
- Levin and colleagues: Levin ML, Goldstein H, Gerhardt PR et al. JAMA 1950;143:336-338 (citation from a reference list; the first name was not verified, hence "M. L. Levin"). "Submitted earlier, published in the same issue" is from Thun. I did not read the sample or the results.
- Müller 1939: Z Krebsforsch 1939;49:57-85 (Davey Smith and Egger). The number reported by Doll-Hill is 86 patients; the germanhistorydocs page says "over a hundred people and their relatives" were interviewed and gives the pages as 62-63. There is a conflict; I used Doll-Hill's report and did not read the full text.
- Schairer-Schöniger: Z Krebsforsch 1943;54:261-269; Int J Epidemiol 2001;30:24-27. Davey Smith and Egger's letter gives the volume as "34"; BMJ 2004 and Morabia say "54," and "54" was used. The part of Morabia I read had no control group or result figures; I left them out.
- Jena funding: Davey Smith (BMJ 2004): 100,000 Reichsmark, Hitler's personal funds. Davey Smith and Egger's letter says "partly"; the difference is unresolved.
- Schrek and colleagues 1950: As reported by Doll-Hill: 14.6 percent of 82 male patients and 23.9 percent of 522 other men had never smoked. I did not verify the citation or publication date.
- 1954 follow-up study: The year rests on the "1954" in Wikipedia and a PubMed entry for a preliminary report on doctors in relation to their smoking habits; I did not read the paper.
- Out of scope: The earlier work of Lickint, Hoffman and Schönherr (their names appear only in the letter), the tobacco industry's reaction, the 1964 report.
Henrietta Lacks and HeLa (1951)
A tissue sample taken from a woman in a hospital in 1951 is today cited as a cell line still multiplying in many laboratories around the world. Why was the cell line named after a woman, and did she know that the sample was being taken? Tomorrow I will tell it after verifying the dates and records.
Author's note: The question "who first showed that smoking causes lung cancer?" sounds as if it expects a single name. The sources I read show that it does not have one; so instead of giving a name I separated the criteria. Where I did not read a full text, I marked it in the fact-check box.
Sources and Further Reading
- Doll R, Hill AB. Smoking and carcinoma of the lung: preliminary report. BMJ 1950;2:739-748 (pages 740-748 read).
- Wynder EL, Graham EA. Tobacco smoking as a possible etiologic factor in bronchiogenic carcinoma: a study of 684 proved cases. JAMA 1950;143(4):329-336 (full text not read; citation from secondary sources).
- Levin ML, Goldstein H, Gerhardt PR et al. Cancer and tobacco: a preliminary report. JAMA 1950;143:336-338 (full text not read).
- Thun MJ. Public Health Classics: Wynder and Graham 1950. Bull World Health Organ 2005 (introduction).
- Davey Smith G, Egger M. Letter on pre-1950 German research on smoking and lung cancer. Bull World Health Organ 2005;83(10):799-800.
- Davey Smith G. Health promotion in Nazi Germany. BMJ 2004;329:1424-1425.
- Morabia A. Evaluation of Schairer and Schöniger 1943 (only the opening was read; citation not verified).
- ASCO Post. "Paving the Way for the Surgeon General's Report," 15 February 2014; LSHTM Library. "No Smoking Day: LSHTM and the link between smoking and lung cancer," 2015.
- Mukherjee S. The Emperor of All Maladies: A Biography of Cancer. New York: Scribner, 2010. (Not separately verified for this piece.)
This piece is day 125 of the "365 Days of Cancer History" series and recounts the studies that showed the link between smoking and lung cancer in 1950 as a historical event. It is popular history; it is not a diagnostic or treatment recommendation. Consult your physician for decisions about your health. Sources have been verified; details that rest only on a secondary source are marked in the fact-check notes.
Was this article helpful?
Your feedback helps us improve our content.
