It seems hardly arguable that one of the greatest public health triumphs of the twentieth century was the linkage of cigarette smoking and cancer. From today’s perspective, buried under volumes of studies verifying the claim, the delay in full acceptance of the observation may seem surprising. Changing ideas had to contend with public perception, industry resistance, habit and addiction. The medical community was called upon to provide its best work in demonstrating, beyond the shadow of a doubt, that the association was not only valid but causal. Reviewing the steps researchers took to prove the claim illustrates two types of clinical research: retrospective and prospective.
The first publications associating cigarette smoking with cancer occurred around 1950, almost simultaneously, in the US, UK and Germany.[1] These early studies used case-control methodology, a retrospective technique.[2] This type of research begins with two groups matched for sex, age, overall health, etc., except for one factor – in this case, those who at the time of the study did have lung cancer and those who did not. Once these groups were defined, the researchers looked back, retrospectively, to see how many people in each group were exposed to a hypothesized cause of lung cancer – cigarette smoking, in this example. With this data, researchers estimated an odds ratio that cigarette smoking contributed to cancer. Odds is the number of patients experiencing an event (cancer) divided by the number of patients who did not experience the event. The odds ratio, then, is the odds of one group (smokers) compared to the odds of another (non-smokers), in ratio form.
Unfortunately, retrospective studies in general are considered less reliable for several reasons: sample error (the selection of subjects may not be as random as assumed), reliance on self-reporting and the assessments of others, and the opportunity for confounding (unmeasured risks contributing to the outcome).[3]
For this reason, prospective studies are favored: group members are chosen in advance of the experimental period and everything that happens to them is recorded. These cannot be accomplished as quickly as case-control studies but are considered more accurate. Two identical groups are identified and followed into the future, one with exposure to a hypothetical risk and one without. This would be called a prospective cohort study [2] and this is just what the lead researchers in the epidemiology of smoking and cancer wisely chose to switch to.
Recognizing the superiority of prospective studies, Sir Richard Doll and Sir Austin Bradford Hill established the British doctor cohort in 1951. They also influenced subsequent initiation of the Nurses' Health Study in the US, followed decades later by the Physicians' Health Study and the Health Professionals Follow-up Study.[4] By 1954 Dr. Doll and Prof. Hill were able to publish the first prospective data linking cigarette smoking to lung cancer.[5] The original cohort[6], the Nurses’ cohort[7], and the Health Professionals Follow-up Study[8] continue to this day, still sources of useful data on epidemiology. The Physicians' Health Study[9] ended after completing its task in 2011.
There are retrospective cohort studies, as well, but these usually look back at prospectively collected data with an eye to predetermined exposures to the hypothetical risk.[2] Large samples sizes are needed if the incidence of measured outcome is rare.[3]
Doll and associates published a fifty-year update on his original prospective cohort in 2004.[10] The additional data gathered during this long period of observation was able to help delineate, among other things, how differences in smoking behavior during a lifetime increased or decreased the risk of cancer.[4]
So, is it better to slowly obtain data by looking ahead with few limitations or seek quicker data by trying to see the forest for the trees in the rear-view mirror? Researchers like Doll and Hill who have had the most impact on medicine and public health chose prospective methods – a slower design with less chance of error and a greater chance of useful outcome.
This series will continue looking at retrospective vs. prospective methods, focusing next time on registries.
[1] Doll R, Hill AB. Smoking and carcinoma of the lung; preliminary report. Br Med J 1950; 2:739–48. [2] Crandon S. Case-control and Cohort Studies: A Brief Overview. Students 4 Best Evidence. https://s4be.cochrane.org/blog/2017/12/06/case-control-and-cohort-studies-overview/. Published December 6, 2017. Accessed December 27, 2020. [3] Nickson C. Retrospective Studies and Chart Reviews. Life in the Fastlane. https://litfl.com/retrospective-studies-and-chart-reviews/. Published November 3, 2020. Accessed December 31, 2020. [4] Jonathan M. Samet, Frank E. Speizer. Sir Richard Doll, 1912–2005, American Journal of Epidemiology, Volume 164, Issue 1, 1 July 2006, Pages 95–100, https://doi.org/10.1093/aje/kwj210 [5] Doll R, Hill AB. The mortality of doctors in relation to their smoking habits: a preliminary report. Br Med J 1954; 1:1451–5. [6] British Doctors Study. Clinical Trial Service Unit & Epidemiological Studies Unit (CTSU): Nuffield Department of Population Health. https://www.ctsu.ox.ac.uk/research/british-doctors-study. Accessed December 30, 2020. [7] Nurses’ Health Study. https://www.nurseshealthstudy.org/. Updated December 13, 2020. Accessed December 30, 2020. [8] Health Professionals Follow-up Study. Harvard T. H. Chan School of Public Health. https://sites.sph.harvard.edu/hpfs/. Accessed December 30, 2020. [9] Physicians' Health Study. https://phs.bwh.harvard.edu/. Updated November 12, 2012. Accessed December 30, 2020. [10] Doll R, Peto R, Boreham J, et al. Mortality in relation to smoking: 50 years' observations on male British doctors. BMJ 2004; 328:1519–27.