June 27, 2022–Kenneth Rothman is Professor of Epidemiology at the Boston University School of Public Health and Distinguished Fellow Emeritus at the Research Triangle Institute. His research includes work on the epidemiology of cancer, cardiovascular disease, birth defects, environmental epidemiology, pharmaceutical products, and methodological, conceptual and ethical issues in epidemiology. Rothman is also the author of two widely used textbooks of epidemiologic methods, Modern Epidemiology and Epidemiology: An Introduction, and is the founding editor of Epidemiology, a leading public health journal. He is the recipient of many awards, including the American Public Health Association’s Abraham Lilienfeld Award for 2002, recognizing excellence in the teaching of epidemiology during the course of a career. We took this opportunity to ask him, as a graduate of our department 50 years ago, to reflect on his time here.
Take us back to spring of 1972… What are your recollections from your time in our department? How would you describe the environment to a visitor (both physical and social)?
Brian MacMahon had been chair for 14 years already, since 1958, but he started as chair at age 35, so he was still in the peak of his career as an educator and cancer epidemiologist. The department under his leadership seemed close-knit; his vision and energy set the tone. Faculty and students brought their bag lunch every day to the department conference room. Several faculty members, including Phil Cole, Dick Monson, Theo Abelin, Olli Miettinen, and some biostatistics faculty – Jane Worcester, Marge Drolette and Jim Warram, in particular – convened at those informal lunches along with doctoral students on the 13th floor of 665 Huntington Avenue. The Kresge building was just being completed and the epidemiology department had not yet moved into their new offices.
One regular activity, run by the doctoral students with encouragement from the faculty, was the John Graunt Literary Society, which met late each Friday afternoon. One of the students – it may have been Bob Hoover – organized snacks and beverages. The meetings were an important part of the week for us, although the conversation only occasionally drifted to John Graunt.
In the spring of 1972, I lived in an apartment across Huntington Avenue from the school. I spent nearly as much time going vertically as horizontally to get to my office. In April of that year my daughter Emily was born, and soon afterward I brought her across the street to meet the department. Brian had stepped out of his office briefly and when he returned, he found Emily sitting quietly on his desk in an infant carrier. It was her first visit to the Harvard Chan School. She returned a few months later for commencement, and a couple of decades later as a Harvard Chan School student.
What was the thesis writing process like for you? How did you pick the topic for your thesis “Epidemiology of Trigeminal Neuralgia?” Were there any particular challenges? Any unexpected surprises? What advice would you give to students just starting the thesis writing process?
My thesis advisor was Richard Monson, who had recently finished his doctor of science at Harvard Chan School and joined the faculty when I entered the program. He advised me to pick a thesis topic that could be completed expeditiously. Trigeminal neuralgia was a disease about which little was known. It is a chronic, progressive condition characterized by paroxysms of intense facial pain. I was able to get permission to review patient charts in two large Boston neurosurgical departments, and from this I was able to conduct three studies, a case-control study of risk factors for trigeminal neuralgia, a cohort study to examine mortality among patients with the disease, and a radiographic study of patients designed to test a theory that a specific anatomical irregularity impinging on the trigeminal nerve caused the neuralgia. I abstracted records and coded the data and wrote the programs for data processing and analysis. Dick Monson’s advice proved to be excellent, and I was able to complete the work in about 18 months. The main surprise was an unexpected finding that came to light about a year or so after the thesis was completed. I was on the Harvard Chan faculty then, working with a student, Tom Beckman. Together we were exploring a lead from the thesis data and learned that the risk factors for the disease differed strikingly according to whether the upper face or lower face was affected. We published this finding in the Lancet, but the explanation for it remains a mystery.
How would you link the work you did in your thesis to what you would go on to do in your later years as an epidemiologist?
It wasn’t the thesis that presaged my later work, but something I pursued on the side while working on my thesis. I came across a paper that had been published a few years earlier in the American Journal of Public Health, by Andrew Z. Keller and a co-author, entitled “The association of alcohol and tobacco with cancer of the mouth and pharynx.” This paper attempted to separate the causal roles of alcohol consumption and tobacco use in oral and pharyngeal cancer. They were mutually confounded because these two risk factors were correlated exposures. But as I read the paper, I was not focused on the confounding between the two causes, but their possible interaction in causing cancer. Did alcohol drinking enhance the carcinogenicity of tobacco smoke? That aspect of the problem had not been considered in their paper. So, I wrote to Dr. Keller and I proposed a new analysis that would measure the effect of joint exposure and compare it with the effects of separate exposure to the two causes. Andy Keller was enthusiastic about the idea and agreed to collaborate. He sent tables that I analyzed, and I drafted a manuscript. Our paper, “The effect of joint exposure to alcohol and tobacco on risk of cancer of the mouth and pharynx”, was published in the Journal of Chronic Diseases in 1972. It was an early effort to distinguish confounding from biologic interaction, not sophisticated by today’s standards, but it got me started on the road to thinking and writing about how to best measure synergy and antagonism, and what causal models I could use to describe these phenomena. My longstanding interest in epidemiologic concepts and methods can be traced back to Keller’s paper and the questions it raised.
What advice would you give to current students in the department? Is there something you wished you’d known during your time as a student?
One thing I have learned is that offering advice based on personal experience is usually an example of the post hoc fallacy, so I’m not inclined to give advice based on my experience. It is difficult to know what choices were important and what events were just a matter of luck. I did learn early on, however, how easy it was to stumble into problems. When I read that paper by Andy Keller on oral and pharyngeal cancer and realized that there was an alternative analysis that would be worth pursuing, I asked for advice from Brian MacMahon about the proper way to seek to collaborate with him. Brian counseled that I should write to the lead author of the paper and explain what I was interested in doing, so I did exactly that. I did not know Dr. Keller, nor his co-author. In retrospect I should have asked some more questions. Had I done so, I might have learned that Dr. Keller was a student and this paper was part of his doctoral thesis. I also might have learned that the paper’s other author, although unknown to me, was a well-known public health researcher and authority, who was Andy Keller’s thesis advisor. If I had known that, I might have written my letter asking about collaboration to Andy’s co-author. Instead, I contacted Andy Keller, and we had a fruitful collaboration, but I learned later that his thesis advisor and co-author was irked at me for not approaching him first. He considered it a slight, one that he would never forgive. He and I interacted many times in the years to come, but I could never patch that crack in our relationship.
Had I approached it differently, I might have avoided some problems, but perhaps my collaboration with Andy would not have developed so smoothly (Andy Keller and I remained good friends until he died). Is this a cautionary tale? I don’t know; I think it’s more of a lesson on how easy it is to create problems as one stumbles through life.
Curious to hear your thoughts on the development of epidemiology from that time onwards…what are some recent developments that you are most excited about?
I imagine that anyone looking at epidemiology as a field would be struck by the rapid rate of development of epidemiologic concepts, methods, and findings during the past 50 years. Should we be surprised by this, given the pace of technological innovation during this time? Today we walk around carrying in our pocket more computing power than was available from a machine that filled a cavernous room when I was a student. Nonetheless, I don’t think that the developments in epidemiology have depended much on advancing technology. For the most part, the methods used to design and analyze epidemiologic studies were not tied to computers or other technological innovations that became available in the last half century. Most of the methods described in the first edition of Modern Epidemiology could be implemented with pencil and paper might have been developed centuries earlier if there had been a stimulus to do it. I believe that two related phenomena spurred growth in epidemiologic thinking and epidemiologic methods during the late 20th century. The first was the wider availability of data. John Graunt was a 17th century haberdasher in London who saw the value of an archive of published weekly reports of deaths, and he mined that resource in his classic book, Natural and Political Observations Mentioned in a Following Index, and Made Upon the Bills of Mortality. In the 20th century, various sources of systematized data on health and disease became available to researchers, first in paper records and later in automated form, making it considerably easier to conduct epidemiologic studies. Equally important, money to fund epidemiologic research enabled researchers to take advantage of these data resources. The combination of the availability of data and funds to conduct research fueled the growth of epidemiology as a field.
Nowadays, the frontier of epidemiologic methods is brimming with rich ideas that provide new insights and solutions to longstanding problems, and have brought new problems to light. These ideas include causal inference methods, directed acyclic graphs, trial emulation, multiple imputation, bias analysis, G-methods, instrumental variable analysis, and many others, all exciting developments that demonstrate the rapid progress in concepts and methods that has characterized the field.