Lilienfeld, David Eugene

In 1945, Jacob Yerushalmy was a statistician in the United States Public Health Service Tuberculosis Control Bureau. In that year he introduced his brother-in-law, Abraham Lilienfeld, to his colleague, Morton Levin, then at the New York State Department of Health. During the ensuing conversation, Levin suggested the brother-in-law join the New York Department of Health for a career in public health. After joining the Department, Lilienfeld began a series of collaborations with Yerushalmy (or Yak, from his Hebrew name Yaakov) that would define various aspects of modern epidemiology, including many of the research and teaching foci in the departments Lilienfeld and Yerushalmy subsequently directed (Lilienfeld at the University of Buffalo and Johns Hopkins University, and Yerushalmy the University of California, Berkeley). This paper examines the association of these 2 public health professionals, my father and uncle.

Yak

Jacob Yerushalmy (Fig. 1) was born in Ottoman-controlled Ekron, near Jerusalem, in 1904, the youngest of 8 children.¹ His father, William Zev Yerushalmy, owned a pharmacy in Jerusalem. When Yak completed gymnasium (high school) in 1924, he went abroad to university; no universities existed in British Mandate Palestine. He started in New York City, though the specific university is not known. In 1925, he moved to Baltimore to undertake a doctorate in mathematics at Johns Hopkins, completing his work under Oscar Zariski, himself a recent addition to the Hopkins faculty. Yerushalmy's thesis topic, Construction of Pencils of Equianharmonic Cubic, reflected his interest in topology and algebraic geometry; he successfully defended in 1930.²

Figure 1

Just as Yerushalmy received his doctorate, the United States entered the Great Depression. Teaching positions were difficult to find. Yerushalmy made do as a National Research Fellow in mathematics at the University of Illinois for the 1930–1931 academic year, and continued as a fellow at Princeton University the following year. Not much is known about this period in Yerushalmy's career, save the absence of skill in poker.³

In 1932, desperate for work, Yerushalmy taught mathematics at Johns Hopkins for a year as an Instructor. One of his students was the son of Lowell Reed, Jr., Chair of the university's biostatistics department. Using this entry to the elder Reed, Yerushalmy complained bitterly of the lack of work available for a Hopkins-trained mathematician. He indicated his willingness to beg, cup in hand, at the center of downtown Baltimore with a sign, “Johns Hopkins Ph.D. in mathematics.” Horrified, Reed directed Yerushalmy to contact the biostatistics department administrator regarding a position as an Assistant Professor. Thus Yerushalmy turned his attention to biostatistics.

Yerushalmy remained at Johns Hopkins until 1935, though there is little evidence of career advancement. At Johns Hopkins, Yerushalmy befriended another member of the biostatistics faculty, Carroll E. Palmer, trained in demography. For a period of almost 2 decades (beginning in the early 1940s), Yerushalmy would work with Palmer (see below). Also of significance, in 1933 Yerushalmy met and, 1 year later eloped with, Eva Zemil, the eldest daughter of a Baltimore businessman. In 1935, with his career stuck, Yerushalmy accepted a position as a biostatistician for the Department of Health in Albany, New York.

His billet allowed Yerushalmy to gain confidence in his biostatistical skills. It also placed him in contact with Dr. Morton L. Levin, the Assistant Director for Cancer Control. The 2 developed a fast friendship, both professional and personal, that would last for almost 4 decades. Yerushalmy kindled in Levin an interest in biostatistics, and encouraged cancer epidemiology studies sponsored by Levin throughout the late 1940s and 1950s. (These studies included the first case–control study on cigarette smoking and lung cancer conducted in the United States).⁴ In turn, Levin showed Yerushalmy the need for biostatistical input into epidemiologic investigations of chronic diseases, such as cancer.

In 1938, another opportunity beckoned to Yerushalmy at the National Institutes of Health (NIH). There, Yerushalmy came to know Harold Dorn, a young demographer leading the First National Cancer Survey. During the 3 years prior to the attack on Pearl Harbor, Dorn began developing ideas for studies following up on survey findings, discussing them with his colleagues. Through such discussions with Dorn, Yerushalmy raised the issue of diagnostic certainty. He reasoned that one often does not truly know if a malignancy was present until it has been shown by biopsy. As a biostatistician, Yerushalmy considered the distribution of biopsy interpretations, along with the possibility of errors.

It was as the NIH that Yerushalmy also began his lifelong interest in child development. He was tasked to investigate childhood mortality. Considering the still-incomplete nature of the national vital statistics system, such an assignment must not have been a desirable one for most ambitious biostatisticians.⁵ Yerushalmy persevered, and his reports showed an irreverent and thoughtful investigator.^6,7

Three years as a statistician at the NIH working with child development data opened another opportunity for Yerushalmy. When Palmer left Johns Hopkins to run the field studies section of the United States Department of Labor Children's Bureau, Yerushalmy left the NIH to work for him, becoming Palmer's Director of the Division of Statistical Research. It was during his tenure at the Department of Labor that Yerushalmy's interest in child development blossomed. This interest would guide his future research, and it would ultimately lead to the formation of the Kaiser-Permanente database in northern California for the conduct of epidemiologic studies (see below). However, the entry of the United States into World War II changed many of the federal government's funding priorities; child development research was not among them, but tuberculosis research was. In early 1943, Yerushalmy's boss moved from the Children's Bureau to the United States Public Health Service Tuberculosis Control Bureau; Yerushalmy followed.

From 1943 until his departure 4 years later to establish a biostatistics department at the University of California, Berkeley School of Public Health, Yerushalmy would serve as the chief statistician at the Tuberculosis Control Bureau. In 1945, Time magazine referred to him as the “Public Health Service's top statistician.”⁸ At the Bureau, Yerushalmy opened 2 areas of research, one (the reproducibility of clinical findings) with which he would be identified for the remainder of his career, and the other (criteria for causal inference in epidemiology) which became the core of the coming debate on cigarette smoking and lung cancer He would also develop a marked skepticism regarding the quality of data used in any epidemiologic query.

In 1943, conventional medical thought held that a physician could diagnose pulmonary tuberculosis on the basis of a chest X-ray. As a nonphysician and earning considerably less than a physician, Yerushalmy was contemptuous of physicians and their techniques. On joining the Tuberculosis Control Bureau, Yerushalmy was immediately skeptical when told of physicians' ability to diagnose tuberculosis by X-ray. Based on his NIH experience, Yerushalmy knew there had to be some misdiagnoses. Working with Palmer, Yerushalmy undertook a series of pioneering studies into the consistency of reading chest X-rays. His findings, published in 1947, indicated that physicians were not only unable to agree on a diagnosis among themselves, but that a given physician would change the reading of the same X-ray when seeing it a second or third time.⁹ In doing so, Yerushalmy developed the concepts of sensitivity and specificity. The findings challenged the medical community, which held the physician's diagnosis as beyond reproach. These findings also had profound implications in the coming development of the randomized clinical trial, indicating the need for centralized laboratories and double-blinded allocation of treatment.

Yerushalmy also began a series of discussions with Palmer on causal inference in epidemiology. The discussions did not develop out of the Bureau's research program, per se, but from Yerushalmy's previous work on childhood mortality. They began with a consideration of what evidence is necessary to make a causal inference. These discussions continued for another decade and a half, and began the development of modern approaches to discerning causality.

In 1947, the Tuberculosis Control Bureau learned of the British randomized trial of streptomycin for treating tuberculosis, directed by the statistician Bradford Hill. Palmer dispatched Yerushalmy to England to learn about the trial. Yerushalmy's trip consisted of visits with Hill, the study coordinating center staff, and selected clinical centers. On his return, during a Sunday visit by both Morton Levin and Abe Lilienfeld, Yerushalmy described Hill as a competent statistician whose data analyses would be solid, but he noted the lack of concern regarding the quality of the data. His conclusion: “I wouldn't believe anything Hill publishes about those data.”

Shortly after his return from England, Yerushalmy was approached about an opportunity in California: the University of California at Berkeley was opening a school of public health and needed someone to organize the biostatistics department. In 1948, Yerushalmy took on the assignment.¹⁰

Abe

Abraham “Abe” Lilienfeld (Fig. 2) was born in New York City on November 13, 1920. His father, Joe, came from a well-to-do family in Galicia (western Ukraine). Joe had trained as a rabbi. As his wife refused to cover her hair (expected of all married Jewish women), Joe could not find a pulpit. The family migrated to the United States in 1914 to escape the draft. They expected victory by the Central Powers, and accordingly maintained their wealth in German marks. At the conflict's conclusion, the family was destitute. Until 1930, Abe, his parents, and his older brother Sam lived on the Lower East Side of Manhattan in New York City, as Joe eked out a living as a bookbinder.

Figure 2

With the Great Depression, employment opportunities for bookbinders disappeared. By then, Sam had entered medical school in England (quotas against Jewish students precluded attending a United States school). Lilienfeld's mother took matters in hand and purchased a run-down chocolate store with a soda fountain. Within a year, the family had made the shop profitable, sold it and purchased another run-down chocolate shop. In another year, that shop was rejuvenated and sold. This pattern continued until Abe entered medical school in 1941, except for a brief interruption in 1932.

In 1932, the Lilienfelds migrated to Palestine in response to the World Zionist Organization's call for the Fifth Aliyah for Jews to return to the Holy Land. However, under British control, much of the Palestinian economy was closed to Jewish involvement. Joe struggled to support his family. After Sam graduated from medical school, he returned to the United States, followed shortly by his family's return to New York City. On their arrival in 1933, Sam informed them of his marriage to a member of a prominent Baltimore family; Sam subsequently established his practice in Baltimore.

For the next 4 years, Abe attended Erasmus High School. He maintained excellent grades in science and mathematics, but found an interest in history. In the evenings, he studied with his father—usually the Talmud, as well as modern philosophers. The Talmudic discussions refined Abe's thinking, forcing him to identify the key variables defining a situation. In 1938, with high school graduation looming, the Lilienfelds discussed the issue of Abe's college education. Abe wanted to study business administration at City College. Given the family's chocolate store business, he thought it a useful field. His parents, however, focused on medicine. With developments in Nazi Germany in mind, they reasoned medical skills were portable; a business was not. Accordingly, Abe applied to 2 colleges: City College and Johns Hopkins. Admitted to both, he packed at the end of the summer of 1938 and moved in with his brother Sam to attend Johns Hopkins.

College opened new vistas to Abe. He took courses in Middle Eastern studies given by Foxwell Albright (doyen of the field), as well as French and German. He enrolled in the premedical curriculum, and he found a particular interest in biophysics. In a bid to minimize the financial drain on his family, he increased his course load and graduated in 3 years. He applied to the Johns Hopkins School of Medicine and was told during the interview that, while he was otherwise qualified, he would be rejected because he was Jewish.

In the fall of 1941, Abe enrolled at the Albany Medical College. He experienced a difficult first semester; his father passed away during that term. In the middle of the semester, Lilienfeld was ready to quit school and return to New York City to help support his mother. At the train station, his roommate persuaded him to at least finish the semester. He did so, and then transferred to the University of Maryland medical school in Baltimore.

After the attack on Pearl Harbor, all United States medical schools were nationalized, all medical students inducted into the military, and classes held year-round. During this time, Lilienfeld was introduced to Lorraine, the youngest daughter of Baltimore businessman Max Zemil. During the next 2 years, as Abe progressed through medical school, he developed a strong relationship with Max; in 1943, having received the approval of Lilienfeld's armed forces commander, Abe and Lorraine married. In marrying Lorraine, Lilienfeld came into contact with Lorraine's brother-in-law, Yak.

Abe graduated from medical school in 1944. For the next year, he persevered through a traditional rotating internship at a Baltimore community hospital. He disliked clinical medicine, particularly the emphasis on performing procedures for income. Having completed his internship, Lilienfeld was assigned by the military to Walter Reed Hospital, then one of the crown jewels in American medicine. He requested assignment in internal medicine, but the only slot open was in otolaryngology; for 6 months, he worked as an otolaryngology resident. Bored and unhappy, Lilienfeld sought transfer to anything other than otolaryngology. When a position opened in the obstetrics-gynecology residency program, he transferred.

While based at Walter Reed, Lilienfeld and his wife lived with the Yerushalmys in Bethesda, Maryland. Lilienfeld's previous experience in British Mandate Palestine prepared him well for this experience; not only conversational in Hebrew, he also knew many of the “local” idioms familiar to Yerushalmy. The 2 conversed on a wide range of topics, not only medicine or family. For example, during one hot evening during the summer of 1945, Lilienfeld realized his neck was particularly hot. In the resulting discussion, he and Yerushalmy considered how many layers of fabric were in a shirt collar worn with a necktie. Their conclusion: more than 20. Accordingly, Lilienfeld resolved that upon discharge from the military service, he would wear clip-on ties; at the time, only bowties were available as clip-ons, and he wore them for the rest of his life.

It was in Washington DC that Yerushalmy introduced Lilienfeld to Morton Levin, who was there for business. Lilienfeld indicated his displeasure with clinical practice and the fee-for-service health care system then dominant in the United States. Levin responded with a description of a career in public health, and he told Lilienfeld that, should Abe decide to leave clinical practice, Levin would be happy to find him a preventive medicine training position at the New York Department of Health.

In 1946, Lilienfeld's residency program assigned him to gynecology at a local community hospital. Soon thereafter, Lilienfeld had lost all interest in clinical practice, the final straw coming during a hysterectomy when the surgeon commented upon delivering the uterus and tossing it into a specimen bucket, “That's my wife's new mink coat.” Lilienfeld remembered seeing an advertisement on the medical staff bulletin board at Walter Reed recruiting physicians for the Public Health Service (PHS). Since few physicians were willing to part with the rewards promised by private practice, the PHS had difficulty recruiting physicians. Lilienfeld immediately contacted the recruitment officer. Within weeks, Lilienfeld was transferred to the PHS. Yerushalmy requested Lilienfeld be assigned to the Tuberculosis Control Bureau, and Lilienfeld found himself reporting to his brother-in-law.

Yerushalmy immediately dispatched Lilienfeld to the field office in Philadelphia. For the remainder of his military service, Lilienfeld worked on tuberculosis control. At the conclusion of his military service, Lilienfeld contacted Levin and asked if Levin had any positions open in his training program. Levin did, and Lilienfeld joined the preventive medicine training program. For 3 months, Lilienfeld was posted to Lake Placid to provide basic public health services. In late summer 1948, Levin directed Lilienfeld to enter the MPH program at the Johns Hopkins School of Hygiene and Public Health.

The 1948–1949 academic year at Johns Hopkins opened new vistas for Lilienfeld, not only the course material but also the faculty. He was taught by Alex Langmuir, William Cochran, Ernest Stebbins, and Phil Sartwell, all of whom had major impact on his career development. He also met a young biostatistician who would become important in his later career—Irwin Bross, a postdoctoral fellow in biostatistics. At the conclusion of the 1-year curriculum, Lilienfeld graduated with his MPH degree and moved his family to Albany, New York for the training program's second year.

The year spent in Albany was a significant one for Lilienfeld. During 1949–1950, he displayed a vision of public health as all-encompassing. Early in the year, on behalf of a legislator interested in cerebral palsy, the New York State Legislature passed a bill directing the state health department to spend 1 million dollars for this disease. The Commissioner directed Levin to oversee the spending. Levin directed Lilienfeld to spend the million dollars, expecting the money to be quickly spent on block grants to construct buildings for cerebral palsy clinics. Instead, Lilienfeld proceeded to develop a program for cerebral palsy in New York State. He included a research component (producing seminal epidemiologic findings), training for physicians and nurses, and the creation of a network of clinics (signing long-term leases to maximize the number of clinics established throughout the state and sponsoring quarterly meetings of the clinic directors) to treat cerebral palsy patients.^11–13 The program established New York State as a center for research and treatment of cerebral palsy for at least the next 2 decades. Lilienfeld's view of public health and epidemiology, with its ultimate focus on the reduction of disease incidence, required a catholic effort, one ably financed with the bounty provided by the legislature.

One component of Lilienfeld's program concerned the epidemiology of cerebral palsy. He discussed possible research designs with Yerushalmy through the mail and at the 1949 Annual Meeting of the American Public Health Association. During World War II, Yerushalmy had spoken with Joseph Berkson, the biostatistician-physician at the Mayo Clinic, about the challenges of using vital statistics data to account for neonatal deaths. Berkson noted the linkage already in use within the Mayo Clinic records.¹⁴ Yerushalmy subsequently began to link birth and death records, but he didn't complete the effort before changing jobs. After the war, Dunn published the concept of record linkage,¹⁵ presumably following one or more discussions with Yerushalmy, although there are no records to document such discussion. Having heard Yerushalmy's thoughts, Lilienfeld linked the cases of cerebral palsy with their birth records.^11,13 Lilienfeld later told Yerushalmy of his success—a key piece of information for Yerushalmy when the latter was asked to consult on a Hawaiian database study in 1954 during a sabbatical in Honolulu. Yerushalmy recognized the advantage that all healthcare on the islands was provided in only a few places, making it possible to link various childhood health and vital records for persons living in the territory. One could then examine a variety of questions dealing with child development.

During the mid 1950s, Lilienfeld continued to publish papers on cerebral palsy and other neurologic conditions, and he suggested that peripartum events were associated with the occurrence of these conditions.^16,17 Officials at the National Institutes of Neurological Diseases and Blindness thought the time appropriate to validate the findings and discussed the idea with Lilienfeld. He suggested a study using record linkage between birth records and medical records. He noted to Yerushalmy the subsequent development of the National Collaborative Perinatal Project. Concurrently, Yerushalmy initiated his own record linkage project, the Child Health and Development Studies, in which vital records were linked with medical records at Kaiser Permanente, a San Francisco-based health maintenance organization. This effort pioneered what subsequently became the Kaiser Permanente linked database, which has supported many epidemiologic studies since its development in the late 1950s.^18,19

At the end of his training program in 1950, Lilienfeld searched for a job. His responsibilities now included 2 children. Although Levin hoped to keep him in Albany, there were no open positions available. Ernest Stebbins, Dean at the Johns Hopkins School of Hygiene and Public Health, intervened, arranging for Lilienfeld to serve as a regional health officer in the Baltimore City Health Department. The epidemiology department at Johns Hopkins had sustained the loss of Alexander Langmuir in 1949 to direct the Epidemic Intelligence Service at the Centers for Disease Control (CDC). The School needed teaching assistants to cover for Langmuir's absence, and Stebbins thought Lilienfeld could function as such until Professor Kenneth Maxcy, the department chair, retired. Lilienfeld would then join the department as a full-time junior faculty member. Accordingly, Lilienfeld joined the Baltimore City Health Department.

Interactions

During their careers, Lilienfeld and Yerushalmy discussed topics ranging from analytical methods to causal inference, including data quality, statistical methods and discerning biologic inferences from epidemiologic data.

Data Quality

Data quality became a lifelong interest of Lilienfeld's; it was a guaranteed topic for questions by any thesis committee that included Lilienfeld. That interest first manifested after Yerushalmy's studies into the accuracy of physician radiograph interpretation, and it solidified after Yak's return from his 1947 trip to England. By the time Lilienfeld arrived in Albany in 1948, the seeds of his interest in data quality had been planted. Indeed, one of Lilienfeld's first publications dealt with the quality of data on birth certificates.²⁰

Bross raised the issue of whether misclassification in a 2 × 2 table biased the odds ratio estimate towards unity.²¹ Concerns about the questions used in Wynder and Graham's study of cigarette smoking and lung cancer stimulated a discussion among Bross, Yerushalmy, and Lilienfeld at an International Biometrics Society meeting in the mid 1950s. Lilienfeld thought the issue worth examining in the field. In particular, no one knew whether misclassification occurred in a study. The test case would be the accuracy of a given man's statement regarding his circumcision status. It turned out men were generally unable to state their status accurately when compared with the assessment of a health care provider, suggesting considerable bias may be present in some data sets.²² Later in his career, Lilienfeld collaborated with Diamond on whether misclassification resulted in an overestimate of the relative risk or an underestimate.^23,24 Not until Baron's 1977 formulation of a general model of misclassification did Lilienfeld consider the issue settled.²⁵

Statistical Methods

The analysis of the polio—gamma globulin study undertaken by the CDC during the 1950s and the analyses of data used by Modan in his postdoctoral study of polycythemia vera exemplify the interaction of Abe and Yak with regard to statistical methods.

In the early 1950s, the CDC invited Lilienfeld to assist in training Epidemic Intelligence Service personnel.²⁶ Following one such training session, Langmuir contacted Lilienfeld about joining CDC, using Lilienfeld's direction of an epidemiologic study of gamma globulin in the prevention of polio as a pilot. The study commenced, and in short order the data were collected and analyzed. Langmuir assembled an advisory board of health officers from New York, Pennsylvania, Illinois, Massachusetts, and Michigan (the most populous states in the country) to vet the study findings. In 1954, Lilienfeld presented the findings to a meeting of this board.

Lilienfeld's presentation began smoothly as he described the study background, study population recruitment, and data collection. When he came to the description of the statistical methods used for data analysis, several of the board members objected. The key method used, the analysis of variance, was one with which they lacked familiarity. Yerushalmy had suggested use of the technique to Lilienfeld in 1953. Concerned about its appropriateness, Lilienfeld consulted with Cochran, who agreed with Yerushalmy's recommendation. At the board meeting, Lilienfeld tried responding to the objections by noting the assent of both Yerushalmy and Cochran. The board members would have none of it, however. Lilienfeld asked for a coffee break.

During the break, Lilienfeld approached Leonard Schuman, who represented Illinois. Lilienfeld had met Schuman at the previous American Public Health Association annual meeting, and Schuman was the only person present besides Langmuir whom Lilienfeld knew. He told Schuman he needed some help. Schuman replied, “Are you sure about the technique? I know nothing about analysis of variance, so I need to be sure it's OK.” “Absolutely. Bill Cochran and Yak Yerushalmy both reviewed its use. They thought it was fine.” “Then don't worry, Abe. Follow my lead.”

After the break, Schuman spoke up. “During the break, I had a chance to review some of the aspects of this important, innovative study,” he began. “I must confess, though, I don't understand all the uproar about the use of analysis of variance for the statistical analysis. It's a common-enough technique, elementary actually. Students in schools of public health know all about it, and we use it all the time in the Illinois Department of Health. It seems perfectly acceptable to me. Why do we need to spend precious time discussing something so elementary when there are important public health policy issues to be addressed?” With those comments, there was no further discussion regarding the statistical analysis stopped and the board approved the findings.^26,27 Thus was the analysis of variance introduced to the CDC, and the professional collaboration between Leonard Schuman and Abe Lilienfeld began.

Another example of the influential interactions of Abe and Yak occurred in the early 1960s. A. Michael Davies, Chair of the Department of Preventive Medicine at the Hebrew University Hadassah Medical School, sent one of his students, Baruch Modan, to Lilienfeld for training in epidemiology. Modan, a hematologist, wanted to study a hematological condition for his dissertation. Lilienfeld identified the outcomes associated with different treatments of polycythemia vera as meriting inquiry. Modan's findings challenged conventional wisdom regarding such treatments, and resulted in the formation of the seminal Polycythemia Vera Study Group.^28,29 Little known, however, is the origin of the statistical techniques involving life tables, which Modan used in his thesis. Although life tables had been widely used in epidemiology for almost a century,^30,31 there had been little attention to the statistical theory underlying their analysis. One of Yerushalmy's colleagues, Chin Long Chiang, had subjected the various parameters of the life table to mathematical analysis. While discussing Modan's thesis with Yerushalmy at the annual meeting of the American Public Health Association, Lilienfeld asked about potential analytical methods. Yerushalmy suggested using some of Chiang's work on life tables, which Modan subsequently did.

Biologic Inferences

Perhaps no area of interaction between Lilienfeld and Yerushalmy was as significant as the discussions the 2 had regarding the derivation of causal inferences from epidemiologic data. Working with his former boss, Carroll Palmer, Yerushalmy initiated one of the key discussions in 1950s chronic disease epidemiology: what evidence is needed to discern an etiologic factor.³¹ The issue arose in the context of the debate on cigarette smoking and lung cancer. Yerushalmy, a smoker (at times, a chain-smoker) and soon to be a paid consultant to the tobacco industry, had reviewed the available information. He did not view it as wholly consistent with an etiologic relationship between cigarette smoking and lung cancer. In 1959, Yerushalmy and Palmer proposed a set of guidelines, modified from the Henle-Koch postulates, as the basis for deriving causal inferences.³² Evans³³ discussed these guidelines as seminal to the subsequent Surgeon General's guidelines and the Hill guidelines. Like Sir Ronald Fisher, Yerushalmy opposed the notion that cigarette smoking caused lung cancer until his dying day.³⁴ Unlike Fisher, Yerushalmy did not dispute the value of observational data more generally. Rather, he viewed the smoking data as not consistently supportive of a causal association.

At professional biostatistical meetings, during the late 1950s and early 1960s, Yerushalmy would often lunch with fellow Johns Hopkins alumnus Joseph Berkson, the physician-statistician from the Mayo Clinic. Yerushalmy and Berkson both denied any relationship between cigarette smoking and lung cancer.³⁵ Their luncheon discussions often focused on the basis for denying the relationship, with Berkson strongly in the camp of Sir Ronald Fisher. Berkson contributed the concept of specificity (the causal factor is specific in its biologic effect) to the Yerushalmy-Palmer guidelines. (Indeed, when evidence of an association between cigarette smoking and bladder cancer began to emerge, Berkson challenged, “What now, do we breathe through our urethras?”)

During the 1950s and 1960s, Abe and Yak often debated the cigarette smoking–lung cancer relationship, usually in a haze of smoke from Yak's cigarette and Abe's pipe. These discussions forced Abe to refine his thinking about the relationship in epidemiologic terms. Stimulated by the discussion into thinking about chronic disease epidemiology more generally, Lilienfeld brought concepts from infectious disease epidemiology (such as incubation periods) to chronic disease settings (such as occupational cancer incidence) and vice versa (such as the use of case–control study designs for epidemiologic studies of infectious diseases). These ideas culminated in his inaugural Wade Hampton Frost Lecture in 1973.^36,37 The discussions also forced him to consider the process of deriving biologic inferences, the subject of another series of papers.^36,38,39 Further, by refining his arguments, Abe was better prepared as an advocate against the tobacco industry; his 1962 paper in The Nation was predicated on arguments he had fleshed out in discussions with Yerushalmy.⁴⁰ (In 1964, when the Surgeon General's Committee was constituted, Lilienfeld was told by friends in the Public Health Service that his 1962 paper had rendered him too “partial” to be appointed to the committee. His “consolation prize” came later in 1964 when he was asked to serve as the Staff Director for the President's Commission on Heart Disease, Cancer, and Stroke.⁴¹)

DISCUSSION

The interactions between Lilienfeld and Yerushalmy were fruitful for both men throughout their careers. In Abe, Yerushalmy found a capable voice for many of his ideas. In Yak, Lilienfeld found a source of useful biostatistical concepts and study design ideas. Through his early interactions with Yerushalmy, Lilienfeld developed an understanding for the statistical perspective in epidemiology. He subsequently was able to form alliances with such statisticians as William Cochran (then Chair of the Biostatistics Department at Johns Hopkins), Jerry Cornfield, Sam Greenhouse, and Max Halperin (at the NIH). Lilienfeld's ability to interact with statisticians bore fruit in the 1960s, including his seminal role in the development of logistic regression and his efforts to save the Framingham Study from closure.^42–44 The latter effort developed in part out of his familiarity with data analysis techniques developed by Jerry Cornfield expressly for Framingham.^45,46 Lilienfeld's efforts to advance biostatistics within epidemiology was recognized by the statistical community, which made him a Fellow in the American Statistical Association.

Yerushalmy's need for a voice in the epidemiologic community may not have been as important in the 1950s as it would become in the 1960s and early 1970s. His disdain for physicians likely contributed in part to the frictions between him and Reuel Stallones, a professor in the epidemiology department at the University of California, Berkeley. Stallones eventually left Berkeley to found the School of Public Health at the University of Texas, Houston. At that point, absent Lilienfeld, Yerushalmy had few avenues left into the epidemiology community. The situation was compounded by Yerushalmy's stand on cigarette smoking and lung cancer. As the 1960s progressed, the epidemiologic evidence for the relationship grew, and, with that growth, those (like Yerushalmy) not recognizing the evidence as increasingly conclusive became intellectually isolated. As much as Lilienfeld needed and benefited from his interactions with Yerushalmy, so too did Yerushalmy need and benefit from those same interactions—a truly symbiotic relationship up until Yerushalmy's death in 1973.

What can we learn from this interaction? There are a number of lessons: advocacy and the role of public health in epidemiology, the formation of modern epidemiology prior to the development of the cigarette smoking–lung cancer controversy, and the significance of community within the field to facilitate its advancement.

Advocacy

Both Lilienfeld and Yerushalmy began their careers focused on public health, and both spent considerable portions of their early careers in public health settings. They were strong advocates for public health. The notion that researchers should not advocate on behalf of public health would likely strike each as foreign; advocacy was a given aspect of membership in the public health community. It is good to be reminded that advocacy by epidemiologists on behalf of public health has a long and respectable history. Advocacy by researchers did not inhibit the development of well-founded health policy in the United States; indeed, it no doubt assisted it.

Formation of Modern Epidemiology Prior to the Cigarette Smoking–Lung Cancer Controversy

Yerushalmy's entry into public health began prior to the post-World War II cigarette smoking–lung cancer controversy, while Lilienfeld's coincided with the development of the issue. Even before the first American case–control studies were published in 1950, Lilienfeld began to focus on noninfectious diseases. The second paper of his career examined the relationship between alcohol consumption and liver cirrhosis.⁴⁷ In general, epidemiology had begun its transformation from an infectious disease orientation to one dominated by chronic diseases.

Community

As noted previously, both Lilienfeld and Yerushalmy conducted their work within the context of public health. They viewed themselves as members of the public health community. While they had many disagreements, personal and professional, during their careers, they remained members of the public health community and continued their interactions as such. At a time of increasing polarization within epidemiology and within the scientific community generally, we can emulate these 2 leaders whose commitment to public health transcended their personal disagreements, even for major issues such as whether cigarette smoking was a cause of lung cancer. For example, Yerushalmy did not allow that disagreement to stop his suggestion to Lilienfeld of analytic techniques for use by Modan. Communication is key the formation and sustenance of a community; both Lilienfeld and Yerushalmy sought to maintain such communication. We would do well to continue to follow their lead.

CONCLUSION

It is popular in the history of medicine and public health to focus on social trends and movements as explanations of historical phenomena. In some instances, such as the promotion of cigarette smoking and the discovery of its subsequent health effects, it is difficult to examine the events in other than a societal context.^48–50 In the present instance, however, the reader may see the effect of 2 individuals, sometimes collaborating, sometimes disagreeing, on the subsequent development of epidemiology. For example, record linkage in the United States can be traced back (at least in part) to Berkson's suggestion through Yerushalmy to Lilienfeld to link birth and death records as a means of creating a population for epidemiologic observation. The later development of the Kaiser-Permanente database built on these efforts. Modern pharmacoepidemiology was one result of the availability of the Kaiser-Permanente and other similar databases developed in the 1970s and early 1980s.⁵¹ Few today are aware of these connections, as little has been published on them.

Although this paper has presented some information about the intellectual output of both men, each could be the subject of dissertations in the history of epidemiology and public health. My observations are not exhaustive; there are likely other interactions I know nothing about. The nature of my relationship with both men (and their relationship with each other) carried with it the lack of formal documentation. Lilienfeld rarely saved correspondence, working drafts of manuscripts, presentations, etc. The same is true for Yerushalmy. Much of the evidence reported in this paper is apocryphal; I trust that this does not diminish the credibility of the findings themselves.

ACKNOWLEDGMENTS

I thank Alfredo Morabia for his helpful suggestions regarding the manuscript.

REFERENCES