Graham, who knew Little from his days at the ASCC, was livid. In a stinging rebuttal written to the editor, he complained, “A causal relationship between heavy cigarette smoking and cancer of the lung is stronger than for the efficacy of vaccination against smallpox, which is only statistical.”
Indeed, like many of his epidemiologist peers, Graham was becoming exasperated with the exaggerated scrutiny of the word cause. That word, he believed, had outlived its original utility and turned into a liability. In 1884, the microbiologist Robert Koch had stipulated that for an agent to be defined as the “cause” of a disease, it would need to fulfill at least three criteria. The causal agent had to be present in diseased animals; it had to be isolated from diseased animals; and it had to be capable of transmitting the disease when introduced into a secondary host. But Koch’s postulates had arisen, crucially, from the study of infectious diseases and infectious agents; they could not simply be “repurposed” for many noninfectious diseases. In lung cancer, for instance, it would be absurd to imagine a carcinogen being isolated from a cancerous lung after months, or years, of the original exposure. Transmission studies in mice were bound to be equally frustrating. As Bradford Hill argued, “We may subject mice, or other laboratory animals, to such an atmosphere of tobacco smoke that they can—like the old man in the fairy story—neither sleep nor slumber; they can neither breed nor eat. And lung cancers may or may not develop to a significant degree. What then?”
Indeed, what then? With Wynder and other coworkers, Graham had tried to expose mice to a toxic “atmosphere of tobacco smoke”—or at least its closest conceivable equivalent. Persuading mice to chain-smoke was obviously unlikely to succeed. So, in an inspired experiment performed in his lab in St. Louis, Graham had invented a “smoking machine,” a contraption that would puff the equivalent of hundreds of cigarettes all day (Lucky Strikes were chosen) and deposit the tarry black residue, through a maze of suction chambers, into a distilling flask of acetone. By serially painting the tar on the skins of mice, Graham and Wynder had found that they could create tumors on the backs of mice. But these studies had, if anything, fanned up even more controversy. Forbes magazine had famously spoofed the research by asking Graham, “How many men distill their tar from their tobacco and paint it on their backs?” And critics such as Little might well have complained that this experiment was akin to distilling an orange to a millionth of a million parts and then inferring, madly, that the original fruit was too poisonous to eat.
Epidemiology, like the old man in Hill’s fairy story, was thus itself huffing against the stifling economy of Koch’s postulates. The classical triad—association, isolation, retransmission—would simply not suffice; what preventive medicine needed was its own understanding of “cause.”
Once again, Bradford Hill, the éminence grise of epidemiology, proposed a solution to this impasse. For studies on chronic and complex human diseases such as cancer, Hill suggested, the traditional understanding of causality needed to be broadened and revised. If lung cancer would not fit into Koch’s straitjacket, then the jacket needed to be loosened. Hill acknowledged epidemiology’s infernal methodological struggle with causation—this was not an experimental discipline at its core—but he rose beyond it. At least in the case of lung cancer and smoking, he argued, the association possessed several additional features:
It was strong: the increased risk of cancer was nearly five- or tenfold in smokers.
It was consistent: Doll and Hill’s study, and Wynder and Graham’s study, performed in vastly different contexts on vastly different populations, had come up with the same link.
It was specific: tobacco was linked to lung cancer—precisely the site where tobacco smoke enters the body.
It was temporal: Doll and Hill had found that the longer one smoked, the greater the increase in risk.
It possessed a “biological gradient”: the more one smoked in quantity, the greater the risk for lung cancer.
It was plausible: a mechanistic link between an inhaled carcinogen and a malignant change in the lung was not implausible.
It was coherent; it was backed by experimental evidence: the epidemiological findings and the laboratory findings, such as Graham’s tar-painting experiments in mice, were concordant.
It behaved similarly in analogous situations: smoking had been correlated with lung cancer, and also with lip, throat, tongue, and esophageal cancer.
Hill used these criteria to advance a radical proposition. Epidemiologists, he argued, could infer causality by using that list of nine criteria. No single item in that list proved a causal relationship. Rather, Hill’s list functioned as a sort of à la carte menu, from which scientists could pick and choose criteria to strengthen (or weaken) the notion of a causal relationship. For scientific purists, this seemed rococo—and, like all things rococo, all too easy to mock: imagine a mathematician or physicist choosing from a “menu” of nine criteria to infer causality. Yet Hill’s list would charge epidemiological research with pragmatic clarity. Rather than fussing about the metaphysical idea about causality (what, in the purest sense, constitutes “cause”?), Hill changed its emphasis to a functional or operational idea. Cause is what cause does, Hill claimed. Often, like the weight of proof in a detective case, the preponderance of small bits of evidence, rather than a single definitive experiment, clinched cause.
Amid this charged and historic reorganization of epidemiology, in the winter of 1956, Evarts Graham suddenly fell ill with what he thought was the flu. He was at the pinnacle of his career, a surgeon in full. His legacy was legion: he had revolutionized lung cancer surgery by stitching together surgical procedures learned from nineteenth-century TB wards. He had investigated mechanisms by which cancer cells arose, using tobacco as his chosen carcinogen. And with Wynder, he had firmly established the epidemiological link between cigarettes and lung cancer.
In the end, though, it was his prior aversion to the theory that he himself had proved that undid Evarts Graham. In January 1957, when the “flu” refused to remit, Graham underwent a battery of tests at Barnes Hospital. An X-ray revealed the cause of his troubles: a large, coarse rind of a tumor clogging the upper bronchioles and both lungs riddled with hundreds of metastatic deposits of cancer. Keeping the identity of the patient hidden, Graham showed his films to a surgical colleague. The surgeon looked at the X-rays and deemed the tumor inoperable and hopeless. Graham then informed him quietly, “[The tumor] is mine.”
On February 14, with his condition deteriorating weekly, Graham wrote to his friend and collaborator the surgeon Alton Ochsner: “Perhaps you have heard that I have recently been a patient at Barnes Hospital because of bilateral bronchogenic carcinoma which sneaked up on me like a thief in the night. . . . You know I quit smoking more than five years ago, but the trouble is that I smoked for 50 years.”
Two weeks later, Graham grew dizzy, nauseated, and confused while shaving. He was brought to Barnes again, to a room a few floors above the operating rooms so beloved by him. He was given intravenous chemotherapy with nitrogen mustard, but to little avail. The “thief” had widely marauded; cancer was growing in his lungs, lymph nodes, adrenal glands, liver, and brain. On February 26, confused, lethargic, and incoherent, he drifted into a coma and died in his room. He was seventy-four years old. By his request, his body was donated to the department of anatomy as an autopsy specimen for other students.
In the winter of 1954, three years before his untimely death, Evarts Graham wrote a strikingly prescient essay in a book entitled Smoking and Cancer. At the end of the essay, Graham wondered about how the spread of tobacco in human societies might be combated in the future. Medicine, he concluded, was not powerful enough to restrict tobacco’s spread. Academic investigators could provide data about risks and argue incessantly about proof and causality, but the solution had to be political. “The obstinacy of [policymakers],” he wrote, “compels one to conclude that it is their own addiction . . . which blinds them. They have eyes to see, but see n
ot because of their inability or unwillingness to give up smoking. All of this leads to the question . . . are the radio and the television to be permitted to continue carrying the advertising material of the cigarette industry? Isn’t it time that the official guardian of the people’s health, the United States Public Health Service, at least make a statement of warning?”
“A statement of warning”
Our credulity would indeed be strained by an assumption that a fatal case of lung cancer could have developed . . . after the alleged smoking by Cooper of Camel cigarettes in reliance upon representations by the defendant in the various forms of advertising.
—Jury verdict on Cooper case, 1956
Certainly, living in America in the last half of the 20th century, one would have to be deaf, dumb and blind not to be aware of the asserted dangers, real or imagined, of cigarette smoking. Yet the personal choice to smoke is . . . the same kind of choice as the driver who downed the beers, and then the telephone pole.
—Open letter from the tobacco industry, 1988
In the summer of 1963, seven years after Graham’s death, a team of three men traveled to East Orange, New Jersey, to visit the laboratory of Oscar Auerbach. A careful man of few words, Auerbach was a widely respected lung pathologist who had recently completed a monumental study comparing lung specimens from 1,522 autopsies of smokers and nonsmokers.
Auerbach’s paper describing the lesions he had found was a landmark in the understanding of carcinogenesis. Rather than initiating his studies with cancer in its full-blown form, Auerbach had tried to understand the genesis of cancer. He had begun not with cancer but with its past incarnation, its precursor lesion—precancer. Long before lung cancer grew overtly and symptomatically out of a smoker’s lung, Auerbach found, the lung contained layer upon layer of precancerous lesions in various states of evolution—like a prehistoric shale of carcinogenesis. The changes began in the bronchial airways. As smoke traveled through the lung, the outermost layers, exposed to the highest concentrations of tar, began to thicken and swell. Within these thickened layers, Auerbach found the next stage of malignant evolution: atypical cells with ruffled or dark nuclei in irregular patches. In a yet smaller fraction of patients, these atypical cells began to show the characteristic cytological changes of cancer, with bloated, abnormal nuclei often caught dividing furiously. In the final stage, these cell clusters broke through the thin lining of the basement membranes and transformed into frankly invasive carcinoma. Cancer, Auerbach argued, was a disease unfolded slowly in time. It did not run, but rather slouched to its birth.
Auerbach’s three visitors that morning were on a field trip to understand that slouch of carcinogenesis as comprehensively as possible. William Cochran was an exacting statistician from Harvard; Peter Hamill, a pulmonary physician from the Public Health Service; Emmanuel Farber,* a pathologist. Their voyage to Auerbach’s laboratory marked the beginning of a long scientific odyssey. Cochran, Hamill, and Farber were three members of a ten-member advisory committee appointed by the U.S. surgeon general. (Hamill was the committee’s medical coordinator.) The committee’s mandate was to review the evidence connecting smoking to lung cancer so that the surgeon general could issue an official report on smoking and lung cancer—the long-due “statement of warning” that Graham had urged the nation to produce.
In 1961, the American Cancer Society, the American Heart Association, and the National Tuberculosis Association sent a joint letter to President Kennedy asking him to appoint a national commission to investigate the link between smoking and health. The commission, the letter recommended, should seek “a solution to this health problem that would interfere least with the freedom of industry or the happiness of individuals.” The “solution,” inconceivably, was meant to be both aggressive and conciliatory—clearly publicizing the link between cancer, lung disease, heart disease, and smoking, yet posing no obvious threat to the freedom of the tobacco industry. Suspecting an insolvable task, Kennedy (whose own political base in the tobacco-rich South was thin) quickly assigned it to his surgeon general, Luther Terry.
Soft-spoken, conciliatory, and rarely combative, Luther Terry was an Alabaman who had picked tobacco as a child. Enthralled from early childhood by the prospect of studying medicine, he had graduated from Tulane University in 1935, then interned in St. Louis, where he had encountered the formidable Evarts Graham in his surgical prime. Terry had moved to the Public Health Service after graduation, then to the NIH in 1953, where, at the Clinical Center, his laboratory had neighbored the clinic buildings where Zubrod, Frei, and Freireich had been waging their battle against leukemia. Terry had thus spent his childhood in the penumbra of tobacco and his academic life in the penumbra of cancer.
Kennedy’s assignment left Terry with three choices. He could quietly skirt the issue—thus invoking the wrath of the nation’s three major medical organizations. He could issue a unilateral statement from the surgeon general’s office about the health risks of tobacco—knowing that powerful political forces would quickly converge to neutralize that report. (In the early sixties, the surgeon general’s office was a little-known and powerless institution; tobacco-growing states and tobacco-selling companies, in contrast, wielded enormous power, money, and influence.) Or he could somehow leverage the heft of science to reignite the link between tobacco and cancer in the public eye.
Hesitantly at first, but with growing confidence—“a reluctant dragon,” as Kenneth Endicott, the NCI director, would characterize him—Terry chose the third path. Crafting a strategy that seemed almost reactionary at first glance, he announced that he would appoint an advisory committee to summarize the evidence on the links between smoking and lung cancer. The committee’s report, he knew, would be scientifically redundant: nearly fifteen years had passed since the Doll and Wynder studies, and scores of studies had validated, confirmed, and reconfirmed their results. In medical circles, the link between tobacco and cancer was such stale news that most investigators had begun to focus on secondhand smoke as a risk factor for cancer. But by “revisiting” the evidence, Terry’s commission would vivify it. It would intentionally create a show trial out of real trials, thus bringing the tragedy of tobacco back into the public eye.
Terry appointed ten members to his committee. Charles LeMaistre, from the University of Texas, was selected as an authority on lung physiology. Stanhope Bayne-Jones, the senior-most member of the committee, was a bearded, white-haired bacteriologist who had moderated several prior committees for the NIH. Louis Fieser, an organic chemist from Harvard, was an expert on chemical carcinogenesis. Jacob Furth from Columbia, a pathologist, was an authority on cancer genetics; John Hickam was a clinical specialist with a particular interest in heart and lung physiology; Walter Burdette, a Utah surgeon; Leonard Schuman, a widely respected epidemiologist; Maurice Seevers, a pharmacologist; William Cochran, a Harvard statistician; Emmanuel Farber, a pathologist who specialized in cell proliferation.
For nine sessions spanning thirteen months, the team met in a sparsely furnished, neon-lit room of the National Library of Medicine, a modern concrete building on the campus of the NIH. Ashtrays filled with cigarette butts littered the tables. (The committee was split exactly five to five among nonsmokers and smokers—men whose addiction was so deep that it could not be shaken even when deliberating the carcinogenesis of smoke.) The committee visited dozens of labs. Data, interviews, opinions, and testimonies were drawn from some 6,000 articles, 1,200 journals, and 155 biologists, chemists, physicians, mathematicians, and epidemiologists. In total, the trials used for the report encompassed studies on about 1,123,000 men and women—one of the largest cohorts ever analyzed in an epidemiological report.
Each member of the committee brought insight to a unique dimension of the puzzle. The precise and meticulous Cochran devised a new mathematical insight to judge the trials. Rather than privilege any particular study, he reasoned, perhaps one could use a method to estimate the relative risk as a composite number through all trials in
the aggregate. (This method, termed meta-analysis, would deeply influence academic epidemiology in the future.) The organic chemist in Fieser was similarly roused: his discussion of chemicals in smoke remains one of the most authoritative texts on the subject. Evidence was culled from animal experiments, from autopsy series, from thirty-six clinical studies, and, crucially, from seven independent prospective trials.
Piece by piece, a highly incontrovertible and consistent picture emerged. The relationship between smoking and lung cancer, the committee found, was one of the strongest in the history of cancer epidemiology—remarkably significant, remarkably conserved between diverse populations, remarkably durable over time, and remarkably reproducible in trial after trial. Animal experiments demonstrating a causal link between smoking and lung cancer were inconclusive at best. But an experiment was not needed—at least not a laboratory experiment in the traditional sense of that word. “The word ‘cause,’” the report read, leaning heavily on Hill’s prior work, “is capable of conveying the notion of a significant, effectual relationship between an agent and an associated disorder or disease in the host. . . . Granted that these complexities were recognized, it is to be noted clearly that the Committee’s considered decision [was] to use the words ‘a cause,’ or ‘a major cause,’ . . . in certain conclusions about smoking and health.”