The Emperor of All Maladies: A Biography of Cancer (46 page)

Wynder and Graham’s trial
followed a simple methodology. Lung cancer patients and a group of control patients without cancer were asked about their history of smoking. The ratio of smokers to nonsmokers within the two groups was measured to estimate whether smokers were overrepresented in lung cancer patients compared to other patients. This setup (called a case-control study) was considered methodologically novel, but the trial itself was thought to be largely unimportant.
When Wynder presented his preliminary ideas
at a conference on lung biology in Memphis, not a single question or comment came from the members of the audience, most of whom had apparently slept through the talk or cared too little about the topic to be roused. In contrast, the presentation that followed Wynder’s, on an obscure disease called pulmonary adenomatosis in sheep, generated a lively, half-hour debate.

Like Wynder and Graham in St. Louis, Doll and Hill could also barely arouse any interest in their study in London. Hill’s department, called the Statistical Unit, was housed in a narrow brick house in London’s Bloomsbury district. Hefty Brunsviga calculators, the precursors of modern computers, clacked and chimed in the rooms, ringing like clocks each time a long division was performed. Epidemiologists from Europe, America, and Australia thronged the statistical seminars. Just a few steps away, on the gilded railings of the London School of Tropical Medicine, the seminal epidemiological discoveries of the nineteenth century—the mosquito as the carrier for malaria, or the sand fly for black fever—were celebrated with plaques and inscriptions.

But many epidemiologists argued that such cause-effect relationships could only be established for infectious diseases, where there was a known pathogen and a known carrier (called a vector) for a disease—the mosquito for malaria or the tsetse fly for sleeping sickness. Chronic, noninfectious diseases such as cancer and diabetes were too complex and too variable to be associated with single vectors or causes, let alone “preventable” causes. The notion that a chronic disease such as lung cancer might have a “carrier” of its own sort, to be gilded and hung like an epidemiological trophy on one of those balconies, was dismissed as nonsense.

In this charged, brooding atmosphere, Hill and Doll threw themselves into work. They were an odd couple, the younger Doll formal, dispas
sionate, and cool, the older Hill lively, quirky, and humorous, a pukka Englishman and his puckish counterpart. The postwar economy was brittle, and the treasury on the verge of a crisis.
When the price of cigarettes was increased
by a shilling to collect additional tax revenues, “tobacco tokens” were issued to those who declared themselves “habitual users.” During breaks in the long hours and busy days, Doll, a “habitual user” himself, stepped out of the building to catch a quick smoke.

Doll and Hill’s study was initially devised as mainly a methodological exercise. Patients with lung cancer (“cases”) versus patients admitted for other illnesses (“controls”) were culled from twenty hospitals in and around London and interviewed by a social worker in a hospital. And since even Doll believed that tobacco was unlikely to be the true culprit, the net of associations was spread widely. The survey included questions about the proximity of gasworks to patients’ homes, how often they ate fried fish, and whether they preferred fried bacon, sausage, or ham for dinner. Somewhere in that haystack of questions, Doll buried a throwaway inquiry about smoking habits.

By May 1, 1948, 156 interviews
had come in. And as Doll and Hill sifted through the preliminary batch of responses, only one solid and indisputable statistical association with lung cancer leapt out: cigarette smoking. As more interviews poured in week after week, the statistical association strengthened. Even Doll, who had personally favored road-tar exposure as the cause of lung cancer, could no longer argue with his own data. In the middle of the survey, sufficiently alarmed, he gave up smoking.

In St. Louis, meanwhile, the Wynder-Graham team had arrived at similar results. (The two studies, performed on two populations across two continents, had converged on almost precisely the same magnitude of risk—a testament to the strength of the association.) Doll and Hill scrambled to get their paper to a journal. In September of that year, their seminal study, “Smoking and Carcinoma of the Lung,” was published in the
British Medical Journal
.
Wynder and Graham had already published their study a few months earlier in the
Journal of the American Medical Association
.

It is tempting to suggest that Doll, Hill, Wynder, and Graham had rather effortlessly proved the link between lung cancer and smoking. But they had, in fact, proved something rather different. To understand that difference—and it is crucial—let us return to the methodology of the case-
control study.

In a case-control study, risk is estimated post hoc—in Doll’s and Wynder’s case by asking patients with lung cancer whether they had smoked. In an often-quoted statistical analogy, this is akin to asking car accident victims whether they had been driving under the influence of alcohol—but interviewing them
after
their accident. The numbers one derives from such an experiment certainly inform us about a potential link between accidents and alcohol. But it does not tell a drinker his or her actual chances of being involved in an accident. It is risk viewed as if from a rearview mirror, risk assessed backward. And as with any distortion, subtle biases can creep into such estimations. What if drivers tend to overestimate (or underestimate) their intoxication at the time of an accident? Or what if (to return to Doll and Hill’s case) the interviewers had unconsciously probed lung cancer victims more aggressively about their smoking habits while neglecting similar habits in the control group?

Hill knew the simplest method to counteract such biases: he had invented it. If a cohort of people could be
randomly
assigned to two groups, and one group forced to smoke cigarettes and the other forced not to smoke, then one could follow the two groups over time and determine whether lung cancer developed at an increased rate in the smoking group. That would prove causality, but such a ghoulish human experiment could not even be conceived, let alone performed on living people, without violating fundamental
principles of medical
ethics.

But what if, recognizing the impossibility of that experiment, one could settle for the next-best option—for a half-perfect experiment? Random assignment aside, the problem with the Doll and Hill study thus far was that it had estimated risk retrospectively. But what if they could set the clocks back and launch their study
before
any of the subjects developed cancer? Could an epidemiologist watch a disease such as lung cancer develop from its moment of inception, much as an embryologist might observe the hatching of an egg?

In the early 1940s, a similar notion had gripped
the eccentric Oxford geneticist Edmund Ford. A firm believer in Darwinian evolution, Ford nonetheless knew that Darwin’s theory suffered from an important limitation: thus far, the evolutionary progression had been inferred indirectly from the fossil record, but never demonstrated directly on a population
of organisms. The trouble with fossils, of course, is that they are fossilized—static and immobile in time. The existence of three fossils A, B, and C, representing three distinct and progressive stages of evolution, might suggest that fossil A
generated
B and fossil B
generated
C. But this proof is retrospective and indirect; that three evolutionary stages exist suggests, but cannot prove, that one fossil had
caused
the genesis of the next.

The only formal method to prove the fact that populations undergo defined genetic changes over time involves capturing that change in the real world in real time—
prospectively
. Ford became particularly obsessed with devising such a prospective experiment to watch Darwin’s cogwheels in motion. To this end, he persuaded several students to tramp through the damp marshes near Oxford collecting moths. Each time a moth was captured, it was marked with a cellulose pen and released back into the wild. Year after year, Ford’s students had returned with galoshes and moth nets, recapturing and studying the moths that they had marked in the prior years and their unmarked descendants—in effect, creating a “census” of wild moths in the field. Minute changes in that cohort of moths, such as shifts in wing markings or variations in size, shape, and color, were recorded each year with great care. By charting those changes over nearly a decade, Ford had begun to watch evolution in action. He had documented gradual changes in the color of moth coats (and thus changes in genes), grand fluctuations in populations and signs of natural selection by moth predators—a macrocosm caught in a marsh.
^*

Both Doll and Hill had followed this work with deep interest.
And the notion of
using a similar cohort of humans occurred to Hill in the winter of 1951—purportedly, like most great scientific notions, while in his bath. Suppose a large group of men could be marked, à la Ford, with some fantastical cellulose pen, and followed, decade after decade after decade. The group would contain some natural mix of smokers and nonsmokers. If smoking truly predisposed subjects to lung cancer (much like bright-winged moths might be predisposed to being hunted by predators), then the smokers would begin to succumb to cancer at an increased rate. By following that cohort over time—by peering into that natural marsh of human pathology—an epidemiologist could calculate the precise relative risk of lung cancer among smokers versus nonsmokers.

But how might one find a large enough cohort? Again, coincidences surfaced. In Britain, efforts to nationalize health care had resulted in a centralized registry of all doctors, containing more than sixty thousand names. Every time a doctor in the registry died, the registrar was notified, often with a relatively detailed description of the cause of death. The result, as Doll’s collaborator and student Richard Peto described it, was the creation of a “fortuitous laboratory” for a cohort study. On October 31, 1951, Doll and Hill mailed out letters to about 59,600 doctors containing their survey. The questions were kept intentionally brief: respondents were asked about their smoking habits, an estimation of the amount smoked, and little else. Most doctors could respond in less than five minutes.

An astonishing number—41,024 of them—wrote back. Back in London, Doll and Hill created a master list of the doctors’ cohort, dividing it into smokers and nonsmokers. Each time a death in the cohort was reported, they contacted the registrar’s office to determine the precise cause of death. Deaths from lung cancer were tabulated for smokers versus nonsmokers. Doll and Hill could now sit back and watch cancer unfold in real time.

In the twenty-nine months between October 1951 and March 1954, 789 deaths were reported in Doll and Hill’s original cohort. Thirty-six of these were attributed to lung cancer. When these lung cancer deaths were counted in smokers versus nonsmokers, the correlation virtually sprang out: all thirty-six of the deaths had occurred in smokers. The difference between the two groups was so significant that Doll and Hill did not even need to apply complex statistical metrics to discern it. The trial designed to bring the most rigorous statistical analysis to the cause of lung cancer barely required elementary mathematics to prove its point.