New York Times
April 18, 2011
On Jan. 21, 1993, the television talk-show host Larry King featured an unexpected guest on his program. It was the evening after Inauguration Day in Washington, and the television audience tuned in expecting political commentary. But King turned, instead, to a young man from Florida, David Reynard, who had filed a tort claim against the cellphone manufacturer NEC and the carrier GTE Mobilnet, claiming that radiation from their phones caused or accelerated the growth of a brain tumor in his wife.
“The tumor was exactly in the pattern of the antenna,” Reynard told King. In 1989, Susan Elen Reynard, then 31, was told she had a malignant astrocytoma, a brain cancer that occurs in about 6,000 adults in America each year. To David Reynard, the shape and size of Susan’s tumor — a hazy line swerving from the left side of her midbrain to the hindbrain — uncannily resembled a malignant shadow of the phone (but tumors, like clouds, can assume the shapes of our imaginations). Suzy, as she was known, held her phone at precisely that angle against her left ear, her husband said. Reynard underwent surgery for her cancer but to little effect. She died in 1992, just short of her 34th birthday. David was convinced that high doses of radiation from the cellphone was the cause.
Reynard v. NEC — the first tort suit in the United States to claim a link between phone radiation and brain cancer — illustrated one of the most complex conceptual problems in cancer epidemiology. In principle, a risk factor and cancer can intersect in three ways. The first is arguably the simplest. When a rare form of cancer is associated with a rare exposure, the link between the risk and the cancer stands out starkly. The juxtaposition of the rare on the rare is like a statistical lunar eclipse, and the association can often be discerned accurately by observation alone. The discipline of cancer epidemiology originated in one such a confluence: in 1775, a London surgeon, Sir Percivall Pott, discovered that scrotal cancer was much more common in chimney sweeps than in the general population. The link between an unusual malignancy and an uncommon profession was so striking that Pott did not even need statistics to prove the association. Pott thus discovered one of the first clear links between an environmental substance — a “carcinogen” — and a particular subtype of cancer.
The opposite phenomenon occurs when a common exposure is associated with a common form of cancer: the association, rather than popping out, disappears into the background, like white noise. This peculiar form of a statistical vanishing act occurred famously with tobacco smoking and lung cancer. In the mid-1930s, smoking was becoming so common and lung cancer so prevalent that it was often impossible to definitively discern a statistical link between the two. Researchers wondered whether the intersection of the two phenomena was causal or accidental. Asked about the strikingly concomitant increases in lung cancer and smoking rates in the 1930s, Evarts Graham, a surgeon, countered dismissively that “the sale of nylon stockings” had also increased. Tobacco thus became the nylon stockings of cancer epidemiology — invisible as a carcinogen to many researchers, until it was later identified as a major cause of cancer through careful clinical studies in the 1950s and 1960s.
But the most complex and most publicly contentious intersection between a risk factor and cancer often occurs in the third instance, when a common exposure is associated with a rare form of cancer. This is cancer epidemiology’s toughest conundrum. The rarity of the cancer provokes a desperate and often corrosive search for a cause (“why, of all people, did I get an astrocytoma?” Susan Reynard must have asked herself). And when patients with brain tumors happen to share a common exposure — in this case, cellphones — the line between cause and coincidence begins to blur. The association does not stand out nor does it disappear into statistical white noise. Instead, it remains suspended, like some sort of peculiar optical illusion that is blurry to some and all too clear to others. (A similarly corrosive intersection of a rare illness, a common exposure and the desperate search for a cause occurred recently in the saga of autism and vaccination. Vaccines are nearly universal, and autism is relatively rare — and many parents, searching to explain why their children became autistic, lunged toward a common culprit: childhood vaccination. An avalanche of panic ensued. It took years of carefully performed clinical trials to finally disprove the link.)
The Florida Circuit Court that heard Reynard v. NEC was quick to discern these complexities. It empathized with David Reynard’s search for a tangible cause for his wife’s cancer. But it acknowledged that too little was known about such cases; “the uncertainty of the evidence . . . the speculative scientific hypotheses and epidemiological studies” made it impossible to untangle cause from coincidence. David Reynard’s claim was rejected in the spring of 1995, three years after it was originally filed. What was needed, the court said, was much deeper and more comprehensive knowledge about cellphones, brain cancer and of the possible intersection of the two.
Allow, then, a thought experiment: what if Susan Reynard was given a diagnosis of astrocytoma in 2011 — but this time, we armed her with the most omniscient of lawyers, the most cutting-edge epidemiological information, the most powerful scientific evidence? Nineteen years and several billion cellphone users later, if Reynard were to reappear in court, what would we now know about a possible link between cellphones and her cancer?
To answer these questions, we need to begin with a more fundamental question: How do we know that anything causes cancer?
The crudest method to capture a carcinogen’s imprint in a real human population is a large-scale population survey. If a cancer-causing agent increases the incidence of a particular cancer in a population, say tobacco smoking and lung cancer, then the overall incidence of that cancer will rise. That statement sounds simple enough — to find a carcinogen’s shadow, follow the trend in cancer incidence — but there are some fundamental factors that make the task complicated.