34

MEDICAL DISPATCH 6 THE CANCER--CLUSTER MYTH When a dozen people in a neighborhood develop tumors, it can’t be coincidence. Or can it? BY ATUL GAWANDE

S it something in the water? During thing seriously wrong with the Western the past two decades, reports of Area,” the artist, Tyler Mercier, told the I cancer clusters—communities in Times. “The neighborhood may be con- which there seems to be an unusual taminated.” In fact, the Los Alamos number of cancers—have soared. The National Laboratory, which was the place names and the suspects vary, but birthplace of the atomic bomb, had the basic story is nearly always the once dumped millions of gallons of ra- same. The Central Valley farming town dioactive and toxic waste in the sur- of McFarland, California, came to na- rounding desert, without providing any tional attention in the eighties after a solid documentation about precisely woman whose child was found to have what was dumped or where. In San Ra- cancer learned of four other children mon, California, a cluster of brain can- with cancer in just a few blocks around cers was discovered at a high-school her home. Soon doctors identified six class reunion. On Long Island, federal, more cases in the town, which had a state, and local officials are currently population of sixty-four hundred. The spending twenty-one million dollars to childhood-cancer rate proved to be four try to find out why towns like West times as high as expected. Suspicion fell Islip and Levittown have elevated rates on groundwater wells that had been of breast cancer. contaminated by pesticides, and law- I myself live in a cancer cluster. A res- suits were filed against six chemical ident in my town—Newton, Massachu- companies. setts—became suspicious of a decades- In 1990, in Los Alamos, New Mex- old dump next to an elementary school ico, a local artist learned of seven cases after her son developed cancer. She of brain cancer among residents of a went from door to door and turned small section of the town’s Western up forty-two cases of cancer within a Area. How could seven cases of brain few blocks of her home. The cluster is cancer in one neighborhood be merely being investigated by the state health

a coincidence? “I think there is some- department. PETER TILL

TNY—2/8/98—PAGE 34 LIVE OPI N40548 4C,133LS 35

No doubt, one reason for the verita- such investigations, epidemiologists ble cluster of cancer clusters in recent tend to be skeptical about their worth. years is the widespread attention that cases like those in McFarland and Los HEN public-health investigators Alamos received, and the ensuing in- W fail to turn up any explanation crease in public awareness and concern. for the appearance of a cancer clus- Another reason, though, is the way in ter, communities can find it frustrating, which states have responded to that even suspicious. After all, these investi- concern: they’ve made available to the gators are highly efficient in tracking public data on potential toxic sites, down the causes of other kinds of dis- along with information from “cancer ease clusters. “Outbreak” stories usually registries” about local cancer rates. The start the same way: someone has an in- result has been to make it easier for tuition that there are just too many peo- people to find worrisome patterns, and, ple coming down with some illness and more and more, they’ve done so. In asks the health department to investi- the late eighties, public-health depart- gate. With outbreaks, though, such in- ments were receiving between thirteen tuitions are vindicated in case after case. hundred and sixteen hundred reports Consider the cluster of American Le- of feared cancer clusters, or “cluster gionnaires who came down with an un- alarms,” each year. Last year, in Massa- usual lung disease in Philadelphia in chusetts alone, the state health depart- 1976; the startling number of limb ment responded to between three thou- deformities among children born to Jap- sand and four thousand cluster alarms. anese women in the sixties; and the ap- Under public pressure, state and fed- pearance of rare Pneumocystis carinii eral agencies throughout the country pneumonia in five young homosexual are engaging in “cancer mapping” men in Los Angeles in 1981. All these to find clusters that nobody has yet clusters prompted what are called “hot- reported. pursuit investigations” by public-health A community that is afflicted with authorities, and all resulted in the defin- an unusual number of cancers quite itive identification of a cause: namely, naturally looks for a cause in the envi- Legionella pneumonitis, or Legionnaires’ ronment—in the ground, the water, disease; mercury poisoning from contam- the air. And correlations are sometimes inated fish; and H.I.V.infection. In fact, found: the cluster may arise after, say, successful hot-pursuit investigations of contamination of the water supply by disease clusters take place almost every a possible carcinogen. The problem is day. A typical recent issue of the Cen- that when scientists have tried to con- ters for Disease Control’s Morbidity and firm such causes, they haven’t been able Mortality Weekly Report described a to. Raymond Richard Neutra, Califor- cluster of six patients who developed nia’s chief environmental health inves- muscle pain after eating fried fish. In- tigator and an expert on cancer clus- vestigation by health authorities iden- ters, points out that among hundreds tified the condition as Haff disease, of exhaustive, published investigations which is caused by a toxin sometimes of residential clusters in the United present in buffalo fish. Four of the cases States, not one has convincingly iden- were traced to a single Louisiana whole- tified an underlying environmental saler, whose suppliers fished the same cause. Abroad, in only a handful of tributaries of the Mississippi River. cases has a neighborhood cancer clus- What’s more, for centuries scientists ter been shown to arise from an en- have succeeded in tracking down the vironmental cause. And only one of causes of clusters of cancer that aren’t these cases ended with the discovery of residential. In 1775, the surgeon Perci- an unrecognized carcinogen. It was in vall Pott discovered a cluster of scrotal- a Turkish village called Karain, where cancer cases among London chimney twenty-five cases of mesothelioma, a sweeps. It was common practice then rare form of lung cancer, cropped up for young boys to do their job naked, among fewer than eight hundred vil- the better to slither down chimneys, lagers. (Scientists traced the cancer and so high concentrations of carcino- to a mineral called erionite, which genic coal dust would accumulate in the is abundant in the soil there.) Given ridges of their scrota. Pott’s chimney the exceedingly poor success rate of sweeps proved to be a classic example of

TNY—2/8/99—PAGE 35 36 THE NEW YORKER, FEBRUARY 8, 1999

be exposed to seventy per cent of the maximum tolerated dose in the course of a full year, or the equivalent. “This kind of exposure is credible as part of chemotherapy or in some work settings,” he wrote in a 1990 paper, “but it must be very rare for most neighborhood and school settings.” For that reason, investi- gations of occupational cancer clusters have been vastly more successful than in- vestigations of residential cancer clusters. Matters are further complicated by the fact that cancer isn’t one disease. What turns a breast cell into breast cancer isn’t what turns a white blood cell into leu- kemia: the precise combination of hits varies. Yet some clusters lump together people with tumors that have entirely different biologies and are unlikely to have the same cause. The cluster in McFarland, for example, involved eleven “Sometimes it’s important to stop whatever break children with nine kinds of cancer. Some you’re taking and just do the work.” of the brain-cancer cases in the Los Alamos cluster were really cancers of • • other organs which had metastasized to the brain. an “occupational” cluster. Scientists have chance combination of successive muta- also been successful in investigating so- tions in these genes—multiple “hits,” as F true neighborhood clusters—that is, called “medical” clusters. In the late cancer biologists put it—to make a cell I local clusters arising from a common nineteen-sixties, for example, the pathol- cancerous rather than simply killing it. A environmental cause—are so rare, why ogist Arthur Herbst was surprised to carcinogen provides one hit. Other hits do we see so many? In a sense, we’re come across eight women between the may come from a genetic defect, a fur- programmed to: nearly all of them are ages of fifteen and twenty-two who had ther environmental exposure, a sponta- the result of almost irresistible errors in clear-cell adenocarcinoma, a type of cer- neous mutation. Even when people have perception. In a pioneering article pub- vical cancer that had never been seen in been subjected to a heavy dose of a car- lished in 1971, the cognitive psycholo- women so young. In 1971, he published cinogen and many cells have been dam- gists Daniel Kahneman and Amos Tver- a study linking the cases to an anti- aged, they will not all get cancer. (For sky identified a systematic error in human miscarriage drug called diethylstilbe- example, DES causes clear-cell adeno- judgment which they called the Belief strol, or DES, which the mothers of carcinoma in only one out of a thou- in the Law of Small Numbers. People these women had taken during preg- sand women exposed to it in utero.) As assume that the pattern of a large popu- nancy. Subsequent studies confirmed a rule, it takes a long time before a cell lation will be replicated in all its subsets. the link with DES, which was taken by receives enough hits to produce the can- But clusters will occur simply through some five million pregnant women be- cer, and so, unlike infections or acute chance. After seeing a long sequence of tween 1938 and 1971. The investiga- toxic reactions, the effect of a carcino- red on the roulette wheel, people find tion of medical and occupational cancer gen in a community won’t be seen for it hard to resist the idea that black is clusters has led to the discovery of dozens years. Besides, in a mobile society like “due”—or else they start to wonder of carcinogens, including asbestos, vinyl ours, cancer victims who seem to be whether the wheel is rigged. We assume chloride, and certain artificial dyes. clustered may not all have lived in an that a sequence of R-R-R-R-R-R is So why don’t hot-pursuit investiga- area long enough for their cancers to somehow less random than, say, R-R- tions of neighborhood cancer clusters have a common cause. B-R-B-B. But the two sequences are yield such successes? For one thing, many To produce a cancer cluster, a carcino- equally likely. (Casinos make a lot of clusters fall apart simply because they gen has to hit a great many cells in a money from the Belief in the Law of violate basic rules of cancer behavior. great many people. A brief, low-level Small Numbers.) Truly random patterns Cancer develops when a cell starts mul- exposure to a carcinogen is unlikely to do often don’t appear random to us. The tiplying out of control, and the process the job. Raymond Richard Neutra has statistician William Feller studied one by which this happens isn’t straight- calculated that for a carcinogen to pro- classic example. During the Germans’ forward. A carcinogen doesn’t just flip duce a sevenfold increase in the occur- intensive bombing of South London in some cancer switch to “on.” Cells have a rence of a cancer (a rate of increase not the Second World War, a few areas variety of genes that keep them func- considered particularly high by epide- were hit several times and others were tioning normally, and it takes an almost miologists) a population would have to not hit at all. The places that were not

TNY—2/8/99—PAGE 36—LIVE OPI GRAPHIC—DO NOT STRIP—COMPARE AGAINST A4343—LINE SCREEN 120 MEDICAL DISPATCH 37 hit seemed to have been deliberately holes, we tend to notice cases first— and water. Thousands, sometimes mil- spared, and, Kahneman says, people be- four cancer patients on one street—and lions, of dollars are spent. And, with all came convinced that those places were then define the population base around those tests, correlations inevitably turn where the Germans had their spies. them. With rare conditions, such as up. Yet, years later, in case after case, When Feller analyzed the statistics of the Haff disease or mercury poisoning, even nothing definite is confirmed. bomb hits, however, he found that the a small clutch of cases really would rep- distribution matched a random pattern. resent a dramatic excess, no matter how “ HE reality is that they’re an abso- Daniel Kahneman himself was in- much Texas sharpshooting we did. But T lute, total, and complete waste of volved in a similar case. “During the most cancers are common enough that taxpayer dollars,” says Alan Bender, an Yom Kippur War, in 1973, I was ap- noticeable residential clusters are bound epidemiologist with the Minnesota De- proached by people in the Israeli Air to occur. Raymond Richard Neutra partment of Health, which investigated Force,” he told me. “They had two points out that, given a typical registry more than a thousand cancer clusters squads that had left base, and when the of eighty different cancers, you could in the state between 1984 and 1995. squads came back one had lost four expect twenty-seven hundred and fifty The problem of perception and poli- planes and the other had lost none. of California’s five thousand census tics, however, remains. If you’re a pub- They wanted to investigate for all kinds tracts to have statistically significant but lic health official, try explaining why of differences between the squadrons, perfectly random elevations of cancer. a dozen children with cancer in one like whether pilots in one squadron had So if you check to see whether your neighborhood doesn’t warrant investi- seen their wives more than in the other. neighborhood has an elevated rate of a gation. According to a national study, I told them to stop wasting their time.” specific cancer, chances are better than health departments have been able to A difference of four lost planes could eas- even that it does—and it almost cer- reassure people by education in more ily have occurred by chance. Yet Kahne- tainly won’t mean a thing. Even when than seventy per cent of cluster alarms. man knew that if Air Force officials in- you’ve established a correlation between Somewhere between one and three per vestigated they would inevitably find a specific cancer and a potential car- cent of alarms, however, result in ex- some measurable differences between cinogen, scientists have hardly any way pensive on-site investigations. And the the squadrons and feel compelled to act to distinguish the “true” cancer cluster cases that are investigated aren’t even on them. that’s worth investigating from the crowd the best-grounded ones: they are the Human beings evidently have a deep- of cluster impostors. cases pushed by the media, enraged cit- seated tendency to see meaning in the One helpful tip-off is an extraordi- izens, or politicians. “Look, you can’t ordinary variations that are bound to narily high cancer rate. In Karain, Tur- just kiss people off,” Bender says. In appear in small samples. For example, key, the incidence of mesothelioma was fact, Minnesota has built such an effec- most basketball players and fans believe more than seven thousand times as high as tive public-response apparatus that it that players have hot and cold streaks expected. In even the most serious clus- has not needed to conduct a formal in shooting. In a paper entitled “The ter alarms that public-health departments cluster investigation in three years. Hot Hand in Basketball,” Tversky and have received, however, the cancer rate has Public-health departments aren’t two colleagues painstakingly analyzed been nowhere near that high. (The law- lavishly funded, and scientists are re- the shooting of individual players in yer Jan Schlichtmann, of “Civil Action” luctant to see money spent on some- more than eighty games played by the fame, is now representing victims of a thing that has proved to be as un- Philadelphia 76ers, the New Jersey cancer cluster in Dover Town- productive as neighborhood Nets, and the New York Knicks during ship, New Jersey, where the cluster alarms or cancer map- the 1980-81 season. It turned out that childhood-cancer rate is thirty ping. Still, public confidence basketball players—even notorious “streak per cent higher than expected.) is poorly served by officials shooters”—have no more runs of hits or This isn’t to say that car- who respond to inquiries with misses than would be expected by chance. cinogens in the local environ- a scientific brushoff and a layer Because of the human tendency to per- ment can’t raise cancer rates; of bureaucracy. To be part of a ceive clusters in random sequences, how- it’s just that such increases dis- cancer cluster is a frightening ever, Tversky and his colleagues found appear in all the background variation thing, and it magnifies our ordinary re- that “no amount of exposure to such se- that occurs in small populations. In sponse when cancer strikes: we want to quences will convince the player, the larger populations, it’s a different story. hold something or someone responsi- coach, or the fan that the sequences are The 1986 Chernobyl disaster exposed ble, even allocate blame. Health offi- in fact random. The more basketball hundreds of thousands of people to cials who understand the fear and anger one watches and plays, the more oppor- radiation; scientists were able to estab- can have impressive success, as the ones tunities one has to observe what appears lish that it caused a more than one- in Minnesota have shown. But there to be streak shooting.” hundred-fold increase in thyroid cancer are times when you cannot maintain In epidemiology, the tendency to among children years later. By contrast, public trust without acting on public isolate clusters from their context is investigating an isolated neighborhood concerns. Science alone won’t put to known as the Texas-sharpshooter fal- cancer cluster is almost always a futile rest questions like the one a McFar- lacy. Like a Texas sharpshooter who exercise. Investigators knock on doors, land mother posed to the Los Angeles shoots at the side of a barn and then track down former residents, and check Times: “How many more of our children draws a bull’s-eye around the bullet medical records. They sample air, soil, must die before something is done?” ♦

TNY—2/8/99—PAGE 37—LIVE SPOT—DO NOT STRIP—10837