Living Downstream

Pekin is the judicial seat of Tazewell County, Illinois. It is situated across the Illinois River and a few miles downstream from Peoria. Just outside of Pekin’s city limits, about two miles west of the house I grew up in, is the unincorporated subdivision of Normandale. The community was created in 1926 to provide housing for factory workers, and its streets are named for the original prewar products that the residents who slept here at night toiled by day to create: Karo Street (after the syrup), Quaker Street (after the paper mill’s round oatmeal boxes), Fleischmann Street (after the yeast).

Normandale is home to 480 people, a popular supper club, a beautiful brick church, and a root-beer stand where I hung out with my best friend in the summers after we learned to drive. Eating onion rings in her father’s car, Gail Williamson and I debated the merits of German versus Latin, big universities versus small colleges, sex versus celibacy. In this parking lot, we decided to settle for nothing short of everything. Gail would go to medical school and play the violin. I would go to graduate school and write poetry. Our present and future boyfriends would just have to understand. Of course. And they also would have to play the guitar.

Normandale is situated on a wedge of land near Dead Lake, a dumping pond for industrial wastes near the river’s east bank. It is flanked on two sides by industry: a foundry, a grain-processing plant, a couple of chemical companies, a coal-burning power plant, and an ethanol distillery. Its third side is bounded by a landfill that operated without state permits until the Illinois Pollution Control Board shut it down in 1988. Twenty rusted barrels leaking an unknown tarry substance were discovered along the blacktop just south of town. This is also Normandale.

The distribution of cancer across space, like its trajectory through time, reveals key clues about its possible causes. For example, if ethnicity played a major role in determining cancer risk, then immigrants should retain the cancer incidence of their homelands. Conversely, if the cancer rates of immigrants come to approximate those of their host country (and this is, in fact, the case), then we have good reason to suspect that environmental agents are at work. If cancer rates are elevated in certain geographic areas—within cities, for example, or in areas of intensive agriculture—we have further leads to pursue. If high rates of cancer follow the course of a river or the path of the prevailing wind or are clustered around a drinking-water well or a certain industrial site, then we have very strong clues indeed.

Paradoxically, the closer we stare at the map of cancer, the more unclear the picture becomes. On the largest scale, when cancer registry data from many nations are pooled and we are looking across whole continents, distinct areas of high and low cancer rates are clearly visible. As we narrow our view to one regional area—a single county or town or, like Normandale, a particular subdivision within a town—our power to discriminate differences decreases. Recall that cancer rates are based on the number of people annually diagnosed for each 100,000 people. Determining whether a cancer cluster exists in a small community of only a few thousand or a few hundred inhabitants is statistically difficult work, and it is at this level where the fiercest arguments fly.

At a global level, fewer arguments arise. The time trends and spatial features of cancer’s occurrence around the globe clearly belie the notion that cancer is a random misfortune. Cancer associates with westernization. Whereas forty years ago, cancer was mostly a disease of wealthy nations, half of all cancers now occur in developing nations, particularly those rapidly industrializing. Some of the ballooning global burden of cancer is attributable to increasing longevity and an aging population, but age-adjusted incidence rates are also rising in many regions. In India, cancer incidence rose by 7 percent between 1983 and 1997. It rose by 12 percent in Latin America. Reporting on these trends is a little-known office of the World Health Organization. Located in Lyon, France, the International Agency for Research on Cancer is charged with the daunting job of monitoring cancer incidence around the world. It does so by collecting registry data from as many countries as possible. Its World Cancer Report 2008 points out that cancer is ascendant in the parts of the world where smoking rates are increasing, diets are westernizing, and rates of obesity are rising. The report expresses frustration about multiple potential exposures to chemical carcinogens, which are also associated with industrialization but about which little is known.

In another global report, the Blacksmith Institute, a New York-based environmental foundation, provides an account of health problems in the world’s worst-polluted places. Cancer is among them. At the top of the unhappy list of the ten most polluted cities in the world stands Sumqayit in Azerbaijan, which had served as a center of petrochemical manufacturing in the former Soviet Union. The cancer rate in Sumqayit is as much as 51 percent higher than the national average. Two of China’s cities—Tianying and Lifen—also appear on this list.

Between 2000 and 2005, coal burning increased in China by 75 percent, making this nation the world’s largest coal producer and consumer. The economic burden of premature death and illness associated with the resulting air pollution—likely the worst among the world’s nations—has been estimated at 3.8 percent of its GDP. What part of this is the cost of cancer is not fully known nor yet fully manifest. In China as a whole, cancer incidence increased by 33 percent between 1973 and 1997. What we know about the geography of cancer’s surge in China comes mainly from the reports of a few heroic journalists. Steven Ribert provides first-hand accounts of apparent cancer clusters in and around the oilfields and petrochemical refineries in northern China. Along the riverbanks of the heavily industrialized Huiai river basin, a team of Chinese journalists has identified swathes of “cancer villages” where the incidence of cancer exceeds twice the national average.

Migrant studies also provide clues to the origins of cancer. When immigrants arrive in their adopted country, they leave behind the cancer rates of their homelands and quickly equilibrate with the rates of their new surroundings. “The most important single conclusion to derive from migrant studies,” states the International Agency for Research on Cancer, “is that, for a group as a whole, it is the new ‘environment’ that determines cancer risk and not the genetic component associated with the ethnic stock of the migrants.” The quotation marks around that stretchy word ‘environment’ call attention to its many elements: dietary habits, cultural attitudes about breastfeeding, social stress, and opportunities for physical activity are all part of our environment. So, too, are chemical pollutants in air, food, and water.

Cancer patterns among migrants to Australia, Canada, Israel, and the United States all illustrate the primacy of environment—as defined expansively—in determining cancer risk. Consider Jewish women who migrate from North Africa, where breast cancer is rare, to Israel, a nation with high incidence. Initially, their breast cancer risk is one-half that of their Israeli counterparts. But risk rises rapidly with duration of stay: within thirty years, African-born and Israeli-born Jews show identical breast cancer rates. Jewish women from the Middle East and Asia also increase their risk of breast cancer upon arrival in Israel, although the pace at which they do so is considerably slower.

Likewise, in the United States, the breast cancer rates of European, Chinese, and Japanese women immigrants all eventually rise to conform to the U.S. rate, but they do so at different speeds. Polish women assume U.S. rates of breast cancer quickly. Japanese women migrating to the U.S. mainland require two generations to achieve our breast cancer rate. First-generation Japanese immigrants show a rate intermediate between that for Japan and the United States; their daughters, however, reflect the U.S. rates completely. Immigrant Hispanic women have lower rates of breast cancer than their U.S.-born counterparts. However, the longer they stay in the United States, the greater their risk for breast cancer.

By 1991, half the homes on Karo Street had a cancer patient residing there. It also seemed to some residents that Normandale’s children were unusually susceptible to eye and ear infections. In one neighborhood, fourteen residents were diagnosed with cancer over a ten-year period. These numbers were calculated by the people themselves and presented to the health department and the local newspaper. A citizens’ group was organized and a letter dispatched to the Tazewell County Health Department requesting an investigation of cancer incidence in their community.

Because we have no nationwide cancer registry, we also have no definitive geography of cancer incidence in the United States. The National Cancer Institute does provide an interactive atlas of cancer mortality—in essence, a collection of customizable cancer death maps. Death from cancer is not randomly distributed in the United States. Shades of red consistently light up the northeast coast, the Great Lakes area, and the mouth of the Mississippi River. For all cancers combined, these are the areas of highest mortality; they are also the areas of the most intense industrial activity. The trend maps show that rates of increase, on the other hand, are actually higher in the parts of the country with lower mortality, indicating that cancer deaths are tending to become more geographically uniform as time passes, possibly due to the growing urbanization of formerly rural areas, the increasing mobility of the population, and the rising use of pesticides. For two cancers, these maps reveal latitudinal patterns: Southern regions predominate in the maps of melanoma, a pattern consistent with exposure to sunlight. Deaths from breast cancer also follow a north-south gradient but with rates higher in the North, especially in the highly industrialized Northeast.

When studying cancer atlases, a reader must keep in mind that these maps display cancer deaths, not cancer diagnoses. Counties with higher levels of contamination may also have worse health care: cancer patients from more polluted, more pesticide-saturated counties may be dying at faster rates simply because they are receiving poorer treatment. On the other hand, the rates of death from other causes—such as cardiovascular or infectious diseases—are not as closely linked to environmental contamination as cancer is. Disparities in health care, then, cannot account for all of the differences in geographic distribution of cancer deaths.

Cancer atlases offer the opportunity to overlay maps of cancer with maps of industrial and agricultural sites to see what patterns exist. In one study, researchers found significant associations between agricultural chemical use and cancer mortality in 1,497 U.S. rural counties. In another, investigators found a close overlap between cancer mortality and environmental contamination: concentrations of industrial toxins were higher in the top-ranked cancer counties than in the rest of the country.

In England, where cancer mortality data have been collected and analyzed for over a century, geographic analysis can be highly sophisticated. In 1997, a team of researchers mapped the home residences of all 22,458 children who had died of leukemia and other cancers in England, Wales, and Scotland between 1953 and 1980. They then created a second map that charted the locations of every potential hazardous site—ranging from power plants to neighborhood auto body shops. They then superimposed the two maps. Their findings reveal that children face an increased risk of cancer if they live within a few miles of certain kinds of industries—especially those involving large-scale use of petroleum or chemical solvents at high temperatures. These include oil refineries, airfields, paint makers, and foundries. The danger was greatest within a few hundred yards and tapered off with distance. Among children who had moved during their short lives, the relationship was stronger for their birth address than it was for their address at the time of their death. This result strongly suggests that very early—probably prenatal—exposures to environmental carcinogens create the threat of cancer in children.

Another way of mapping cancer is to examine how it distributes itself among people of various occupations. Just as cancer is not scattered uniformly across the physical landscape, neither does it afflict with an even hand the landscape of work. Understanding occupational cancers is important not only because people spend so many hours of their lives in the workplace but also because it yields critical clues about cancers beyond the factory wall and the office door. Released into air or water, hauled away as toxic waste, or mixed into consumer products, most cancer-causing agents in the workplace ultimately become part of the general environment in which we all live. Workplace carcinogens are largely identical to those agents that cause cancer in the general population. Indeed, the near half of the substances now classified as known human carcinogens by the International Agency for Research on Cancer were first identified in studies of workers. Let’s look at farmers first.

Farmers from industrialized countries around the world exhibit consistently higher rates of many of the same cancers that are also on the rise among the general population. In other words, farmers die more often from the same types of tumors that are also afflicting, with increasing frequency, the rest of us. These include multiple myeloma, melanoma, and prostate cancer. Farmers also suffer from rates of non-Hodgkin lymphoma and brain cancers higher than those of the general population—although these excesses are more modest. In spite of lower overall mortality and lower rates of heart disease (and lower overall cancer occurrences), farmers also die significantly more often than the general public from Hodgkin disease, leukemia, and cancers of the lip and stomach. Likewise, migrant farmworkers suffer excess rates of multiple myeloma, as well as of stomach, prostate, and testicular cancers. These results are consistent with the geographic patterns revealed within cancer atlases: death rates from multiple myeloma are highest in rural farming areas. Found in the central Corn and Wheat Belt region of the United States are high rates of leukemia and ly mphoma.

Additional clues about farm chemicals and cancer are emerging from the ongoing Agricultural Health Study. Begun in 1993 and sponsored by the National Cancer Institute, this investigation has been following a cohort of fifty-seven thousand farmers in Iowa and North Carolina. Spouses and children of these farmers are also enrolled in the study. Among the findings to date is the good news that the overall cancer rate among the study’s farmers appears significantly lower than that of non-farmers. Farmers use less tobacco and enjoy higher levels of physical activity and lower levels of diabetes than their off-farm counterparts. And yet, certain cancers nevertheless stalk these farmers and their families with greater frequency than the general public. Prostate cancer is one.

The Agricultural Health Study is revealing other patterns as well: One pesticide (permethrin) shows associations with bone marrow cancer (multiple myeloma). Two different weed killers show associations with pancreatic cancer. Parental pesticide application is linked to lymphoma in children. Children whose fathers do not use gloves when handling pesticides are also at higher risk for leukemia. Farm children whose fathers used atrazine recently had higher concentrations of this herbicide in their urine than farm children living where atrazine had not been recently applied. Women who use pesticides have longer menstrual cycles and later age at menopause. Yet pesticide use does not associate with breast cancer risk. On the other hand, women whose homes were located closest to areas of pesticide application do suffer modestly elevated rates of breast cancer.

Occupational studies of other professions reveal still more associations. Elevated cancer rates are found among painters, welders, asbestos workers, plastics manufacturers, dye and fabric makers, miners, printers, and radiation workers. Workers exposed to formaldehyde are more likely to contract leukemia. Firefighters are twice as likely to develop testicular cancer and suffer elevated rates of non-Hodgkin lymphoma, prostate cancer, and cancer of the bone marrow. Barbers and hairdressers have elevated risks for bladder cancer. Finnish women exposed to solvents and gasoline on the job also show increased risks of bladder cancer, as well as liver cancer. Finnish women workers exposed to diesel exhaust have higher rates of ovarian cancer—and risk of disease rose as exposure increased. Taiwanese women electronics workers exposed to chlorinated solvents in one particularly contaminated factory had increased incidence of breast cancer. There are ongoing concerns about the breast cancer risks of those who work in beauty parlors or nail salons.

People who work in a number of white-collar jobs are also at higher risk: for example, chemists, chemical engineers, dentists and dental assistants, and—perhaps most ironically—chemotherapy nurses. (Many of the chemicals used to treat cancer are themselves carcinogenic, as the high rate of adult cancers among childhood leukemia survivors attests.)

As we saw with farmworkers, the children of adults who work in specific occupations also have higher rates of cancer. Childhood brain cancers and leukemias are consistently associated with parental exposure to paint, petroleum products, solvents, and pesticides. Some exposures may occur before birth. Children can also be exposed when these materials are carried into the home on their parents’ clothes and shoes, through breast milk (which can be contaminated directly or through maternal contact with the father’s clothing), or even through exhaled air: because solvents are, in part, cleared by the lungs, parents can expose their children to carcinogens simply by breathing on them. In this way, a father’s home-coming kiss and work-clothed embrace can contaminate his child.

In response to the questions raised by the people of Normandale, two health studies were quickly conducted—one by the state health department and one by the county. Neither involved mapping disease patterns, identifying pollution sources, estimating actual exposures, locating those who had moved away, or, for those who had died, interviewing their next-of-kin. No blood, urine, or fat samples were collected to test for the presence of contaminants. In fact, the study design did not require public health officials even to set foot in Normandale.

In the first study, the Illinois Department of Public Health pulled up from its computerized cancer registry banks all the cancers diagnosed in Pekin’s ZIP code area—as reported to the Illinois Cancer Registry between 1986 and 1989. From these data, researchers calculated an actual cancer rate for the whole town. Based on the statewide rates, researchers then generated an expected number of cancer cases for a hypothetical town the size of Pekin. Cancers were categorized by location in the body (colon, ovaries, breast, and so forth), the actual numbers were compared to the expected numbers, and . . . no statistically significant differences were found.

If cancer-causing chemicals in the environment play a significant role in actually causing cancer, then we should expect to find high rates of the disease in areas where carcinogens are highly concentrated. The industrial workplace, where such chemicals are manufactured or used, is one such area. Hazardous waste sites, where such chemicals are dumped, are another.

Quite a few of us are included in the population of the potentially exposed. By 1990, the EPA had tallied up 32,645 sites of past chemical waste dumping in need of cleanup. Some of these are actual hazardous waste landfills, but many are former manufacturing sites where drums full of chemicals have simply been abandoned. The names of the most notorious appear on the EPA’s National Priorities List. These are the so-called Superfund sites, named for the super fund of money put together by Congress in 1980 to clean them up. In 2009, the Superfund list contained 1,331 sites. One-quarter of the U.S. population lives within four miles of one of them. Among those living within one mile are an estimated 1.1 million children under the age of six. Currently, Illinois is home to fifty Superfund sites.

Most of these sites did not exist before the end of World War II, when most plastics, solvents, detergents, pesticides—and all the unwanted by-products of petrochemical manufacturing—made their debut on the planet. Poor and dispossessed children have lived cheek-by-jowl with carcinogenic waste ever since soot-encrusted chimney sweeps in eighteenth-century England were discovered to be at high risk for scrotal cancer. But those of us born after World War II are the first generation to grow up in such large numbers near such large amounts and diverse assortments of manufactured chemical refuse. Between the late 1950s and the late 1980s, more than 750 million tons of toxic chemical wastes were discarded.

Several large studies have detected elevated cancer rates around hazardous waste sites. One of them was conducted in New Jersey, a petite state with an astonishing 133 Superfund sites. Researchers asked whether cancer mortality was associated with environmental factors of various kinds, including the location of toxic waste dumps. Their results showed that communities near toxic waste sites had significantly elevated mortality from stomach and colon cancers. Additionally, in twenty-one different New Jersey counties, breast cancer mortality among white women rose as the distance from residence to dump site shrank. However, many of the clusters of excess cancer occurred in heavily industrialized counties so that air pollution from these sources confounded the results. Thus, a woman with breast cancer in northeastern New Jersey cannot know with certainty whether she is dying because of the air wafting down from the factory stacks or because of the water contaminated by the dump site.

In another large study, researchers scoured the United States for counties that met two criteria: first, their hazardous waste sites had contaminated the groundwater, and second, this groundwater served as the sole source of the drinking water for the residents. Meeting these qualifications were 593 waste sites in 339 counties in forty-nine states. Next, researchers obtained for each of these 339 counties ten years’ worth of cancer mortality data and compared them to cancer mortality data from counties without hazardous waste sites.

Here are the results: men living in hazardous waste counties suffered significantly higher mortality from cancers of the lung, bladder, esophagus, colon, and stomach than did their contemporaries residing in counties without such sites. Women living in hazardous waste counties suffered significantly higher mortality from lung, breast, bladder, colon, and stomach cancers. Indeed, counties with hazardous waste sites were 6.5 times more likely to have elevated breast cancer rates than counties without such sites.

Other studies corroborated these results. Looking specifically at breast cancer, researchers found that mortality rates at the county level were significantly correlated with Superfund sites. Counties with the highest breast cancer mortality had four times as many facilities that treated and stored toxic waste than the national average.

Studies such as these two are considered preliminary rather than definitive because possible confounding factors could not be controlled. These include the possibility that residents living in counties with hazardous waste facilities are getting more cancers not because of the dumps but because they work for the companies that create the waste or because they smoke more and drink harder.

Among other things, the term ecological fallacy refers to the temptation of assuming that all associations are causative when one examines statistical patterns. My statistics professor was fond of telling the story of the boy and the department store escalator: the boy wondered what caused the escalator to move. After hours of observation, he concluded the escalator ran on the energy generated by the revolving door, because when the door ceased turning at the close of the day, the escalator stopped.

Ecological fallacy became a real issue for me when I started work as a field biologist. In Minnesota, I wanted to know why pine trees were failing to reproduce. Absence of new seedlings was correlated with high population levels of deer—but also with low frequency of forest fires and high populations of hazel shrubs. Which, if any, was the root cause of the problem and which were the confounders? Or, if fire, hazel, and deer all conspired to contribute to the demise of the pines, how exactly did they do so? Once I had established the pattern, I needed to design experiments that would uncover causal mechanisms. I found this work very exciting.

But as a woman with cancer who grew up in a county with numerous hazardous waste sites, several carcinogen-emitting industries, and public water wells that, from time to time, show detectable levels of toxic chemicals, I am less concerned about whether the cancer in my community is more directly connected to the dump sites, the air emissions, the occupational exposures, or the drinking water. I am more concerned that the uncertainty over details is being used to call into doubt the fact that profound connections do exist between human health and the environment. I am more concerned that uncertainty is too often parlayed into an excuse to do nothing until more research can be conducted.

I call an old high school teacher who has served a long term on the city council. The questions raised in Normandale have him thinking about other issues, such as emissions from the hospital incinerator and diesel exhaust from the produce trucks that rumble through town after the harvest. I ask him about the results of the Normandale investigation.

Epidemiologists investigate patterns of disease in human populations. They look at the world through a wide-angle lens. While the focus of medicine is the treatment and prevention of diseases in individuals, epidemiology attempts to explain and prevent the occurrence of disease in large groups.

One type of investigation is what epidemiologists call ecological studies. In these, investigators compare the frequency of a given disease (e.g., cancer) in populations that differ in some factor of interest (e.g., the presence or absence of a leaking hazardous waste site). Statistics are then used to determine whether the frequency of disease is significantly different in the two types of communities. Researchers can often complete ecological studies without ever talking directly to any of the human subjects or assessing their exposure levels to the contaminants in question. The studies in Pekin and Normandale were ecological studies. As strange as it sounds to ecologists, the word ecological is used by epidemiologists simply to mean a descriptive, rather than an analytical, approach. Ecological studies, like circumstantial evidence, provide the weakest demonstration of proof.

Included in epidemiology’s analytical category are two basic study designs. One is the case-control study. Here, a group of diseased people are identified (the cases) and compared to a group of people drawn from the larger population (the controls). The point of comparison is their exposure to possible disease-causing agents. Mary Wolff ’s study of DDT and breast cancer, discussed in Chapter One, is an example. Her cases were women with breast cancer; her controls were women without breast cancer (matched for age, menopausal status, and other variables of personal history); exposures were assessed by measuring blood levels of DDT and PCBs. Her results showed that women with breast cancer had significantly higher DDT levels than women without breast cancer.

Closely related to the case-control study is the cohort study, in which people are classified as exposed or unexposed and are followed through time until disease or death occurs. (The farmers and their families monitored in the ongoing Agricultural Health Study form one such cohort.) In this way, we compare the rate of disease in people known to be exposed to a possible carcinogen to disease rates in unexposed persons. The ratio of the two is known as relative risk.

One needs to understand a bit about the inner workings of cancer epidemiology in order to understand why the topic of individual cancer clusters is such a vexing one. Determining from ecological studies that communities near hazardous waste sites tend to suffer from an excessive risk of cancer is one kind of investigation. Determining that any one particular community has an elevated cancer risk due to any one particular waste site is a very different kind of project. The second kind is the one most people are interested in. We live in particular communities, not general ones, and our concerns are about the health of the particular people in our families and neighborhoods. Indeed, almost all cancer cluster studies are initiated by alert citizens contacting their health departments to request such investigations. Their phone calls and letters often tell of “cancer streets,” along which the prevalence of cancer seems extraordinarily high, or of growing numbers of neighborhood children afflicted with disease. This is exactly what happened in Normandale.

In spite of public concern, many public health officials become downright apoplectic when the subject of community-level cancer clusters is raised. Some consider the investigation of alleged clusters a disparaged practice and lament the inability of common people to grasp the statistical concept of randomness. Within the medical literature, publications advise health authorities on how best to deal with citizen requests for cluster studies. They are often overtly dismissive. Change a noun or two and some of them could double as guidelines for how to deal with people who wish to report a U.F.O. sighting. Typically, the message relayed back to those vigilant citizens seeking explanations is that their questions are misguided. Too rarely are they told that the tools of epidemiology are just too blunt to provide answers.

One problem with cancer cluster studies is that investigations of individual communities have limited power to identify existing problems. In this context, the word power refers to the ability to detect a significantly increased cancer rate if indeed the increase really exists. The word significantly also has a particular meaning. Significance is a statistical standard that limits a finding to only those increases in cancer rates we are reasonably sure did not occur by chance. Reasonably sure is traditionally defined as 95 percent sure, so 95 percent is accepted as the conventional cutoff for significance. If I roll a pair of dice six times and they always come up sixes, I can be more than 95 percent certain that this event is not due to chance. The finding is statistically significant. I conclude the dice are loaded. However, if I roll only one die one time and I get a six, the finding is not considered significant. By chance alone, the odds of this outcome are 16.7 percent. The dice may indeed be fraudulent, but my test does not have the power to say so.

Looking for a cancer cluster in a single, small community is like rolling the dice only once. Before the possibility that the cluster has occurred by chance can be ruled out, cancer rates in small communities must reach extraordinarily high levels—sometimes as high as eight to twenty times higher than levels for the surrounding areas. Because of the small sample size, lesser increases will not attain sufficient power for the study to be conclusive.

The second problem with cancer cluster studies is that there often remain no unexposed populations to use as a comparison group. In cluster studies, epidemiologists look for an increase over and above some background level, but if the people in the background are also becoming increasingly contaminated, the researchers are paddling a boat in a moving stream. Differences are harder to see.

Suppose, for example, we want to know whether people living near a particular hazardous waste dump are getting cancer because of it. Suppose the chemicals wafting into the air and trickling into the groundwater include several pesticides, some vinyl chloride, and an industrial solvent called trichloroethylene (TCE), classified by the EPA as a probable human carcinogen. With such contents, this dump would be utterly typical: TCE is the most frequently reported substance at Superfund sites, vinyl chloride is not far behind, and half of all hazardous waste sites contain pesticides. We have already seen that almost all of us experience chronic incremental exposure to vinyl chloride and pesticides from our air, food, and water.

Most of us are also exposed regularly to molecules of TCE. Used by industry to degrease metal parts, TCE is now estimated to be in 34 percent of the nation’s drinking water. Most processed foods contain traces as well. TCE is also found in paint removers, spot removers, cosmetics, and rug cleaners. An estimated 3.5 million workers are exposed to TCE on the job. Not so long ago, TCE was also used as an obstetrical anesthetic, a fumigant for grain, an ingredient in typewriter correction fluid, and a coffee decaffeinater. These uses have been phased out, but there is still sufficient release of TCE into the general environment to ensure that traces of this vaporized metal degreaser persist in the ambient air that we all breathe—including detectable amounts in the air above the Arctic Circle. Therefore, if we design a study that compares cancer rates between people living near this hypothetical dump site and a control group of people drawn from the general population, our results might reveal little about why either group is getting cancer. The cluster group and comparison group share exposures, and there is no unexposed control. As one nurse has observed, “To the public, it is no consolation that we are all exposed to environmental pollutants equally, and therefore not at increased risk of cancer because of it when compared with anybody else.”

At least two additional problems with cluster studies exist, and they both have to do with the nature of cancer. First, cancer usually requires a long period of time to develop after exposure occurs. This lag time makes exposure assessment very difficult. Researchers must rely on old, incomplete records—which may not exist at all—or people’s memories, which may be imperfect. Second, the origins of cancer are multiple, often resulting from exposures to combinations of substances—vinyl chloride plus heavy drinking, for example. These two features of the disease dilute and confuse epidemiologists’ finest efforts to understand its causes in any given community. In a case-control study, some of our cases may have contracted cancer from prenatal exposures, some from the dump site, some from their jobs, some from pesticide residues, and some from a combination of these. Furthermore, unexposed people migrate into the community, and exposed people move out. Epidemiologists cannot ask people living near carcinogens to stay put for ten years so that they may conduct a decent cohort study.

In 1953, New York City police reported to the health department that eleven homeless men in one neighborhood had all been discovered to be very ill and that all of them had turned sky blue. This particular skin color is the hallmark symptom of a disease called methemoglobinemia. Eleven cases in one neighborhood is thousands of times above background level. Knowing that this disease is associated with the ingestion of sodium nitrite, epidemiologists interviewed the men about their eating habits and discovered all had frequented a particular neighborhood diner and all had used the saltshaker there. The said shaker was impounded, laboratory tests were run, and the discovery was made that sodium nitrite had indeed been substituted for sodium chloride. The cook had made a mistake. Mystery solved.

Now imagine that cancer made people turn blue. And further imagine a skid row saltshaker containing a powerful chemical carcinogen that eleven customers unwittingly sprinkle over their food and that eventually causes them to develop cancer. In spite of their telltale color, the reason for their disease would probably never be uncovered. Because of the delay between exposure and onset of disease, at least ten years would pass before any of the eleven turned blue, and some of them would undoubtedly move away during this time. Because cancer is a disease with multiple causes, other drifters with blue complexions, who contracted cancer for unrelated reasons, would move into the area. The saltshaker itself would be long gone. Thus, although a cluster of people did indeed contract cancer from a single, identifiable source, a study of all blue-faced people in the neighborhood would not likely be able to establish the fact.

To overcome the limitations of basic epidemiology, state-of-the art cluster studies now incorporate geographic information systems (GIS) mapping and exposure assessment. GIS tools can create visually compelling pictures of potential cancer clusters. These spatial patterns can then be statistically tested for randomness. Exposure assessment can take the form of biomonitoring, which consists of collecting samples—urine or blood, for example—from bodies of the potentially exposed people themselves and testing them for the presence of particular contaminants. (Or it can involve sending the household dust bunnies off to the chemistry lab.) Analytic methods now allow for the direct testing of at least three hundred chemicals. Results can then be compared with baseline human exposure data, collected by the Centers for Disease Control and Prevention.

Even with all these methodological upgrades, the work of investigating cancer clusters suffers bedeviling problems. GIS mapping was designed to provide snapshots of locational information for businesses, not track chronic disease patterns. It lacks a temporal dimension. Cancer data are often aggregated by ZIP code—a geographic unit intended to speed mail delivery, not serve to analyze disease statistics by geography. It lacks standardization. And rarely can biomonitoring detect exposures from years earlier that may be influencing cancer risk now. But the biggest obstacle standing in the way of a good faith, due-diligence cluster investigation is not technological. As identified by the Pew Environmental Health Commission and reiterated in a recent issue of American Journal of Public Health, the biggest hindrance to cancer cluster inquiry is ignorance. Our nation lacks basic knowledge about the toxic properties of chemicals in commerce and, until 2002, lacked any system for environmental health tracking at all. At the state level, there is no consistent job title or agency with responsibility for following up on cancer cluster reports by citizens. There are no standard protocols for doing so, no rapid response teams at the ready, nor any systematic record-keeping on prior investigations. We track pizza deliveries and overnight packages more closely than we track toxic chemicals or cancer diagnoses.

Sometimes cluster studies go forward in spite of this paralyzing incapacity. Sometimes they yield instructive results. An investigation of twenty-five bladder cancer clusters in Florida revealed that advanced bladder cancer aggregated near arsenic-contaminated drinking water wells. Bladder cancer mortality was found elevated among men living in Clinton County, Pennsylvania, near a 46-acre chemical dump containing benzene and aromatic amines. An investigation in Sugar Creek, Missouri, found significantly elevated rates of Hodgkin lymphoma in a benzene-contaminated town where an oil refinery with a history of leaks now sits abandoned. In Endicott, New York, researchers verified elevated rates of lymphoma among former workers in a computer manufacturing plant. The plumes of TCE-contaminated water near the plant are now receiving close scrutiny as are the cancer rates of community members. In Ohio, state officials pinpointed a cluster of childhood cancers in Sandusky County. No one has an explanation yet, but the inquiry continues. Says the chief of Ohio’s cancer control program, “We owe it to these kids, parents, and future generations of kids to try to find out.”

Recall that statistical significance is traditionally defined as a less than 5 percent probability that any observed differences are attributable to chance. Curiously, state officials conducting this study chose 1 percent, rather than 5 percent, as their cutoff level for significance. This is an unusually strict measure, which, not surprisingly, causes differences to disappear. Two excesses in the Pekin area did in fact attain statistical significance at the usual 5 percent level: ovarian cancer and lymphoma.

No mention of the study’s statistical methodology was made in the newspaper account headlined AREA CANCER RATES NORMAL. Nor was any discussion devoted to what normal meant in this context. Tazewell County’s toxic emissions are high, but in comparison to those in the rest of the state, they are not off the chart. Thus, statistics aside, a discovery that Pekin’s cancer rates are rising in tandem with the rest of the state’s says nothing about whether our cancers, or anyone else’s in Illinois, are or are not attributable to environmental exposures.

One of these places is Long Island, New York. In 1994, the state health department released the results of a case-control study of Long Island women that showed that women with breast cancer were more likely to live near a chemical plant than women without the disease. Breast cancer risk rose with number of facilities: the more chemical plants in the community, the higher the incidence of breast cancer. And the closer a woman lived to one of these plants, the greater her chance of developing breast cancer.

This study was the first to show that breast cancer has links to air pollution. It was undertaken as a reaction to an earlier study that had dismissed environmental links to breast cancer in Long Island, concluding instead that breast cancer incidence in Long Island correlated with affluent lifestyles. The Centers for Disease Control reviewed these findings and, in 1992, recommended no further follow-up. Nevertheless, when women found various flaws in the study’s design, they took matters into their own hands. Some began creating their own maps and others began petitioning Congress to order a federal investigation. In the fall of 1993, a group of advocates hosted their own scientific conference, the first in the nation to bring together scientists and women with cancer to design a program of study. It was within this atmosphere that the 1994 study linking breast cancer and proximity to chemical plants emerged.

In the meantime and after considerable pressure, the U.S. Congress directed two federal agencies to begin a multi-million-dollar study: the Long Island Breast Cancer Study Project. This project is actually a collection of ten different study projects, the results of which are still coming in. Teams of researchers investigated such environmental features as aircraft emissions, pesticide practices, and plumes of contaminants in groundwater. Exposures were measured directly. Blood from thousands of Long Island women with and without breast cancer was analyzed for organochlorine pesticide residues and industrial chemicals.

At this writing, here is what we know from the Long Island Breast Cancer Study Project. First, there is no evidence for a link between adult exposure to individual organochlorine chemicals and breast cancer risk. Women with breast cancer had body burdens of DDT that were no higher, on average, than in women without the disease. With the benefit of hindsight, we now know that this is not a surprising result: as discussed in Chapter One, coherent evidence now indicates that the operative risk factor associated with DDT is early-life exposure when breast tissue is developing rapidly. The Long Island study measured levels in adult women after the time of diagnosis and missed this period of developmental sensitivity. It also looked for associations with individual chemicals rather than chemical mixtures. The importance of developmental exposures and real-life mixtures were not fully appreciated at the time the Long Island Study was designed. By the time the results were released in 2002, awareness of the study’s design limitations were widely acknowledged. And yet, these negative results were overgeneralized in the media as proof that pesticides of all kinds are not linked to breast cancer . . . and then still later as proof that the very act of searching for environmental links to cancer was a fool’s errand and a waste of tax dollars. Like the story of the fish that got away, the fish got larger with every retelling.

Meanwhile, positive findings have quietly emerged from the Long Island study. Most notably, breast cancer risk is significantly higher among women who report using pesticides in their lawns and gardens. And it is also higher among women whose blood cells show signs of DNA damage caused by the inhalation of polycyclic aromatic hydrocarbons. Also known as soot.

Across Long Island Sound lie the shores of Connecticut and Rhode Island. Just beyond, past Buzzards Bay, Cape Cod emerges from the Massachusetts coastline like a girl’s arm, curled and slender. As a midwestern adolescent, I became enchanted with Henry David Thoreau’s account of walking the length of this narrowest of peninsulas as it unfurled into the Atlantic. Cape Cod seemed to me a place of danger, beauty, and wild escape.

In the 1980s, year-round residents of the Upper Cape began agitating for an investigation into the relationship between environmental hazards and cancer rates. It seemed to them that cancer in their isolated communities was unusually common, and they were also aware of many environmental hazards, including pesticide use in cranberry bogs and golf courses as well as groundwater and air contamination from a nearby military reservation. Many could recall how the entire Cape had been drenched with DDT for several years during the 1950s in a failed campaign to eradicate the gypsy moth.

Residents were correct about the cancer rates. Records from the state cancer registry revealed excesses in prostate, colon, and lung cancers, along with elevations in pancreas, kidney, and bladder cancers. Of the ten towns with the highest breast cancer incidence in Massachusetts, seven are located on the Cape, and nearly all the Cape’s towns have higher-than-average rates of breast cancer.

Persistent citizen pressure led to two studies: an Upper Cape study completed in 1991 by two Boston University epidemiologists and another, the Cape Cod Breast Cancer and the Environment Study, which began in 1994 with a $1.2-million appropriation from the state legislature and continues to this day under the direction of scientists at the Silent Spring Institute in Newton. Let’s look first at the ongoing study.

Of particular interest to Silent Spring investigators is the underground aquifer that supplies nearly all the area’s drinking water. Covered with sandy soil, the aquifer is vulnerable to all manner of contamination—from pesticides to septic tank effluent to jet fuel and solvents spilled at the military base. Ironically, environmental regulations that protect the Cape’s coastal marine sanctuary mean that all waste water is discharged onto land, where it trickles through the sand and into the groundwater. Many of the chemicals contained in this waste are believed to play a role in breast cancer. Plumes of contaminated groundwater and areas where pesticides were sprayed were mapped and compared to maps of breast cancer incidence on the Cape. Residential history was factored in, so distinguishing between newcomers and old-timers. Using GIS techniques combined with models that allow for spatial data to be analyzed together with temporal data, researchers have created an animated picture of breast cancer’s shadow as it moved across the Cape over a forty-seven-year period.

Here is what Silent Spring researchers have learned so far: the space-time maps revealed an association between breast cancer risk and living in some areas near the military reservation from 1947 to 1956, but most of the elevated incidence was far from the military base. While the ground-water aquifer used for drinking water is contaminated with wastewater that contains hormonally active agents, limited evidence showed no connection between breast cancer and contaminated drinking water. Nevertheless, risk of breast cancer rises with duration of residency on the Cape, leaving open the question of causes in the ongoing study.

While researchers in the ongoing study are focusing specifically on breast cancer, those in the 1991 Cape study, by contrast, were looking at nine different cancers. Organized in case-control fashion, the study’s cases comprised Upper Cape residents diagnosed with cancer between 1983 and 1986, and the controls were a random sample drawn from the entire population of Upper Cape residents. Exposures were assessed through interviews. In this way, potential confounding factors such as smoking and other lifestyle habits could be uncovered and corrected for. The study was impressively thorough: when dealing with people who had already died from their disease, researchers matched them with nonliving controls—people who had died from other diseases and whose names were selected randomly from death certificate registries. To gather exposure information on cases and controls no longer alive, researchers interviewed their next-of-kin.

After three years of research, the study’s chief investigators reached the following conclusion:

In summary, this inquiry was begun because of concern about the generally increased cancer rates in the Upper Cape region along with the presence of known or suspected environmental hazards. After an extensive review of environmental factors it is clear that there was ample cause for concern.

While low statistical power prevented researchers from explaining all of the cancer increases, several interesting results emerged. The rates of both lung and breast cancer were elevated among residents living near the gun and mortar positions on the reservation. One possible explanation is airborne exposure to the military’s chemical propellant, dinitrotoluene, which is used for firing artillery. Classified as a probable human carcinogen, dinitrotoluene has been shown to cause breast cancer in laboratory animals. The study also yielded evidence for an increase in brain cancer among people living close to cranberry bogs, and it revealed elevations in leukemia and bladder cancer among those whose homes were fed by a particular type of water distribution pipe.

These water pipes had long been under suspicion. In the late 1960s, a new innovation in cement water pipes was introduced into New England: pipes with plastic liners that improved the water’s taste. A large number were laid in the Upper Cape, which was then undergoing rapid development. In manufacturing these pipes, workers applied vinyl paste to the inside surface, using a solvent called tetrachloroethylene. For reasons known only to organic chemists, tetrachloroethylene is more commonly referred to as perchloroethylene, PCE, or simply perc. Like its chemical cousin trichloroethylene, perchloroethylene is classified by the International Agency for Research on Cancer as a probable human carcinogen.

The assumption by the manufacturers of these waters pipes was that all the solvent would evaporate during the curing process. It did not. In fact, substantial quantities remained and slowly leached into the drinking water. Thus, the drinking water of the Upper Cape is contaminated not only by chemicals leaking from the land’s surface into the sole-source aquifer from which public water supplies are drawn, but also, in some areas, by the pipes carrying this water into individuals’ residences. The knowledge that Upper Cape water pipes were shedding perc into the drinking water was not a new discovery. This phenomenon had been known since the 1970s, but perc was not a substance regulated in drinking water during that decade. In 1980, plastic-lined water pipes were finally banned for use.

The link between perchloroethylene and bladder cancer was also not a new discovery. Perc is a familiar substance to almost all of us. Since the 1930s, it has been the chemical of choice for dry-cleaning clothes. Compared to the general population, dry cleaners have twice the rate of esophageal cancer and twice the rate of bladder cancer. Thus, a discovery of a bladder cancer cluster among the folk of the Upper Cape should come as no surprise. Further studies of the Upper Cape’s water pipes, published in 1993, showed that people’s actual exposure to perc varied widely, depending on the length, shape, size, and age of the water pipe, the pattern of water flow, and the person’s length of residence in that house. For those people with highest exposure, bladder cancer risk was four times higher and leukemia nearly twice as high when compared to people without such pipes.

The Journal of the American Water Works Association first reported on the problem of perchloroethylene leaching from drinking water pipes in 1983. The following words, written by scientists studying the Upper Cape, were published exactly ten years later:

In conclusion, we have found evidence for an association between PCE-contaminated public drinking water and leukemia and bladder cancer. In some EPA surveys, 14-26 percent of groundwater and 38 percent of surface water sources have some degree of PCE contamination. Thus, its carcinogenic potential is a matter of significant public health concern.

The second study in Normandale supported the first one. Because it could not provide cancer incidence data on any scale smaller than ZIP code level, the state health department turned the remainder of the investigation over to the county. County officials promised to conduct a door-to-door survey with the goal of determining whether “the cancer cases in Normandale are out of sync with the rest of the ZIP code.” They did not. Instead, questionnaires, which recipients were asked to fill out and mail back, were sent to 184 Normandale residences. Sixty-seven completed forms came back—a 37.5 percent response rate—and among these, eight cases of cancer were described.

The headline on March 6, 1992, announced, STUDY: NO CANCER CLUSTER, and the accompanying story read, in part:

The Tazewell County Health Department found no significant cancer problem [in Normandale], officials said Thursday in announcing results of the department’s cancer survey of the 40-acre subdivision. . . . The findings put an end to five months of investigation by state and county health officials into some residents’ fears that they were living amidst a cancer cluster.

One does not need to be an epidemiologist to see the glaring problems with this study. First, the numbers are too small to draw conclusions one way or the other. Second, there is no way of knowing whether respondents represent a random sample of the community. Perhaps responding households were, on average, healthier or better educated than nonresponding households. Perhaps households providing care to a family member with cancer were more likely to misplace the mailing or were more likely to be too grief-stricken or too overwhelmed to sit down and fill out answers to lengthy questions. Perhaps they were too busy fighting with insurance companies or planning funerals to write up a detailed family history and remember to mail it. Perhaps families with cancer were more likely to be out of town. Perhaps illiteracy prevented some from responding. Perhaps those angriest about the county’s broken promises chose to boycott the questionnaire. Or perhaps, conversely, families with cancer patients paid more attention to the questionnaire. In short, without direct human contact, no one can know why nonresponding households, the majority, remained silent.

Furthermore, anyone who had lived alone and died from cancer had no chance to be counted at all. The local newspaper obtained county death certificates showing that at least five cancer deaths in the community were never reported to the county’s survey. These included one case of liver cancer, two cases of breast cancer, one case of leukemia, and one of ovarian cancer.

These questions were not lost on the people of Normandale, many of whom expressed doubts about the study’s validity. Still the inhabitants of Normandale are not the residents of Cape Cod nor the women of Long Island. They are not positioned to reject the results of a county investigation and insist on a multi-million-dollar federal study. They have no friends in Congress. They are unlikely to invite world-renowned scientists to convene proceedings in the parking lot of the A & W root-beer stand.

The citizens of Cape Cod and Long Island have struggled mightily to bring scientific attention to the link between cancers and environmental contamination in their communities. Still, the resources they command are starkly different from those among Normandale’s residents. My meetings with the breast cancer activists of Long Island have taken place on college campuses and convention hotels. I have spoken with the cancer activists of Cape Cod in a beachfront conference center. When I met with a community leader in my own hometown, we held our discussion in the back room of an auto repair shop and towing company.