by Dan Fagin
It took a few months for Molholt to get his bearings amid the avalanche of paper, but he was encouraged by what he read. From a toxicological standpoint, he thought, the families had a pretty good case. There was solid, if indirect, evidence that many of the compounds dumped at Ciba and Reich Farm were capable of inducing cancer. Not one compound out of the hundreds Molholt identified in the wastes dumped at Ciba or Reich Farm had been conclusively shown to cause leukemia or nervous system cancer, the two categories of cancer that were most highly elevated in Toms River.8 But there was plenty of evidence that more than a dozen of those compounds might cause tumors in either category.9
The wild card was SAN trimer, which was now being scrutinized for the first time. Toxicity testing of this obscure waste product was supposed to have started in early 1997 but was delayed for more than a year because Union Carbide had none on hand—indeed, it had not had any since the 1970s, when it stopped combining styrene and acrylonitrile to make plastics. Instead, the company had to synthesize a small amount of SAN trimer from material purchased from one of its competitors, Bayer. The families wanted the EPA to test this newly synthesized trimer, but Union Carbide wanted its contractors to conduct the tests, and the EPA did not insist on maintaining control. Despite being responsible for the safety of more than sixty thousand chemicals in commerce, the EPA almost never conducted its own toxicity tests—and still does not. Instead, it typically restricts its oversight to evaluating tests conducted by manufacturers or manufacturers’ contractors.
A contractor for Union Carbide finished the first toxicity tests on SAN trimer in mid-1999 and concluded that trimer was toxic to rats only at high doses and that it was probably not mutagenic. Actually, the very first round of tests suggested that trimer might damage bacterial DNA and induce chromosomal aberrations in hamsters. But the company had a different lab redo the tests, using a purer sample and rats instead of hamsters, and got the results it was hoping for: no evidence of mutagenicity.10 These were not the tests that mattered, however. Toms River was a cancer cluster; the key question was whether styrene acrylonitrile trimer was a carcinogen. Linda Gillick was determined not to allow Union Carbide to provide that answer, too. She convinced New Jersey’s congressional delegation to petition the federal National Toxicology Program to take over cancer testing of the trimer, and in 1999 the agency agreed.
The twenty-year-old NTP, based in North Carolina, was the most prominent exception to the leave-it-to-industry attitude toward chemical testing that prevailed elsewhere in government. It was a research program that tended to operate like a reform school, taking on the most troubled cases—the most toxic compounds in commerce. The agency specialized in a costly test known as the two-year bioassay, a distant descendant of the cancer induction tests Katsusaburo Yamagiwa had devised for his rabbits in 1913. The NTP’s standard bioassay called for rats or mice to be fed, injected, or gassed with a suspect compound at one of four dose levels: high, medium, low, and none. After twenty-four months, the survivors were asphyxiated and autopsied to see if animals that received the highest doses had the most health problems, especially malignant tumors. The process was so complex and required so many steps before and after the two-year bioassay that it often took eight years or longer for the agency to render a final verdict. By 1999, the NTP had finished screening more than five hundred compounds, most of them chosen because they were already suspected to cause cancer.11 Slightly more than half of the time, the NTP concluded, a screened chemical was carcinogenic in rats or mice.12
The four hundred mice or rats in a typical two-year bioassay (fifty males and fifty females per dose level) were supposed to serve as surrogates for a much larger population of humans. But it was a highly imperfect surrogacy. The rodents had to be genetically identical to ensure that chemical exposure, not individual variation, was responsible for any detected tumors. Yet they were supposed to be stand-ins for a genetically diverse human population in which some individuals were much more susceptible than others. Just fifty animals were typically tested for each sex and dose, yet they were supposed to be an indicator of risk for a condition so rare that fewer than one in six thousand children were diagnosed with any type of cancer each year. The animals were kept in a carefully controlled environment in which they were exposed to just one hazardous chemical at a time, yet they were supposed to be a surrogate for a human population exposed to hundreds of potentially hazardous compounds every day, though almost always at very low levels.
The NTP resolved these contradictions by giving rodents much higher doses than humans would confront in a real-world environment, even in Toms River. The agency had no other choice; the process was already so complex that a typical two-year bioassay cost at least $2 million, with about half of the cost associated with handling the animals. An improved system, one based on real-world exposure levels, would require tens of thousands of rats or mice instead of a few hundred, at astronomical cost. The only practical choice was to give very large doses to small groups of rodents, even if that would make the studies less realistic and more vulnerable to criticism from industry.
To Linda Gillick and the rest of the TEACH parents, it was obvious what the NTP should do: It should feed those four hundred rodents a more potent version of the same cocktail of hundreds of contaminants that the families of Toms River had unwittingly drunk via the Parkway wells. True, the contaminant concentrations in that mixture were low—their collective concentration was probably a few hundred parts per billion, though no one knew for sure—but tens of thousands of people, including pregnant women and children, had drunk that low-level mixture for many years. If four hundred genetically identical rodents over just two years were supposed to be surrogates for a much larger and more diverse human population exposed over a much longer period of time, then those animals should get higher doses of the same mixture the families had drunk, the parents thought.
But the NTP was not in the business of testing chemical mixtures. Its two-year bioassay was designed to look for the all-important dose-response relationship, but how could anyone establish three consistent, controlled doses of a mélange of more than 250 compounds, most of them unidentified? How could anyone tell which compounds were benign and which were dangerous? When Union Carbide objected to the families’ proposal to test the entire Parkway mixture, NTP managers quickly sided with the company. The agency had always conducted its toxicology tests one chemical at a time and it was not going to change now, not even for a very high-profile case like Toms River. The New Jersey Department of Environmental Protection had already given up trying to find out exactly what was in the Parkway water mixture, and now the National Toxicology Program would not even try to find out whether the mixture was carcinogenic.
Instead, everything would come down to styrene acrylonitrile trimer, a chemical that under normal circumstances would be too obscure even to be considered for testing by the NTP. But Toms River, of course, was anything but a typical case. An entire state congressional delegation had taken the very unusual step of asking for the tests, and the NTP was not about to say no to the same lawmakers who set its budget every year. Besides, SAN trimer did have a disturbing lineage: Both of its key ingredients, styrene and acrylonitrile, had already been linked to cancer in factory studies or animal tests.13
The NTP did agree to one change that made the four hundred caged rats slightly closer surrogates for the families of Toms River. Normally, the agency’s two-year tests involved only adult rodents, which made sense when the program was looking at compounds that were mainly risks to factory workers. But in Toms River, the fear was that SAN trimer might be triggering cancer in children before they were even born. So for this study, the NTP decided to conduct a multigenerational test. Pregnant rats would be fed daily doses of trimer, and after birth the dosing would continue for three weeks while the mothers nursed. The pups would then be fed trimer-laced food every day for two years until they were euthanized and autopsied. The Toms River parents liked this idea: They had been dosed w
hile pregnant, so the rats should be, too.
Still, there were many reasons to doubt whether the rat study would end in a meaningful result. A key concern was that so little SAN trimer was available for testing: only twenty-one kilograms of the smelly, brown gel. This was a major handicap because the most alarming kinds of cancers in Toms River—leukemias and brain tumors—were extremely difficult to find in rodents. Fischer 344 rats, the kind the NTP chose for the trimer tests, were especially poor models for human leukemia because they were naturally prone to a different type of leukemia almost never found in people.14 The NTP picked Fischer rats anyway because they were the best hope of finding brain tumors.15 With more money and more trimer, the NTP could have tested two kinds of rats, or rats and mice, and thus had a chance of finding leukemias, too. It could even have doubled the number of rodents tested and thus increased the tests’ ability to detect even rare brain tumors. But funds were scarce, and Union Carbide was not providing any more trimer.
What the families had envisioned as a broad inquiry into the carcinogenicity of the Parkway well water had narrowed drastically to just one group of cancers—brain and central nervous system tumors—and just one compound: styrene acrylonitrile trimer. With each narrowing of the investigative lens, the chances of discovering a carcinogenic effect diminished. And now the families would have to wait eight years or longer for any results at all—almost certainly far too long to influence the outcome of the legal case. Floyd Genicola’s discovery of SAN trimer in Parkway water back in 1996 had been electrifying, but now the slow-grinding realities of contested science had set in. Linda Gillick and the TEACH families faced a very long wait for a very uncertain result. They would have to pin their hopes elsewhere.
The first, very preliminary results of the Toms River case-control studies were released in December of 1999, and they were tantalizing. Under pressure to show progress after three years, Jerry Fagliano and his colleagues had finished analyzing the questionnaire, looking for differences in the responses of the forty case families compared to the 159 healthy controls.16 Interview-only studies were a weak form of epidemiology because they depended on fallible memories. If everyone’s memory were equally unreliable, this might not be a major problem. But families struck by traumatic illness tended to have much more detailed memories about past exposures than control-group families who had been spared. This well-known phenomenon was known as recall bias. The air and water distribution models, which were still under construction in 1999, would minimize the distorting effects of recall bias because their exposure estimates were based not on memories but on reconstructions of historical emissions patterns. Since the models were not ready, Fagliano warned the families not to make too much of the interview results.
Still, the results were intriguing. The case and control families had been asked more than two hundred questions, on topics ranging from miscarriage history to hot dog consumption. They were asked about so many possible risk factors that it was likely that for a few of those questions the differences between the cases and controls were going to look significant even if they were really due to nothing but chance. As it turned out, though, there were only a few examples of these flukes.17 Instead, almost all of the potential nonenvironmental causes explored in the interviews—including prominent theories such as maternal smoking, alcohol consumption, and family history of cancer—were no more common in case families than in control families. Case and control families tended to have similar diets and were similarly likely to have used nail polish, pesticides, and paint thinner, among many other possible risks. By process of elimination, water and air pollution now loomed larger than ever as possible explanations for the cluster. This was especially true because when the interviewers asked the study families whether they drank tap or bottled water, children who later developed cancer were 8 percent more likely to have drunk tap water and 60 percent more likely to have been born to mothers who drank tap water during their pregnancies. In fact, for both mothers and kids, there was a dose-response relationship: Cancer risk was highest for those who drank more than five glasses of tap water per day, lowest for those who drank none, and in between for those who had just a few glasses.18 As usual, Fagliano kept his feelings to himself, but he could not help being excited about this preliminary finding, even if it was influenced by recall bias. “It suggested we were on the right track,” he later explained.
The lawyers for the families were excited too; the drinking water correlation was another bit of leverage they could use in possible settlement negotiations. By early 2000, the tone of their meetings in Princeton had shifted again. No longer just exchanging information, the opposing attorneys were now focused on convincing each other of the correctness of their arguments. The stridency of their initial meetings was gone for good—as Schlichtmann had hoped, the lawyers knew each other too well now for overt hostility—but both sides also knew that the time for casual research was ending, especially because the case-control study was at last approaching completion. Fagliano had told Gillick’s committee that the study results would be available by the end of 2001—almost six years after the rowdy public meeting that had launched the investigation.
From the start, the case-control study (now two studies: the interview study and the birth record study) had loomed over the lawyers’ meetings as an omnipotent, if seldom discussed, presence. If, in the end, the study results linked the cluster to air pollution from Ciba or to water pollution from Reich Farm and Ciba via United Water’s pipes, then the families would obviously have a very strong case for a huge payout, either from a settlement or—years later—a jury verdict. On the other hand, if the study did not find an association between pollution and cancer, the companies would surely offer nothing more than a token amount, and the families would have to decide whether to take it or risk a court battle premised on a greatly weakened case. A crucial question for both sides, then, was whether to take an all-or-nothing gamble by waiting for the case-control study results or instead try to negotiate a settlement before the study was completed.
“It got to the point where it was clear that the case-control study really was going to come to an end, and that we on the plaintiff’s side were not going to go away,” said lawyer Mark Cuker. “So the discussion became, how do we resolve this?” William Warren, a lawyer for Union Carbide, made it clear that his client would not settle the case merely out of sympathy for the families or to avoid bad publicity. He and his counterparts at Ciba and United Water hinted that a settlement was possible, but first the families’ lawyers would have to show that they could present a legitimate case for causation.
Essentially, what the companies wanted was a mini-trial that would be a surrogate for the real thing, but with no judge or jury—not even a mediator. Union Carbide, in particular, had resisted mediation because it implied that a settlement was in the offing. However, the company’s combative general counsel, Robert Butler, did have a soft spot for Eric Green, a law professor at Boston University and one of the best-known mediators in the United States. (He would soon achieve even more notoriety for brokering the massive 2001 antitrust settlement between Microsoft and federal prosecutors.) So when Schlichtmann heard that Green was going to be in Princeton for a different case on the same day the Toms River lawyers were meeting there, he convinced Butler and a few others to interrupt Green in the middle of a solitary lunch at the Forrestal Hotel (he had a napkin under his chin and was busy tucking into a steak) to see if he would be willing to consider getting involved in the Toms River case.
They were not looking for a mediator to engineer a settlement, the lawyers told Green. Instead, they wanted a “facilitator” to organize and oversee a series of meetings at which experts from each side would present information and opinions about the key issues, just as they would do if they were witnesses at a trial. Instead of trying to persuade a judge, jury, or mediator, however, the lawyers and their experts would be trying to convince their counterparts on the other side that they could present a winning c
ase if there ever were a trial. Later on, the lawyers told him, Green might play a more formal role in brokering a settlement—but that would depend on the persuasiveness of the expert presentations. Each side would first have to be convinced that a deal was preferable to the risk of waiting for the case-control study results or going to trial. Though irritated at having his lunch interrupted by the ever-pushy Schlichtmann, Green was intrigued; he had never been involved with anything quite like this. The lawyers agreed to hire him, splitting his fee, which would ultimately top $100,000.
Almost immediately, there were complications. In May of 2000, four other lawyers sued Ciba on behalf of six hundred residents, mostly of Oak Ridge, who lived downwind from the factory’s smokestacks or atop its groundwater plumes. About two hundred had cancer or other health issues they blamed on the factory.19 (A second class-action suit, seeking compensation from Ciba for diminished property values in Oak Ridge, was filed eight months later on behalf of about seven hundred homeowners.)20 Cuker, Berezofsky, and Schlichtmann had been dreading the prospect of other lawyers entering the fray because they feared that it would shut down the ongoing exchange of information and end any chance of a settlement. Union Carbide was already considering pulling out of the talks because it was being acquired by Dow Chemical, which had a history of taking a hard line in lawsuits. After a tense few weeks of uncertainty, the company lawyers said that they would keep participating while reserving the right to drop out at any time.