 |
||
 |
|
|
 |
||
 |
||
|
|
|
 |
|||||||
 |
|||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Expertise |
||
|
|
|
|
|
|
|
|
|
|
|
|
 |
|
|
||||||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Special Report
Judging the Judges
How well and consistently do judges carry out their 'gatekeeping' role? How well do they understand science? And is this court activity assessed in any way?
Types of Causation Data: What data qualifies to prove chemical causation of injury in humans?
By Thomas F. Schrager, Ph.D.
It is generally understood and accepted by courts that human epidemiological data is the best source of information of health effects of chemicals in humans. In many instances this is an understandable and legitimate conclusion: showing an increase in an adverse health effect in a population of exposed individuals is a direct and obvious way to establish general causation (that the chemical can, under certain conditions, cause an effect).
But in the 'black box' approach of toxicity, this is only one possible scenario. What if epidemiological evidence is not available? What if available epidemiological evidence is negative? Does negative epidemiological evidence automatically set a standard of non causation that other data must 'disprove'? Did the epidemiological study have sufficient statistical power to prove what it set out to prove? Was it properly constructed and interpreted so that the absence of an effect can be considered proof of absence, rather than an indication that the study was not properly designed?
As the National Academies of Science points out in a report of a Workshop on Science in the Courtroom (p. 18), a study of 300, 000 petroleum workers showed that exposure to benzene (in the petroleum) did not lead to an increased incidence of non Hodgkin's lymphoma. However, another study showed distinct chromosomal abnormalities in those exposed to benzene, specifically characteristic of non Hodgkin's lymphoma? So does benzene cause non Hodgkin's lymphoma or not?
The real question for the court is not that one, or figuring out what the answer might be, but whether that question and the surrounding data will or will not be presented to a jury to decide. Do the findings of a population study involving 300, 000 people take precedence and if so, why? What is it about the study that carries such weight, or about possible flaws in the study that carry so little weight, that a jury is precluded from even considering this and other complex questions. Preventing the jury from hearing this conflicting data means that the court has ruled that the epidemiological data is definitive, something probably few scientists would feel comfortable concluding. More so, such preclusion might result in summary judgement in favor of the other side.
What accounts for this? Is epidemiological evidence the best and should it supercede--even the absence of it--any other type of evidence? Part of the answer is obvious--what you see is what you get. It's an idea of causation that we all grow up with and that is reinforced through the nature of traumatic injury, something that can be easily seen and visualized. And if we can't see it or visualize it, it must not actually occur. However much we are aware of this way of looking at things, it is hard to put aside this orientation when dealing with the 'black box' of chemical exposure and injury, in which so much of what happens is unseen and not necessarily understood.
Even in the case of cigarette smoking, which everyone agrees causes lung cancer, this information is derived from the increase risk to smoking populations versus non smoking populations--a ten fold increase, clearly a dramatic effect. And yet ninety percent of smokers will never develop lung cancer and some non smokers will, so when an individual smoker develops lung cancer can we really say his or her cancer was 'caused' by the smoking? How do we know? Maybe this smoker was one of the ninety percent that would not get it and maybe the cause was among that that occurs in non smokers.
The consensus among the scientific community, as reflected in such highly respected consensus type agencies as the International Agency for Research on Cancer (IARC) and the National Toxicology Program (NTP) is reflected in the determination of chemicals that are deemed to cause or likely to cause cancer in humans. In civil court cases, that is, with a standard of more likely than not, both of these designations surpass that threshold.
(continued)
The guidelines for NTP categorization of human carcinogen include any type of human data, including epidemiological, clinical case study and cell studies; guidelines for IARC categorization include both human data but also, in exceptional cases when evidence is less than sufficient in humans but is sufficient in animals, combined with strong indication of shared mechanism of action, the animal / mechanistic data can be used to establish human carcinogen. Key in both of these cases is the absence of solid epidemiological data, or weak epidemiological data (which can include negative data in poorly designed experiments, which do not count).
The next set of categories, probable human carcinogen and reasonably anticipated to be a human carcinogen--which most closely equates to the legal standard of more likely than not--includes insufficient human data or human data which could be interpreted to mean human carcinogen but could also be interpreted to be due to chance or other causes, combined with adequate animals studies and strong studies showing similar mechanisms of action in humans and animals. The animal data and the mechanistic data alone are not sufficient to support a conclusion of human carcinogen but together, the fit of these data with each other do support such a conclusion.
In no cases in either the top category or the second category (probable or reasonably anticipated human carcinogen) did the use of animal data for inclusion in the category subsequently give way to a downgrading when additional data, human or animal become available. This suggests that not only was the basis of categorization based on animal data accurate, but it was also conservative. That is, it is possible to use this kind of data and follow strict rules and not over do it or end up with unnecessary and un scientific speculation. In terms of reliability and error rate, this was highly scientific. In the other direction, additional data led to the upgrade of eight compounds from either probable or possible human carcinogen to category 1 (human carcinogen). In other words, with each change due to new data, human data or additional animal data confirmed, strongly, the existing animal data such that the categorization could be increased.
Of the 55 substances deemed 'probable human carcinogens', 38 or 70 percent had been previously rated possible human carcinogen and had been upgraded with additional data. Again, all data that changed status confirmed and strengthened the categorization. Not a single compound rated as probable human carcinogen was downgraded to possible human carcinogen, based on additional data becoming available. In other words, in the probable human category, which is between definite and possible human carcinogen, all additional data and all change of categorization was in the movement of strenthening and confirmation and not to a weakening due to contradiction.
Only in the 'possible' category, which includes both insufficient human and animal data, were eight compounds downgraded into category 3, which includes compounds that cannot be categorized due to lack of information or contradictory information. This category includes approximately 500 compounds, about ten times as many as in either the human carcinogen category or in the probable human carcinogen category. (Contrary to popular belief, most chemicals do NOT cause cancer in laboratory animals even if given in massive doses.)
It is important to continue educating the court about these matters and to keep using real examples to demonstrate the point.