Joseph L. Gastwirth
Department of Statistics and Economics
The George Washington University
Statistical Issues in the Use and Interpretation of Epidemiological
Evidence in Product Liability Cases
Epidemiologic studies demonstrating or refuting an increased risk of disease
from exposure to a potentially toxic chemical or drug are useful evidence in
product liability litigation, environmental regulation and the determination
of a worker's eligibility for compensation due to exposure in the workplace.
Different laws may require different amounts of scientific evidence and the
strength of the association usually measured by the relative risk, to
convince the fact-finder. The role of epidemiologic findings in product
liability cases to prove either causation or "failure to warn" of a risk
will be described. The evidence in two cases, Key Pharmaceu-ticals and
Wells v. Ortho will be reviewed and the usefulness of a "reverse" Cornfield
inequality to assess the effect of an omitted variable will be illustrated.
The Wells case has been heavily criticized, e.g. in the N.Y. Times and Wall
St. Journal and the New England Journal of Medicine. After our review of
all the studies, especially those bearing on the "failure to warn" issue,
the listener can decide for themselves whether the trial judge and appellate
court opinions truly deserve the strong criticism they have received. Some
suggestions for increasing the usefulness of scientific evidence in legal
proceedings will be mentioned. If time permits a simple, approximate way of
updating risk estimates as studies accumulate will be discussed and
illustrated on the studies of Reye's syndrome, which led to the notification
of the public about the connection between the use of aspiring to treat flu
and chicken pox in children and the illness. Since then, the incidence of
Reye syndrome has dropped substantially.
Back to Colloquium Series