BERGERINTRO.DOC 4/23/2004 12:32 PM SCIENCE FOR JUDGES II INTRODUCTION Margaret A. Berger* This issue of the Journal of Law and Policy contains a second installment of articles about science-related questions that arise in the litigation context. As previously explained, these essays are expanded and edited versions of presentations made to federal and state judges at programs funded by the Common Benefit Trust established in the Silicone Breast Implant Products Liability Litigation.1 These conferences are held at Brooklyn Law School under the auspices of its Center for Health Law and Policy, in collaboration with the Federal Judicial Center, the National Center for State Courts, and the National Academies of Science=s Panel on Science, Technology and Law. Science for Judges II focused on two principal topics: (1) the practice of epidemiology and its role in judicial proceedings; and (2) the production of science through the regulatory process of administrative agencies. Epidemiology has played a significant role in toxic tort actions in proving causation, often the most crucial issue in dispute. A failure to prove causation means a victory for the defense. Many courts consider epidemiologic evidence the “gold standard” of proof, and some judges go so far as to hold that a plaintiff cannot prevail in proving causation in the absence of confirmatory epidemiologic studies.2 The three papers on epidemiology by * Suzanne J. and Norman Miles Professor of Law, Brooklyn Law School. Professor Berger is the Director of the Science for Judges Program. 1 See Margaret A. Berger, Introduction, Science for Judges, 12 J. L. & POL’Y 1 (2003). 2 See, e.g., Wade-Greaux v. Whitehall Labs., Inc., 874 F. Supp. 1441, 1480 (D.V.I. 1994). 485 BERGERINTRO.DOC 4/23/2004 12:32 PM 486 JOURNAL OF LAW AND POLICY extremely well-credentialed scientists should therefore be of considerable interest to anyone concerned with toxic tort litigation. The first, by Professor John Concato of the Yale University School of Medicine, provides an overview of different research designs that epidemiologists employ in conducting studies. The second, by Professors Joseph Lau and John Ioannidis of the Tufts-New England Medical Center, discusses how and when multiple epidemiologic studies can be combined. The final paper by Professor James Robins of the Harvard School of Public Health expresses skepticism about hinging compensation in toxic tort actions on proof of causation derived from epidemiologic data. The second set of papers deals with science produced by administrative agencies. An introductory comment by Professor Richard Merrill of the University of Virginia Law School explains that both the Federal Drug Administration (FDA) and the Environmental Protection Agency (EPA)—which are responsible for regulating the great majority of products that become the subject of toxic tort litigation—require scientific studies and make scientific assessments in the course of their work. The science that is produced may subsequently become relevant in court proceedings when, for instance, a plaintiff claims that taking a drug approved by the FDA caused adverse health effects. Papers by Dr. Michael Friedman, formerly with the FDA, and Robert Sussman, Esq., formerly with the EPA, discuss the respective roles of these agencies in creating scientific information. Professor Wendy Wagner of the University of Texas Law School writes of a relatively new phenomenon: the importation into regulatory decision-making of a new approach that has its roots in the Daubert3 test used by federal courts in determining the The notion that one can accurately extrapolate from animal data to humans to prove causation without supportive positive epidemiologic studies is scientifically invalid because it is inconsistent with several universally accepted and tested scientific principles. Id. 3 In Daubert v. Merrrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993), the first of the Supreme Court=s recent cases on the admissibility of expert testimony, the Court imposed an obligation on federal district judges to screen scientific opinions proffered by an expert to ensure scientific reliability before BERGERINTRO.DOC 4/23/2004 12:32 PM SCIENCE FOR JUDGES II 487 admissibility of expert testimony. In the final paper, Professor Sheila Jasanoff of the John F. Kennedy School of Government at Harvard University and Dogan Parese provide a comparative perspective by contrasting the policies that drive the treatment of asbestos claims in the United States with those that lead Great Britain and the Netherlands to manage the compensation of asbestos victims through administrative rather than judicial processing. These brief descriptions of the papers contained in this issue of the Journal provide a glimpse of the complexity and importance of the scientific and policy issues that courts encounter when handling toxic tort litigation. It is the hope of the organizers of the Science for Judges programs that these papers will prove useful to judges and lawyers who deal with the daunting questions that arise at the intersection of science and the law. allowing the expert to testify. CONCATOMACRO WITH FIGURES COMPLETE.DOC 4/23/2004 12:33 PM OVERVIEW OF RESEARCH DESIGN IN EPIDEMIOLOGY John Concato, M.D., M.S., M.P.H.* INTRODUCTION The objectives of this paper are to first, provide background regarding a conceptual model of causality in epidemiology; second, describe common types of structure (architecture) used for research design in epidemiology, including descriptive, cohort, case-control, and cross-sectional studies; third, review frequently- encountered formats for reporting the results of such studies; and fourth, discuss the strengths and limitations of strategies used in epidemiology. The first and fourth objectives represent a big- picture assessment for interpreting epidemiologic studies; whereas the second and third objectives promote a nuts-and-bolts understanding of the studies themselves. I. CONCEPTUAL MODEL OF CAUSALITY A discussion of research design involves considerations of the concept of causality, i.e., what causes disease.1 In this context, * The author is an Associate Professor of Medicine in the Department of Internal Medicine at Yale University School of Medicine in New Haven, CT, and Director of the Clinical Epidemiology Research Center at VA Connecticut Healthcare System in West Haven, CT. The conclusions and opinions expressed in this article are those of the author and do not represent any official or unofficial position of Yale University or the U.S. Department of Veterans Affairs. Presented at the second Science for Judges Symposium, Brooklyn Law School, Brooklyn, NY, November 7, 2003. 1 See, e.g., Sir Austin Bradford-A.B. Hill, The Environment and Disease: Association or Causation?, 58 PROC. R. SOC. MED. 295-300 (1965). 489 CONCATOMACRO WITH FIGURES COMPLETE.DOC 4/23/2004 12:33 PM 490 JOURNAL OF LAW AND POLICY biological phenomena can be considered either deterministic or probabilistic. A deterministic situation exists when the exposure can be linked conclusively to an outcome (e.g., when a particular genetic rearrangement causes sickle cell anemia). In contrast, a probabilistic phenomenon occurs when exposure is said to be associated with an outcome (e.g., if hypertension (high blood pressure) is associated with stroke). In the first example, the genetic problem is always found with the disease, and vice versa. In the second example, hypertension increases the probability of stroke; but some patients with hypertension do not suffer a stroke, and some patients with stroke do not have antecedent hypertension. A major role of epidemiological research design is to provide information for, or against, a probabilistic association. A conceptual (intellectual) model2 has been developed for this purpose, and is often described with terms such as cause-effect research (Figure 1). The entity being evaluated as a possible cause of a disease, or other endpoint, is referred to as an exposure, but is not limited to environmental exposures, and can include a person’s age, sex, personal habits (such as cigarette smoking), ingested medications, etc. The entity being assessed as a possible disease or other endpoint is referred to as an outcome, and can include the development of a disease, a quality of life measurement, death, etc. FIGURE 1—CONCEPTUAL MODEL FOR CAUSE-EFFECT RESEARCH Baseline state → Exposure → Outcome Condition prior Factor(s) that Disease to exposure may lead or entity or outcome to outcome of interest Example: Healthy adults → Hypertension → Stroke The model can be applied to the association of hypertension and stroke, addressing the question of whether hypertension can 2 ALVAN R. FEINSTEIN, CLINICAL EPIDEMIOLOGY: THE ARCHITECTURE OF CLINICAL RESEARCH 50 (Saunders 1985). CONCATOMACRO WITH FIGURES COMPLETE.DOC 4/23/2004 12:33 PM RESEARCH DESIGN IN EPIDEMIOLOGY 491 “cause” stroke (in a probabilistic, not a deterministic, sense). In this example, adults, either with or without hypertension, do or do not develop subsequent stroke. The challenge involves determining whether stroke is more common among patients with hypertension; and if so, whether the corresponding evidence (in the form of data) is strong enough, and stable enough, to confirm that an association exists and is “real.” A useful aspect of cause and effect studies is the ability to summarize the association of interest with a simplified schematic (2 × 2 table) showing the relationship of exposure and outcome (Table 1A). Although most