The Need to Examine the Role of Epidemiology in Proving Causation
     The Supreme Court's Daubert ruling shifts the test of admissibility of scientific expert testimony away from the "general acceptance" Frye test to the broader "helpful scientific knowledge" test of the Federal Rules of Evidence, and also gives the trial judge an enhanced gatekeeping responsibility when determining admissibility.⁽¹⁾Daubert requires judges to pay close attention to the methodology of proposed expert witnesses, requiring them to enter into a morass of epidemiology, toxicology, statistics, and scientific inferences. But great judges do not necessarily make great scientists, as Chief Justice Rehnquist alludes to in Daubert.⁽²⁾ Epidemiological evidence in particular can be "boring and confusing." ⁽³⁾ Understandably, judges have looked for a bright line indicator of scientific reliability. Increasingly, epidemiology has garnered great attention as a possible "sine qua non"⁽⁴⁾ for proof of causation.⁽⁵⁾ Epidemiological results in the form of Relative Risks, Attributable Risk Proportions, and Odds Ratios have become "talismanic."⁽⁶⁾ Surely Daubert requires a judge to examine proposed expert testimony, and the bases for that testimony, with greater precision than previously required. But Daubert does not specifically mention epidemiology as a necessary element of causation, and epidemiology is no philosopher's stone that will turn unreliable science into gold.⁽⁷⁾ In examining the proffered evidence of causation, judges must remain mindful of the power and limits of statistical evidence and decide whether the methodology underlying expert testimony is both reliable and relevant.

A Brief Introduction to Epidemiology
     Epidemiology is the study of the "causes, distribution, and control of disease in populations."⁽⁸⁾ An epidemiological study can be used to "statistically measure the relation between the exposure to a chemical and the manifestation of disease" by comparing the incidence of disease that would be expected to occur by chance in a certain group with the actual incidence of disease in that group.⁽⁹⁾ If the statistical analysis shows a significantly greater incidence of the disease, it suggests that there is a factor other than chance that explains the greater frequency. If the members of the particular group are a more or less random sample except for one characteristic, then the epidemiology supports the inference that the common characteristic causes the condition.⁽¹⁰⁾

The Argument why Epidemiology is Necessary to Prove Causation
     Proponents of epidemiology have argued that where the etiology⁽¹¹⁾ of a disease is unknown, epidemiological studies are "highly persuasive"⁽¹²⁾ and "critical"⁽¹³⁾ and can be "indispensable"⁽¹⁴⁾ to showing causation. The argument is that a plaintiff should not be allowed to argue that X caused her particular injury (specific causation) without first showing that X can cause that injury (general causation). Thus, a plaintiff must produce statistical analysis showing that those who have come into contact with the agent in question generally have a greater risk of contracting the disease than expected. Other evidence of causation (animal studies, chemical structure analysis, and differential diagnosis) is deemed useless unless epidemiology shows a significant relationship between the agent and disease. These considerations led Judge Weinstein to conclude "sound epidemiological studies are the only useful studies having any bearing on causation."⁽¹⁵⁾

     Epidemiology can shift the burdens of production and persuasion, and, depending on which party introduces the studies and on how conclusive they are, can either prove a plaintiff's case or show that there is no question of fact in summary judgment.⁽¹⁶⁾ Defendants can rely on negative epidemiological results to establish a rebuttable presumption against causation.⁽¹⁷⁾ The burden of production cannot be met by expert testimony unless the experts impeach the negative epidemiology or provide positive epidemiology. Otherwise, the expert testimony is without "reliable basis," presumably failing the reliability demand of Daubert.⁽¹⁸⁾

The Argument why Epidemiology is Not Necessary to Prove Causation
     One objection to the claim that epidemiology is required to prove causation is that no one method in particular is needed to show causation, and so certainly "the legal standard of proof doesn't require plaintiffs to prove causation only through statistically significant epidemiology."⁽¹⁹⁾ The Fourth Circuit in Benedi v. McNeil - P.P.C. Inc.⁽²⁰⁾ recognized that Daubert neither requires the plaintiff to produce epidemiology nor the plaintiff's scientific experts to base their conclusions on epidemiology. However, for this to be effective during trial, the experts must have utilized sound methodologies such as those used by doctors "day in and day out" in diagnosing their patients. ⁽²¹⁾ The Third Circuit has stated in In re Paoli R.R. Yard PCB Litigation⁽²²⁾ and DeLuca v. Merrell Dow Pharmaceuticals, Inc.⁽²³⁾ that no one particular piece of evidence is necessary to show causation. Furthermore, if a plaintiff's expert does use epidemiology, it does not have to meet any certain level of statistical significance.⁽²⁴⁾

     Even if a party does present epidemiological evidence, the study may be limited or flawed in ways which caution exclusive reliance on it. According to some, epidemiology is but one tool, and not necessarily "more probative" than toxicology or animal studies.⁽²⁵⁾ In their effort to sort through technical and complex scientific evidence, fact finders "may be misled by numerical exertions of probability."⁽²⁶⁾ At best this may result in confusion and boredom.⁽²⁷⁾ At worst, the court may embark on its own statistical quest backed by whatever (inexpert) computational prowess it possesses. In Bazemore v. Friday,⁽²⁸⁾ the desire to sort through statistical evidence led a court to represent as "dispositive evidence for the defendant" statistical evidence which, if correctly computed, actually favored the plaintiff.⁽²⁹⁾ Otherwise compelling epidemiological evidence may suffer from any one of a number of flaws that reduce their probative value.⁽³⁰⁾ Ethical constraints on testing human subjects force epidemiological studies to often rely on "incomplete and unreliable data" or on an "inadequate control group."⁽³¹⁾ A study may also be poorly designed, externally invalid, or misinterpreted.⁽³²⁾ Negative epidemiological results, such as those relied on by Judge Weinstein, are particularly prone to methodological flaws, and thus may "carry little weight."⁽³³⁾ Finally, there is often a lack of research. Epidemiological evidence of carcinogeneity is available for less than 1% of chemicals currently in use.⁽³⁴⁾ Although the Agent Orange litigation is replete with epidemiological studies, none are conclusive, and it is unlikely that "any convincing epidemiological study of Agent Orange could be done at any cost."⁽³⁵⁾ Rigid commitment to epidemiology will unfairly penalize those "plaintiffs unfortunate enough to be injured by something not already the subject of basic research"⁽³⁶⁾ and those without the financial means to fund their own research. Rigidly requiring epidemiological evidence "leaves entirely to defendants the range between what jurors . . . can rationally believe, and what statisticians can prove." ⁽³⁷⁾

The Argument That Epidemiology Must Show a Doubling of the Risk to Prove Causation
     One statistic generated by epidemiology is the Relative Risk, the ratio between a studied population's actual and expected rates of disease. For example, if 10% of Americans got food poisoning last year we would expect a random sample population of 100 to have 10 cases of food poisoning. Now suppose that out of 100 consumers of Ptomaine brand cheese there are 20 cases of food poisoning. The ratio of actual cases to expected cases (20:10) equals 2, and so the Relative Risk here is 2.0. Since we can attribute 10 of these cases to "chance," or the "background risk" of food poisoning, it seems that, all things being equal, we can attribute the other 10 cases to the cheese.⁽³⁸⁾ So for our case, the chance that a particular cheese-eater's food poisoning is due to chance is equal to the chance that it is due to the cheese. Since a plaintiff must prove the defendant 'more likely than not' caused his injury, it might seem that "a Relative Risk greater than 2.0 would permit an inference that an individual plaintiff's disease was 'more likely than not' caused by the implicated agent."⁽³⁹⁾ This view requiring a Relative Risk greater than 2.0 was suggested by the district court on remand in Daubert, and was recently used to grant summary judgment in a breast implant case in Oregon.⁽⁴⁰⁾ Since a Relative Risk of 1.0 implies that all incidents of disease are due to chance, it has been suggested that causation cannot be shown if the plaintiff's epidemiology results in a range of Relative Risk includes 1.0.⁽⁴¹⁾ At first blush, a Relative Risk of 2.0 or less would not seem to support causation.

The Argument That Epidemiology Does Not Need to Show a Doubling of the Risk to Prove Causation
     Some courts have allowed findings of causation even when the epidemiology most favorable to the plaintiffs shows a Relative Risk of less than 2.0. One way to justify this decision is simply to note that the legal standard of causation may be more liberal than the scientific standard.⁽⁴²⁾ A better justification is to realize that epidemiology is just one factor scientists (and jurists) should weigh when attempting to determine causation -- the "most reliable results" occur when epidemiology works in tandem with other disciplines.⁽⁴³⁾ The Second Circuit has declined to enforce a minimum "floor" for Relative Risks, choosing to allow plaintiffs to prove causation by a combination of Relative Risk and other evidence and allowing a carefully instructed jury to decide whether relative risk between 1.0 and 2.0 shows causation.⁽⁴⁴⁾ A risk of 2.0 or greater is not always dispositive of causation -- it is just "one piece of evidence, not a password to causation."⁽⁴⁵⁾ The equation of Relative Risk at 2.0 with the legal standard of 'more likely than not' is "inappropriate," since epidemiologists rely on more than the Relative Risk finding to determine causation.⁽⁴⁶⁾ Without further corroboration, "a Relative Risk of 2.0 offers weak evidence of causation in a study population and very incomplete evidence for specific, individual causation."⁽⁴⁷⁾ Even a risk as high as 8.0 may not establish causation.⁽⁴⁸⁾

Conclusion
     In making a ruling whether or not to admit testimony put forth by an expert witness, a judge is often presented with epidemiological studies aimed at proving causation. The issues discussed above necessarily shape and influence a judge's decision whether or not to allow such studies to be considered by a jury in its deliberation of a question of causation.

1. Daubert v. Merrell Dow Pharmaceuticals, Inc., 113 S.Ct. 2786 (1993).
2. Id. at 2800.
3. Wayne Roth-Nelson & Kathey Verdeal, Risk Evidence in Toxic Torts, 2 Envtl. Law. 441 (1996).
4. Translated roughly as "the essential component."
5. Roth-Nelson & Verdeal, supra note 3 at 427.
6. John M. Conley & David W. Peterson, The Science of Gatekeeping: The Federal Judicial Center's New Reference Manual on Scientific Evidence, 74 N.C. L. Rev. 1183, 1219 (1996).
7. Peter H. Shuck, Agent Orange On Trial 235 (1986).
8. American Heritage College Dictionary (3rd ed. 1993).
9. Roth-Nelson & Verdeal, supra note 3 at 405.
10. Id. at 407.
11. Etiology can be defined as the cause of a disease or disorder.
12. In re "Agent Orange" Product Liability Litigation, MDL No. 381, 611 F. Supp. 1223, 1240 (E.D.N.Y. 1985), aff'd, 818 F. 2d 187 (2d Cir. 1987), cert. den. sub nom, Lombardi v. Dow, 487 U.S. 1234 (1988).
13. 611 F.Supp. at 1239.
14. In re Joint Eastern & Southern District Asbestos Litigation, 52 F 3d 1124, 1128 (2d Cir. 1995).
15. 611 F.Supp. at 1231.
16. Conley & Peterson, supra note 6, at 1221; 611 F. Supp. at 1241.
17. 611 F.Supp. at 1259.
18. Id. at 1234.
19. Roth-Nelson & Verdeal, supra note 3 at 427.
20. 66 F. 3d 1378 (4th Cir. 1995).
21. G. Michael Fenner, The Daubert Handbook: The Case, Its Essential Dilemma, and Its Progeny, 29 Creighton L. Rev. 939, 1009 (1996).
22. 916 F. 2d 829, 862 (3rd Cir. 1990), cert. den. sub nom, General Elec. Co. v. Knight, 499 U.S. 961 (1991).
23. 911 F. 2d 941 (3rd Cir. 1990).
24. Roth-Nelson & Verdeal, supra note 3 at 428.
25. Schuck, supra note 7 at 237.
26. Roth-Nelson & Verdeal, supra note 3 at 438.
27. Id. at 441.
28. 751 F. 2d 662 (4th Cir. 1984), aff'd in part, vacated in part, and remanded, 478 U.S. 385 (1986) (per curiam), on remand, 848 F. 2d 476 (4th Cir. 1988).
29. Conley & Peterson, supra note 6 at 1207.
30. Roth-Nelson & Verdeal, supra note 3 at 416.
31. Schuck, supra note 7 at 235.
32. Roth-Nelson & Verdeal, supra note 3 at 423-4, 427.
33. Id. at 434.
34. Id. at 416.
35. Schuck, supra note 7 at 236.
36. Conley & Peterson, supra note 3 at 1197.
37. Charles Nesson, Agent Orange Meets the Blue Bus: Factfinding at the Frontier of Knowledge, 66 B.U. L. Rev. 521, 526 (1986).
38. It is worth noting that epidemiological data can only definitively show a correlation. Causation might be inferred from it in a variety of ways, one of which is the ruling out of any other cause for the increase shared at a high rate by Ptomaine-eaters.
39. Conley & Peterson, supra note 6 at 1221.
40. Id. at 1197; Hall v. Baxter Healthcare Corp., 947 F.Supp. 1387 (D. Or. 1996).
41. 52 F.3d at 1133.
42. Roth-Nelson & Verdeal, supra note 3 at 427.
43. Conley & Peterson, supra note 6 at 1222.
44. 52 F.3d at 1128, 1134.
45. Roth-Nelson & Verdeal, supra note 3 at 425.
46. Id. at 421-22.
47. Id. at 422.
48. Id. at 425, 429.