We are 30-plus years into the “Daubert” era, in which federal district courts are charged with gatekeeping the relevance and reliability of scientific evidence. Not surprisingly, given the lawsuit industry’s propensity on occasion to use dodgy science, the burden of awakening the gatekeepers from their dogmatic slumber often falls upon defense counsel in civil litigation. It therefore behooves defense counsel to speak carefully and accurately about the grounds for Rule 702 exclusion of expert witness opinion testimony.
In the context of medical causation opinions based upon epidemiologic evidence, the first obvious point is that whichever party is arguing for exclusion should distinguish between excluding an expert witness’s opinion and prohibiting an expert witness from relying upon a particular study. Rule 702 addresses the exclusion of opinions, whereas Rule 703 addresses barring an expert witness from relying upon hearsay facts or data unless they are reasonably relied upon by experts in the appropriate field. It would be helpful for lawyers and legal academics to refrain from talking about “excluding epidemiological evidence under FRE 702.”[1] Epidemiologic studies are rarely admissible themselves, but come into the courtroom as facts and data relied upon by expert witnesses. Rule 702 is addressed to the admissibility vel non of opinion testimony, some of which may rely upon epidemiologic evidence.
Another common lawyer mistake is the over-generalization that epidemiologic research provides “gold standard” of general causation evidence.[2] Although epidemiology is often required, it not “the medical science devoted to determining the cause of disease in human beings.”[3] To be sure, epidemiologic evidence will usually be required because there is no genetic or mechanistic evidence that will support the claimed causal inference, but counsel should be cautious in stating the requirement. Glib statements by courts that epidemiology is not always required are often simply an evasion of their responsibility to evaluate the validity of the proffered expert witness opinions. A more careful phrasing of the role of epidemiology will make such glib statements more readily open to rebuttal. In the absence of direct biochemical, physiological, or genetic mechanisms that can be identified as involved in bringing about the plaintiffs’ harm, epidemiologic evidence will be required, and it may well be the “gold standard” in such cases.[4]
When epidemiologic evidence is required, counsel will usually be justified in adverting to the “hierarchy of epidemiologic evidence.” Associations are shown in studies of various designs with vastly differing degrees of validity; and of course, associations are not necessarily causal. There are thus important nuances in educating the gatekeeper about this hierarchy. First, it will often be important to educate the gatekeeper about the distinction between descriptive and analytic studies, and the inability of descriptive studies such as case reports to support causal inferences.[5]
There is then the matter of confusion within the judiciary and among “scholars” about whether a hierarchy even exists. The chapter on epidemiology in the Reference Manual on Scientific Evidence appears to suggest the specious position that there is no hierarchy.[6] The chapter on medical testimony, however, takes a different approach in identifying a normative hierarchy of evidence to be considered in evaluating causal claims.[7] The medical testimony chapter specifies that meta-analyses of randomized controlled trials sit atop the hierarchy. Yet, there are divergent opinions about what should be at the top of the hierarchical evidence pyramid. Indeed, the rigorous, large randomized trial will often replace a meta-analysis of smaller trials as the more definitive evidence.[8] Back in 2007, a dubious meta-analysis of over 40 clinical trials led to a litigation frenzy over rosiglitazone.[9] A mega-trial of rosiglitazone showed that the 2007 meta-analysis was wrong.[10]
In any event, courts must purge their beliefs that once there is “some” evidence in support of a claim, their gatekeeping role is over. Randomized controlled trials really do trump observational studies, which virtually always have actual or potential confounding in their final analyses.[11] While disclaimers about the unavailability of randomized trials for putative toxic exposures are helpful, it is not quite accurate to say that it is “unethical to intentionally expose people to a potentially harmful dose of a suspected toxin.”[12] Such trials are done all the time when there is an expected therapeutic benefit that creates at least equipoise between the overall benefit and harm at the outset of the trial.[13]
At this late date, it seems shameful that courts must be reminded that evidence of associations does not suffice to show causation, but prudence dictates giving the reminder.[14] Defense counsel will generally exhibit a Pavlovian reflex to state that causality based upon epidemiology must be viewed through a lens of “Bradford Hill criteria.”[15] Rhetorically, this reflex seems wrong given that Sir Austin himself noted that his nine different considerations were “viewpoints,” not criteria. Taking a position that requires an immediate retreat seems misguided. Similarly, urging courts to invoke and apply the Bradford Hill considerations must be accompanied the caveat that courts must first apply Bradford Hill’s predicate[16] for the nine considerations:
“Disregarding then any such problem in semantics we have this situation. Our observations reveal an association between two variables, perfectly clear-cut and beyond what we would care to attribute to the play of chance. What aspects of that association should we especially consider before deciding that the most likely interpretation of it is causation?”[17]
Courts should be mindful that the language from the famous, often-cited paper was part of an after-dinner address, in which Sir Austin was speaking informally. Scientists will understand that he was setting out a predicate that calls for
(1) an association, which is
(2) “perfectly clear cut,” such that bias and confounding are excluded, and
(3) “beyond what we would care to attribute to the play of chance,” with random error kept to an acceptable level, before advancing to further consideration of the nine viewpoints commonly recited.
These predicate findings are the basis for advancing to investigate Bradford Hill’s nine viewpoints; the viewpoints do not replace or supersede the predicates.[18]
Within the nine viewpoints, not all are of equal importance. Consistency among studies, a particularly important consideration, implies that isolated findings in a single observational study will rarely suffice to support causal conclusions. Another important consideration, the strength of the association, has nothing to do with “statistical significance,” which is a predicate consideration, but reminds us that large risk ratios or risk differences provides some evidence that the association does not result from unmeasured confounding. Eliminating confounding, however, is one of the predicate requirements for applying the nine factors. As with any methodology, the Bradford Hill factors are not self-executing. The annals of litigation provide all-too-many examples of undue selectivity, “cherry picking,” and other deviations from the scientist’s standard of care.
Certainly lawyers must steel themselves against recommending the “carcinogen” hazard identifications advanced by the International Agency for Research on Cancer (IARC). There are several problematic aspects to the methods of IARC, not the least of which is IARC’s fanciful use of the word “probable.” According to the IARC Preamble, “probable” has no quantitative meaning.[19] In common legal parlance, “probable” typically conveys a conclusion that is more likely than not. Another problem arises from the IARC’s labeling of “probable human carcinogens” made in some cases without any real evidence of carcinogenesis in humans. Regulatory pronouncements are even more diluted and often involved little more than precautionary principle wishcasting.[20]
[1] Christian W. Castile & and Stephen J. McConnell, “Excluding Epidemiological Evidence Under FRE 702,” For The Defense 18 (June 2023) [Castile]. Although these authors provide an interesting overview of the subject, they fall into some common errors, such as failing to address Rule 703. The article is worth reading for its marshaling recent case law on the subject, but I detail of its errors here in the hopes that lawyers will speak more precisely about the concepts involved in challenging medical causation opinions.
[2] Id. at 18. In re Zantac (Ranitidine) Prods. Liab. Litig., No. 2924, 2022 U.S. Dist. LEXIS 220327, at *401 (S.D. Fla. Dec. 6, 2022); see also Horwin v. Am. Home Prods., No. CV 00-04523 WJR (Ex), 2003 U.S. Dist. LEXIS 28039, at *14-15 (C.D. Cal. May 9, 2003) (“epidemiological studies provide the primary generally accepted methodology for demonstrating a causal relation between a chemical compound and a set of symptoms or disease” *** “The lack of epidemiological studies supporting Plaintiffs’ claims creates a high bar to surmount with respect to the reliability requirement, but it is not automatically fatal to their case.”).
[3] See, e.g., Siharath v. Sandoz Pharm. Corp., 131 F. Supp. 2d 1347, 1356 (N.D. Ga. 2001) (“epidemiology is the medical science devoted to determining the cause of disease in human beings”).
[4] See, e.g., Lopez v. Wyeth-Ayerst Labs., No. C 94-4054 CW, 1996 U.S. Dist. LEXIS 22739, at *1 (N.D. Cal. Dec. 13, 1996) (“Epidemiological evidence is one of the most valuable pieces of scientific evidence of causation”); Horwin v. Am. Home Prods., No. CV 00-04523 WJR (Ex), 2003 U.S. Dist. LEXIS 28039, at *15 (C.D. Cal. May 9, 2003) (“The lack of epidemiological studies supporting Plaintiffs’ claims creates a high bar to surmount with respect to the reliability requirement, but it is not automatically fatal to their case”).
[5] David A. Grimes & Kenneth F. Schulz, “Descriptive Studies: What They Can and Cannot Do,” 359 Lancet 145 (2002) (“…epidemiologists and clinicians generally use descriptive reports to search for clues of cause of disease – i.e., generation of hypotheses. In this role, descriptive studies are often a springboard into more rigorous studies with comparison groups. Common pitfalls of descriptive reports include an absence of a clear, specific, and reproducible case definition, and interpretations that overstep the data. Studies without a comparison group do not allow conclusions about cause of disease.”).
[6] Michael D. Green, D. Michal Freedman, and Leon Gordis, “Reference Guide on Epidemiology,” Reference Manual on Scientific Evidence 549, 564n.48 (citing a paid advertisement by a group of scientists, and misleadingly referring to the publication as a National Cancer Institute symposium) (citing Michele Carbone et al., “Modern Criteria to Establish Human Cancer Etiology,” 64 Cancer Res. 5518, 5522 (2004) (National Cancer Institute symposium [sic] concluding that “[t]here should be no hierarchy [among different types of scientific methods to determine cancer causation]. Epidemiology, animal, tissue culture and molecular pathology should be seen as integrating evidences in the determination of human carcinogenicity.”).
[7] John B. Wong, Lawrence O. Gostin & Oscar A. Cabrera, “Reference Guide on Medical Testimony,” in Reference Manual on Scientific Evidence 687, 723 (3d ed. 2011).
[8] See, e.g., J.M. Elwood, Critical Appraisal of Epidemiological Studies and Clinical Trials 342 (3d ed. 2007).
[9] See Steven E. Nissen & Kathy Wolski, “Effect of Rosiglitazone on the Risk of Myocardial Infarction and Death from Cardiovascular Causes,” 356 New Engl. J. Med. 2457 (2007). See also “Learning to Embrace Flawed Evidence – The Avandia MDL’s Daubert Opinion” (Jan. 10, 2011).
[10] Philip D. Home, et al., “Rosiglitazone evaluated for cardiovascular outcomes in oral agent combination therapy for type 2 diabetes (RECORD): a multicentre, randomised, open-label trial,” 373 Lancet 2125 (2009).
[11] In re Zantac (Ranitidine) Prods. Liab. Litig., No. 2924, 2022 U.S. Dist. LEXIS 220327, at *402 (S.D. Fla. Dec. 6, 2022) (“Unlike experimental studies in which subjects are randomly assigned to exposed and placebo groups, observational studies are subject to bias due to the possibility of differences between study populations.”)
[12] Castile at 20.
[13] See, e.g., Benjamin Freedman, “Equipoise and the ethics of clinical research,” 317 New Engl. J. Med. 141 (1987).
[14] See, e.g., In Re Onglyza (Saxagliptin) & Kombiglyze Xr (Saxagliptin & Metformin) Prods. Liab. Litig., No. 5:18-md-2809-KKC, 2022 U.S. Dist. LEXIS 136955, at *127 (E.D. Ky. Aug. 2, 2022); Burleson v. Texas Dep’t of Criminal Justice, 393 F.3d 577, 585-86 (5th Cir. 2004) (affirming exclusion of expert causation testimony based solely upon studies showing a mere correlation between defendant’s product and plaintiff’s injury); Beyer v. Anchor Insulation Co., 238 F. Supp. 3d 270, 280-81 (D. Conn. 2017); Ambrosini v. Labarraque, 101 F.3d 129, 136 (D.C. Cir. 1996).
[15] Castile at 21. See In re Zoloft (Sertraline Hydrochloride) Prods. Liab. Litig., 26 F. Supp. 3d 449, 454-55 (E.D. Pa. 2014).
[16] “Bradford Hill on Statistical Methods” (Sept. 24, 2013); see also Frank C. Woodside, III & Allison G. Davis, “The Bradford Hill Criteria: The Forgotten Predicate,” 35 Thomas Jefferson L. Rev. 103 (2013).
[17] Austin Bradford Hill, “The Environment and Disease: Association or Causation?” 58 Proc. Royal Soc’y Med. 295, 295 (1965).
[18] Castile at 21. See, e.g., In re Onglyza (Saxagliptin) & Kombiglyze XR (Saxagliptin & Metformin) Prods. Liab. Litig., No. 5:18-md-2809-KKC, 2022 U.S. Dist. LEXIS 1821, at *43 (E.D. Ky. Jan. 5, 2022) (“The analysis is meant to apply when observations reveal an association between two variables. It addresses the aspects of that association that researchers should analyze before deciding that the most likely interpretation of [the association] is causation”); Hoefling v. U.S. Smokeless Tobacco Co., LLC, 576 F. Supp. 3d 262, 273 n.4 (E.D. Pa. 2021) (“Nor would it have been appropriate to apply them here: scientists are to do so only after an epidemiological association is demonstrated”).
[19] IARC Monographs on the Identification of Carcinogenic Hazards to Humans – Preamble 31 (2019) (“The terms probably carcinogenic and possibly carcinogenic have no quantitative significance and are used as descriptors of different strengths of evidence of carcinogenicity in humans.”).
[20] “Improper Reliance upon Regulatory Risk Assessments in Civil Litigation” (Mar. 19, 2023).
