Most epidemiologic studies are not admissible. Such studies involve many layers of hearsay evidence, measurements of exposures, diagnoses, records, and the like, which cannot be “cross-examined.” Our legal system allows expert witnesses to rely upon such studies, although clearly inadmissible, when “experts in the particular field would reasonably rely on those kinds of facts or data in forming an opinion on the subject.” Federal Rule of Evidence 703. One of the problems that judges face in carrying out their gatekeeping duties is to evaluate whether challenged expert witnesses have reasonably relied upon particular studies and data. Judges, unlike juries, have an obligation to explain their decisions, and many expert witness gatekeeping decisions by judges fall short by failing to provide citations to the contested studies at issue in the challenge. Sometimes the parties may be able to discern what is being referenced, but the judicial decision has a public function that goes beyond speaking to the litigants before the court. Without full citations to the studies that underlie an expert witness’s opinion, the communities of judges, lawyers, scientists, and others cannot evaluate the judge’s gatekeeping. Imagine a judicial opinion that vaguely referred to a decision by another judge, but failed to provide a citation? We would think such an opinion to be a miserable failure of the judge’s obligation to explain and justify the resolution of the matter, as well as a case of poor legal scholarship. The same considerations should apply to the scientific studies relied upon by an expert witness, whose opinion is being discussed in a judicial opinion.
Judge Sarah Vance’s opinion in Burst v. Shell Oil Co., C. A. No. 14–109, 2015 WL 3755953 (E.D. La. June 16, 2015) [cited as Burst], is a good example of judicial opinion writing, in the context of deciding an evidentiary challenge to an expert witness’s opinion, which satisfies the requirements of judicial opinion writing, as well as basic scholarship. The key studies relied upon by the challenged expert witness are identified, and cited, in a way that permits both litigants and non-litigants to review Her Honor’s opinion, and evaluate both the challenged expert witness’s opinion, and the trial judge’s gatekeeping performance. Citations to the underlying studies creates the delicious possibility that the trial judge might actually have read the papers to decide the admissibility question. On the merits, Judge Vance’s opinion in Burst also serves as a good example of judicial scrutiny that cuts through an expert witness’s hand waving and misdirection in the face of inadequate, inconsistent, and insufficient evidence for a causal conclusion.
Burst is yet another case in which plaintiff claimed that exposure to gasoline caused acute myeloid leukemia (AML), one of several different types of leukemia[1]. The claim is fraught with uncertainty and speculation in the form of extrapolations between substances, from high to low exposures, and between diseases.
Everyone has a background exposure to benzene from both natural and anthropogenic sources. Smoking results in approximately a ten-fold elevation of benzene exposure. Agency for Toxic Substances and Disease Registry (ATSDR) Public Health Statement – Benzene CAS#: 71-43-2 (August 2007). Gasoline contains small amounts of benzene, on the order of 1 percent or less. U.S. Environmental Protection Agency (EPA), Summary and Analysis of the 2011 Gasoline Benzene Pre-Compliance Report (2012).
Although gasoline has always contained benzene, the quantitative difference in levels of benzene exposure involved in working with concentrated benzene and with gasoline has led virtually all scientists and regulatory agencies to treat the two exposures differently. Benzene exposure is a known cause of AML; gasoline exposure, even in occupational contexts, is not taken to be a known cause of AML. Dose matters.
Although the reviews of the International Agency for Research on Cancer (IARC) are sometimes partisan, incomplete, and biased towards finding carcinogenicity, the IARC categorizes benzene as a known human carcinogen, in large part because of its known ability to cause AML, but regards the evidence for gasoline as inadequate for making causal conclusions. IARC, Monographs on the Evaluation of Carcinogenic Risks to Humans, Vol. 45, Occupational Exposures in Petroleum Refining; Crude Oil and Major Petroleum Fuels (1989) (“There is inadequate evidence for the carcinogenicity in humans of gasoline.”) (emphasis in original)[2].
To transmogrify a gasoline case into a benzene case, plaintiff called upon Peter F. Infante, a fellow of the white-hat conspiracy, Collegium Ramazzini, and an adjunct professor at George Washington University School of Public Health and Health Services. Previously, Dr. Infante was Director of OHSA’s Office of Standards Review (OSHA). More recently, Infante is known as the president and registered agent of Peter F. Infante Consulting, LLC, in Falls Church, Virginia, and a go-to expert witness for plaintiffs in toxic tort litigation[3].
In the Burst case, Infante started out in trouble, by claiming that he had he “followed the methodology of the International Agency for Research on Cancer (IARC) and of the Occupational Safety and Health Administration (OSHA) in evaluating epidemiological studies, case reports and toxicological studies of benzene exposure and its effect on the hematopoietic system.” Burst at *4. Relying upon the IARC’s methodology might satisfy some uncritical courts, but here the IARC itself sharply distinguished its characterizations of benzene and gasoline in separate reviews. Infante’s opinion ignored this divide, although it ultimately had to connect gasoline exposure to the claimed injury[4].
Judge Vance found that Infante’s proffered opinions ransacked the catalogue of expert witness errors. Infante:
- relied upon studies of benzene exposure and diseases other than the outcome of interest, AML. Burst at *4, *10, *13.
- relied upon studies of benzene exposure rather than gasoline exposure. Burst at *9.
- relied upon studies that assessed outcomes in groups with multiple exposures, which studies were hopelessly confounded. Burst at *7.
- failed to acknowledge the inconsistency of outcomes in the studies of the relevant exposure, gasoline. Burst at *9.
- relied upon studies that lacked adequate exposure measurements and characterizations, which lack was among the reasons that the ATSDR declined to label gasoline a carcinogen. Burst at *12.
- relied upon studies that did not report statistically significant associations between gasoline exposure and AML. Burst at *10, *12
- cherry picked studies and failed to explain contrary results. Burst at *10.
- cherry picked data from within studies that did not otherwise support his conclusion. Burst at *10.
- interpreted studies at odds with how the authors of published papers interpreted their own studies. Burst at *10.
- failed to reconcile conflicting studies. Burst at *10.
- manipulated data without sufficient explanation or justification. Burst at *14.
- failed to conduct an appropriate analysis of the entire dataset, along the lines of Sir Austin Bradford Hill’s nine factors. Burst at *10.
The manipulation charge is worth further discussion because it reflects upon the trial court’s acumen and the challenged witness’s deviousness. Infante combined the data from two exposure subgroups from one study[5] to claim that the study actually had a statistically significant association. The trial court found that Dr. Infante failed to explain or justify the recalculation. Burst at *14. At the pre-trial hearing, Dr. Infante offered that he performed the re-calculation on a “sticky note,” but failed to provide his calculations. The court might also have been concerned about the misuse of claiming statistical significance in a post-hoc, non-prespecified analysis that would have clearly raised a multiple comparisons issue. Infante also combined two separate datasets from an unpublished study (the Spivey study for Union Oil), which the court found problematic for his failure to explain and justify the aggregation of data. Id. This recalculation raises the issue whether the two separate datasets could be appropriately combined.
For another study[6], Infante adjusted the results based upon his assessment that the study was biased by a “healthy worker effect[7].” Burst at *15. Infante failed to provide any explanation of how he adjusted for the healthy worker effect, thus giving the court no basis for evaluating the reliability of his methodology. Perhaps more telling, the authors of this study acknowledged the hypothetical potential for healthy worker bias, but chose not to adjust for it because their primary analyses were conducted internally within the working study population, which fully accounted for the potential bias[8].
The court emphasized that it did not question whether combining datasets or adjusting for bias was accepted or proper methodology; rather it focused its critical scrutiny on Infante’s refusal or failure to explain and justify his post-hoc “manipulations of published data.” Burst at *15. Without a showing that AML is more common among non-working, disabled men, the health worker adjustment could well be questioned.
In the final analysis, Infante’s sloppy narrative review could not stand in the face of obviously inconsistent epidemiologic data. Burst at *16. The trial court found that Dr. Infante’s methodology of claiming reliance upon multiple studies, which did not reliably (validly) support his claims or “fit” his conclusions, failed to satisfy the requirements of Federal Rule of Evidence 702. The analytical gap between the data and the opinion were too great. Id. at *8. Infante’s opinion fell into the abyss[9].
[1] See, e.g., Castellow v. Chevron USA, 97 F. Supp. 2d 780, 796 (S.D.Tex.2000) (“Plaintiffs here have not shown that the relevant scientific or medical literature supports the conclusion that workers exposed to benzene, as a component of gasoline, face a statistically significant risk of an increase in the rate of AML.”); Henricksen v. Conoco Phillips Co., 605 F.Supp.2d 1142, 1175 (E.D.Wa. 2009) (“None of the studies relied upon have concluded that gasoline has the same toxic effect as benzene, and none have concluded that the benzene component of gasoline is capable of causing AML.”); Parker v. Mobil Oil Corp., 7 N.Y.3d 434, 450 (N.Y.2006) (“[N]o significant association has been found between gasoline exposure and AML. Plaintiff’s experts were unable to identify a single epidemiologic study finding an increased risk of AML as a result of exposure to gasoline.”).
[2] See also ATSDR Toxicological Profile for Gasoline (1995) (concluding “there is no conclusive evidence to support or refute the carcinogenic potential of gasoline in humans or animals based on the carcinogenicity of one of its components, benzene”); ATSDR, Public Health Statement for Automotive Gasoline (June 1995) (“[However, there is no evidence that exposure to gasoline causes cancer in humans. There is not enough information available to determine if gasoline causes birth defects or affects reproduction.”).
[3] See, e.g., Harris v. CSX Transp., Inc., 753 SE 2d 275, 232 W. Va. 617 (2013); Henricksen v. ConocoPhillips Co., 605 F. Supp. 2d 1142 (E.D. Wash. 2009); Roney v. GENCORP, Civil Action No. 3: 05-0788 (S.D.W. Va. Sept. 18, 2009); Chambers v. Exxon Corp., 81 F. Supp. 2d 661 (M.D. La. 2000).
[4] Judge Vance did acknowledge that benzene studies were relevant to Infante’s causation opinion, but emphasized that such studies could not suffice to show that all gasoline exposures could cause AML. Burst at *10 (citing Dickson v. Nat’l Maint. & Repair of Ky., Inc., No. 5:08–CV–00008, 2011 WL 12538613, at *6 (W.D. Ky. April 28, 2011) (“Benzene may be considered a causative agent despite only being a component of the alleged harm.”).
[5] L. Rushton & H. Romaniuk, “A Case-Control Study to Investigate the
Risk of Leukaemia Associated with Exposure to Benzene in Petroleum Marketing and Distribution Workers in the United Kingdom,” 54 Occup. & Envt’l Med. 152 (1997).
[6] Otto Wong, et al., “Health Effects of Gasoline Exposure. II. Mortality Patterns of Distribution Workers in the United States,” 101 Envt’l Health Persp. 6 (1993).
[7] Burst at *15, citing and quoting from John Last, A Dictionary of Epidemiology (3d ed.1995) (“Workers usually exhibit lower overall death rates than the general population because the severely ill and chronically disabled are ordinarily excluded from employment.”).
[8] Wong, supra.
[9] In a separate opinion, Judge Vance excluded a physician, Dr. Robert Harrison, who similarly opined that gasoline causes AML, and Mr. Burst’s AML, without the benefit of sound science to support his opinion. Burst v. Shell Oil Co., C. A. No. 14–109, 2015 WL 2015 WL 3620111 (E.D. La. June 9, 2015).
