The Federal Judicial Center’s Reference Manual on Scientific Evidence has avoided any clear, consistent guidance on the issue of case reports. The Second Edition waffled:
“Case reports lack controls and thus do not provide as much information as controlled epidemiological studies do. However, case reports are often all that is available on a particular subject because they usually do not require substantial, if any, funding to accomplish, and human exposure may be rare and difficult to study. Causal attribution based on case studies must be regarded with caution. However, such studies may be carefully considered in light of other information available, including toxicological data.”
F.J.C. Reference Manual on Scientific Evidence at 474-75 (2d ed. 2000). Note the complete lack of discussion of base-line risk, prevalence of exposure, and external validity of the “toxicological data.”
The second edition’s more analytically acute and rigorous chapter on statistics generally acknowledged the unreliability of anecdotal evidence of causation. See David Kaye & David Freedman, “Reference Guide on Statistics,” in F.J.C. Reference Manual on Scientific Evidence 91 – 92 (2d ed. 2000).
The Third Edition of the Reference Manual is even less coherent. Professor Berger’s introductory chapter[1] begrudgingly acknowledges, without approval, that:
“[s]ome courts have explicitly stated that certain types of evidence proffered to prove causation have no probative value and therefore cannot be reliable.59”
The chapter on statistical evidence, which had been relatively clear in the second edition, now states that controlled studies may be better but case reports can be helpful:
“When causation is the issue, anecdotal evidence can be brought to bear. So can observational studies or controlled experiments. Anecdotal reports may be of value, but they are ordinarily more helpful in generating lines of inquiry than in proving causation.14”
Reference Manual at 217 (3d ed. 2011). The “generally” is given no context or contour for readers. These authors fail to provide any guidance on what will come from anecdotal evidence, or when and why anecdotal reports may do more than merely generating “lines of inquiry.”
In Matrixx Initiatives Inc. v. Siracusano, 131 S. Ct. 1309 (2011), the Supreme Court went out of its way, way out of its way, to suggest that statistical significance was not always necessary to support conclusions of causation in medicine. Id. at 1319. The Court cited three Circuit court decisions to support its suggestion, but two of three involved specific causation inferences from so-called differential etiologies. General causation was assumed in those two cases, and not at issue[2]. The third case, the notorious Wells v. Ortho Pharmaceutical Corp., 788 F. 2d 741, 744–745 (11th Cir. 1986), was also cited in support of the suggestion that statistical significance was not necessary, but in Wells, the plaintiffs’ expert witnesses actually relied upon studies that claimed at least nominal statistical significance. Wells was and remains representative of what is possible and results when trial judges ignore the constraints of study validity. The Supreme Court, in any event, abjured any intent to specify “whether the expert testimony was properly admitted in those cases [Wells and others],” and the Court made no “attempt to define here what constitutes reliable evidence of causation.” 131 S. Ct. at 1319.
The causal claim in Siracusano involved anosmia, loss of the sense of smell, from the use of Zicam, zinc gluconate. The case arose from a motion to dismiss the complaint; no evidence was ever presented or admitted. No baseline risk of anosmia was pleaded; nor did plaintiffs allege that any controlled study demonstrated an increased risk of anosmia from nasal instillation of zinc gluconate. There were, however, clinical trials conducted in the 1930s, with zinc sulfate for poliomyelitis prophylaxis, which showed a substantial incidence of anosmia in the treated children[3]. Matrixx tried to argue that this evidence was unreliable, in part because it involved a different compound, but this argument (1) in turn demonstrated a factual issue that required discovery and perhaps a trial, and (2) traded on a clear error in asserting that the zinc in zinc sulfate and zinc gluconate were different, when in fact they are both ionic compounds that result in zinc ion exposure, as the active constituent.
The position stridently staked out in Matrixx Initiatives is not uncommon among defense counsel in tort cases. Certainly, similar, unqualified statements, rejecting the use of case reports for supporting causal conclusions, can be found in the medical literature[4].
When the disease outcome has an expected value, a baseline rate, in the exposed population, then case reports simply confirm what we already know: cases of the disease happen in people regardless of their exposure status. For this reason, medical societies, such as the Teratology Society, have issued guidances that generally downplay or dismiss the role that case reports may have in the assessment and determination of causality for birth defects:
“5. A single case report by itself is not evidence of a causal relationship between an exposure and an outcome. Combinations of both exposures and adverse developmental outcomes frequently occur by chance. Common exposures and developmental abnormalities often occur together when there is no causal link at all. Multiple case reports may be appropriate as evidence of causation if the exposures and outcomes are both well-defined and low in incidence in the general population. The use of multiple case reports as evidence of causation is analogous to the use of historical population controls: the co-occurrence of thalidomide ingestion in pregnancy and phocomelia in the offspring was evidence of causation because both thalidomide use and phocomelia were highly unusual in the population prior to the period of interest. Given how common exposures may be, and how common adverse pregnancy outcome is, reliance on multiple case reports as the sole evidence for causation is unsatisfactory.”
The Public Affairs Committee of the Teratology Society, “Teratology Society Public Affairs Committee Position Paper Causation in Teratology-Related Litigation,” 73 Birth Defects Research (Part A) 421, 423 (2005).
When the base rate for the outcome is near zero, and other circumstantial evidence is present, some commentators insist that causality may be inferred from well-documented case reports:
“However, we propose that some adverse drug reactions are so convincing, even without traditional chronological causal criteria such as challenge tests, that a well documented anecdotal report can provide convincing evidence of a causal association and further verification is not needed.”
Jeffrey K. Aronson & Manfred Hauben, “Drug safety: Anecdotes that provide definitive evidence,” 333 Brit. Med. J. 1267, 1267 (2006) (Dr. Hauben was medical director of risk management strategy for Pfizer, in New York, at the time of publication). But which ones are convincing, and why?
* * * * * * * * *
Dr. David Schwartz, in a recent blog post, picked up on some of my discussion of the gadolinium case reports (see here and there), and posited the ultimate question: when are case reports sufficient to show causation? David Schwartz, “8 Examples of Causal Inference Without Data from Controlled Studies” (Dec. 14, 2014).
Dr. Schwartz discusses several causal claims, all of which gave rise to litigation at some point, in which case reports or case series played an important, if not dispositive, role:
- Gadolinium-based contrast agents and NSF
- Amphibole asbestos and malignant mesothelioma
- Ionizing radiation and multiple cancers
- Thalidomide and teratogenicity
- Rezulin and acute liver failure
- DES and clear cell vaginal adenocarcinoma
- Vinyl chloride and angiosarcoma
- Manganese exposure and manganism
Dr. Schwartz’s discussion is well worth reading in its entirety, but I wanted to emphasize some of Dr. Schwartz’s caveats. Most of the exposures are rare, as are the outcomes. In some cases, the outcomes occur almost exclusively with the identified exposures. All eight examples pose some danger of misinterpretation. Gadolinium-based contrast agents appear to create a risk of NSF only in the presence of chronic renal failure. Amphibole asbestos, and most importantly, crocidolite causes malignant mesothelioma after a very lengthy latency period. Ionizing radiation causes some cancers that are all-too common, but the presence of multiple cancers in the same person, after a suitable latency period, is distinctly uncommon, as is the level of radiation needed to overwhelm bodily defenses and induce cancers. Thalidomide was associated by case reports fairly quickly with phocomelia, which has an extremely low baseline risk. Other birth defects were not convincingly demonstrated by the case series. Rezulin, an oral antidiabetic medication, was undoubtedly causally responsible for rare cases of acute liver failure. Chronic liver disease, however, which is common among type 2 diabetic patients, required epidemiologic evidence, which never materialized[5].
Manganism, by definition, is the cause of manganism, but extremely high levels of manganese exposure, and specific speciation of the manganese, are essential to the causal connection. Manganism raises another issue often seen in so-called signature diseases: diagnostic accuracy. Unless the diagnostic criteria have perfect (100%) specificity, with no false-positive diagnoses, then once again, we expect false-positive cases to appear when the criteria are applied to large numbers of people. In the welding fume litigation, where plaintiffs’ counsel and physicians engaged in widespread, if not wanton, medico-legal screenings, it was not surprising that they might find occasional cases that appeared to satisfy their criteria. Of course, the more the criteria are diluted to accommodate litigation goals, the more likely there will be false positive cases.[6]
Dr. Schwartz identifies some common themes and important factors in identifying the bases for inferring causality from uncontrolled evidence:
“(a) low or no background rate of the disease condition;
(b) low background rate of the exposure;
(c) a clear understanding of the mechanism of action.”
These factors and perhaps others should not be confused with strict criteria here. The exemplar cases suggest a family resemblance of overlapping factors that help support the inference, even against the most robust skepticism.
In litigation, defense counsel typically argue that analytical epidemiology is always necessary, and plaintiffs’ counsel claim epidemiology is never needed. The truth is more nuanced and conditional, but the great majority of litigated cases do require epidemiology for health effects because the claimed harms are outcomes that have an expected incidence or prevalence in the exposed population irrespective of exposure.
[1] Reference Manual on Scientific Evidence at 23 (3d ed. 2011) (citing “Cloud v. Pfizer Inc., 198 F. Supp. 2d 1118, 1133 (D. Ariz. 2001) (stating that case reports were merely compilations of occurrences and have been rejected as reliable scientific evidence supporting an expert opinion that Daubert requires); Haggerty v. Upjohn Co., 950 F. Supp. 1160, 1164 (S.D. Fla. 1996), aff’d, 158 F.3d 588 (11th Cir. 1998) (“scientifically valid cause and effect determinations depend on controlled clinical trials and epidemiological studies”); Wade-Greaux v. Whitehall Labs., Inc., 874 F. Supp. 1441, 1454 (D.V.I. 1994), aff’d, 46 F.3d 1120 (3d Cir. 1994) (stating there is a need for consistent epidemiological studies showing statistically significant increased risks).”)
[2] Best v. Lowe’s Home Centers, Inc., 563 F. 3d 171, 178 (6th Cir 2009); Westberry v. Gislaved Gummi AB, 178 F. 3d 257, 263–264 (4th Cir. 1999).
[3] There may have been a better argument for Matrixx in distinguishing the method and place of delivery of the zinc sulfate in the polio trials of the 1930s, but when Matrixx’s counsel was challenged at oral argument, he asserted simply, and wrongly, that the two compounds were different.
[4] Johnston & Hauser, “The value of a case report,” 62 Ann. Neurology A11 (2007) (“No matter how compelling a vignette may seem, one must always be concerned about the reliability of inference from an “n of one.” No statistics are possible in case reports. Inference is entirely dependent, then, on subjective judgment. For a case meant to suggest that agent A leads to event B, the association of these two occurrences in the case must be compared to the likelihood that the two conditions could co-occur by chance alone …. Such a subjective judgment is further complicated by the fact that case reports are selected from a vast universe of cases.”); David A. Grimes & Kenneth F. Schulz, “Descriptive studies: what they can and cannot do,” 359 Lancet 145, 145, 148 (2002) (“A frequent error in reports of descriptive studies is overstepping the data: studies without a comparison group allow no inferences to be drawn about associations, causal or otherwise.”) (“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 and disease.”); Troyen A. Brennan, “Untangling Causation Issues in Law and Medicine: Hazardous Substance Litigation,” 107 Ann. Intern. Med. 741, 746 (1987) (recommending that testifying physicans “[a]void anecdotal evidence; clearly state the opposing side is relying on anecdotal evidence and why that is not good science.”).
[5] See In re Rezulin, 2004 WL 2884327, at *3 (S.D.N.Y. 2004).
[6] This gaming of diagnostic criteria has been a major invitation to diagnostic invalidity in litigation over asbestosis and silicosis in the United States.
