Reference Manual on Scientific Evidence – 3rd Edition is Past Its Expiry

INTRODUCTION

The new, third edition of the Reference Manual on Scientific Evidence was released to the public in September 2011, as a joint production of the National Academies of Science, and the Federal Judicial Center. Within a year of its publication, I wrote that the Manual needed attention on several key issues. Now that there is a committee working on the fourth edition, I am reprising the critique, slightly modified, in the hope that it may make a difference for the fourth edition.

The Development Committee for the third edition included Co-Chairs, Professor Jerome Kassirer, of Tufts University School of Medicine, and the Hon. Gladys Kessler, who sits on the District Court for the District of Columbia.  The members of the Development Committee included:

  • Ming W. Chin, Associate Justice, The Supreme Court of California
  • Pauline Newman, Judge, Court of Appeals for the Federal Circuit
  • Kathleen O’Malley, Judge, Court of Appeals for the Federal Circuit (formerly a district judge on the Northern District of Ohio)
  • Jed S. Rakoff, Judge, Southern District of New York
  • Channing Robertson, Professor of Engineering, Stanford University
  • Joseph V. Rodricks, Principal, Environ
  • Allen Wilcox, Senior Investigator, Institute of Environmental Health Sciences
  • Sandy L. Zabell, Professor of Statistics and Mathematics, Weinberg College of Arts and Sciences, Northwestern University

Joe S. Cecil, Project Director, Program on Scientific and Technical Evidence, in the Federal Judicial Center’s Division of Research, who shepherded the first two editions, served as consultant to the Committee.

With over 1,000 pages, there was much to digest in the third edition of the Reference Manual on Scientific Evidence (RMSE 3d).  Much of what is covered was solid information on the individual scientific and technical disciplines covered.  Although the information is easily available from other sources, there is some value in collecting the material in a single volume for the convenience of judges and lawyers.  Of course, given that this information is provided to judges from an ostensibly neutral, credible source, lawyers will naturally focus on what is doubtful or controversial in the RMSE. To date, there have been only a few reviews and acknowledgments of the new edition.[1]

Like previous editions, the substantive scientific areas were covered in discrete chapters, written by subject matter specialists, often along with a lawyer who addresses the legal implications and judicial treatment of that subject matter.  From my perspective, the chapters on statistics, epidemiology, and toxicology were the most important in my practice and in teaching, and I have focused on issues raised by these chapters.

The strengths of the chapter on statistical evidence, updated from the second edition, remained, as did some of the strengths and flaws of the chapter on epidemiology.  In addition, there was a good deal of overlap among the chapters on statistics, epidemiology, and medical testimony.  This overlap was at first blush troubling because the RMSE has the potential to confuse and obscure issues by having multiple authors address them inconsistently.  This is an area where reviewers of the upcoming edition should pay close attention.

I. Reference Manual’s Disregard of Study Validity in Favor of the “Whole Tsumish”

There was a deep discordance among the chapters in the third Reference Manual as to how judges should approach scientific gatekeeping issues. The third edition vacillated between encouraging judges to look at scientific validity, and discouraging them from any meaningful analysis by emphasizing inaccurate proxies for validity, such as conflicts of interest.[2]

The Third Edition featured an updated version of the late Professor Margaret Berger’s chapter from the second edition, “The Admissibility of Expert Testimony.”[3]  Berger’s chapter criticized “atomization,” a process she describes pejoratively as a “slicing-and-dicing” approach.[4]  Drawing on the publications of Daubert-critic Susan Haack, Berger rejected the notion that courts should examine the reliability of each study independently.[5]  Berger contended that the “proper” scientific method, as evidenced by works of the International Agency for Research on Cancer, the Institute of Medicine, the National Institute of Health, the National Research Council, and the National Institute for Environmental Health Sciences, “is to consider all the relevant available scientific evidence, taken as a whole, to determine which conclusion or hypothesis regarding a causal claim is best supported by the body of evidence.”[6]

Berger’s contention, however, was profoundly misleading.  Of course, scientists undertaking a systematic review should identify all the relevant studies, but some of the “relevant” studies may well be insufficiently reliable (because of internal or external validity issues) to answer the research question at hand. All the cited agencies, and other research organizations and researchers, exclude studies that are fundamentally flawed, whether as a result of bias, confounding, erroneous data analyses, or related problems.  Berger cited no support for her remarkable suggestion that scientists do not make “reliability” judgments about available studies when assessing the “totality of the evidence.”

Professor Berger, who had a distinguished career as a law professor and evidence scholar, died in November 2010.  She was no friend of Daubert,[7] but remarkably her antipathy had outlived her.  Berger’s critical discussion of “atomization” cited the notorious decision in Milward v. Acuity Specialty Products Group, Inc., 639 F.3d 11, 26 (1st Cir. 2011), which was decided four months after her passing.[8]

Professor Berger’s contention about the need to avoid assessments of individual studies in favor of the whole “tsumish” must also be rejected because Federal Rule of Evidence 703 requires that each study considered by an expert witness “qualify” for reasonable reliance by virtue of the study’s containing facts or data that are “of a type reasonably relied upon by experts in the particular field forming opinions or inferences upon the subject.”  One of the deeply troubling aspects of the Milward decision is that it reversed the trial court’s sensible decision to exclude a toxicologist, Dr. Martyn Smith, who outran his headlights on issues having to do with a field in which he was clearly inexperienced – epidemiology.

Scientific studies, and especially epidemiologic studies, involve multiple levels of hearsay.  A typical epidemiologic study may contain hearsay leaps from patient to clinician, to laboratory technicians, to specialists interpreting test results, back to the clinician for a diagnosis, to a nosologist for disease coding, to a national or hospital database, to a researcher querying the database, to a statistician analyzing the data, to a manuscript that details data, analyses, and results, to editors and peer reviewers, back to study authors, and on to publication.  Those leaps do not mean that the final results are untrustworthy, only that the study itself is not likely admissible in evidence.

The inadmissibility of scientific studies is not problematic because Rule 703 permits testifying expert witnesses to formulate opinions based upon facts and data, which are not independently admissible in evidence. The distinction between relied upon and admissible studies is codified in the Federal Rules of Evidence, and in virtually every state’s evidence law.

Referring to studies, without qualification, as admissible in themselves is usually wrong as a matter of evidence law.  The error has the potential to encourage carelessness in gatekeeping expert witnesses’ opinions for their reliance upon inadmissible studies.  The error is doubly wrong if this approach to expert witness gatekeeping is taken as license to permit expert witnesses to rely upon any marginally relevant study of their choosing.  It is therefore disconcerting that the RMSE 3d failed to make the appropriate distinction between admissibility of studies and admissibility of expert witness opinion that has reasonably relied upon appropriate studies.

Consider the following statement from the chapter on epidemiology:

“An epidemiologic study that is sufficiently rigorous to justify a conclusion that it is scientifically valid should be admissible, as it tends to make an issue in dispute more or less likely.”[9]

Curiously, the advice from the authors of the epidemiology chapter, by speaking to a single study’s validity, was at odds with Professor Berger’s caution against slicing and dicing. The authors of the epidemiology chapter seemed to be stressing that scientifically valid studies should be admissible.  Their footnote emphasized and confused the point:

See DeLuca v. Merrell Dow Pharms., Inc., 911 F.2d 941, 958 (3d Cir. 1990); cf. Kehm v. Procter & Gamble Co., 580 F. Supp. 890, 902 (N.D. Iowa 1982) (“These [epidemiologic] studies were highly probative on the issue of causation—they all concluded that an association between tampon use and menstrually related TSS [toxic shock syndrome] cases exists.”), aff’d, 724 F.2d 613 (8th Cir. 1984). Hearsay concerns may limit the independent admissibility of the study, but the study could be relied on by an expert in forming an opinion and may be admissible pursuant to Fed. R. Evid. 703 as part of the underlying facts or data relied on by the expert. In Ellis v. International Playtex, Inc., 745 F.2d 292, 303 (4th Cir. 1984), the court concluded that certain epidemiologic studies were admissible despite criticism of the methodology used in the studies. The court held that the claims of bias went to the studies’ weight rather than their admissibility. Cf. Christophersen v. Allied-Signal Corp., 939 F.2d 1106, 1109 (5th Cir. 1991) (“As a general rule, questions relating to the bases and sources of an expert’s opinion affect the weight to be assigned that opinion rather than its admissibility. . . .”).”[10]

This footnote, however, that studies relied upon by an expert in forming an opinion may be admissible pursuant to Rule 703, was unsupported by and contrary to Rule 703 and the overwhelming weight of case law interpreting and applying the rule.[11] The citation to a pre-Daubert decision, Christophersen, was doubtful as a legal argument, and managed to engender much confusion

Furthermore, Kehm and Ellis, the cases cited in this footnote by the authors of the epidemiology chapter, both involved “factual findings” in public investigative or evaluative reports, which were independently admissible under Federal Rule of Evidence 803(8)(C). See Ellis, 745 F.2d at 299-303; Kehm, 724 F.2d at 617-18.  As such, the cases hardly support the chapter’s suggestion that Rule 703 is a rule of admissibility for epidemiologic studies.

Here the RMSE 3d, in one sentence, confused Rule 703 with an exception to the rule against hearsay, which would prevent the statistically based epidemiologic studies from being received in evidence.  The point is reasonably clear, however, that the studies “may be offered” in testimony to explain an expert witness’s opinion. Under Rule 705, that offer may also be refused. The offer, however, is to “explain,” not to have the studies admitted in evidence.  The RMSE 3d was certainly not alone in advancing this notion that studies are themselves admissible.  Other well-respected evidence scholars have lapsed into this error.[12]

Evidence scholars should not conflate admissibility of the epidemiologic (or other) studies with the ability of an expert witness to advert to a study to explain his or her opinion.  The testifying expert witness really should not be allowed to become a conduit for off-hand comments and opinions in the introduction or discussion section of relied upon articles, and the wholesale admission of such hearsay opinions undermines the trial court’s control over opinion evidence.  Rule 703 authorizes reasonable reliance upon “facts and data,” not every opinion that creeps into the published literature.

II. Toxicology for Judges

The toxicology chapter, “Reference Guide on Toxicology,” in RMSE 3d was written by Professor Bernard D. Goldstein, of the University of Pittsburgh Graduate School of Public Health, and Mary Sue Henifin, a partner in the Princeton, New Jersey office of Buchanan Ingersoll, P.C.

  1. Conflicts of Interest

At the question and answer session of the Reference Manual’s public release ceremony, in September 2011, one gentleman rose to note that some of the authors were lawyers with big firm affiliations, which he supposed must mean that they represent mostly defendants.  Based upon his premise, he asked what the review committee had done to ensure that conflicts of interest did not skew or distort the discussions in the affected chapters.  Dr. Kassirer and Judge Kessler responded by pointing out that the chapters were peer reviewed by outside reviewers, and reviewed by members of the supervising review committee.  The questioner seemed reassured, but now that I have looked at the toxicology chapter, I am not so sure.

The questioner’s premise that a member of a large firm will represent mostly defendants and thus have a pro-defense bias was probably a common perception among unsophisticated lay observers.  For instance, some large firms represent insurance companies intent upon denying coverage to product manufacturers.  These counsel for insurance companies often take the plaintiffs’ side of the underlying disputed issue in order to claim an exclusion to the contract of insurance, under a claim that the harm was “expected or intended.”  Similarly, the common perception ignores the reality of lawyers’ true conflict:  although gatekeeping helps the defense lawyers’ clients, it takes away legal work from firms that represent defendants in the litigations that are pretermitted by effective judicial gatekeeping.  Erosion of gatekeeping concepts, however, inures to the benefit of plaintiffs, their counsel, as well as the expert witnesses engaged on behalf of plaintiffs in litigation.

The questioner’s supposition in the case of the toxicology chapter, however, is doubly flawed.  If he had known more about the authors, he would probably not have asked his question.  First, the lawyer author, Ms. Henifin, despite her large firm affiliation, has taken some aggressive positions contrary to the interests of manufacturers.[13]  As for the scientist author of the toxicology chapter, Professor Goldstein, the casual reader of the chapter may want to know that he has testified in any number of toxic tort cases, almost invariably on the plaintiffs’ side.  Unlike the defense lawyer, who loses business revenue, when courts shut down unreliable claims, plaintiffs’ testifying or consulting expert witnesses stand to gain by minimalist expert witness opinion gatekeeping.  Given the economic asymmetries, the reader must thus want to know that Professor Goldstein was excluded as an expert witness in some high-profile toxic tort cases.[14]  There do not appear to be any disclosures of Professor Goldstein’s (or any other scientist author’s) conflicts of interests in RMSE 3d.  Having pointed out this conflict, I would note that financial conflicts of interest are nothing really compared with ideological conflicts of interest, which often propel scientists into service as expert witnesses to advance their political agenda.

  1. Hormesis

One way that ideological conflicts might be revealed is to look for imbalances in the presentation of toxicologic concepts.  Most lawyers who litigate cases that involve exposure-response issues are familiar with the “linear no threshold” (LNT) concept that is used frequently in regulatory risk assessments, and which has metastasized to toxic tort litigation, where LNT often has no proper place.

LNT is a dubious assumption because it claims to “know” the dose response at very low exposure levels in the absence of data.  There is a thin plausibility for LNT for genotoxic chemicals claimed to be carcinogens, but even that plausibility evaporates when one realizes that there are DNA defense and repair mechanisms to genotoxicity, which must first be saturated, overwhelmed, or inhibited, before there can be a carcinogenic response. The upshot is that low exposures that do not swamp DNA repair and tumor suppression proteins will not cause cancer.

Hormesis is today an accepted concept that describes a dose-response relationship that shows a benefit at low doses, but harm at high doses. The toxicology chapter in the Reference Manual has several references to LNT but none to hormesis.  That font of all knowledge, Wikipedia reports that hormesis is controversial, but so is LNT.  This is the sort of imbalance that may well reflect an ideological bias.

One of the leading textbooks on toxicology describes hormesis[15]:

“There is considerable evidence to suggest that some non-nutritional toxic substances may also impart beneficial or stimulatory effects at low doses but that, at higher doses, they produce adverse effects. This concept of “hormesis” was first described for radiation effects but may also pertain to most chemical responses.”

Similarly, the Encyclopedia of Toxicology describes hormesis as an important phenomenon in toxicologic science[16]:

“This type of dose–response relationship is observed in a phenomenon known as hormesis, with one explanation being that exposure to small amounts of a material can actually confer resistance to the agent before frank toxicity begins to appear following exposures to larger amounts.  However, analysis of the available mechanistic studies indicates that there is no single hormetic mechanism. In fact, there are numerous ways for biological systems to show hormetic-like biphasic dose–response relationship. Hormetic dose–response has emerged in recent years as a dose–response phenomenon of great interest in toxicology and risk assessment.”

One might think that hormesis would also be of great interest to federal judges, but they will not learn about it from reading the Reference Manual.

Hormesis research has come into its own.  The International Dose-Response Society, which “focus[es] on the dose-response in the low-dose zone,” publishes a journal, Dose-Response, and a newsletter, BELLE:  Biological Effects of Low Level Exposure.  In 2009, two leading researchers in the area of hormesis published a collection of important papers:  Mark P. Mattson and Edward J. Calabrese, eds., Hormesis: A Revolution in Biology, Toxicology and Medicine (2009).

A check in PubMed shows that LNT has more “hits” than “hormesis” or “hermetic,” but still the latter phrases exceed 1,267 references, hardly insubstantial.  In actuality, there are many more hermetic relationships identified in the scientific literature, which often fails to identify the relationship by the term hormesis or hermetic.[17]

The Reference Manual’s omission of hormesis was regrettable.  Its inclusion of references to LNT but not to hormesis suggests a biased treatment of the subject.

  1. Questionable Substantive Opinions

Readers and litigants would fondly hope that the toxicology chapter would not put forward partisan substantive positions on issues that are currently the subject of active litigation.  Fervently, we would hope that any substantive position advanced would at least be well documented.

For at least one issue, the toxicology chapter disappointed significantly.  Table 1 in the chapter presents a “Sample of Selected Toxicological End Points and Examples of Agents of Concern in Humans.” No documentation or citations are provided for this table.  Most of the exposure agent/disease outcome relationships in the table are well accepted, but curiously at least one agent-disease pair, which is the subject of current litigation, is wildly off the mark:

“Parkinson’s disease and manganese[18]

If the chapter’s authors had looked, they would have found that Parkinson’s disease is almost universally accepted to have no known cause, at least outside court rooms.  They would also have found that the issue has been addressed carefully and the claimed relationship or “concern” has been rejected by the leading researchers in the field (who have no litigation ties).[19]  Table 1 suggests a certain lack of objectivity, and its inclusion of a highly controversial relationship, manganese-Parkinson’s disease, suggests a good deal of partisanship.

  1. When All You Have Is a Hammer, Everything Looks Like a Nail

The substantive area author, Professor Goldstein, is not a physician; nor is he an epidemiologist.  His professional focus on animal and cell research appeared to color and bias the opinions offered in this chapter:[20]

“In qualitative extrapolation, one can usually rely on the fact that a compound causing an effect in one mammalian species will cause it in another species. This is a basic principle of toxicology and pharmacology.  If a heavy metal, such as mercury, causes kidney toxicity in laboratory animals, it is highly likely to do so at some dose in humans.”

Such extrapolations may make sense in regulatory contexts, where precauationary judgments are of interest, but they hardly can be said to be generally accepted in controversies in scientific communities, or in civil actions over actual causation.  There are too many counterexamples to cite, but consider crystalline silica, silicon dioxide.  Silica causes something resembling lung cancer in rats, but not in mice, guinea pigs, or hamsters.  It hardly makes sense to ask juries to decide whether the plaintiff is more like a rat than a mouse.

For a sober second opinion to the toxicology chapter, one may consider the views of some well-known authors:

“Whereas the concordance was high between cancer-causing agents initially discovered in humans and positive results in animal studies (Tomatis et al., 1989; Wilbourn et al., 1984), the same could not be said for the reverse relationship: carcinogenic effects in animals frequently lacked concordance with overall patterns in human cancer incidence (Pastoor and Stevens, 2005).”[21]

III. New Reference Manual’s Uneven Treatment of Causation and of Conflicts of Interest

The third edition of the Reference Manual on Scientific Evidence (RMSE) appeared to get off to a good start in the Preface by Judge Kessler and Dr. Kassirer, when they noted that the Supreme Court mandated federal courts to:

“examine the scientific basis of expert testimony to ensure that it meets the same rigorous standard employed by scientific researchers and practitioners outside the courtroom.”

RMSE at xiii.  The preface faltered, however, on two key issues, causation and conflicts of interest, which are taken up as an introduction to the third edition.

  1. Causation

The authors reported in somewhat squishy terms that causal assessments are judgments:

“Fundamentally, the task is an inferential process of weighing evidence and using judgment to conclude whether or not an effect is the result of some stimulus. Judgment is required even when using sophisticated statistical methods. Such methods can provide powerful evidence of associations between variables, but they cannot prove that a causal relationship exists. Theories of causation (evolution, for example) lose their designation as theories only if the scientific community has rejected alternative theories and accepted the causal relationship as fact. Elements that are often considered in helping to establish a causal relationship include predisposing factors, proximity of a stimulus to its putative outcome, the strength of the stimulus, and the strength of the events in a causal chain.”[22]

The authors left the inferential process as a matter of “weighing evidence,” but without saying anything about how the scientific community does its “weighing.” Language about “proving” causation is also unclear because “proof” in scientific parlance connotes a demonstration, which we typically find in logic or in mathematics. Proving empirical propositions suggests a bar set so high such that the courts must inevitably acquiesce in a very low threshold of evidence.  The question, of course, is how low can and will judges go to admit evidence.

The authors thus introduced hand waving and excuses for why evidence can be weighed differently in court proceedings from the world of science:

“Unfortunately, judges may be in a less favorable position than scientists to make causal assessments. Scientists may delay their decision while they or others gather more data. Judges, on the other hand, must rule on causation based on existing information. Concepts of causation familiar to scientists (no matter what stripe) may not resonate with judges who are asked to rule on general causation (i.e., is a particular stimulus known to produce a particular reaction) or specific causation (i.e., did a particular stimulus cause a particular consequence in a specific instance). In the final analysis, a judge does not have the option of suspending judgment until more information is available, but must decide after considering the best available science.”[23]

But the “best available science” may be pretty crummy, and the temptation to turn desperation into evidence (“well, it’s the best we have now”) is often severe.  The authors of the Preface thus remarkable signalled that “inconclusive” is not a judgment open to judges charged with expert witness gatekeeping.  If the authors truly meant to suggest that judges should go with whatever is dished out as “the best available science,” then they have overlooked the obvious:  Rule 702 opens the door to “scientific, technical, or other specialized knowledge,” not to hunches, suggestive but inconclusive evidence, and wishful thinking about how the science may turn out when further along.  Courts have a choice to exclude expert witness opinion testimony that is based upon incomplete or inconclusive evidence. The authors went fairly far afield to suggest, erroneously, that the incomplete and the inconclusive are good enough and should be admitted.

  1. Conflicts of Interest

Surprisingly, given the scope of the scientific areas covered in the RMSE, the authors discussed conflicts of interest (COI) at some length.  Conflicts of interest are a fact of life in all endeavors, and it is understandable counsel judges and juries to try to identify, assess, and control them.  COIs, however, are weak proxies for unreliability.  The emphasis given here was, however, undue because federal judges were enticed into thinking that they can discern unreliability from COI, when they should be focused on the data, inferences, and analyses.

What becomes fairly clear is that the authors of the Preface set out to use COI as a basis for giving litigation plaintiffs a pass, and for holding back studies sponsored by corporate defendants.

“Conflict of interest manifests as bias, and given the high stakes and adversarial nature of many courtroom proceedings, bias can have a major influence on evidence, testimony, and decisionmaking. Conflicts of interest take many forms and can be based on religious, social, political, or other personal convictions. The biases that these convictions can induce may range from serious to extreme, but these intrinsic influences and the biases they can induce are difficult to identify. Even individuals with such prejudices may not appreciate that they have them, nor may they realize that their interpretations of scientific issues may be biased by them. Because of these limitations, we consider here only financial conflicts of interest; such conflicts are discoverable. Nonetheless, even though financial conflicts can be identified, having such a conflict, even one involving huge sums of money, does not necessarily mean that a given individual will be biased. Having a financial relationship with a commercial entity produces a conflict of interest, but it does not inevitably evoke bias. In science, financial conflict of interest is often accompanied by disclosure of the relationship, leaving to the public the decision whether the interpretation might be tainted. Needless to say, such an assessment may be difficult. The problem is compounded in scientific publications by obscure ways in which the conflicts are reported and by a lack of disclosure of dollar amounts.

Judges and juries, however, must consider financial conflicts of interest when assessing scientific testimony. The threshold for pursuing the possibility of bias must be low. In some instances, judges have been frustrated in identifying expert witnesses who are free of conflict of interest because entire fields of science seem to be co-opted by payments from industry. Judges must also be aware that the research methods of studies funded specifically for purposes of litigation could favor one of the parties. Though awareness of such financial conflicts in itself is not necessarily predictive of bias, such information should be sought and evaluated as part of the deliberations.”[24]

All in all, rather misleading advice.  Financial conflicts are not the only conflicts that can be “discovered.”  Often expert witnesses will have political and organizational alignments, which will show deep-seated ideological alignments with the party for which they are testifying.  For instance, in one silicosis case, an expert witness in the field of history of medicine testified, at an examination before trial, that his father suffered from a silica-related disease.  This witness’s alignment with Marxist historians and his identification with radical labor movements made his non-financial conflicts obvious, although these COI would not necessarily have been apparent from his scholarly publications alone.

How low will the bar be set for discovering COI?  If testifying expert witnesses are relying upon textbooks, articles, essays, will federal courts open the authors/hearsay declarants up to searching discovery of their finances? What really is at stake here is that the issues of accuracy, precision, and reliability are lost in the ad hominem project of discovery COIs.

Also misleading was the suggestion that “entire fields of science seem to be co-opted by payments from industry.”  Do the authors mean to exclude the plaintiffs’ lawyer lawsuit industry, which has become one of the largest rent-seeking organizations, and one of the most politically powerful groups in this country?  In litigations in which I have been involved, I have certainly seen plaintiffs’ counsel, or their proxies – labor unions, federal agencies, or “victim support groups” provide substantial funding for studies.  The Preface authors themselves show an untoward bias by their pointing out industry payments without giving balanced attention to other interested parties’ funding of scientific studies.

The attention to COI was also surprising given that one of the key chapters, for toxic tort practitioners, was written by Dr. Bernard D. Goldstein, who has testified in toxic tort cases, mostly (but not exclusively) for plaintiffs.[25]  In one such case, Makofsky, Dr. Goldstein’s participation was particularly revealing because he was forced to explain why he was willing to opine that benzene caused acute lymphocytic leukemia, despite the plethora of published studies finding no statistically significant relationship.  Dr. Goldstein resorted to the inaccurate notion that scientific “proof” of causation requires 95 percent certainty, whereas he imposed only a 51 percent certainty for his medico-legal testimonial adventures.[26] Dr. Goldstein also attempted to justify the discrepancy from the published literature by adverting to the lower standards used by federal regulatory agencies and treating physicians.  

These explanations were particularly concerning because they reflect basic errors in statistics and in causal reasoning.  The 95 percent derives from the use of the coefficient of confidence in confidence intervals, but the probability involved there is not the probability of the association’s being correct, and it has nothing to do with the probability in the belief that an association is real or is causal.  (Thankfully the RMSE chapter on statistics got this right, but my fear is that judges will skip over the more demanding chapter on statistics and place undue weight on the toxicology chapter.)  The reference to federal agencies (OSHA, EPA, etc.) and to treating physicians was meant, no doubt, to invoke precautionary principle concepts as a justification for some vague, ill-defined, lower standard of causal assessment.  These references were really covert invitations to shift the burden of proof.

The Preface authors might well have taken their own counsel and conducted a more searching assessment of COI among authors of Reference Manual.  Better yet, the authors might have focused the judiciary on the data and the analysis.

  1. Reference Manual on Scientific Evidence (3d edition) on Statistical Significance

How does the new Reference Manual on Scientific Evidence treat statistical significance?  Inconsistently and at times incoherently.

  1. Professor Berger’s Introduction

In her introductory chapter, the late Professor Margaret A. Berger raised the question what role statistical significance should play in evaluating a study’s support for causal conclusions[27]:

“What role should statistical significance play in assessing the value of a study? Epidemiological studies that are not conclusive but show some increased risk do not prove a lack of causation. Some courts find that they therefore have some probative value,62 at least in proving general causation.63

This seems rather backwards.  Berger’s suggestion that inconclusive studies do not prove lack of causation seems nothing more than a tautology. Certainly the failure to rule out causation is not probative of causation. How can that tautology support the claim that inconclusive studies “therefore” have some probative value? Berger’s argument seems obviously invalid, or perhaps text that badly needed a posthumous editor.  And what epidemiologic studies are conclusive?  Are the studies individually or collectively conclusive?  Berger introduced a tantalizing concept, which was not spelled out anywhere in the Manual.

Berger’s chapter raised other, serious problems. If the relied-upon studies are not statistically significant, how should we understand the testifying expert witness to have ruled out random variability as an explanation for the disparity observed in the study or studies?  Berger did not answer these important questions, but her rhetoric elsewhere suggested that trial courts should not look too hard at the statistical support (or its lack) for what expert witness testimony is proffered.

Berger’s citations in support were curiously inaccurate.  Footnote 62 cites the Cook case:

“62. See Cook v. Rockwell Int’l Corp., 580 F. Supp. 2d 1071 (D. Colo. 2006) (discussing why the court excluded expert’s testimony, even though his epidemiological study did not produce statistically significant results).”

Berger’s citation was disturbingly incomplete.[28] The expert witness, Dr. Clapp, in Cook did rely upon his own study, which did not obtain a statistically significant result, but the trial court admitted the expert witness’s testimony; the court denied the Rule 702 challenge to Clapp, and permitted him to testify about a statistically non-significant ecological study. Given that the judgment of the district court was reversed

Footnote 63 is no better:

“63. In re Viagra Prods., 572 F. Supp. 2d 1071 (D. Minn. 2008) (extensive review of all expert evidence proffered in multidistricted product liability case).”

With respect to the concept of statistical significance, the Viagra case centered around the motion to exclude plaintiffs’ expert witness, Gerald McGwin, who relied upon three studies, none of which obtained a statistically significant result in its primary analysis.  The Viagra court’s review was hardly extensive; the court did not report, discuss, or consider the appropriate point estimates in most of the studies, the confidence intervals around those point estimates, or any aspect of systematic error in the three studies.  At best, the court’s review was perfunctory.  When the defendant brought to light the lack of data integrity in McGwin’s own study, the Viagra MDL court reversed itself, and granted the motion to exclude McGwin’s testimony.[29]  Berger’s chapter omitted the cautionary tale of McGwin’s serious, pervasive errors, and how they led to his ultimate exclusion. Berger’s characterization of the review was incorrect, and her failure to cite the subsequent procedural history, misleading.

  1. Chapter on Statistics

The Third Edition’s chapter on statistics was relatively free of value judgments about significance probability, and, therefore, an improvement over Berger’s introduction.  The authors carefully described significance probability and p-values, and explain[30]:

“Small p-values argue against the null hypothesis. Statistical significance is determined by reference to the p-value; significance testing (also called hypothesis testing) is the technique for computing p-values and determining statistical significance.”

Although the chapter conflated the positions often taken to be Fisher’s interpretation of p-values and Neyman’s conceptualization of hypothesis testing as a dichotomous decision procedure, this treatment was unfortunately fairly standard in introductory textbooks.  The authors may have felt that presenting multiple interpretations of p-values was asking too much of judges and lawyers, but the oversimplification invited a false sense of certainty about the inferences that can be drawn from statistical significance.

Kaye and Freedman, however, did offer some important qualifications to the untoward consequences of using significance testing as a dichotomous outcome[31]:

“Artifacts from multiple testing are commonplace. Because research that fails to uncover significance often is not published, reviews of the literature may produce an unduly large number of studies finding statistical significance.111 Even a single researcher may examine so many different relationships that a few will achieve statistical significance by mere happenstance. Almost any large dataset—even pages from a table of random digits—will contain some unusual pattern that can be uncovered by diligent search. Having detected the pattern, the analyst can perform a statistical test for it, blandly ignoring the search effort. Statistical significance is bound to follow.

There are statistical methods for dealing with multiple looks at the data, which permit the calculation of meaningful p-values in certain cases.112 However, no general solution is available, and the existing methods would be of little help in the typical case where analysts have tested and rejected a variety of models before arriving at the one considered the most satisfactory (see infra Section V on regression models). In these situations, courts should not be overly impressed with claims that estimates are significant. Instead, they should be asking how analysts developed their models.113

This important qualification to statistical significance was omitted from the overlapping discussion in the chapter on epidemiology, where it was very much needed.

  1. Chapter on Multiple Regression

The chapter on regression did not add much to the earlier and later discussions.  The author asked rhetorically what is the appropriate level of statistical significance, and answers:

“In most scientific work, the level of statistical significance required to reject the null hypothesis (i.e., to obtain a statistically significant result) is set conventionally at 0.05, or 5%.47

Daniel Rubinfeld, “Reference Guide on Multiple Regression,” in RMSE3d 303, 320.

  1. Chapter on Epidemiology

The chapter on epidemiology[32] mostly muddled the discussion set out in Kaye and Freedman’s chapter on statistics.

“The two main techniques for assessing random error are statistical significance and confidence intervals. A study that is statistically significant has results that are unlikely to be the result of random error, although any criterion for ‘significance’ is somewhat arbitrary. A confidence interval provides both the relative risk (or other risk measure) found in the study and a range (interval) within which the risk likely would fall if the study were repeated numerous times.”

The suggestion that a statistically significant study has results unlikely due to chance, without reminding the reader that the finding is predicated on the assumption that there is no association, and that the probability distribution was correct, and came close to crossing the line in committing the transposition fallacy so nicely described and warned against in the statistics chapter. The problem was that “results” is ambiguous as between the data as extreme or more so than what was observed, and the point estimate of the mean or proportion in the sample, and the assumptions that lead to a p-value were not disclosed.

The suggestion that alpha is “arbitrary,” was “somewhat” correct, but this truncated discussion was distinctly unhelpful to judges who are likely to take “arbitrary“ to mean “I will get reversed.”  The selection of alpha is conventional to some extent, and arbitrary in the sense that the law’s setting an age of majority or a voting age is arbitrary.  Some young adults, age 17.8 years old, may be better educated, better engaged in politics, better informed about current events, than 35 year olds, but the law must set a cut off.  Two year olds are demonstrably unfit, and 82 year olds are surely past the threshold of maturity requisite for political participation. A court might admit an opinion based upon a study of rare diseases, with tight control of bias and confounding, when p = 0.051, but that is hardly a justification for ignoring random error altogether, or admitting an opinion based upon a study, in which the disparity observed had a p = 0.15.

The epidemiology chapter correctly called out judicial decisions that confuse “effect size” with statistical significance[33]:

“Understandably, some courts have been confused about the relationship between statistical significance and the magnitude of the association. See Hyman & Armstrong, P.S.C. v. Gunderson, 279 S.W.3d 93, 102 (Ky. 2008) (describing a small increased risk as being considered statistically insignificant and a somewhat larger risk as being considered statistically significant.); In re Pfizer Inc. Sec. Litig., 584 F. Supp. 2d 621, 634–35 (S.D.N.Y. 2008) (confusing the magnitude of the effect with whether the effect was statistically significant); In re Joint E. & S. Dist. Asbestos Litig., 827 F. Supp. 1014, 1041 (S.D.N.Y. 1993) (concluding that any relative risk less than 1.50 is statistically insignificant), rev’d on other grounds, 52 F.3d 1124 (2d Cir. 1995).”

Actually this confusion is not understandable at all.  The distinction has been the subject of teaching since the first edition of the Reference Manual, and two of the cited cases post-date the second edition.  The Southern District of New York asbestos case, of course, predated the first Manual.  To be sure, courts have on occasion badly misunderstood significance probability and significance testing.   The authors of the epidemiology chapter could well have added In re Viagra, to the list of courts that confused effect size with statistical significance.[34]

The epidemiology chapter appropriately chastised courts for confusing significance probability with the probability that the null hypothesis, or its complement, is correct[35]:

“A common error made by lawyers, judges, and academics is to equate the level of alpha with the legal burden of proof. Thus, one will often see a statement that using an alpha of .05 for statistical significance imposes a burden of proof on the plaintiff far higher than the civil burden of a preponderance of the evidence (i.e., greater than 50%).  See, e.g., In re Ephedra Prods. Liab. Litig., 393 F. Supp. 2d 181, 193 (S.D.N.Y. 2005); Marmo v. IBP, Inc., 360 F. Supp. 2d 1019, 1021 n.2 (D. Neb. 2005) (an expert toxicologist who stated that science requires proof with 95% certainty while expressing his understanding that the legal standard merely required more probable than not). But see Giles v. Wyeth, Inc., 500 F. Supp. 2d 1048, 1056–57 (S.D. Ill. 2007) (quoting the second edition of this reference guide).

Comparing a selected p-value with the legal burden of proof is mistaken, although the reasons are a bit complex and a full explanation would require more space and detail than is feasible here. Nevertheless, we sketch out a brief explanation: First, alpha does not address the likelihood that a plaintiff’s disease was caused by exposure to the agent; the magnitude of the association bears on that question. See infra Section VII. Second, significance testing only bears on whether the observed magnitude of association arose  as a result of random chance, not on whether the null hypothesis is true. Third, using stringent significance testing to avoid false-positive error comes at a complementary cost of inducing false-negative error. Fourth, using an alpha of .5 would not be equivalent to saying that the probability the association found is real is 50%, and the probability that it is a result of random error is 50%.”

The footnotes went on to explain further the difference between alpha probability and burden of proof probability, but somewhat misleadingly asserted that “significance testing only bears on whether the observed magnitude of association arose as a result of random chance, not on whether the null hypothesis is true.”[36]  The significance probability does not address the probability that the observed statistic is the result of random chance; rather it describes the probability of observing at least as large a departure from the expected value if the null hypothesis is true.  Of course, if this cumulative probability is sufficiently low, then the null hypothesis is rejected, and this would seem to bear upon whether the null hypothesis is true.  Kaye and Freedman’s chapter on statistics did much better at describing p-values and avoiding the transposition fallacy.

When they stayed on message, the authors of the epidemiology chapter were certainly correct that significance probability cannot be translated into an assessment of the probability that the null hypothesis, or the obtained sampling statistic, is correct.  What these authors omitted, however, was a clear statement that the many courts and counsel who have misstated this fact do not create any worthwhile precedent, persuasive or binding.

The epidemiology chapter ultimately failed to help judges in assessing statistical significance:

“There is some controversy among epidemiologists and biostatisticians about the appropriate role of significance testing.85 To the strictest significance testers, any study whose p-value is not less than the level chosen for statistical significance should be rejected as inadequate to disprove the null hypothesis. Others are critical of using strict significance testing, which rejects all studies with an observed p-value below that specified level. Epidemiologists have become increasingly sophisticated in addressing the issue of random error and examining the data from a study to ascertain what information they may provide about the relationship between an agent and a disease, without the necessity of rejecting all studies that are not statistically significant.86 Meta-analysis, as well, a method for pooling the results of multiple studies, sometimes can ameliorate concerns about random error.87  Calculation of a confidence interval permits a more refined assessment of appropriate inferences about the association found in an epidemiologic study.88

Id. at 578-79.  Mostly true, but again rather unhelpful to judges and lawyers.  Some of the controversy, to be sure, was instigated by statisticians and epidemiologists who would elevate Bayesian methods, and eliminate the use of significance probability and testing altogether. As for those scientists who still work within the dominant frequentist statistical paradigm, the chapter authors divided the world up into “strict” testers and those critical of “strict” testing.  Where, however, is the boundary? Does criticism of “strict” testing imply embrace of “non-strict” testing, or of no testing at all?  I can sympathize with a judge who permits reliance upon a series of studies that all go in the same direction, with each having a confidence interval that just misses excluding the null hypothesis.  Meta-analysis in such a situation might not just ameliorate concerns about random error, it might eliminate them.  But what of those scientists critical of strict testing?  This certainly does not suggest or imply that courts can or should ignore random error; yet that is exactly what happened in the early going in In re Viagra Products Liab. Litig.[37]  The epidemiology chapter’s reference to confidence intervals was correct in part; they permit a more refined assessment because they permit a more direct assessment of the extent of random error in terms of magnitude of association, as well as the point estimate of the association obtained from and conditioned on the sample.  Confidence intervals, however, do not eliminate the need to interpret the extent of random error; rather they provide a more direct assessment and measurement of the standard error.

V. Power in the Reference Manual for Scientific Evidence

The Third Edition treated statistical power in three of its chapters, those on statistics, epidemiology, and medical testimony.  Unfortunately, the treatments were not always consistent.

The chapter on statistics has been consistently among the most frequently ignored content of the three editions of the Reference Manual.  The third edition offered a good introduction to basic concepts of sampling, random variability, significance testing, and confidence intervals.[38]  Kaye and Freedman provided an acceptable non-technical definition of statistical power[39]:

“More precisely, power is the probability of rejecting the null hypothesis when the alternative hypothesis … is right. Typically, this probability will depend on the values of unknown parameters, as well as the preset significance level α. The power can be computed for any value of α and any choice of parameters satisfying the alternative hypothesis. Frequentist hypothesis testing keeps the risk of a false positive to a specified level (such as α = 5%) and then tries to maximize power. Statisticians usually denote power by the Greek letter beta (β). However, some authors use β to denote the probability of accepting the null hypothesis when the alternative hypothesis is true; this usage is fairly standard in epidemiology. Accepting the null hypothesis when the alternative holds true is a false negative (also called a Type II error, a missed signal, or a false acceptance of the null hypothesis).”

The definition was not, however, without problems.  First, it introduced a nomenclature issue likely to be confusing for judges and lawyers. Kaye and Freeman used β to denote statistical power, but they acknowledge that epidemiologists use β to denote the probability of a Type II error.  And indeed, both the chapters on epidemiology and medical testimony used β to reference Type II error rate, and thus denote power as the complement of β, or (1- β).[40]

Second, the reason for introducing the confusion about β was doubtful.  Kaye and Freeman suggested that statisticians usually denote power by β, but they offered no citations.  A quick review (not necessarily complete or even a random sample) suggests that many modern statistics texts denote power as (1- β).[41]   At the end of the day, there really was no reason for the conflicting nomenclature and the likely confusion it would engenders.  Indeed, the duplicative handling of statistical power, and other concepts, suggested that it is time to eliminate the repetitive discussions, in favor of one, clear, thorough discussion in the statistics chapter.

Third, Kaye and Freeman problematically refer to β as the probability of accepting the null hypothesis when elsewhere they more carefully instructed that a non-significant finding results in not rejecting the null hypothesis as opposed to accepting the null.  Id. at 253.[42]

Fourth, Kaye and Freeman’s discussion of power, unlike most of their chapter, offered advice that is controversial and unclear:

“On the other hand, when studies have a good chance of detecting a meaningful association, failure to obtain significance can be persuasive evidence that there is nothing much to be found.”[43]

Note that the authors left open what a legal or clinically meaningful association is, and thus offered no real guidance to judges on how to evaluate power after data are collected and analyzed.  As Professor Sander Greenland has argued, in legal contexts, this reliance upon observed power (as opposed to power as a guide in determining appropriate sample size in the planning stages of a study) was arbitrary and “unsalvageable as an analytic tool.”[44]

The chapter on epidemiology offered similar controversial advice on the use of power[45]:

“When a study fails to find a statistically significant association, an important question is whether the result tends to exonerate the agent’s toxicity or is essentially inconclusive with regard to toxicity.93 The concept of power can be helpful in evaluating whether a study’s outcome is exonerative or inconclusive.94  The power of a study is the probability of finding a statistically significant association of a given magnitude (if it exists) in light of the sample sizes used in the study. The power of a study depends on several factors: the sample size; the level of alpha (or statistical significance) specified; the background incidence of disease; and the specified relative risk that the researcher would like to detect.95  Power curves can be constructed that show the likelihood of finding any given relative risk in light of these factors. Often, power curves are used in the design of a study to determine what size the study populations should be.96

Although the authors correctly emphasized the need to specify an alternative hypothesis, their discussion and advice were empty of how that alternative should be selected in legal contexts.  The suggestion that power curves can be constructed was, of course, true, but irrelevant unless courts know where on the power curve they should be looking.  The authors were also correct that power is used to determine adequate sample size under specified conditions; but again, the use of power curves in this setting is today rather uncommon.  Investigators select a level of power corresponding to an acceptable Type II error rate, and an alternative hypothesis that would be clinically meaningful for their research, in order to determine their sample size. Translating clinical into legal meaningfulness is not always straightforward.

In a footnote, the authors of the epidemiology chapter noted that Professor Rothman has been “one of the leaders in advocating the use of confidence intervals and rejecting strict significance testing.”[46] What the chapter failed, however, to mention is that Rothman has also been outspoken in rejecting post-hoc power calculations that the epidemiology chapter seemed to invite:

“Standard statistical advice states that when the data indicate a lack of significance, it is important to consider the power of the study to detect as significant a specific alternative hypothesis. The power of a test, however, is only an indirect indicator of precision, and it requires an assumption about the magnitude of the effect. In planning a study, it is reasonable to make conjectures about the magnitude of an effect to compute study-size requirements or power. In analyzing data, however, it is always preferable to use the information in the data about the effect to estimate it directly, rather than to speculate about it with study-size or power calculations (Smith and Bates, 1992; Goodman and Berlin, 1994; Hoening and Heisey, 2001). Confidence limits and (even more so) P-value functions convey much more of the essential information by indicating the range of values that are reasonably compatible with the observations (albeit at a somewhat arbitrary alpha level), assuming the statistical model is correct. They can also show that the data do not contain the information necessary for reassurance about an absence of effect.”[47]

The selective, incomplete scholarship of the epidemiology chapter on the issue of statistical power was not only unfortunate, but it distorted the authors’ evaluation of the sparse case law on the issue of power.  For instance, they noted:

“Even when a study or body of studies tends to exonerate an agent, that does not establish that the agent is absolutely safe. See Cooley v. Lincoln Elec. Co., 693 F. Supp. 2d 767 (N.D. Ohio 2010). Epidemiology is not able to provide such evidence.”[48]

Here the authors, Green, Freedman, and Gordis, shifted the burden to the defendant and then go to an even further extreme of making the burden of proof one of absolute certainty in the product’s safety.  This is not, and never has been, a legal standard. The cases they cited amplified the error. In Cooley, for instance, the defense expert would have opined that welding fume exposure did not cause parkinsonism or Parkinson’s disease.  Although the expert witness had not conducted a meta-analysis, he had reviewed the confidence intervals around the point estimates of the available studies.  Many of the point estimates were at or below 1.0, and in some cases, the upper bound of the confidence interval excluded 1.0.  The trial court expressed its concern that the expert witness had inferred “evidence of absence” from “absence of evidence.”  Cooley v. Lincoln Elec. Co., 693 F. Supp. 2d 767, 773 (N.D. Ohio 2010).  This concern, however, was misguided given that many studies had tested the claimed association, and that virtually every case-control and cohort study had found risk ratios at or below 1.0, or very close to 1.0.  What the court in Cooley, and the authors of the epidemiology chapter in the third edition have lost sight of, is that when the hypothesis is repeatedly tested, with failure to reject the null hypothesis, and with point estimates at or very close to 1.0, and with narrow confidence intervals, then the claimed association is probably incorrect.[49]

The Cooley court’s comments might have had some validity when applied to a single study, but not to the impressive body of exculpatory epidemiologic evidence that pertained to welding fume and Parkinson’s disease.  Shortly after the Cooley case was decided, a published meta-analysis of welding fume or manganese exposure demonstrated a reduced level of risk for Parkinson’s disease among persons occupationally exposed to welding fumes or manganese.[50]

VI. The Treatment of Meta-Analysis in the Third Edition

Meta-analysis is a statistical procedure for aggregating data and statistics from individual studies into a single summary statistical estimate of the population measurement of interest.  The first meta-analysis is typically attributed to Karl Pearson, circa 1904, who sought a method to overcome the limitations of small sample size and low statistical power.  Statistical methods for meta-analysis in epidemiology and the social sciences, however, did not mature until the 1970s.  Even then, the biomedical scientific community remained skeptical of, if not out rightly hostile to, meta-analysis until relatively recently.

The hostility to meta-analysis, especially in the context of observational epidemiologic studies, was colorfully expressed by two capable epidemiologists, Samuel Shapiro and Alvan Feinstein, as late as the 1990s:

“Meta-analysis begins with scientific studies….  [D]ata from these studies are then run through computer models of bewildering complexity which produce results of implausible precision.”

* * * *

“I propose that the meta-analysis of published non-experimental data should be abandoned.”[51]

The professional skepticism about meta-analysis was reflected in some of the early judicial assessments of meta-analysis in court cases.  In the 1980s and early 1990s, some trial judges erroneously dismissed meta-analysis as a flawed statistical procedure that claimed to make something out of nothing.[52]

In In re Paoli Railroad Yard PCB Litigation, Judge Robert Kelly excluded plaintiffs’ expert witness Dr. William Nicholson and his testimony based upon his unpublished meta-analysis of health outcomes among PCB-exposed workers.  Judge Kelly found that the meta-analysis was a novel technique, and that Nicholson’s meta-analysis was not peer reviewed.  Furthermore, the meta-analysis assessed health outcomes not experienced by any of the plaintiffs before the trial court.[53]

The Court of Appeals for the Third Circuit reversed the exclusion of Dr. Nicholson’s testimony, and remanded for reconsideration with instructions.[54]  The Circuit noted that meta-analysis was not novel, and that the lack of peer-review was not an automatic disqualification.  Acknowledging that a meta-analysis could be performed poorly using invalid methods, the appellate court directed the trial court to evaluate the validity of Dr. Nicholson’s work on his meta-analysis. On remand, however, it seems that plaintiffs chose – wisely – not to proceed with Nicholson’s meta-analysis.[55]

In one of many squirmishes over colorectal cancer claims in asbestos litigation, Judge Sweet in the Southern District of New York was unimpressed by efforts to aggregate data across studies.  Judge Sweet declared that:

“no matter how many studies yield a positive but statistically insignificant SMR for colorectal cancer, the results remain statistically insignificant. Just as adding a series of zeros together yields yet another zero as the product, adding a series of positive but statistically insignificant SMRs together does not produce a statistically significant pattern.”[56]

The plaintiffs’ expert witness who had offered the unreliable testimony, Dr. Steven Markowitz, like Nicholson, another foot soldier in Dr. Irving Selikoff’s litigation machine, did not offer a formal meta-analysis to justify his assessment that multiple non-significant studies, taken together, rule out chance as a likely explanation for an aggregate finding of an increased risk.

Judge Sweet was quite justified in rejecting this back of the envelope, non-quantitative meta-analysis.  His suggestion, however, that multiple non-significant studies could never collectively serve to rule out chance as an explanation for an overall increased rate of disease in the exposed groups is completely wrong.  Judge Sweet would have better focused on the validity issues in key studies, the presence of bias and confounding, and the completeness of the proffered meta-analysis.  The Second Circuit reversed the entry of summary judgment, and remanded the colorectal cancer claim for trial.[57]  Over a decade later, with even more accumulated studies and data, the Institute of Medicine found the evidence for asbestos plaintiffs’ colorectal cancer claims to be scientifically insufficient.[58]

Courts continue to go astray with an erroneous belief that multiple studies, all without statistically significant results, cannot yield a statistically significant summary estimate of increased risk.  See, e.g., Baker v. Chevron USA, Inc., 2010 WL 99272, *14-15 (S.D.Ohio 2010) (addressing a meta-analysis by Dr. Infante on multiple myeloma outcomes in studies of benzene-exposed workers).  There were many sound objections to Infante’s meta-analysis, but the suggestion that multiple studies without statistical significance could not yield a summary estimate of risk with statistical significance was not one of them.

In the last two decades, meta-analysis has emerged as an important technique for addressing random variation in studies, as well as some of the limitations of frequentist statistical methods.  In 1980s, articles reporting meta-analyses were rare to non-existent.  In 2009, there were over 2,300 articles with “meta-analysis” in their title, or in their keywords, indexed in the PubMed database of the National Library of Medicine.[59]

The techniques for aggregating data have been studied, refined, and employed extensively in thousands of methods and application papers in the last decade. Consensus guideline papers have been published for meta-analyses of clinical trials as well as observational studies.[60]  Meta-analyses, of observational studies and of randomized clinical trials, routinely are relied upon by expert witnesses in pharmaceutical and so-called toxic tort litigation.[61]

The second edition of the Reference Manual on Scientific Evidence gave very little attention to meta-analysis; the third edition did not add very much to the discussion.  The time has come for the next edition to address meta-analyses, and criteria for their validity or invalidity.

  1. Statistics Chapter

The statistics chapter of the third edition gave scant attention to meta-analysis.  The chapter noted, in a footnote, that there are formal procedures for aggregating data across studies, and that the power of the aggregated data will exceed the power of the individual, included studies.  The footnote then cautioned that meta-analytic procedures “have their own weakness,”[62] without detailing what that weakness is. The time has come to spell out the weaknesses so that trial judges can evaluate opinion testimony based upon meta-analyses.

The glossary at the end of the statistics chapter offers a definition of meta-analysis:

“meta-analysis. Attempts to combine information from all studies on a certain topic. For example, in the epidemiological context, a meta-analysis may attempt to provide a summary odds ratio and confidence interval for the effect of a certain exposure on a certain disease.”[63]

This definition was inaccurate in ways that could yield serious mischief.  Virtually all meta-analyses are, or should be, built upon a systematic review that sets out to collect all available studies on a research issue of interest.  It is a rare meta-analysis, however, that includes “all” studies in its quantitative analysis.  The meta-analytic process involves a pre-specification of inclusionary and exclusionary criteria for the quantitative analysis of the summary estimate of risk.  Those criteria may limit the quantitative analysis to randomized trials, or to analytical epidemiologic studies.  Furthermore, meta-analyses frequently and appropriately have pre-specified exclusionary criteria that relate to study design or quality.

On a more technical note, the offered definition suggests that the summary estimate of risk will be an odds ratio, which may or may not be true.  Meta-analyses of risk ratios may yield summary estimates of risk in terms of relative risk or hazard ratios, or even of risk differences.  The meta-analysis may combine data of means rather than proportions as well.

  1. Epidemiology Chapter

The chapter on epidemiology delved into meta-analysis in greater detail than the statistics chapter, and offered apparently inconsistent advice.  The overall gist of the chapter, however, can perhaps best be summarized by the definition offered in this chapter’s glossary:

“meta-analysis. A technique used to combine the results of several studies to enhance the precision of the estimate of the effect size and reduce the plausibility that the association found is due to random sampling error.  Meta-analysis is best suited to pooling results from randomly controlled experimental studies, but if carefully performed, it also may be useful for observational studies.”[64]

It is now time to tell trial judges what “careful” means in the context of conducting and reporting and relying upon meta-analyses.

The epidemiology chapter appropriately noted that meta-analysis can help address concerns over random error in small studies.[65]  Having told us that properly conducted meta-analyses of observational studies can be helpful, the chapter then proceeded to hedge considerably[66]:

“Meta-analysis is most appropriate when used in pooling randomized experimental trials, because the studies included in the meta-analysis share the most significant methodological characteristics, in particular, use of randomized assignment of subjects to different exposure groups. However, often one is confronted with nonrandomized observational studies of the effects of possible toxic substances or agents. A method for summarizing such studies is greatly needed, but when meta-analysis is applied to observational studies – either case-control or cohort – it becomes more controversial.174 The reason for this is that often methodological differences among studies are much more pronounced than they are in randomized trials. Hence, the justification for pooling the results and deriving a single estimate of risk, for example, is problematic.175

The stated objection to pooling results for observational studies was certainly correct, but many research topics have sufficient studies available to allow for appropriate selectivity in framing inclusionary and exclusionary criteria to address the objection.  The chapter went on to credit the critics of meta-analyses of observational studies.  As they did in the second edition of the Reference Manual, the authors in the third edition repeated their cites to, and quotes from, early papers by John Bailar, who was then critical of such meta-analyses:

“Much has been written about meta-analysis recently and some experts consider the problems of meta-analysis to outweigh the benefits at the present time. For example, John Bailar has observed:

‘[P]roblems have been so frequent and so deep, and overstatements of the strength of conclusions so extreme, that one might well conclude there is something seriously and fundamentally wrong with the method. For the present . . . I still prefer the thoughtful, old-fashioned review of the literature by a knowledgeable expert who explains and defends the judgments that are presented. We have not yet reached a stage where these judgments can be passed on, even in part, to a formalized process such as meta-analysis.’

John Bailar, “Assessing Assessments,” 277 Science 528, 529 (1997).”[67]

Bailar’s subjective preference for “old-fashioned” reviews, which often cherry picked the included studies is, well, “old fashioned.”  More to the point, it is questionable science, and a distinctly minority viewpoint in the light of substantial improvements in the conduct and reporting of systematic reviews and meta-analyses of observational studies.  Bailar may be correct that some meta-analyses should have never left the protocol stage, but the third edition of the Reference Manual failed to provide the judiciary with the tools to appreciate the distinction between good and bad meta-analyses.

This categorical rejection, cited with apparent approval, is amplified by a recitation of some real or apparent problems with meta-analyses of observational studies.  What is missing is a discussion of how many of these problems can be and are dealt with in contemporary practice[68]:

“A number of problems and issues arise in meta-analysis. Should only published papers be included in the meta-analysis, or should any available studies be used, even if they have not been peer reviewed? Can the results of the meta-analysis itself be reproduced by other analysts? When there are several meta-analyses of a given relationship, why do the results of different meta-analyses often disagree? The appeal of a meta-analysis is that it generates a single estimate of risk (along with an associated confidence interval), but this strength can also be a weakness, and may lead to a false sense of security regarding the certainty of the estimate. A key issue is the matter of heterogeneity of results among the studies being summarized.  If there is more variance among study results than one would expect by chance, this creates further uncertainty about the summary measure from the meta-analysis. Such differences can arise from variations in study quality, or in study populations or in study designs. Such differences in results make it harder to trust a single estimate of effect; the reasons for such differences need at least to be acknowledged and, if possible, explained.176 People often tend to have an inordinate belief in the validity of the findings when a single number is attached to them, and many of the difficulties that may arise in conducting a meta-analysis, especially of observational studies such as epidemiologic ones, may consequently be overlooked.177

The epidemiology chapter authors were entitled to their opinion, but their discussion left the judiciary uninformed about current practice, and best practices, in epidemiology.  A categorical rejection of meta-analyses of observational studies is at odds with the chapter’s own claim that such meta-analyses can be helpful if properly performed. What was needed, and is missing, is a meaningful discussion to help the judiciary determine whether a meta-analysis of observational studies was properly performed.

  1. Medical Testimony Chapter

The chapter on medical testimony is the third pass at meta-analysis in the third edition of the Reference Manual.  The second edition’s chapter on medical testimony ignored meta-analysis completely; the new edition addresses meta-analysis in the context of the hierarchy of study designs[69]:

“Other circumstances that set the stage for an intense focus on medical evidence included

(1) the development of medical research, including randomized controlled trials and other observational study designs;

(2) the growth of diagnostic and therapeutic interventions;141

(3) interest in understanding medical decision making and how physicians reason;142 and

(4) the acceptance of meta-analysis as a method to combine data from multiple randomized trials.143

This language from the medical testimony chapter curiously omitted observational studies, but the footnote reference (note 143) then inconsistently discussed two meta-analyses of observational, rather than experimental, studies.[70]  The chapter then provided even further confusion by giving a more detailed listing of the hierarchy of medical evidence in the form of different study designs[71]:

3. Hierarchy of medical evidence

With the explosion of available medical evidence, increased emphasis has been placed on assembling, evaluating, and interpreting medical research evidence.  A fundamental principle of evidence-based medicine (see also Section IV.C.5, infra) is that the strength of medical evidence supporting a therapy or strategy is hierarchical.  When ordered from strongest to weakest, systematic review of randomized trials (meta-analysis) is at the top, followed by single randomized trials, systematic reviews of observational studies, single observational studies, physiological studies, and unsystematic clinical observations.150 An analysis of the frequency with which various study designs are cited by others provides empirical evidence supporting the influence of meta-analysis followed by randomized controlled trials in the medical evidence hierarchy.151 Although they are at the bottom of the evidence hierarchy, unsystematic clinical observations or case reports may be the first signals of adverse events or associations that are later confirmed with larger or controlled epidemiological studies (e.g., aplastic anemia caused by chloramphenicol,152 or lung cancer caused by asbestos153). Nonetheless, subsequent studies may not confirm initial reports (e.g., the putative association between coffee consumption and pancreatic cancer).154

This discussion further muddied the water by using a parenthetical to suggest that meta-analyses of randomized clinical trials are equivalent to systematic reviews of such studies — “systematic review of randomized trials (meta-analysis).” Of course, systematic reviews are not meta-analyses, although they are usually a necessary precondition for conducting a proper meta-analysis.  The relationship between the procedures for a systematic review and a meta-analysis are in need of clarification, but the judiciary will not find it in the third edition of the Reference Manual.

CONCLUSION

The idea of the Reference Manual was important to support trial judge’s efforts to engage in gatekeeping in unfamiliar subject matter areas. In its third incarnation, the Manual has become a standard starting place for discussion, but on several crucial issues, the third edition was unclear, imprecise, contradictory, or muddled. The organizational committee and authors for the fourth edition have a fair amount of work on their hands. There is clearly room for improvement in the Fourth Edition.


[1] Adam Dutkiewicz, “Book Review: Reference Manual on Scientific Evidence, Third Edition,” 28 Thomas M. Cooley L. Rev. 343 (2011); John A. Budny, “Book Review: Reference Manual on Scientific Evidence, Third Edition,” 31 Internat’l J. Toxicol. 95 (2012); James F. Rogers, Jim Shelson, and Jessalyn H. Zeigler, “Changes in the Reference Manual on Scientific Evidence (Third Edition),” Internat’l Ass’n Def. Csl. Drug, Device & Biotech. Comm. Newsltr. (June 2012).  See Schachtman “New Reference Manual’s Uneven Treatment of Conflicts of Interest.” (Oct. 12, 2011).

[2] The Manual did not do quite so well in addressing its own conflicts of interest.  See, e.g., infra at notes 7, 20.

[3] RSME 3d 11 (2011).

[4] Id. at 19.

[5] Id. at 20 & n. 51 (citing Susan Haack, “An Epistemologist in the Bramble-Bush: At the Supreme Court with Mr. Joiner,” 26 J. Health Pol. Pol’y & L. 217–37 (1999).

[6] Id. at 19-20 & n.52.

[7] Professor Berger filed an amicus brief on behalf of plaintiffs, in Rider v. Sandoz Pharms. Corp., 295 F.3d 1194 (11th Cir. 2002).

[8] Id. at 20 n.51. (The editors noted misleadingly that the published chapter was Berger’s last revision, with “a few edits to respond to suggestions by reviewers.”). I have written elsewhere of the ethical cloud hanging over this Milward decision. SeeCarl Cranor’s Inference to the Best Explanation” (Feb. 12, 2021); “From here to CERT-ainty” (June 28, 2018); “The Council for Education & Research on Toxics” (July 9, 2013) (CERT amicus brief filed without any disclosure of conflict of interest). See also NAS, “Carl Cranor’s Conflicted Jeremiad Against Daubert” (Sept. 23, 2018).

[9] RMSE 3d at 610 (internal citations omitted).

[10] RMSE 3d at 610 n.184 (emphasis, in bold, added).

[11] Interestingly, the authors of this chapter seem to abandon their suggestion that studies relied upon “might qualify for the learned treatise exception to the hearsay rule, Fed. R. Evid. 803(18), or possibly the catchall exceptions, Fed. R. Evid. 803(24) & 804(5),” which was part of their argument in the Second Edition.  RMSE 2d at 335 (2000).  See also RMSE 3d at 214 (discussing statistical studies as generally “admissible,” but acknowledging that admissibility may be no more than permission to explain the basis for an expert’s opinion, which is hardly admissibility at all).

[12] David L. Faigman, et al., Modern Scientific Evidence:  The Law and Science of Expert Testimony v.1, § 23:1,at 206 (2009) (“Well conducted studies are uniformly admitted.”).

[13] See Richard M. Lynch and Mary S. Henifin, “Causation in Occupational Disease: Balancing Epidemiology, Law and Manufacturer Conduct,” 9 Risk: Health, Safety & Environment 259, 269 (1998) (conflating distinct causal and liability concepts, and arguing that legal and scientific causal criteria should be abrogated when manufacturing defendant has breached a duty of care).

[14]  See, e.g., Parker v. Mobil Oil Corp., 7 N.Y.3d 434, 857 N.E.2d 1114, 824 N.Y.S.2d 584 (2006) (dismissing leukemia (AML) claim based upon claimed low-level benzene exposure from gasoline), aff’g 16 A.D.3d 648 (App. Div. 2d Dep’t 2005).  No; you will not find the Parker case cited in the Manual‘s chapter on toxicology. (Parker is, however, cited in the chapter on exposure science even though it is a state court case.).

[15] Curtis D. Klaassen, Casarett & Doull’s Toxicology: The Basic Science of Poisons 23 (7th ed. 2008) (internal citations omitted).

[16] Philip Wexler, Bethesda, et al., eds., 2 Encyclopedia of Toxicology 96 (2005).

[17] See Edward J. Calabrese and Robyn B. Blain, “The hormesis database: The occurrence of hormetic dose responses in the toxicological literature,” 61 Regulatory Toxicology and Pharmacology 73 (2011) (reviewing about 9,000 dose-response relationships for hormesis, to create a database of various aspects of hormesis).  See also Edward J. Calabrese and Robyn B. Blain, “The occurrence of hormetic dose responses in the toxicological literature, the hormesis database: An overview,” 202 Toxicol. & Applied Pharmacol. 289 (2005) (earlier effort to establish hormesis database).

[18] Reference Manual at 653

[19] See e.g., Karin Wirdefeldt, Hans-Olaf Adami, Philip Cole, Dimitrios Trichopoulos, and Jack Mandel, “Epidemiology and etiology of Parkinson’s disease: a review of the evidence.  26 European J. Epidemiol. S1, S20-21 (2011); Tomas R. Guilarte, “Manganese and Parkinson’s Disease: A Critical Review and New Findings,” 118 Environ Health Perspect. 1071, 1078 (2010) (“The available evidence from human and non­human primate studies using behavioral, neuroimaging, neurochemical, and neuropathological end points provides strong sup­port to the hypothesis that, although excess levels of [manganese] accumulation in the brain results in an atypical form of parkinsonism, this clini­cal outcome is not associated with the degen­eration of nigrostriatal dopaminergic neurons as is the case in PD [Parkinson’s disease].”)

[20] RMSE3ed at 646.

[21] Hans-Olov Adami, Sir Colin L. Berry, Charles B. Breckenridge, Lewis L. Smith, James A. Swenberg, Dimitrios Trichopoulos, Noel S. Weiss, and Timothy P. Pastoor, “Toxicology and Epidemiology: Improving the Science with a Framework for Combining Toxicological and Epidemiological Evidence to Establish Causal Inference,” 122 Toxciological Sciences 223, 224 (2011).

[22] RMSE3d at xiv.

[23] RMSE3d at xiv.

[24] RMSE3d at xiv-xv.

[25] See, e.g., Parker v. Mobil Oil Corp., 7 N.Y.3d 434, 857 N.E.2d 1114, 824 N.Y.S.2d 584 (2006); Exxon Corp. v. Makofski, 116 SW 3d 176 (Tex. Ct. App. 2003).

[26] Goldstein here and elsewhere has confused significance probability with the posterior probability required by courts and scientists.

[27] Margaret A. Berger, “The Admissibility of Expert Testimony,” in RMSE3d 11, 24 (2011).

[28] Cook v. Rockwell Int’l Corp., 580 F. Supp. 2d 1071, 1122 (D. Colo. 2006), rev’d and remanded on other grounds, 618 F.3d 1127 (10th Cir. 2010), cert. denied, ___ U.S. ___ (May 24, 2012).

[29] In re Viagra Products Liab. Litig., 658 F. Supp. 2d 936, 945 (D. Minn. 2009). 

[31] Id. at 256 -57.

[32] Michael D. Green, D. Michal Freedman, and Leon Gordis, “Reference Guide on Epidemiology,” in RMSE3d 549, 573.

[33] Id. at 573n.68.

[34] See In re Viagra Products Liab. Litig., 572 F. Supp. 2d 1071, 1081 (D. Minn. 2008).

[35] RSME3d at 577 n81.

[36] Id.

[37] 572 F. Supp. 2d 1071, 1081 (D. Minn. 2008).

[38] David H. Kaye & David A. Freedman, “Reference Guide on Statistics,” in RMSE3ed 209 (2011).

[39] Id. at 254 n.106

[40] See Michael D. Green, D. Michal Freedman, and Leon Gordis, “Reference Guide on Epidemiology,” in RMSE3ed 549, 582, 626 ; John B. Wong, Lawrence O. Gostin, and Oscar A. Cabrera, Abogado, “Reference Guide on Medical Testimony,” in RMSE3ed 687, 724.  This confusion in nomenclature is regrettable, given the difficulty many lawyers and judges seem have in following discussions of statistical concepts.

[41] See, e.g., Richard D. De Veaux, Paul F. Velleman, and David E. Bock, Intro Stats 545-48 (3d ed. 2012); Rand R. Wilcox, Fundamentals of Modern Statistical Methods 65 (2d ed. 2010).

[42] See also Daniel Rubinfeld, “Reference Guide on Multiple Regression,“ in RMSE3d 303, 321 (describing a p-value > 5% as leading to failing to reject the null hypothesis).

[43] RMSE3d at 254.

[44] See Sander Greenland, “Nonsignificance Plus High Power Does Not Imply Support Over the Alternative,” 22 Ann. Epidemiol. 364, 364 (2012).

[45] Michael D. Green, D. Michal Freedman, and Leon Gordis, “Reference Guide on Epidemiology,” RMSE3ed 549, 582.

[46] RMSE3d at 579 n.88.

[47] Kenneth Rothman, Sander Greenland, and Timothy Lash, Modern Epidemiology 160 (3d ed. 2008).  See also Kenneth J. Rothman, “Significance Questing,” 105 Ann. Intern. Med. 445, 446 (1986) (“[Simon] rightly dismisses calculations of power as a weak substitute for confidence intervals, because power calculations address only the qualitative issue of statistical significance and do not take account of the results already in hand.”).

[48] RMSE3d at 582 n.93; id. at 582 n.94 (“Thus, in Smith v. Wyeth-Ayerst Labs. Co., 278 F.Supp. 2d 684, 693 (W.D.N.C. 2003), and Cooley v. Lincoln Electric Co., 693 F. Supp. 2d 767, 773 (N.D. Ohio 2010), the courts recognized that the power of a study was critical to assessing whether the failure of the study to find a statistically significant association was exonerative of the agent or inconclusive.”).

[49] See, e.g., Anthony J. Swerdlow, Maria Feychting, Adele C. Green, Leeka Kheifets, David A. Savitz, International Commission for Non-Ionizing Radiation Protection Standing Committee on Epidemiology, “Mobile Phones, Brain Tumors, and the Interphone Study: Where Are We Now?” 119 Envt’l Health Persp. 1534, 1534 (2011) (“Although there remains some uncertainty, the trend in the accumulating evidence is increasingly against the hypothesis that mobile phone use can cause brain tumors in adults.”).

[50] James Mortimer, Amy Borenstein, and Lorene Nelson, “Associations of welding and manganese exposure with Parkinson disease: Review and meta-analysis,” 79 Neurology 1174 (2012).

[51] Samuel Shapiro, “Meta-analysis/Smeta-analysis,” 140 Am. J. Epidem. 771, 777 (1994).  See also Alvan Feinstein, “Meta-Analysis: Statistical Alchemy for the 21st Century,” 48 J. Clin. Epidem. 71 (1995).

[52] Allen v. Int’l Bus. Mach. Corp., No. 94-264-LON, 1997 U.S. Dist. LEXIS 8016, at *71–*74 (suggesting that meta-analysis of observational studies was controversial among epidemiologists).

[53] 706 F. Supp. 358, 373 (E.D. Pa. 1988).

[54] In re Paoli R.R. Yard PCB Litig., 916 F.2d 829, 856-57 (3d Cir. 1990), cert. denied, 499 U.S. 961 (1991); Hines v. Consol. Rail Corp., 926 F.2d 262, 273 (3d Cir. 1991).

[55] SeeThe Shmeta-Analysis in Paoli,” (July 11, 2019).

[56] In re Joint E. & S. Dist. Asbestos Litig., 827 F. Supp. 1014, 1042 (S.D.N.Y. 1993).

[57] 52 F.3d 1124 (2d Cir. 1995).

[58] Institute of Medicine, Asbestos: Selected Cancers (Wash. D.C. 2006).

[59] See Michael O. Finkelstein and Bruce Levin, “Meta-Analysis of ‘Sparse’ Data: Perspectives from the Avandia CasesJurimetrics J. (2011).

[60] See Donna Stroup, et al., “Meta-analysis of Observational Studies in Epidemiology: A Proposal for Reporting,” 283 J. Am. Med. Ass’n 2008 (2000) (MOOSE statement); David Moher, Deborah Cook, Susan Eastwood, Ingram Olkin, Drummond Rennie, and Donna Stroup, “Improving the quality of reports of meta-analyses of randomised controlled trials: the QUOROM statement,” 354 Lancet 1896 (1999).  See also Jesse Berlin & Carin Kim, “The Use of Meta-Analysis in Pharmacoepidemiology,” in Brian Strom, ed., Pharmacoepidemiology 681, 683–84 (4th ed. 2005); Zachary Gerbarg & Ralph Horwitz, “Resolving Conflicting Clinical Trials: Guidelines for Meta-Analysis,” 41 J. Clin. Epidemiol. 503 (1988).

[61] See Finkelstein & Levin, supra at note 59. See also In re Bextra and Celebrex Marketing Sales Practices and Prod. Liab. Litig., 524 F. Supp. 2d 1166, 1174, 1184 (N.D. Cal. 2007) (holding that reliance upon “[a] meta-analysis of all available published and unpublished randomized clinical trials” was reasonable and appropriate, and criticizing the expert witnesses who urged the complete rejection of meta-analysis of observational studies).

[62] RMSE 3d at 254 n.107.

[63] Id. at 289.

[64] Reference Guide on Epidemiology, RSME3d at 624.  See also id. at 581 n. 89 (“Meta-analysis is better suited to combining results from randomly controlled experimental studies, but if carefully performed it may also be helpful for observational studies, such as those in the epidemiologic field.”).

[65] Id. at 579; see also id. at 607 n. 171.

[66] Id. at 607.

[67] Id. at 607 n.177.

[68] Id. at 608.

[69] RMSE 3d at 722-23.

[70] Id. at 723 n.143 (“143. … Video Software Dealers Ass’n v. Schwarzenegger, 556 F.3d 950, 963 (9th Cir. 2009) (analyzing a meta-analysis of studies on video games and adolescent behavior); Kennecott Greens Creek Min. Co. v. Mine Safety & Health Admin., 476 F.3d 946, 953 (D.C. Cir. 2007) (reviewing the Mine Safety and Health Administration’s reliance on epidemiological studies and two meta-analyses).”).

[71] Id. at 723-24.