The incretin-mimetic litigation involved claims that the use of Byetta, Januvia, Janumet, and Victoza medications causes pancreatic cancer. All four medications treat diabetes mellitus through incretin hormones, which stimulate or support insulin production, which in turn lowers blood sugar. On Planet Earth, the only scientists who contend that these medications cause pancreatic cancer are those hired by the lawsuit industry.
The cases against the manufacturers of the incretin-mimetic medications were consolidated for pre-trial proceedings in federal court, pursuant to the multi-district litigation (MDL) statute, 28 US Code § 1407. After years of MDL proceedings, the trial court dismissed the cases as barred by the doctrine of federal preemption, and for good measure, excluded plaintiffs’ medical causation expert witnesses from testifying.[1] If there were any doubt about the false claiming in this MDL, the district court’s dismissals were affirmed by the Ninth Circuit.[2]
The district court’s application of Federal Rule of Evidence 702 to the plaintiffs’ expert witnesses’ opinion is an important essay in patho-epistemology. The challenged expert witnesses provided many examples of invalid study design and interpretation. Of particular interest, two of the plaintiffs’ high-volume statistician testifiers, David Madigan and Martin Wells, proffered their own meta-analyses of clinical trial safety data. Although the current edition of the Reference Manual on Scientific Evidence[3] provides virtually no guidance to judges for assessing the validity of meta-analyses, judges and counsel do now have other readily available sources, such as the FDA’s Guidance on meta-analysis of safety outcomes of clinical trials.[4] Luckily for the Incretin-Mimetics pancreatic cancer MDL judge, the misuse of meta-analysis methodology by plaintiffs’ statistician expert witnesses, David Madigan and Martin Wells was intuitively obvious.
Madigan and Wells had a large set of clinical trials at their disposal, with adverse safety outcomes assiduously collected. As is the case with many clinical trial safety outcomes, the trialists will often have a procedure for blinded or unblinded adjudication of safety events, such as pancreatic cancer diagnosis.
At deposition, Madigan testified that he counted only adjudicated cases of pancreatic cancer in his meta-analyses, which seems reasonable enough. As discovery revealed, however, Madigan employed the restrictive inclusion criteria of adjudicated pancreatic cancer only to the placebo group, not to the experimental group. His use of restrictive inclusion criteria for only the placebo group had the effect of excluding several non-adjudicated events, with the obvious spurious inflation of risk ratios. The MDL court thus found with relative ease that Madigan’s “unequal application of criteria among the two groups inevitably skews the data and critically undermines the reliability of his analysis.” The unexplained, unjustified change in methodology revealed Madigan’s unreliable “cherry-picking” and lack of scientific rigor as producing a result-driven meta-analyses.[5]
The MDL court similarly found that Wells’ reports “were marred by a selective review of data and inconsistent application of inclusion criteria.”[6] Like Madigan, Wells cherry picked studies. For instance, he excluded one study, EXSCEL, on grounds that it reported “a high pancreatic cancer event rate in the comparison group as compared to background rate in the general population….”[7] Wells’ explanation blatantly failed, however, given that the entire patient population of the clinical trial had diabetes, a known risk factor for pancreatic cancer.[8]
As Professor Ioannidis and others have noted, we are awash in misleading meta-analyses:
“Currently, there is massive production of unnecessary, misleading, and conflicted systematic reviews and meta-analyses. Instead of promoting evidence-based medicine and health care, these instruments often serve mostly as easily produced publishable units or marketing tools. Suboptimal systematic reviews and meta-analyses can be harmful given the major prestige and influence these types of studies have acquired. The publication of systematic reviews and meta-analyses should be realigned to remove biases and vested interests and to integrate them better with the primary production of evidence.”[9]
Whether created for litigation, like the Madigan-Wells meta-analyses, or published in the “peer-reviewed” literature, courts will have to up their game in assessing the validity of such studies. Published meta-analyses have grown exponentially from the 1990s to the present. To date, 248,886 meta-analyses have been published, according the National Library of Medicine’s Pub-Med database. Last year saw over 35,000 meta-analyses published. So far, this year, 20,416 meta-analyses have been published, and we appear to be on track to have a bumper crop.
The data analytics from Pub-Med provide a helpful visual representation of the growth of meta-analyses in biomedical science.
Count of Publications with Keyword Meta-analysis in Pub-Med Database
In 1979, the year I started law school, one meta-analysis was published. Lawyers could still legitimately argue that meta-analyses involved novel methodology that had not been generally accepted. The novelty of meta-analysis wore off sometime between 1988, when Judge Robert Kelly excluded William Nicholson’s meta-analysis of health outcomes among PCB-exposed workers, on grounds that such analyses were “novel,” and 1990, when the Third Circuit reversed Judge Kelly, with instructions to assess study validity.[10] Fortunately, or not, depending upon your point of view, plaintiffs dropped Nicholson’s meta-analysis in subsequent proceedings. A close look at Nicholson’s non-peer reviewed calculations shows that he failed to standardize for age or sex, and that he merely added observed and expected cases, across studies, without weighting by individual study variance. The trial court never had the opportunity to assess the validity vel non of Nicholson’s ersatz meta-analysis.[11] Today, trial courts must pick up on the challenge of assessing study validity of meta-analyses relied upon by expert witnesses, regulatory agencies, and systematic reviews.
[1] In re Incretin-Based Therapies Prods. Liab. Litig., 524 F.Supp.3d 1007 (S.D. Cal. 2021).
[2] In re Incretin-Based Therapies Prods. Liab. Litig., No. 21-55342, 2022 WL 898595 (9th Cir. Mar. 28, 2022) (per curiam)
[3] “The Treatment of Meta-Analysis in the Third Edition of the Reference Manual on Scientific Evidence” (Nov. 15, 2011).
[4] Food and Drug Administration, Center for Drug Evaluation and Research, “Meta-Analyses of Randomized Controlled Clinical Trials to Evaluate the Safety of Human Drugs or Biological Products – (Draft) Guidance for Industry” (Nov. 2018); Jonathan J. Deeks, Julian P.T. Higgins, Douglas G. Altman, “Analysing data and undertaking meta-analyses,” Chapter 10, in Julian P.T. Higgins, James Thomas, Jacqueline Chandler, Miranda Cumpston, Tianjing Li, Matthew J. Page, and Vivian Welch, eds., Cochrane Handbook for Systematic Reviews of Interventions (version 6.3 updated February 2022); Donna F. Stroup, Jesse A. Berlin, Sally C. Morton, Ingram Olkin, G. David Williamson, Drummond Rennie, David Moher, Betsy J. Becker, Theresa Ann Sipe, Stephen B. Thacker, “Meta-Analysis of Observational Studies: A Proposal for Reporting,” 283 J. Am. Med. Ass’n 2008 (2000); David Moher, Alessandro Liberati, Jennifer Tetzlaff, and Douglas G Altman, “Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement,” 6 PLoS Med e1000097 (2009).
[5] In re Incretin-Based Therapies Prods. Liab. Litig., 524 F.Supp.3d 1007, 1037 (S.D. Cal. 2021). See In re Lipitor (Atorvastatin Calcium) Mktg., Sales Practices & Prods. Liab. Litig. (No. II) MDL2502, 892 F.3d 624, 634 (4th Cir. 2018) (“Result-driven analysis, or cherry-picking, undermines principles of the scientific method and is a quintessential example of applying methodologies (valid or otherwise) in an unreliable fashion.”).
[6] In re Incretin-Based Therapies Prods. Liab. Litig., 524 F.Supp.3d 1007, 1043 (S.D. Cal. 2021).
[7] Id. at 1038.
[8] See, e.g., Albert B. Lowenfels & Patrick Maisonneuve, “Risk factors for pancreatic cancer,” 95 J. Cellular Biochem. 649 (2005).
[9] John P. Ioannidis, “The mass production of redundant, misleading, and conflicted systematic reviews and meta-analyses,” 94 Milbank Quarterly 485 (2016).
[10] In re Paoli R.R. Yard PCB Litig., 706 F. Supp. 358, 373 (E.D. Pa. 1988), rev’d and remanded, 916 F.2d 829, 856-57 (3d Cir. 1990), cert. denied, 499 U.S. 961 (1991). See also Hines v. Consol. Rail Corp., 926 F.2d 262, 273 (3d Cir. 1991).
[11] “The Shmeta-Analysis in Paoli” (July 11, 2019). See James A. Hanley, Gilles Thériault, Ralf Reintjes and Annette de Boer, “Simpson’s Paradox in Meta-Analysis,” 11 Epidemiology 613 (2000); H. James Norton & George Divine, “Simpson’s paradox and how to avoid it,” Significance 40 (Aug. 2015); George Udny Yule, “Notes on the theory of association of attributes in statistics,” 2 Biometrika 121 (1903).
