Zantac MDL Decision Highlights Need for Rigorous and Objective Approach in Bradford Hill Analyses

As detailed in our prior post, the U.S. District Court for the Southern District of Florida recently granted summary judgment against more than 2,450 plaintiffs with cases pending in the Zantac MDL, ruling that they had failed to produce reliable expert testimony necessary to support their claims under Rule 702.^[1] As part of its decision, the court excluded the opinions of Plaintiffs’ epidemiology experts for failing to conduct appropriate Bradford Hill analyses—a commonly used methodology in assessing causality in products liability and mass tort proceedings. As part of our continuing commentary on the Zantac decision, this review examines the court’s rationale for doing so and highlights potential avenues for Daubert challenges to experts conducting Bradford Hill analyses.

The Bradford Hill Criteria

The Bradford Hill criteria, also known as Hill’s criteria, are a set of nine principles used to determine causality between an exposure and an outcome in epidemiology. The criteria were first proposed by the British epidemiologist Sir Austin Bradford Hill in 1965 to evaluate the strength of evidence for a causal relationship between two variables.^[2] The criteria are:

1. Strength of association: Refers to the degree of the relationship between the exposure and the outcome, typically provided as a measure of increased or decreased risk, such risk ratios. The stronger the association, the more likely it is that the exposure is causally related to the outcome. For further discussion on the interplay between risk ratios, the Bradford Hill criteria, and general causation, please see our prior analysis.
2. Temporality: Refers to the order in which the exposure and the outcome occur. If the exposure does not precede the outcome, it cannot be causally related.
3. Consistency: Refers to whether different persons, in different places and at different times, have repeatedly observed the relationship.
4. Biological gradient: Refers to the dose-response relationship between the exposure and the outcome. If increasing exposure is associated with increased risk of the outcome, this can support a causal relationship.
5. Coherence: Refers to the consistency of the relationship between the exposure and the outcome with the existing knowledge. If the relationship tracks what is already known on the natural history and biology of the outcome, it is more likely to be causally related.
6. Biological plausibility: Refers to the biological mechanisms that may explain the relationship. If there is a plausible biological explanation for the relationship, it is more likely to be causally related.
7. Analogy: Refers to similarities between the relationship under investigation and relationships that have been established previously. If the relationship is similar to already-established relationships, it is more likely to be causally related.
8. Experimental evidence: Refers to the results of experimental studies that examine the relationship. If the relationship is supported by experimental evidence, it is more likely to be causally related.
9. Specificity: Refers to the specificity of the relationship between the exposure and the outcome to a particular disease or condition. If the relationship is specific to a particular outcome, it is more likely to be causally related.

It is important to note that the Bradford Hill criteria are not a strict set of rules, and no single criterion, by itself, is enough to establish causality.^[3] Rather, they provide a framework for considering the evidence and making an overall assessment of causality.

The Zantac Decision

On a more general level, the court explained that because Bradford Hill analyses involve review of “various qualities of a data set—such as strength or consistency,” along with the “credit[ing] or discount[ing] of the data selected for input,” courts often “describe these methodologies as ‘flexible methodologies.’”^[4] In other words, Bradford Hill is a “flexible” methodology because it “can be implemented in multiple ways.”^[5] Thus, even though “as a general matter” Bradford Hill analyses are “standard practice in epidemiology” and can be a reliable methodology, the court observed that “the reality is that every case-specific application” of the Bradford Hill criteria “is distinct and should be analyzed for reliability.”^[6]

The court also observed that for a Bradford Hill analysis to be reliable and admissible, “there must be a scientific method of weighting that is used and explained.”^[7] On the other hand, an expert who “does not explain the amount of weight that was assigned to each piece of evidence” engages in “in a conclusion-oriented process,” which the court noted is a “malleable and vague approach” in “tension with the first principles under Daubert.”^[8] Stated otherwise, engaging in a “conclusion-oriented process” makes it “all too easy for an expert to manipulate the Bradford Hill factors to support a desired conclusion of causation, and far too hard for an ensuing expert to replicate and rigorously test the expert’s analytic approach.”^[9]

Here, Plaintiffs’ two epidemiology experts employed the Bradford Hill criteria to determine whether ranitidine use and cancer were causally related. However, the court found that their application of the Hill criteria was lacking in key respects, thus justifying exclusion of their opinions.

As for the first of Plaintiffs’ epidemiologists, the court recognized that while her analysis “weigh[ed] how strongly the various [Bradford Hill] factors influenced her conclusion (such as assigning high weight to her observation of consistency of an association),” the actual discussion of the factors “[wa]s general, brief, and lacking in specificity as to how the data influenced her analysis.”^[10] In particular, the expert did not “identify many of the studies she weighs or credits by name.”^[11] As a result, the court “struggled to identify the studies that served as inputs into [Plaintiffs’ expert’s] final analysis.”^[12] For example, the expert stated that liver cancer risk was “highly apparent in more studies than not,” but did not specify which studies underpinned that statement.^[13]

Critically, the expert also “omit[ted] the weight she assigns to the study she relies upon,”^[14] and the court therefore could not “even say for certain (for some cancers) what studies influence[d]” the expert’s “ultimate conclusion.”^[15] The court also noted that the expert’s deposition testimony did not provide any more clarity, as the expert “routinely evaded germane, fair questions that would have clarified how [she] conduct[ed] her analysis.”^[16]

The Court similarly held that Plaintiffs’ second epidemiologist did “not select the data inputs for her Bradford Hill analysis and reach her ultimate conclusions based upon objective, scientific criteria” for four reasons.^[17] First, the expert’s weighting of studies “such as the length of time in databases, sample sizes, and good ascertainment of exposure” was inconsistently applied across the body of literature considered in her review.^[18] As an example, the court noted that it did not believe that objective application of the expert’s criteria “could result in a dietary study of Uruguayan men’s consumption of meat. . .being assigned greater weight than almost all of the available epidemiology on ranitidine.”^[19]

Second, the expert “omit[ted] and all but ignore[d] the study authors’ conclusions in the studies she relies upon, the majority of which contradict her opinion and, at the very least, do not support her opinion.”^[20] Third, the expert selectively relied on “data from studies that fit with her ultimate opinion while simultaneously ignoring data in the very same studies that do not fit with her opinions.”^[21] Fourth and lastly, the court found that the expert’s opinions and criticisms about individual studies were both “internally inconsistent,” and inconsistent with Plaintiffs’ other epidemiologist.^[22] Thus, after “[r]eviewing [the expert’s] opinion in its totality,” the court found that the expert “utilize[d] a conclusion-oriented process to select the inputs (to assign weights to the various studies) into her Bradford Hill analysis.”^[23]

In sum, the Zantac decision makes clear that experts engaging in a proper Bradford Hill analysis must detail both the “data inputs” (i.e., scientific studies) used in their analysis and the application of the weighting criteria used to assess that data. Put another way, an expert must explain a case-specific Bradford Hill analysis with sufficient specificity “for an ensuing expert to replicate and rigorously test the expert’s analytic approach.”^[24] Additionally, selective use of only certain data supporting the expert’s Bradford Hill analysis, while “simultaneously ignoring data in the very same studies,” is a “conclusion-oriented” methodology that justifies exclusion of those opinions.^[25] Likewise, inconsistent application of an expert’s weighting criteria—or weighting certain studies in a manner contrary to the conclusions reached by the studies’ own authors—goes against standard scientific practice and a proper Bradford Hill analysis.^[26]

Strategic Takeaways

Companies defending product liability and mass tort lawsuits should carefully analyze plaintiff experts employing Bradford Hill analyses for any deficiencies, such as those identified in the Zantac decision, that may merit exclusion of their opinions.

---

^{[1] In re Zantac (Ranitidine) Prod. Liab. Litig., No. 20-MD-2924, ---F. Supp. 3d---, 2022 WL 17480906, at *4, *6, *167 (S.D. Fla. Dec. 6, 2022).}

^{[2] Hill, A.B., The Environment and Disease: Association or Causation? Proc R Soc Med. 1965 May; 58(5):295–300.}

^{[3] See id. at 299 (“None of my nine viewpoints can bring indisputable evidence for or against the cause-and-effect hypothesis and none can be required as a sine qua non.”).}

^{[4] In re Zantac, 2022 WL 17480906, at *126 (citation omitted).}

^{[5] Id.}

^{[6] Id. (quotations omitted).}

^{[7] Id. (citation omitted).}

^{[8] Id. (quotations omitted).}

^{[9] Id. (citation omitted).}

^{[10] Id. at *125.}

^{[11] Id. at *122.}

^{[12] Id.}

^{[13] Id. at *125.}

^{[14] Id. at *113.}

^{[15] Id. at *126.}

^{[16] Id. at *122.}

^{[17] Id. at *138.}

^{[18] Id.}

^{[19] Id.}

^{[20] Id. at *139.}

^{[21] Id.}

^{[22] Id.}

^{[23] Id.}

^{[24] Id. at *126 (citation omitted).}

^{[25] Id. at *139.}

^{[26] See id.}