Second Circuit: Daubert Gatekeeping Cannot Displace Scientific Disagreement in Bradford Hill–Based General Causation; OTC Pregnancy Warning Rules Do Not Preempt Supplemental Risk Warnings

I. Introduction

Rutledge v. Walgreen Co. and Phippen v. Walgreen Co. are tandem appeals from the Southern District of New York in the nationwide acetaminophen (Tylenol/generics) MDL. Plaintiffs—children and their parents/guardians—alleged state-law failure-to-warn claims, contending that prenatal ingestion of acetaminophen can cause attention-deficit/hyperactivity disorder (ADHD) and/or autism spectrum disorder (ASD), and that defendants (manufacturers, pharmacies, and retailers) should have warned of those risks.

The district court (Cote, J.) excluded all plaintiffs’ proffered general-causation experts under Rule 702 and Daubert v. Merrell Dow Pharms., Inc., then granted summary judgment in hundreds of cases. In Phippen, after the initial exclusions, plaintiffs offered an additional expert (Dr. Ness) focused on ADHD; her testimony was also excluded, and summary judgment entered.

The Second Circuit’s opinion addresses two recurring mass-tort flashpoints: (1) how far a district court may go in scrutinizing expert causation opinions in the face of active scientific debate, and (2) whether FDA OTC pregnancy-warning regulations preempt state-law claims seeking additional warnings.

II. Summary of the Opinion

  • Rule 702 / Daubert: The court held the district court exceeded its discretion by excluding the testimony of Drs. Baccarelli, Hollander, and Pearson, because their opinions employed generally accepted methodologies and reflected interpretive choices within the range of reasonable scientific disagreement.
  • The court held the district court did not abuse its discretion in excluding Drs. Cabrera and Louie, finding methodological/application defects: Cabrera did not adequately explain weighting/synthesis in a Bradford Hill analysis; Louie made misstatements and unsupported extrapolations regarding dose/duration thresholds.
  • Phippen remand: Because the appellate ruling restored key expert testimony, the court vacated and remanded without deciding the admissibility of Dr. Ness, leaving reconsideration to the district court if needed.
  • Preemption: The court affirmed the district court’s rejection of preemption, holding FDA rules requiring a general pregnancy warning for systemically absorbed OTC drugs do not bar manufacturers from adding supplemental, specific warnings about plausible risks.
  • Disposition: Judgments VACATED and REMANDED.

III. Analysis

A. Precedents Cited (and How They Shaped the Decision)

1. The Daubert framework and appellate posture

  • Daubert v. Merrell Dow Pharms., Inc. supplied the baseline: the district court is a gatekeeper ensuring expert testimony is both reliable and relevant, focusing on methodology rather than mere conclusions.
  • Kumho Tire Co. v. Carmichael anchored the “intellectual rigor” test—courtroom reasoning should mirror the rigor used in the relevant field.
  • Amorgianos v. Nat'l R.R. Passenger Corp. provided the Second Circuit’s operational phrasing: a “rigorous examination” of data, methods, and application to the case.
  • In re Mirena IUS Levonorgestrel-Related Prods. Liab. Litig. (No. II) supplied both the necessity of general causation evidence in complex products cases and the abuse-of-discretion standard (“manifestly erroneous”).

2. Methodology versus conclusions, and “analytical gap” policing

  • Gen. Elec. Co. v. Joiner was central to the boundary line: courts may exclude expert opinions connected to data only by ipse dixit, and may act where there is “too great an analytical gap” between studies and conclusions. The Second Circuit invoked this principle to uphold exclusion of Louie (dose/duration threshold inference).
  • Ruggiero v. Warner-Lambert Co. reinforced that inadequate data/method/studies cannot support the conclusion; it underwrote the court’s willingness to affirm exclusion where application failures were concrete.

3. Bradford Hill as a recognized causation-structuring tool

  • Sarkees v. E.I. Dupont De Nemours & Co. and In re Zoloft (Setraline Hydrochloride) Prods. Liab. Litig. were cited for the proposition that Bradford Hill is a generally accepted method courts often recognize for assessing whether observed associations support causal inference.
  • The opinion also relied on MDL-era Bradford Hill “weighting” jurisprudence: Daniels-Feasel v. Forest Pharmaceuticals, Inc. (and its reliance on Zoloft) and In re Mirena IUS Levonorgestrel-Related Prods. Liab. Litig. (No. II) for the requirement that, when an expert purports to apply Bradford Hill as a methodology, the expert should explain how criteria are weighed (supporting exclusion of Cabrera).

4. Expert qualification and use of literature reviews

  • United States v. Tin Yat Chin supplied the qualification comparator: match the expert’s superior knowledge to the subject of testimony.
  • McCullock v. H.B. Fuller Co. supported the proposition that experts can reliably opine based on reviewing other studies (not only by conducting original research).

5. Preemption doctrine in drug labeling

  • Wyeth v. Levine framed “impossibility pre-emption” as a demanding defense and recognized continuing manufacturer responsibility for label adequacy in the post-approval setting (used here by analogy to reject an “FDA-exclusive” warning theory).
  • PLIVA, Inc. v. Mensing was cited to describe the NDA system and label approval rigidity (a contrast with OTC monographs).
  • Gibbons v. Bristol-Myers Squibb Co., Marentette v. Abbott Laboratories, and Medtronic, Inc. v. Lohr anchored the presumption against preemption in areas of traditional state police powers.
  • Safe Haven Home Care, Inc. v. United States Dep't of Health & Hum. Servs. was cited for the interpretive point that unambiguous regulations do not require deference to agency interpretation—used to keep focus on regulatory text.

6. Risk ratios and causation thresholds (what courts should not rigidify)

  • Hardeman v. Monsanto Co. (and noting its partial abrogation by Monsanto Co. v. Durnell on other grounds) was used to reject a categorical demand for an adjusted odds ratio above 2.0 as a prerequisite for admissible causation testimony.

B. Legal Reasoning

1. The opinion’s core evidentiary holding: reliability is not correctness

The court repeatedly returned to a key distinction: Rule 702 asks whether expert opinions are reliably reached, not whether they are ultimately correct. The panel emphasized that when qualified experts use accepted methods and reach conclusions within the spectrum of scientific disagreement, a court crosses the line if it excludes testimony because it prefers a different interpretation. That principle drove reinstatement of Baccarelli, Hollander, and Pearson.

2. Transdiagnostic (multiple-outcomes) Bradford Hill analysis is not per se unreliable

A pivotal move was the court’s acceptance—at the admissibility stage—that an epidemiologist may structure a single Bradford Hill analysis across related neurodevelopmental outcomes (ADHD/ASD and symptom endpoints), where such an approach is used in the field. The district court’s concern that ADHD and ASD have “undeniably distinct” diagnostic criteria was treated as a weight/persuasion point, not a reliability defect, given record evidence that scientists (and even regulators) sometimes evaluate neurobehavioral outcomes together.

3. Courts must not redefine Bradford Hill factors to impose unattested “rules”

The panel faulted the district court for effectively imposing extra-scientific requirements on certain Bradford Hill considerations: for example, criticizing dose-response assessment because studies lacked precise dosage data, even though observational studies often use exposure proxies (e.g., duration) and the field treats such proxies as relevant. Similarly, the panel rejected the notion that biological plausibility requires mechanistic certainty; the criterion tolerates hypotheses consistent with current knowledge.

4. “Cherry-picking” is not established merely because an expert chooses among competing studies

The court refined the litigation usage of “cherry-picking”: it is not cherry-picking for an expert to prefer one study over another if the expert provides a coherent scientific explanation. The opinion applied this to genetic confounding debates (including sibling-control studies) and negative-control reasoning, holding that the district court treated reasonable scientific dispute as methodological impropriety.

5. Where application defects are concrete, exclusion remains appropriate

The court’s partial affirmance underscored that it was not weakening gatekeeping: Cabrera’s Bradford Hill discussion was excludable for failing to explain weighting/synthesis across factors (preventing meaningful review), and Louie’s opinion was excludable where it mischaracterized studies and leapt from “average risk in a >28 days group” to a “threshold” conclusion without a justified inferential bridge—classic Joiner “analytical gap” territory.

6. Preemption: OTC pregnancy warning text is a floor, not a ceiling

On preemption, the panel adopted a straightforward textual analysis of 21 C.F.R. § 201.63 and related “exact language” provisions. These rules require a verbatim general warning (“If pregnant or breast-feeding, ask a health professional before use.”), but do not forbid additional, more specific warnings. Because manufacturers could comply with both federal requirements and any state-law duty to add a supplemental ADHD/ASD warning, defendants failed to show “impossibility” preemption under Wyeth v. Levine.

C. Impact

1. A meaningful constraint on “over-gatekeeping” in contested science MDLs

The decision will likely be cited by plaintiffs in toxic tort and pharmaceutical MDLs to argue that district courts may not exclude epidemiological causation opinions simply because the literature is mixed or because the court would weigh Bradford Hill factors differently. It re-centers the inquiry on whether the expert is practicing science as scientists do—especially important where randomized trials are not feasible.

2. Practical guidance for litigants building (or attacking) Bradford Hill opinions

  • Permitted (if grounded): transdiagnostic framing; symptom endpoints; dose-response using exposure proxies; plausibility via hypothesized mechanisms.
  • Vulnerable (if not explained): Bradford Hill “checklists” without weighting/synthesis; threshold opinions built from grouped exposure categories; misstating what studies actually measured.

3. Preemption ramifications for OTC monograph products

The preemption holding signals that FDA-mandated OTC warning language (including pregnancy warnings) does not automatically create an exclusive federal warning regime insulating manufacturers from state failure-to-warn suits, particularly where federal law does not expressly bar supplemental warnings.

IV. Complex Concepts Simplified

  • General causation vs. specific causation: General causation asks whether a product is capable of causing a condition in the population; specific causation asks whether it caused this plaintiff’s condition. The MDL staged discovery around general causation first.
  • Bradford Hill criteria: A set of considerations (consistency, strength, dose-response, temporality, plausibility, etc.) used to assess whether an observed association is likely causal. It is not a mechanical checklist; experts must explain how the pieces fit together.
  • Confounding (including genetic confounding): A third factor (e.g., inherited traits) that could explain an observed association without the exposure being causal. Sibling-control and negative-control designs try to address confounding, but they have limitations.
  • Risk ratio and confidence interval: A risk ratio compares outcome frequency in exposed vs. unexposed groups; a confidence interval describes statistical uncertainty around the estimate. A wide interval may reduce certainty, but does not automatically make the study unusable.
  • OTC monograph system: For many OTC drugs, FDA sets category-wide conditions (active ingredients, some labeling requirements) without preapproving every product label’s full text. The court treated the pregnancy warning regulation as mandatory baseline language, not an exclusive label.
  • Impossibility preemption: State law is preempted only if it is impossible to comply with both state and federal law. Here, defendants could keep the exact federally required warning and add a supplemental warning, so impossibility was not shown.

V. Conclusion

The Second Circuit’s principal contribution is a sharpened Rule 702 boundary in epidemiology-heavy mass torts: when qualified experts apply generally accepted scientific methodologies and their interpretive choices fall within genuine scientific dispute, exclusion for “unreliability” risks becoming an impermissible judicial resolution of the scientific debate. At the same time, the court reaffirmed that Daubert remains a real filter where experts misstate evidence or fail to explain inferential steps.

On preemption, the court clarified that FDA’s OTC pregnancy-warning rules require specific baseline language but do not bar manufacturers from adding supplemental pregnancy-related risk warnings, defeating an “impossibility” defense at this stage and preserving traditional state failure-to-warn claims.