When a Warning Label Becomes a Constitutional Problem
Glyphosate | Roundup™ | Monsanto v. Durnell | SCOTUS
Well, for the first time, I submitted an opinion piece to The New York Times. After three days of silence, I’ll take the hint and share it here. What follows is the submission, lightly revised. I’ll be listening on Monday, possibly live, and taking notes.
Dear Editors,
On April 27, the Supreme Court will hear oral argument in Monsanto v. Durnell. Protesters chanting and carrying “People v. Poison” signs will be on the steps. Influencers will post. Politicians will grandstand. And nearly every brief and news story will frame this as a glyphosate case. It may be much more than that. I’m a toxicologist who has followed the glyphosate saga for years. That billions of dollars have changed hands in the face of worldwide regulatory consensus is remarkable on its own. My attached draft (~1,000 words) argues there are additional litigation and constitutional arguments that deserve attention before argument day.
Happy to revise for length or angle on short notice.
When a Warning Label Becomes a Constitutional Problem
On April 27, the Supreme Court will hear arguments in Monsanto v. Durnell, described as a dispute over pesticide labeling. That framing understates what is at stake. The Court is being asked whether a state may force a manufacturer to carry a cancer warning that EPA has repeatedly, explicitly, and on scientific grounds declined to require. The answer will determine not just the fate of glyphosate litigation, but whether constitutional limits on state regulatory power retain meaning when regulatory science becomes politically negotiable.
FIFRA sections are straightforward: states cannot impose labeling requirements “in addition to or different from” those required under federal law. EPA approved Roundup’s label without a cancer warning. A Missouri jury required one. The labels differ. The statute forbids that difference. One analysis ends there. The temptation to litigate whose toxicology is correct is a losing strategy. The moment you accept that frame, you have conceded that correctness determines the outcome. It doesn’t.
Preemption is structural, rooted in Congressional intent for uniformity. It operates on the relationship between two labels, not on the quality of the EPA’s reasoning. This matters especially post-Loper Bright as deference to EPA is no longer available. The textual argument is important. Registration is not a defense and the statute contains no exception for agency error. States cannot impose a different warning even if EPA made the wrong call. The only question that matters is whether the labels differ. They do. Everything else is noise.
A ruling could resolve this case, but would not resolve the underlying problem, which is how the warning got there in the first place. Consider the cancer warnings of the State of California and the theory that drove the Missouri jury to the same conclusion. These did not originate with any domestic regulatory body, but the International Agency for Research on Cancer, which classified glyphosate as “probably carcinogenic to humans” in 2015 in Monograph 112. That classification has since anchored billions of dollars in tort liability. It should not have.
IARC’s Monograph program is a hazard identification exercise. It asks whether a substance can, under some conditions, interact with biological systems in ways relevant to cancer. It does not ask whether harm occurs at exposure levels humans actually encounter. That second question is what risk assessment answers, and risk assessment is the function that determines whether a warning is scientifically warranted. The distinction is not semantic. Hazard without exposure context produces no actionable information about actual human risk.
The program’s architecture compounds this problem in ways that should disqualify it from any evidentiary role in litigation. IARC’s own 2012 internal framework document states explicitly that the monograph program does not perform quantitative risk characterizations. The EPA, the European Food Safety Authority, Health Canada, and Australia’s APVMA each independently evaluated glyphosate’s carcinogenic potential, concluded the evidence did not meet threshold for a cancer warning at relevant human exposure levels. That is not a divided scientific community. It is the consensus of agencies actually chartered to answer the question IARC disclaims answering.
IARC’s methodology excludes unpublished GLP-compliant regulatory guideline studies, omits dose–response analysis, and provides no avenue for challenge or appeal. By contrast, guideline studies specified by the Organisation for Economic Co-operation and Development are the product of decades of collaboration among academia, industry, and regulators, and form the standardized toxicological foundation for product registration. Under the Federal Rules of Evidence, expert testimony must be grounded in methods that reliably connect all relevant evidence to a conclusion. A process that excludes the most rigorous data, sidelines qualified scientific expertise, and disclaims quantitative risk assessment fails that test. Even so, it has underwritten billions in settlements and verdicts—despite deficiencies that should preclude its use as expert evidence in any court.
The methodological problems run deeper still. IARC’s elimination of its Group 4 category, formerly reserved for evidence of non-carcinogenicity, made the classification system a one-way ratchet: it can identify carcinogens but has no architecture for recognizing substances that reduce cancer incidence, promotion, or progression that would be aligned with the IARC mission. The recent pivot toward Key Characteristics of Carcinogens add to asymmetry by anchoring classifications in mechanistic concordance without dose-response context. Mechanistic signals detectable only at exposures humans will never sustain are not evidence of risk at exposures they actually encounter. A framework structurally incapable of registering resolution, only concern, and with no classificatory space for chemoprevention does not producing science fit for automatic regulatory consequences.
California’s mechanism for converting IARC’s output into law compounds the problem structurally. Under Health and Safety Code once IARC classifies a substance as Group 1 or 2A, California’s warning obligation follows automatically, without independent scientific review, without notice-and-comment rulemaking, ministerially, and without forum in which regulated parties can challenge the underlying monograph. The predicate for compelled speech being fixed by a body not subject to U.S. law or domestic legal accountability.
Additional constitutional defects, largely absent from the briefing before the Supreme Court, warrant closer scrutiny. Under Proposition 65, California has effectively delegated its police power to a body it cannot direct, override, or hold accountable. This creates a nondelegation problem with no discernible limiting principle and no meaningful administrative procedure. Regulated parties are denied any opportunity to contest the scientific predicate before it becomes binding law. There is no notice, no hearing, and no corrective mechanism in the process.
The consequences do not stop at California’s borders. In practice, its labeling mandates propagate into national markets, burdening interstate commerce and, in some cases, directly conflicting with federal regulatory determinations that govern both domestic and foreign commerce. Finally, the compelled speech problem is not theoretical. The Ninth Circuit has recognized that such warnings cannot be required where they are not purely factual and uncontroversial, a standard these labels fail to meet in light of the governing federal scientific record.
In California, the selection, timing, and even the determination of “no significant risk” levels may themselves be arbitrary and capricious. Warning inflation is not incidental; it erodes the signal value of warnings that truly matter.
Regulatory science draws its authority from consistency, transparency, reproducibility, and the mutual acceptance of data. Alignment across state, federal, and international jurisdictions promotes comparability, reduces duplicative effort, and strengthens confidence in shared standards. That kind of coherence builds trust, which is already in short supply, rather than deepening polarization and ideological divides.
A ruling in Monsanto v. Durnell would resolve only a narrow issue within the broader glyphosate litigation. It would not answer the more fundamental questions of how warning obligations should be determined or why glyphosate-related misinformation has proven so durable. Nor would such a decision confer blanket immunity. Claims grounded in design defect or negligence would remain available, and any preemption defense would still need to be litigated and proven on the merits. Assertions to the contrary, including claims by political figures that such a ruling would grant sweeping immunity, reflect either a misunderstanding or a misstatement of well-settled legal principles. In short, the larger dispute will persist well beyond the Court’s decision.
What a ruling would affirm is a more fundamental proposition: federal regulatory determinations carry governing weight. States cannot override them through automatic statutory triggers that bypass independent scientific review, and plaintiffs cannot evade them through failure-to-warn theories that function as indirect end-runs. The point is not to insulate any product or defendant, but to preserve a coherent system in which the same scientific record yields consistent legal consequences. That coherence depends not only on federal primacy, but on alignment with international risk assessments and the mutual acceptance of data developed under shared, validated methods. Regulatory judgments should not fracture at state lines, diverge across jurisdictions without scientific cause, or be reinvented case by case. Uniformity grounded in evidence, harmonization across borders, and mutual assurance of data are what allow the law to track science and what give warnings their credibility.
Brian H. Mathison, Ph.D.
Toxicologist