Methylene Chloride: Why a Federal Court Threw Out EPA’s Near-Ban on a Toxic Solvent

The Facts

In May 2024, EPA finalized a rule that prohibited 40 of the 53 ways methylene chloride is used in commerce. For the remaining 13, it capped worker exposure at 16 parts per million (ppm) over 15 minutes. The Occupational Safety and Health Administration (OSHA) has enforced a limit of 125 ppm for the same 15 minutes since 1997.

On September 15, 2026, a Fifth Circuit panel vacated the rule, along with the risk determination it rested on, and sent both back to EPA. The court found errors of law and a lack of substantial evidence. It also found that EPA had gone further than the statute allows in choosing to prohibit uses rather than limit exposure.

The Chemical

Methylene chloride (MC) is a volatile liquid solvent used for decades in adhesives and sealants, automotive products, and paint and coating removers. Its particular advantage, the court notes, is that it strips paint or coatings without damaging the surface underneath. The opinion lists furniture refinishing and the manufacture of energy-efficient air conditioner coolants and EV batteries among its essential uses. According to EPA’s rule, up to about 35 percent of production volume went to pharmaceutical uses, which the Toxic Substances Control Act (TSCA) does not regulate.

The court is equally direct about the hazards. Over-exposure, it says, can cause injury or even death, and MC has carcinogenic properties. The health effects EPA identified range from death at extreme inhaled concentrations to temporary neurotoxicity and potential effects on the liver and heart.

The Parties

Three petitioners challenged the rule as too aggressive: the American Chemistry Council, East Fork Enterprises, and Epic Paint Company (the “Industry Petitioners”). East Fork and Epic Paint make products that contain MC, and the Council’s members include MC manufacturers. The Sierra Club filed a separate petition arguing the opposite: that EPA did not go far enough to protect fenceline communities (those adjacent to facilities that emit MC), people genetically susceptible to cancer, and people at high elevations concerned about ozone depletion. The petitions were consolidated.

The Rule

Under the 2024 rule, the prohibitions covered using, manufacturing, importing, processing, and distributing MC across 40 conditions of use, including products containing MC. The 13 uses that survived include domestic manufacturing, importing, use as a reactant, use as a laboratory chemical, paint and coating removal from certain aircraft or spacecraft components, and disposal. Each would operate under a Workplace Chemical Protection Program requiring exposure limits, monitoring and reporting, and respiratory and dermal protection. For a few commercial or industrial uses, the prohibition was deferred for a few years.

The rule also barred retailers from distributing MC or any MC-containing product, and defined “retailer” to include any person who distributes or makes available products to at least one consumer. EPA acknowledged in the rulemaking that the rule would have a very severe impact on antique architectural work and commercial furniture refinishing.

The Two Questions Every TSCA Rule Must Answer

The case depends on how TSCA is built, so the structure is worth stating plainly. Congress enacted TSCA in 1976 and amended it in 2016. Under the amended law, EPA regulating a chemical means answering two separate questions in order.

First: does this chemical present an “unreasonable risk of injury” to health or the environment “under the conditions of use”? The statute defines conditions of use as the circumstances under which a chemical is intended, known, or reasonably foreseen to be manufactured, processed, distributed, used, or disposed of. At this stage EPA may not consider costs or other non-risk factors. It must consider the likely duration, intensity, frequency, and number of exposures, and it must rely on the best available science and the weight of the scientific evidence.

Second: if there is unreasonable risk, what is the least EPA must do to eliminate it? The statute directs EPA to apply one or more of seven listed measures “to the extent necessary” so the chemical no longer presents that risk. The menu runs from recordkeeping and notice requirements to outright prohibition. At this stage EPA must factor in the chemical’s benefits, the reasonably ascertainable economic consequences of the rule, and whether feasible substitutes will be reasonably available.

Courts review these decisions under a demanding test. TSCA requires that EPA’s action be supported by “substantial evidence in the rulemaking record taken as a whole.” In plain terms, EPA must weigh the entire record, including the parts that cut against its conclusion, and reach a result a reasonable person could accept. The opinion describes this standard, citing other courts, as “fairly rigorous and more searching” than the general standard under the Administrative Procedure Act.

This court has applied that scrutiny to EPA before. In 1991 it vacated EPA’s 1989 rule banning most asbestos products, for lack of substantial evidence that EPA had adequately considered less burdensome options.

“Congress did not enact TSCA as a zero-risk statute.”

Corrosion Proof Fittings v. EPA (5th Cir. 1991), quoted in the opinion at 19

How the Rule Was Built

Same Science, New Rules

The turn in this case comes between the June 2020 evaluation and the November 2022 revision. According to the court, EPA did not amend any of its fundamental findings or its scientific analysis between the two. It changed two policies. Those two changes drove the rule that followed.

Change One: The “Whole Chemical” Shortcut

In 2020, EPA made a separate determination for each use. In 2022, it declared that MC posed unreasonable risk “as a whole chemical substance,” regardless of the variation among uses. The practical effect was that every use had to be regulated. Six uses that EPA had previously found did not present unreasonable risk were recharacterized as risky, with no change to the underlying science for those uses. The court rejected EPA’s argument that this was harmless error, because EPA either prohibited or stringently regulated those six uses.

The court held that TSCA does not allow this. Its reasoning: the statute asks whether activities involving a chemical present unreasonable risk, and the phrase “conditions of use” appears in at least thirteen subsections of the relevant section. The court concluded that the statute focuses on determining and regulating risk according to the separate conditions of use. It added that EPA’s own 2017 procedural rule required a determination “under each condition of use,” and that EPA had promised rules would apply only to uses presenting unreasonable risk. EPA’s later reading of that rule as permitting whole-chemical findings was, in the court’s view, wrong. The court noted that EPA had issued use-by-use determinations for several years before discovering a different interpretation.

Change Two: Assuming Nobody Wears Protective Equipment

The 2020 evaluation assumed workers use personal protective equipment (PPE), meaning respirators, gloves, protective clothing, and eye protection, consistent with a longstanding OSHA regulation. The 2022 revision assumed they do not. The court held that PPE use is a “condition of use” that EPA must account for when the evidence shows it is used, and that moving PPE to the later risk-management stage contradicted both the statutory definition and EPA’s duty to consider the intensity of exposure.

The court also found no factual support for the reversal. It observed that EPA’s own explanations pulled in different directions:

Between those two statements, EPA reversed a default assumption of compliance while saying it was not asserting widespread noncompliance. The court called these explanations “remarkably self-contradictory” and held the new assumption contrary to law, arbitrary and capricious, and lacking substantial evidence.

How EPA Got to 16 ppm and 2 ppm

The court described EPA’s method, in the opinion’s own simplified terms, as four steps: a hazard assessment, an exposure assessment, a risk characterization, and a risk determination. The mechanics matter, because the court found problems in them.

EPA started with a “point of departure,” an estimate of the exposure level linked to an adverse effect in a study. It divided that by the exposure workers actually experience under each use, producing a “margin of exposure.” It then compared that margin to a “benchmark,” a number built by multiplying uncertainty factors meant to cover things like differences between animals and humans, differences among people, and reliance on studies that observed effects rather than no effects. If the margin fell below the benchmark, EPA deemed the risk unreasonable. To set an exposure limit, EPA divided the point of departure by the benchmark. The court’s phrase for the benchmark was “a fudge factor.” EPA defended its approach as conventional.

The Acute Limit: A Study of Twelve People

EPA’s 15-minute point of departure was 478 ppm, derived from a double-blind study of twelve people (the “Putz study”) who experienced a temporary, modest impairment, about a 7 percent reduction in peripheral vision, at 195 ppm. EPA converted that result to a 15-minute equivalent. It then applied a factor of ten for human variability and a factor of three for relying on a study that observed an effect rather than none, for a composite factor of thirty. Dividing 478 by 30 and rounding up produced 16 ppm. Forty-seven conditions of use had a margin of exposure below thirty.

The court found this unsupported. Its reasons, in sequence: EPA relied exclusively on the Putz study, whose effect EPA itself described as “small magnitude.” EPA called the vision effect a “precursor” to nervous-system depression, but the court noted a precursor is not in itself an adverse health effect. Meanwhile EPA discounted other evidence in the record. The Winneke study reported minimal adverse effects at 500 ppm for 3.8 hours, with participants exposed to as much as 800 ppm without severe effects. A study by Soden reported no deaths or adverse health effects among workers exposed to 475 ppm for eight hours a day over at least ten years.

EPA responded by emphasizing other Putz results: after four hours, participants showed a 36 percent reduction in hand-eye coordination, a 17 percent reduction in peripheral vision, and a 17 percent reduction in auditory vigilance. The court did not say Putz should have been excluded. It said the problem was reliance on it alone.

The court also placed the numbers against real-world harm. EPA’s own evaluation listed about a dozen fatalities, all in situations involving concentrations of at least 1,711 ppm, near the 2,300 ppm level rated “immediately dangerous to life or health.” OSHA’s 125 ppm limit had been in place for a quarter century and was reaffirmed after reconsideration in 2010. The court said EPA paid no attention to that standard in choosing a limit nearly one-tenth of it. It held the 15-minute limit was not supported by substantial evidence.

The Chronic Limit: One Rat Study Against Three Human Studies

For long-term exposure, EPA started from a study of 180 female rats (the “Nitschke study”). Rats exposed to 500 ppm for six hours a day for two years developed liver irregularities such as lesions. At 200 ppm, they did not. EPA estimated the exposure producing a ten percent risk of such effects in rats, used a pharmacokinetic model at its most conservative estimate to calculate a human-equivalent concentration, and applied a benchmark of ten (two factors of three, rounded up). The opinion states that conventionally dividing the point of departure by the benchmark would have produced a limit below 0.5 ppm, which it called impossible. EPA instead set the limit at 2 ppm.

The Industry Petitioners pointed to three epidemiological studies of workers regularly exposed to MC, submitted during the comment period. In two, median exposures were as high as 475 ppm. EPA rated all three “medium quality” and “acceptable,” the same rating it gave the study its own limit rested on. A commenter retained Dr. Jonathon Borak, a Yale clinical professor of medicine, to review them.

The court found EPA’s responses inadequate. EPA pointed to increased bilirubin in two studies, but the court noted bilirubin is neither a necessary nor a sufficient marker of adverse liver conditions and has many possible causes. EPA said the human data did not provide clear evidence of adverse liver effects; the court asked whether that was “a finding worthy of note.” And EPA did not respond to Dr. Borak’s analysis at all, though an agency must respond to significant comments.

The court drew on its 1983 decision in Gulf South Insulation, which vacated a Consumer Product Safety Commission ban on urea-formaldehyde foam insulation that rested on a study of 240 rats. There the court said it was “not good science to rely on a single experiment.” The Industry Petitioners cited that precedent in their brief. According to the court, EPA did not respond. The court held the eight-hour limit unsupported by substantial evidence and arbitrary and capricious.

“Unreasonable Risk” Is Not “Any Risk”

Underneath these holdings sits a legal ruling with reach beyond MC. EPA argued that Congress gave it authority to make the technical determination of what counts as an “unreasonable risk,” and that courts should defer to its expertise. The court disagreed with the premise. It called “unreasonable risk” not a factual finding but a prudential baseline set by the statute, and said that after the Supreme Court’s 2024 decision in Loper Bright, courts are the independent interpreters of statutory language.

The court then held that EPA’s approach was more conservative than the statutory term connotes. It compared it to the precautionary principle, the idea that regulators may act against a potential harm even before it is shown to be likely. Congress writes that principle into some statutes; the court cited the Clean Air Act’s requirement of an “adequate margin of safety.” TSCA contains no such language.

The court summarized its view of EPA’s analysis in a passage that is pointed even by judicial standards. It said EPA “has gone with the most extreme position” at every juncture of its scientific analysis: discounting human studies, assuming no PPE use, relying on a rat study with extreme exposures, and using the smallest temporary effects as its starting point, then applying inflated benchmarks. These are the court’s characterizations of the record. EPA defended its methodology as conventional, based on best available science and the weight of evidence.

When Uncertainty Became a Reason to Ban

The court said it could have stopped once it found the rule’s risk analysis defective. It went on to address a second problem: EPA’s decision to prohibit most uses rather than adopt exposure limits closer to OSHA’s. The Industry Petitioners, the court noted, did not assert that MC presents no unreasonable risk. The dispute was over how far EPA could go.

EPA’s stated reasoning for prohibition, as quoted in the opinion, was this:

EPA also said it worried that without a ban, users would keep trying and failing to meet the new limits, risking harm to workers, and it regretted the potential for MC use to increase in a sector that had already moved away from it. The court called these speculative rationales that misconstrue EPA’s authority. In its reading, EPA may regulate “to the extent necessary” to eliminate actual, identified unreasonable risk. It cannot ban a chemical because it is uncertain whether a risk will remain reasonable.

Three further points from the court:

• Burden of proof. EPA invited companies to comment and show they could meet the new limits, and many did not. The court held that treating silence as support shifted the burden to regulated entities. The statute places the burden on EPA to show that a risk is unreasonable and that its remedy goes only as far as necessary.
• OSHA citations. EPA cited forty-four OSHA citations for MC exposure violations between October 2022 and September 2023 as evidence of high noncompliance. The court found that bare citation counts prove little without a discussion of their nature, such as their degree, whether they were willful, or whether they involved recordkeeping failures.
• Alternatives. When a rule would substantially prevent a use, TSCA requires EPA to consider whether technically and economically feasible substitutes will be reasonably available. The court found EPA focused on whether an alternative performs a similar function, ignored the economic feasibility of alternatives, and did not account for higher costs, lower productivity, or adverse environmental consequences of switching.

The court’s answer was that it was practicable for EPA to take seriously the numerous specific comments explaining that no feasible alternatives exist for many uses.

What EPA Decided, What It Argued, and What the Court Ordered

EPA’s determination was that MC posed unreasonable risk to health and, per the 2020 evaluation, no unreasonable risk to the environment. Its rule followed from the revised health determination. In court, EPA defended its methodology as conventional, argued it had applied best available science and weight of evidence, and argued that courts should defer to its technical expertise. It also argued that its 2022 change was harmless error. As noted above, it told the court shortly before oral argument that it no longer defended the whole-chemical and no-PPE interpretations.

The court’s disposition was to grant the Industry Petitioners’ petition, vacate the challenged rule and the associated risk determination, deny the Sierra Club’s petition, and remand to EPA for proceedings consistent with the opinion. The 2019 consumer paint-remover rule, which the opinion describes as unchallenged, was not part of the ruling.

The Sierra Club’s Case

The Sierra Club came at the rule from the opposite direction, arguing it was too weak. The court’s vacatur, it noted, “vitiates” EPA’s response to those arguments, but it addressed them anyway and found no merit. It established the Club’s standing to petition based on a member who lives two miles from a facility that releases MC.

Fenceline Communities

EPA excluded air and drinking-water exposures from the scope of its initial risk evaluation. After the Ninth Circuit challenge and the 2021 remand, EPA ran a “screening-level” fenceline assessment using a draft methodology. It concluded that most risk was adequately addressed by the MC rule and other laws such as the Clean Air Act, but it did not amend the risk evaluation’s scope, saying it was still developing the science.

The court held that TSCA leaves EPA latitude to exclude exposure pathways it lacks a reliable method to assess. The dataset lacked release-point locations and the number of release days, and EPA acknowledged that some assumptions likely overestimated air emissions. The court distinguished a Ninth Circuit decision the Club relied on but added that it was not deciding whether it agreed with that court’s reading of the statute. It wrote that it would “leave this debate for another day.” It also held that EPA is authorized to address “unreasonable,” not merely “elevated,” risk, and that the draft methodology, marked “Public Comment Draft,” was not settled policy EPA had to justify departing from.

Genetic Susceptibility to Cancer

EPA acknowledged that roughly one-third of the U.S. population is genetically susceptible to cancer. It used a conservative estimate based on the 95th percentile of the general population, reasoning that risk to the subpopulation is reflected within the general population’s risk. The court found the Club offered no significant response, and rejected the argument that the Clean Air Act’s one-in-a-million cancer risk benchmark governs under TSCA. It added that it did not opine on what level of carcinogenic risk coincides with unreasonable risk.

Ozone Depletion

The court held the Club lacked standing on the ozone claim. It described the injury theory as a chain of conjectural steps: inadequate regulation might leave too much MC emitted, which might break down atmospheric ozone, which might raise ultraviolet exposure, which might raise cancer risk. The court also noted that EPA has consistently found MC is not an ozone-depleting substance since 1994. It acknowledged that EPA noted in 2016, under a different authority, that recent research indicated increasing industrial MC emissions “could have a detectible impact on the ozone layer,” but called that statement the exception.

What the Court Did Not Decide

Much of the public reading of a ruling like this will flatten it into “court sides with industry.” The opinion is narrower, and the limits matter.

• It did not find MC safe. The Industry Petitioners did not argue it presents no unreasonable risk, according to the court, and the opinion opens by acknowledging its hazards.
• It did not set a number. It said “unreasonable risk” must allow for some risk and that the relevant question is how much risk is unreasonable. It did not answer that question.
• It did not decide what level of carcinogenic risk is unreasonable. It noted the cancer aspect played little role in EPA’s final determination and no role in the Industry Petitioners’ briefing.
• It did not resolve whether the Ninth Circuit’s reading of EPA’s authority to exclude exposure pathways is correct.
• It did not bar EPA from regulating MC. It remanded for further proceedings, and the court said that once a legally compliant unreasonable-risk determination exists, EPA must follow the statute’s menu of permissible measures.
• It did not address what governs MC in the meantime. The opinion is silent on interim consequences.

What a Legitimate Fix Looks Like

The recommendations below are this publication’s analysis. They are grounded in failure modes the court identified, but they are not findings of the court, and none is guaranteed to prevent a repeat.

Regulatory Track

• Return to use-by-use determinations, as the statute and EPA’s own procedural rule required, so that a use with no documented unreasonable risk is not regulated because of its neighbors.
• Where evidence shows workers use PPE, count it. Where EPA believes they do not, produce the evidence. The court faulted speculation, not the idea that noncompliance could be shown.
• Answer significant comments in writing, especially expert analyses that cut against the preferred study. The court held EPA failed here on Dr. Borak’s review.
• Explain which factors beyond the margin-of-exposure comparison drive a risk determination. The court noted EPA said the comparison is not always dispositive but never identified what else contributed.
• Coordinate with OSHA. The court treated OSHA’s 1997 limits as a benchmark EPA ignored, but nothing in the opinion establishes that those limits are adequate for current science. If EPA believes they are not, it needs a record that says why.
• Finish the fenceline methodology. The draft EPA used was not built for TSCA risk determinations, and the court upheld the exclusion of pathways EPA could not yet assess. A finalized method would give communities a defined path into future evaluations.

Legislative Track

• Congress could clarify whether “unreasonable risk” under TSCA carries any precautionary margin. The court read the statute as having none, and elsewhere in the U.S. Code Congress has legislated a margin of safety explicitly.
• Congress could address what EPA should do with exposure pathways it cannot yet measure. The court noted the statutory deadline and the risk evaluation rule did not let EPA stop the clock to gather more data.
• Congress could clarify how TSCA workplace rules should interact with OSHA standards, including which agency’s exposure limits should control where they diverge by a factor of ten.

Corporate Governance Track

• Companies and trade groups arguing that their workplaces comply with OSHA should be ready to document it: PPE programs, exposure monitoring, and corrective actions. The court held that EPA cannot treat silence as proof of noncompliance, but a thin record could invite the next agency to fill the gap.
• Firms that receive OSHA citations for MC exposure could disclose what the violations involved and how they were corrected. The court found EPA’s citation count uninformative because the nature of the violations was undiscussed.
• Users and manufacturers who say no feasible alternatives exist should put specific cost, productivity, and environmental data on the record. The court held that EPA had specific comments on this before it and did not take them seriously.

What to Watch

The unresolved question is a number. The Fifth Circuit has said what the statute does not allow: a whole-chemical shortcut, a blanket assumption that no worker wears protective equipment, exposure limits built on a single study while contrary evidence sat in the record, and a ban justified by uncertainty. It has not said what the statute does allow. For a chemical the court agrees can injure and kill at high doses, and about which it declined to say what level of cancer risk counts as unreasonable, that determination is now EPA’s to make again, on a record that has to answer the evidence the court says it skipped.