Edwin Hardeman v. Monsanto Company

FOR PUBLICATION

UNITED STATES COURT OF APPEALS
FOR THE NINTH CIRCUIT

EDWIN HARDEMAN, Nos. 19-16636
Plaintiff-Appellee/ 19-16708
Cross-Appellant,
D.C. Nos.
v. 3:16-cv-00525-VC
3:16-md-02741-VC
MONSANTO COMPANY,
Defendant-Appellant/
Cross-Appellee. OPINION

Appeal from the United States District Court
for the Northern District of California
Vince G. Chhabria, District Judge, Presiding

Argued and Submitted October 23, 2020
San Francisco, California

Filed May 14, 2021

Before: Michael D. Hawkins, N. Randy Smith, and
Ryan D. Nelson, Circuit Judges.

Opinion by Judge R. Nelson;
Dissent by Judge N.R. Smith
2 HARDEMAN V. MONSANTO

SUMMARY *

Pesticides / Punitive Damages

The panel affirmed the district court’s judgment in favor
of Edwin Hardeman in his action alleging that Monsanto’s
pesticide, Roundup, caused his non-Hodgkin’s lymphoma.

Roundup is pesticide with the active ingredient
glyphosate. Since 2015, thousands of cancer victims sued
Monsanto in state and federal court. This appeal arose out
of the first bellwether trial for the federal cases consolidated
in a multidistrict litigation. The jury awarded Hardeman
$5,267.634.10 in compensatory damages, and $75 million in
punitive damages. The district court reduced the punitive
damages award to $20 million.

The panel held that Hardeman’s state failure-to-warn
claims based on Roundup’s labeling were consistent with the
Federal Insecticide, Fungicide, and Rodenticide Act
(“FIFRA”) and thus were neither expressly nor impliedly
preempted. Specifically, the panel affirmed the district
court’s conclusion that Hardeman’s state failure-to-warn
claims were “equivalent to” and “fully consistent with”
FIFRA and therefore not expressly preempted. Bates v. Dow
Agrosciences LLC, 554 U.S. 431, 449 (2005). In addition,
because Monsanto could comply with both FIFRA and
California law, FIFRA did not impliedly preempt
Hardeman’s state failure-to-warn claims.

*
This summary constitutes no part of the opinion of the court. It
has been prepared by court staff for the convenience of the reader.
 HARDEMAN V. MONSANTO 3

The panel held that the district court ultimately applied
the correct standard from Daubert v. Merrell Dow
Pharmaceuticals, Inc., 509 U.S. 579 (1993), and did not
abuse its discretion in admitting Hardeman’s expert
testimony. Despite its incorrect assumption that this court
was more permissive than others in admitting Daubert
testimony, the district court still employed the correct legal
standard for reliability when it admitted Hardeman’s expert
testimony. The panel held further that the district court did
not abuse its discretion in concluding that Hardeman’s
experts reliably based their general causation opinions on
epidemiological evidence showing a connection between
glyphosate and cancer. The panel also held that the district
court did not abuse its discretion in admitting Hardeman’s
expert testimony on specific causation to show that
Hardeman’s cancer was caused by glyphosate, rather than
some other factor. Here, Hardeman’s experts reliably used
differential diagnosis because they ruled in glyphosate based
on the epidemiological evidence supporting the general
causation opinions and ruled out alternate causes, such as
idiopathy and Hepatitis C (HCV).

The panel held that the district court did not abuse its
discretion in admitting the International Agency for
Research on Cancer’s classification of glyphosate as
probably carcinogenic and three regulatory rejections of that
classification by excluding evidence from other regulatory
bodies. The panel held further that even if these evidentiary
decisions were erroneous, any error was harmless because it
was more probable than not that the admission of the
evidence did not affect the jury’s verdict.

The panel held that the district court’s jury instruction on
causation was erroneous because it was inconsistent with the
4 HARDEMAN V. MONSANTO

Judicial Council of California Civil Jury Instructions and
California case law, but it was harmless error.

The panel held that the district court properly denied
Monsanto judgment as a matter of law because evidence
showed the carcinogenic risk of glyphosate was knowable at
the time of Hardeman’s exposure.

The panel held that evidence supported a punitive
damages award, punitive damages were properly reduced,
and the reduced award – while close to the outer limit – was
constitutional. Specifically, the panel held that punitive
damages were permissible under California law because
substantial evidence was presented that Monsanto acted with
malice by, among other things, ignoring Roundup’s
carcinogenic risks. The panel held that the jury’s $75 million
punitive damages award was “grossly excessive” given the
mitigating factors found by the district court. However,
considering the evidence of Monsanto’s reprehensibility, the
district court’s reduced $20 million punitive damages award
(a 3.8 to 1 damages ratio), while at the outer limits of
constitutional propriety, ultimately comported with due
process.

The panel cautioned that although this appeal involved a
bellwether trial, many of its holdings were fact-specific, and
different Roundup cases may present different
considerations, leading to different results.

Judge N.R. Smith dissented to section VII.B, concerning
punitive damages. He would hold that Monsanto’s low
degree of reprehensibility cannot constitutionally justify the
district court’s substantial punitive damages award. The
facts found by the district court did not support a 3.8:1 ratio
to compensatory damages.
 HARDEMAN V. MONSANTO 5

COUNSEL

Seth P. Waxman (argued) and Paul R.Q. Wolfson, Wilmer
Cutler Pickering Hale and Dorr LLP, Washington, D.C.;
Thomas G. Sprankling and Henry J. Becker, Wilmer Cutler
Pickering Hale and Dorr LLP, Palo Alto, California; Leon T.
Kenworthy, Clair H. Chung, James Barton, Samuel M.
Strongin, and Rafael J. Gallardo Hevia, Wilmer Cutler
Pickering Hale and Dorr LLP, Washington, D.C.; Brian L.
Stekloff and Rakesh Kilaru, Wilkinson Walsh and Eskovitz
LLP, Washington, D.C.; Philip J. Perry and Richard P.
Bress, Latham & Watkins LLP, Washington, D.C.;
Michael X. Imbroscio and David M. Zionts, Covington &
Burling LLP, Washington, D.C.; Lee Marshall, Bryan Cave
Leighton Paisner LLP, San Francisco, California; for
Defendant-Appellant/Cross-Appellee.

David J. Wool (argued) and Aimee H. Wagstaff, Andrus
Wagstaff PC, Lakewood, Colorado; Leslie A. Brueckner,
Public Justice, Oakland, California; Jennifer A. Moore,
Moore Law Group PLLC, Louiseville, Kentucky; for
Plaintiff-Appellee/Cross-Appellant.

Jonathan D. Brightbill (argued) and Eric Grant, Deputy
Assistant Attorneys General; Jennifer Scheller Neumann,
Varudhini Chilakamarri, and Matthew R. Oakes, Attorneys;
Environment and Natural Resources Division, United States
Department of Justice, Washington, D.C.; Erin S. Koch and
Amber L. Aranda, Attorneys, EPA Office of General
Counsel, Washington, D.C.; for Amicus Curiae United
States.

Andrew Wiener (argued), Laura Zuckerman, and Dennis
Ragen, Deputy Attorneys General; Harrison M. Pollack,
Supervising Deputy Attorney General; Office of the
6 HARDEMAN V. MONSANTO

Attorney General, Oakland, California; for Amicus Curiae
State of California.

Shannen W. Coffin and Sara Beth Watson, Steptoe &
Johnson LLP, Washington, D.C., for Amicus Curiae
CropLife America.

Laura W. Brill, Nicholas F. Daum, and Sharon S. Song,
Kendall Brill & Kelly LLP, Los Angeles, California, for
Amicus Curiae Genentech Inc.

Douglas J. Peterson, Attorney General; Justin D. Lavene,
Maegan L. Woita, and Joshua E. Dethlefsen, Assistant
Attorneys General; Office of the Attorney General, Lincoln,
Nebraska; Lawrence G. Wasden, Attorney General of Idaho;
Jeff Landry, Attorney General of Louisiana; Wayne
Stenehjem, Attorney General of North Dakota; Jason
Ravnsborg, Attorney General of South Dakota; Ken Paxton,
Attorney General of Texas; and Sean D. Reyes, Attorney
General of Utah; for Amici Curiae States of Nebraska,
Idaho, Louisiana, North Dakota, South Dakota, Texas, and
Utah.

Curtis A. Cole, Cassidy C. Davenport, and Scott M.
Klausner, Cole Pedroza LLP, San Marino, California, for
Amici Curiae California Medical Association, California
Dental Association, and California Hospital Association.

William R. Stein, Eric S. Parnes, Stephen R. Halpin III, and
J. Chesley Burruss, Hughes Hubbard & Reed LLP,
Washington, D.C.; Theodore V.H. Mayer, Hughes Hubbard
& Reed LLP, New York, New York; Steven P. Lehotsky and
Michael B. Schon, U.S. Chamber Litigation Center,
Washington, D.C.; James C. Stansel and Melissa B. Kimmel,
Pharmaceutical Research and Manufacturers of America,
 HARDEMAN V. MONSANTO 7

Washington, D.C.; for Amici Curiae Chamber of Commerce
of the United States of America, and Pharmaceutical
Research and Manufacturers of America.

Adina H. Rosenbaum and Allison M. Zieve, Public Citizen
Litigation Group, Washington, D.C., for Amicus Curiae
Public Citizen.

Matthew W.H. Wessler and Larkin Turner, Gupta Wessler
PLLC, Washington, D.C.; Bruce Stern, President, American
Association for Justice, Washington, D.C.; for Amicus
Curiae American Association for Justice.

Ashley Keller, Travis Lenker, and Warren Postman, Keller
Lenkner LLC, Chicago, Illinois; Ernest A. Young, Apex,
North Carolina; for Amici Curiae Public Law Scholars.

Ryan D. Talbott, Center for Food Safety, Portland, Oregon,
for Amici Curiae Center for Food Safety and Center for
Biological Diversity.

Carrie Apfel, Earthjustice, Washington, D.C.; Alexis
Andiman and Peter Lehner, Earthjustice, New York, New
York; Patti Goldman, Earthjustice, Seattle, Washington; for
Amici Curiae California Rural Legal Assistance Foundation,
Farmworker Association of Florida, Farmworker Justice,
Migrant Clinicians Network, Pesticide Action Network,
United Farm Workers, and UFW Foundation.

Melanie Benesh and Caroline Leary, Environmental
Working Group, Washington, D.C., for Amicus Curiae
Environmental Working Group.
8 HARDEMAN V. MONSANTO

OPINION

R. NELSON, Circuit Judge:

Monsanto Company manufactures Roundup, a pesticide
with the active ingredient glyphosate. Since 2015, thousands
of cancer victims have sued Monsanto in state and federal
court, alleging that Roundup caused their non-Hodgkin’s
lymphoma. This appeal arises out of the first bellwether trial
for the federal cases consolidated in a multidistrict litigation.

The jury returned a verdict in favor of plaintiff Edwin
Hardeman, awarding him $5,267,634.10 in compensatory
damages and $75 million in punitive damages. The district
court reduced the jury’s punitive damages award to
$20 million.

Monsanto appeals, arguing the Federal Insecticide,
Fungicide, and Rodenticide Act (“FIFRA”) preempts
Hardeman’s failure-to-warn claims; the district court made a
series of evidentiary and jury instruction errors; the district
court erred in denying judgment as a matter of law; and the
punitive damages award violates California law and the Due
Process Clause. Hardeman cross-appeals, arguing the jury’s
$75 million punitive damages award was constitutional.

We affirm the district court and hold that (1) Hardeman’s
state failure-to-warn claims are not preempted by FIFRA;
(2) the district court ultimately applied the correct standard
from Daubert v. Merrell Dow Pharmaceuticals, Inc.,
509 U.S. 579 (1993), and did not abuse its discretion in
admitting Hardeman’s expert testimony; (3) the district court
did not abuse its discretion in admitting the International
Agency for Research on Cancer’s classification of
glyphosate as probably carcinogenic and three regulatory
rejections of that classification but excluding evidence from
 HARDEMAN V. MONSANTO 9

other regulatory bodies; (4) the district court’s jury
instruction on causation, though erroneous, was harmless;
(5) Monsanto was properly denied judgment as a matter of
law because evidence shows the carcinogenic risk of
glyphosate was knowable at the time of Hardeman’s
exposure; and (6) evidence supports a punitive damages
award, punitive damages were properly reduced, and the
reduced award—while close to the outer limits—is
constitutional.

I

A

Under FIFRA, the United States Environmental
Protection Agency (“EPA”) enforces “the use, . . . sale[,] and
labeling[] of pesticides.” Bates v. Dow Agrosciences LLC,
544 U.S. 431, 437 (2005) (citation omitted). A state may
“not impose or continue in effect any requirements for
labeling or packaging in addition to or different from those”
required by FIFRA. 7 U.S.C. § 136v(b).

FIFRA requires pesticide manufacturers to register their
products with EPA. 7 U.S.C. § 136a(a). EPA makes
registration determinations after considering available
scientific data, § 136a(c)(1)(F), (c)(2)(A); 40 C.F.R.
§ 158.500, and FIFRA requires EPA to re-review a
pesticide’s registration, including its effects on human
health, every fifteen years, § 136a(g)(1)(A). FIFRA states,
however, that “[i]n no event shall registration of an article be
construed as a defense for the commission of any offense
under this subchapter.” § 136a(f)(2). Rather, “[a]s long as
no cancellation proceedings are in effect,” registration of a
pesticide is merely “prima facie evidence that the pesticide,
its labeling and packaging comply with the registration
provisions of the subchapter.” Id.
10 HARDEMAN V. MONSANTO

EPA can also institute cancellation proceedings,
7 U.S.C. § 136d(b), or take other enforcement action against
the manufacturer of a registered pesticide if the agency
determines the product is “misbranded.” Bates, 544 U.S.
at 439. Remedies for misbranding include civil and criminal
penalties. Id. at 439 n.11 (citing 7 U.S.C. § 136l). A duly
registered pesticide can be misbranded if the label “does not
contain adequate instructions for use, or if its label omits
necessary warnings or cautionary statements.” Bates,
544 U.S. at 438 (citation omitted). “Because it is unlawful
under the statute to sell a pesticide that is registered but
nevertheless misbranded, manufacturers have a continuing
obligation to adhere to FIFRA’s labeling requirements.” Id.
(citations omitted). This obligation includes a duty to seek
approval to amend a label that does not contain all
“necessary warnings or cautionary statements.” Id. (citation
omitted).

Starting in 1974, EPA registered pesticides containing
glyphosate, the active ingredient in Roundup. 1 EPA,
Glyphosate Proposed Interim Registration Review Decision
4 (Apr. 2019) (“Proposed Interim Registration Review”). In
1985, an EPA review of a mouse study found “[g]lyphosate
was oncogenic in male mice,” causing rare tumors. EPA
classified glyphosate as a possible human carcinogen. Since
then, however, EPA has repeatedly approved the use of
glyphosate as a pesticide, each time concluding that it is not
likely to be carcinogenic to humans. See Nat’l Fam. Farm
Coal. v. EPA, 966 F.3d 893, 905 (9th Cir. 2020).

Though commonly referred to as an herbicide, Roundup is defined
1

as a pesticide under 7 U.S.C. § 136(t), (u). Roundup contains
glyphosate, water, and other ingredients called “surfactants.”
 HARDEMAN V. MONSANTO 11

In the early 1990s, EPA reevaluated glyphosate’s effects
on human health as part of its regular review of glyphosate’s
registration. After considering numerous carcinogenicity
studies in rats and mice—including new evidence submitted
by Monsanto—EPA changed its designation of glyphosate
to a “Group E carcinogen” signifying “evidence of non-
carcinogenicity in humans.”

In 2015, a working group at the International Agency for
Research on Cancer (“IARC”), an agency of the World
Health Organization, issued a report classifying glyphosate
as a “Group 2A” agent, meaning it is “probably carcinogenic
to humans” based on glyphosate’s “limited evidence” of
cancer in humans and “sufficient evidence” of cancer in
experimental animals. IARC’s classification was a “hazard
identification,” the first step of a public health assessment
designed to identify cancer hazards. That hazard
determination asked whether glyphosate “is capable of
causing cancer under some circumstances,” but did not
include a “risk assessment” gauging the carcinogenic effects
from real-world human exposure. Since IARC’s
classification, other national and international agencies
charged with reviewing pesticides—such as the European
Union’s European Chemicals Agency (“ECA”), European
Food Safety Authority (“EFSA”), and the national health
authorities of Australia, Canada, Germany, and New
Zealand—have reported that scientific evidence does not
show glyphosate causes cancer.

When the IARC report was released, EPA was
conducting its registration review of glyphosate, during
which it examined various scientific studies, including those
IARC considered. In 2017, EPA published its proposed
conclusion: Glyphosate was not likely to be carcinogenic to
humans. But, that same year, pursuant to Proposition 65,
12 HARDEMAN V. MONSANTO

California law categorized glyphosate as a chemical known
to the state to cause cancer. Cal. Off. of Env’t Health Hazard
Assessment, Glyphosate, (“Glyphosate Proposition 65”),
https://oehha.ca.gov/proposition-65/chemicals/glyphosate.
That classification triggered a state law requirement to attach
a warning label to glyphosate products. See id.; Cal. Health
& Safety Code § 25249.6.

In April 2019—one month after the jury verdict in this
case—EPA noted that commenters “expressed concerns that
glyphosate formulations are more toxic than glyphosate
alone.” Proposed Interim Registration Review at 10. EPA
explained that “there are few research projects that have
attempted to directly compare technical grade glyphosate to
the formulations under the same experimental design,” but
“[i]f at any time, information becomes available that
indicates adverse human health effects of concern for
exposure to glyphosate or its formulations, EPA intends to
review it and determine the appropriate regulatory action.”
Id. at 11.

About five months after the jury verdict, EPA issued a
letter to all registrants of glyphosate-containing products.
Letter from Michael L. Goodis, EPA, Office of Pesticide
Programs (Aug. 7, 2019) (“2019 letter”). The 2019 letter
was not the product of any formal proceeding, was not
published in the Federal Register, did not cite any new
scientific findings, and took no position on whether
Roundup causes cancer. Instead, this letter challenged
California’s inclusion of glyphosate in Proposition 65 as
contrary to “EPA’s determination that glyphosate is ‘not
likely to be carcinogenic to humans.’” Id. at 1. Given this
determination, EPA “considers the Proposition 65 warning
language” that glyphosate is carcinogenic “to constitute a
false and misleading statement” that violates FIFRA’s
 HARDEMAN V. MONSANTO 13

prohibition against “misbranded” substances. Id. 1–2 (citing
§ 136(q)(1)(A)). The letter concluded with EPA instructing
registrants to remove such warning statements from labels of
glyphosate-based pesticides. Id. at 2.

B

In 2016, Hardeman sued Monsanto alleging that his use
of Roundup—which started in the 1980s and ended in
2012—led to his diagnosis of non-Hodgkin’s lymphoma
(“NHL”) in early 2015. Hardeman’s case is one of
approximately 5,000 in federal court alleging that Roundup
causes NHL. The Judicial Panel on Multidistrict Litigation
consolidated those cases for pretrial proceedings in the
Northern District of California. Hardeman’s case was the
first of these consolidated cases to go to trial.

NHL is a cancer that affects white blood cells in the
immune system. Approximately 70% or more of NHL cases
are idiopathic, meaning they develop for unknown reasons.
However, some causes of NHL—such as hepatitis C
(“HCV”)—are well established. Hardeman had HCV for
25 to 40 years before developing NHL.

Hardeman alleged Monsanto’s failure to warn him of the
carcinogenic risks of Roundup caused his NHL. Monsanto
moved to dismiss, arguing that Hardeman’s claims were
preempted by FIFRA given EPA’s registration of
glyphosate, approval of the Roundup label, and
classification of glyphosate as non-carcinogenic. The
district court denied Monsanto’s motion. Monsanto raised
preemption again in a motion for summary judgment, which
the district court likewise denied.

The district court bifurcated the pretrial proceedings.
The first phase addressed “general causation”—whether
14 HARDEMAN V. MONSANTO

glyphosate can cause NHL at exposure levels humans might
experience. The second phase addressed “specific
causation”—whether Hardeman’s exposure to Roundup
caused his NHL.

The district court granted in part and denied in part
Monsanto’s motion to exclude Hardeman’s general
causation experts, allowing three of Hardeman’s experts to
testify—Dr. Portier, Dr. Ritz, and Dr. Weisenburger. These
experts introduced their general causation opinions with
scientific evidence from epidemiology (study of disease in
human populations), toxicology (animal studies), and
genotoxicology (cell studies); applied the Bradford Hill
criteria; 2 and used meta-analyses that combined and
analyzed the results of case-control studies.

The district court, however, acknowledged that
significant problems with Hardeman’s experts’ analyses
made it a “very close question” whether their testimony was
admissible to support general causation. In re Roundup
Prods. Liab. Litig., 390 F. Supp. 3d 1102, 1108 (N.D. Cal.
2018). The district court interpreted the Ninth Circuit’s
approach to Daubert as requiring “slightly more room for
deference to experts in close cases than might be appropriate
in some other Circuits.” Id. at 1113 (citations omitted).
Ultimately, the district court concluded Hardeman’s three

2
The Bradford Hill criteria are nine factors generally accepted as
relevant to assessing causation, such as: (1) the strength of the
association; (2) consistency; (3) specificity; (4) temporality;
(5) biological gradient or dose response; (6) biological plausibility;
(7) coherence with other scientific knowledge; (8) experimental
evidence; and (9) analogy. See In re Roundup Prods. Liab. Litig., 390 F.
Supp. 3d 1102, 1116 (N.D. Cal. 2018) (citing Austin Bradford Hill, The
Environment and Disease: Association or Causation?, 58 Proceedings
of the Royal Society of Medicine 295 (1965)).
 HARDEMAN V. MONSANTO 15

experts’ opinions were relevant and reliable, satisfying
Federal Rule of Evidence 702 and Daubert.

The district court later denied Monsanto’s motion to
exclude Hardeman’s specific causation experts.
Hardeman’s experts performed differential diagnosis, a
methodology by which a physician “rules in” all potential
causes of a disease, “rules out” those for “which there is no
plausible evidence of causation, and then determines the
most likely cause among those that cannot be excluded.”
Wendell v. GlaxoSmithKline LLC, 858 F.3d 1227, 1234 (9th
Cir. 2017). Hardeman’s experts considered various risk
factors beyond Roundup exposure that could explain his
disease, including age, race, obesity, hepatitis B (“HBV”),
and HCV, as well as idiopathic origin—i.e., no known cause.
They concluded Roundup caused Hardeman’s NHL by
ruling in Roundup based on general causation expert
opinions and ruling out HCV and idiopathy 3 as causes of
Hardeman’s NHL. The district court admitted the experts’
opinions, noting this circuit affords experts “wide latitude in
how they practice their art when offering causation
opinions.” In re Roundup, 358 F. Supp. 3d at 960 (citing
Wendell, 858 F.3d at 1237).

Monsanto requested a bifurcated trial, with the first
phase addressing whether Roundup caused Hardeman’s
cancer (without reference to any regulatory decisions
regarding glyphosate or Roundup) and the second phase

3
As to idiopathy, the district court held that “[i]t is sufficient for a
qualified expert, in reliance on his clinical experience, review of a
plaintiff[’s] medical records, and evaluation of the general causation
evidence, to conclude that an ‘obvious and known risk factor[]’ is the
cause of that plaintiff’s disease.” In re Roundup Prods. Liab. Litig.,
358 F. Supp. 3d 956, 960 (N.D. Cal. 2019) (quoting Wendell, 858 F.3d
at 1235).
16 HARDEMAN V. MONSANTO

addressing liability and damages (where the jury could see
some of that evidence). Monsanto moved to exclude all
evidence regarding IARC’s report, which detailed the
agency’s classification of glyphosate as probably
carcinogenic, as irrelevant and likely to confuse and distract
the jury. But if IARC evidence were admitted, Monsanto
argued, the district court should admit evidence that
numerous regulatory agencies around the world concluded
that glyphosate is safe.

Ultimately, the district court excluded IARC’s report but
admitted IARC’s classification of glyphosate as probably
carcinogenic to mitigate the prejudice caused to Hardeman
due to bifurcation of the trial. The district court also
admitted conclusions from EPA, EFSA, and ECA that
glyphosate was safe but excluded conclusions from other
regulatory bodies as cumulative.

At trial, Hardeman’s experts testified that his exposure
to glyphosate caused his NHL. Monsanto’s experts testified
that little evidence links glyphosate to cancer in humans and
that Hardeman’s HCV most likely caused his cancer or his
cancer was idiopathic.

The district court issued a “substantial factor” causation
instruction. The jury was instructed that, to rule for
Hardeman, it must find that glyphosate exposure was a but-
for cause of his cancer or one of two or more factors that
independently could have caused his cancer.

After Phase One (on causation), the jury returned a
verdict that Roundup exposure was a “substantial factor” in
causing Hardeman’s NHL. After Phase Two (on liability
and damages), the jury found that Monsanto failed to warn
about Roundup’s NHL risk and Hardeman was entitled to
punitive damages. The jury awarded Hardeman
 HARDEMAN V. MONSANTO 17

$5,267,634.10 in compensatory damages and $75 million in
punitive damages.

In post-trial motions, Monsanto argued that the district
court improperly excluded evidence of foreign regulatory
approvals of glyphosate, which allegedly deprived the jury
of the scope of evidence reinforcing Monsanto’s view of the
science. The district court explained that such evidence
about foreign regulators would have been cumulative under
Federal Rule of Evidence 403 and denied Monsanto’s
motion to overturn the verdict and for judgment as a matter
of law. But the district court reduced the punitive damages
award of $75 million to $20 million. These appeals
followed.

II

Whether Hardeman’s state claims are preempted is
reviewed de novo. Nathan Kimmel, Inc. v. DowElanco,
275 F.3d 1199, 1203 (9th Cir. 2002). Monsanto argues that
Hardeman’s failure-to-warn claims are preempted by
FIFRA, under which states cannot “impose . . . any
requirements for labeling or packaging in addition to or
different from” the requirements in FIFRA itself. § 136v(b)
(emphasis added); see also U.S. Const. art. VI, cl. 2 (federal
law “shall be the supreme Law of the Land . . . any Thing in
the Constitution or Laws of any State to the Contrary
notwithstanding”). We conclude that Hardeman’s failure-
to-warn claims based on Roundup’s labeling are consistent
with FIFRA and thus are neither expressly nor impliedly
preempted.

A

FIFRA does not expressly preempt Hardeman’s claims
because FIFRA’s requirement that a pesticide not be
18 HARDEMAN V. MONSANTO

misbranded is consistent with, if not broader than,
California’s common law duty to warn. Bates employs a
two-part test to determine whether FIFRA preempts a state
law claim. 544 U.S. at 444. First, the state law must be a
requirement “for labeling or packaging.” Id. (quoting
§ 136v(b)). Second, the state law must impose a labeling or
packaging requirement that is “in addition to or different
from” those required under FIFRA. Id. (quoting § 136v(b)).
Because Hardeman’s complaint is based on Monsanto’s
failure to provide an adequate warning on a label under
California law, part one of this test is satisfied.

As to part two of the Bates test, “a state-law labeling
requirement is not pre-empted by § 136v(b) if it is equivalent
to, and fully consistent with, FIFRA’s misbranding
provisions.” 544 U.S. at 447. State law is “equivalent to”
and “fully consistent with” FIFRA where both impose
“parallel requirements,” meaning that a violation of the state
law is also a violation of FIFRA. Id.; see also id. at 454
(“[A] manufacturer should not be held liable under a state
labeling requirement subject to § 136v(b) unless the
manufacturer is also liable for misbranding as defined by
FIFRA.”). Thus, if a violation of California’s duty to warn
would also be a violation of FIFRA’s misbranding provision,
then they impose parallel requirements fully consistent with
each other. Id. at 454 (“To survive pre-emption, the state-
law requirement need not be phrased in the identical
language as its corresponding FIFRA requirement . . . .”).
To that end, elements of California’s duty to warn and
FIFRA’s misbranding provision are compared below.

FIFRA’s misbranding provision requires a pesticide
label “contain a warning or caution statement which may be
necessary and if complied with . . . is adequate to protect
health and the environment.” § 136(q)(1)(G). Similarly,
 HARDEMAN V. MONSANTO 19

California common law requires a manufacturer to warn
either of any health risk 4 that is “known or knowable” (in
strict liability) or those risks “a reasonably prudent
manufacturer would have known and warned about” (in
negligence). Conte v. Wyeth, Inc., 85 Cal. Rptr. 3d 299, 310
(Ct. App. 2008). Thus, FIFRA—which requires a warning
“necessary” and “adequate to protect health”—is broader
than California’s requirement under negligence (no warning
needed if unreasonable to do so) 5 and is, at minimum,
consistent with California’s requirement under strict liability
(no warning needed if risk not known or knowable). See id.;
§ 136(q)(1)(G). Because FIFRA’s misbranding
requirements parallel those of California’s common law
duty, Hardeman’s failure-to-warn claims effectively enforce
FIFRA’s requirement against misbranding and are thus not
expressly preempted. See § 136(q)(1)(G); Bates, 544 U.S.
at 447–48 (citing favorably Justice O’Connor’s explanation
in Medtronic, 518 U.S. 470, that “a state cause of action that
seeks to enforce a federal requirement ‘does not impose a
requirement that is “different from, or in addition to,”
requirements under federal law’”).

4
Because a risk of cancer is a risk contemplated by FIFRA as
“necessary” and “adequate to protect health,” § 136(q)(1)(G), (x), (bb),
we need not address the possibility that California common law may
require a manufacturer to warn of a risk not contemplated by FIFRA’s
misbranding provision.
5
Though “it may be necessary as a matter of [state] law to prove
that th[e] violations were the result of negligent conduct . . . such
additional elements of the state-law cause of action would make the state
requirements narrower, not broader, than the federal requirement. While
such a narrower requirement might be ‘different from’ the federal rules
in a literal sense, such a difference would surely provide a strange reason
for finding pre-emption of a state rule insofar as it duplicates the federal
rule.” Medtronic, Inc. v. Lohr, 518 U.S. 470, 495 (1996).
20 HARDEMAN V. MONSANTO

Monsanto, however, argues that because EPA repeatedly
registered Roundup for sale without a cancer warning on the
label, a jury’s decision that Roundup should include such a
warning would effectively impose a requirement “in
addition to or different from” that required by FIFRA, and
so the state law is preempted. Granted, EPA is highly
involved in the pesticide registration process, which includes
approval of product labels. And EPA will not register a
pesticide unless it determines that the label “compl[ies]
with” FIFRA’s “requirements.” § 136a(c)(5)(B). But this
argument misses the point for two reasons.

First, EPA’s approval of a label—one step in a larger
registration process—is not conclusive of FIFRA
compliance. FIFRA specifies:

In no event shall registration of an article be
construed as a defense for the commission of
any offense under this subchapter. As long
as no cancellation proceedings are in effect
registration of a pesticide shall be prima facie
evidence that the pesticide, its labeling and
packaging comply with the registration
provisions of the subchapter.

§ 136a(f)(2) (emphasis added). 6 Because EPA has not
instituted any cancellation proceedings against Monsanto,

6
Section 136a(f)(2) distinguishes this case from Riegel v.
Medtronic, Inc., 552 U.S. 312 (2008), which held that the Medical
Device Amendments (“MDA”) to the Federal Food, Drug, and Cosmetic
Act (“FDCA”) expressly preempted claims challenging the safety and
effectiveness of a medical device that received premarket approval from
the Food and Drug Administration (“FDA”). Id. at 315, 330. Like
FIFRA’s preemption provision, the MDA preempts certain state
requirements that are different from, or in addition to, certain federal
 HARDEMAN V. MONSANTO 21

EPA’s approval of Roundup’s label is prima facie evidence
of FIFRA compliance. See id. And looking at FIFRA
holistically, this makes sense—if mere EPA approval of a
label were determinative of FIFRA compliance, then
FIFRA’s misbranding provision and regulations imposing a
duty to report “additional factual information regarding
unreasonable adverse effects” would serve no purpose.
§ 136d(a)(2); see also § 136(q)(1) (detailing when a
pesticide is misbranded); 40 C.F.R. § 159.152 (imposing
duty to report additional information on adverse effects). So
even though EPA approved Roundup’s label, a judge or jury
could disagree and find that same label violates FIFRA. And
because EPA’s labeling determinations are not dispositive of
FIFRA compliance, they similarly are not conclusive as to
which common law requirements are “in addition to or
different from” the requirements imposed by FIFRA. See
§ 136v(b); cf. Bates, 544 U.S. at 451 (“Private remedies that
enforce federal misbranding requirements would seem to
aid, rather than hinder, the functioning of FIFRA.”); Indian
Brand Farms, Inc. v. Novartis Crop Prot. Inc., 617 F.3d 207,
222 (3d Cir. 2010) (explaining that Bates “established that
mere inconsistency between the duty imposed by state law
and the content of a manufacturer’s labeling approved by the
EPA at registration did not necessarily mean that the state
law duty was preempted”).

Second, the EPA actions that Monsanto alleges preempt
Hardeman’s claims do not carry the force of law. As noted
in Bates, “[a] requirement is a rule of law that must be
obeyed.” 544 U.S. at 445. To establish requirements that

requirements. See 21 U.S.C. § 360k(a)(1). But the MDA does not
contain a provision like FIFRA’s § 136a(f)(2), which clarifies that the
agency’s approval of a label is not determinative of compliance with the
statute.
22 HARDEMAN V. MONSANTO

can preempt state law under § 136v(b), agency action must
have the force of law. See Wyeth v. Levine, 555 U.S. 555,
576, 580 (2009). In other words, only where there is a
relevant EPA action carrying the force of law are state
failure-to-warn claims prohibited from imposing
requirements inconsistent with that action. 7 Monsanto tries
to circumvent this caveat by arguing that although EPA’s
approval of Roundup’s label was not a rulemaking, it
happened “in the context of [a] registration process” that
“has the hallmarks of formal agency action.” See § 136a;
40 C.F.R. § 155.50(b)–(c). But, as explained above, FIFRA
expressly states that EPA’s decision to approve a label
during the registration process raises only a rebuttable
presumption that the pesticide and its label comply with
FIFRA. § 136a(f)(2). It would defy logic to say a rebuttable
presumption carries the force of law necessary to have
preemptive effect, as doing so would deny any ability to
rebut the presumption.

Nor does EPA’s 2019 letter, sent after the conclusion of
Hardeman’s trial to all registrants of products containing
glyphosate, carry the force of law. Generally, “Congress
contemplates administrative action with the effect of law
when it provides for a relatively formal administrative
procedure tending to foster the fairness and deliberation that

7
Monsanto relies on Bates’s explanation that a failure-to-warn
claim alleging that a label should have stated “DANGER” instead of
“CAUTION” would be preempted “because it is inconsistent with
40 C.F.R. § 156.64 (2004), which specifically assigns these warnings to
particular classes of pesticides based on their toxicity.” 544 U.S. at 453.
But this example deals with agency action that has the force of law—
FIFRA regulation 40 C.F.R. § 156.64. Here, however, neither EPA’s
approval of Roundup’s label during registration nor EPA’s 2019 letter
carries the force of law necessary to preempt Hardeman’s failure-to-warn
claims.
 HARDEMAN V. MONSANTO 23

should underlie a pronouncement of such force.” United
States v. Mead Corp., 533 U.S. 218, 230 (2001). But the
2019 letter—stating that EPA believes any pesticide label
with a cancer warning due to the presence of glyphosate will
be misbranded—did not follow any “formal administrative
procedure” that would give the letter the force of law. 8 See
id. The 2019 letter was issued without any written notice,
gave no hearing or opportunity to respond, and lacked any
sort of dispute-resolution process. See Merck Sharp &
Dohme Corp. v. Albrecht, 139 S. Ct. 1668, 1679 (2019).
Instead, the 2019 letter is similar to the letter in Fellner v.
Tri-Union Seafoods, LLC, which lacked preemptive effect
because the FDA “merely expressed an informal policy
opinion in a letter, and it did so only after [the plaintiff’s]
injuries were allegedly suffered.” 539 F.3d 237, 255 (3d Cir.
2008). 9

8
EPA’s 2017 determination that glyphosate is not carcinogenic does
not magically give the “force of law” to this 2019 letter on misbranding.
EPA’s 2017 determination was given in the context of glyphosate
“undergoing Registration Review” after evaluating glyphosate’s
carcinogenic potential. EPA, Revised Glyphosate Issue Paper:
Evaluation of Carcinogenic Potential 12 (Dec. 2017) (“Registration
Review also allows the agency to incorporate new science.”). Even if
the 2017 determination stems from more formal procedures, it is not
necessarily at odds with the future failure-to-warn claim because it was
made as part of EPA’s registration decision, which only supports
presumptive (not conclusive) compliance with FIFRA. See § 136a(f)(2).
9
In contrast, EPA’s cancellation proceedings, for example, may
have the force of law given that § 136d(b) lays out a formal notice and
hearing process, and no comparable prima facie evidence restriction
applies. See § 136a(f)(2) (stating that registration is “prima facie
evidence” of FIFRA compliance “[a]s long as no cancellation
proceedings are in effect”). But no cancellation proceedings were in
effect here.
24 HARDEMAN V. MONSANTO

Thus, we affirm the district court’s conclusion that
Hardeman’s state failure-to-warn claims are “equivalent to”
and “fully consistent with” FIFRA and therefore not
expressly preempted. See Bates, 544 U.S. at 449 (“The long
history of tort litigation against manufacturers of poisonous
substances adds force to the basic presumption against pre-
emption. If Congress had intended to deprive injured parties
of a long available form of compensation, it surely would
have expressed that intent more clearly.”). The Supreme
Court decided Bates over fifteen years ago, and regulatory
preemption in other contexts has developed considerably in
the interim. For FIFRA preemption, however, currently
Bates controls.

B

Because Monsanto could comply with both FIFRA and
California law, FIFRA did not impliedly preempt
Hardeman’s state failure-to-warn claims.

1

A state failure-to-warn claim is impliedly preempted if
the relevant federal and state laws “irreconcilably conflict.”
Merck, 139 S. Ct. at 1679 (quoting Rice v. Norman Williams
Co., 458 U.S. 654, 659 (1982)). “[S]tate and federal law
conflict where it is impossible for a private party to comply
with both state and federal requirements.” PLIVA, Inc. v.
Mensing, 564 U.S. 604, 618 (2011) (internal quotation marks
and citation omitted). To demonstrate an “irreconcilabl[e]
conflict,” Monsanto must present “clear evidence” that
(1) the agency was “fully informed” of “the justifications for
the warning” the plaintiff demands, (2) the agency has
“informed the . . . manufacturer that [it] would not approve
changing the . . . label to include that warning,” and (3) the
agency’s action “carr[ies] the force of law.” Merck, 139 S.
 HARDEMAN V. MONSANTO 25

Ct. at 1678–79. However, because EPA’s actions—such as
registering Roundup, approving Roundup’s label, and
issuing the 2019 letter—do not have the force of law,
Monsanto fails part (3) of Merck’s “clear evidence” of
“irreconcilabl[e] conflict” test and cannot show preemption.
See supra Section II.A.

2

Monsanto also argues that Hardeman’s claims are
impliedly preempted because, under EPA’s regulations,
Monsanto could not have unilaterally changed Roundup’s
label, making it impossible for Monsanto to comply with
both FIFRA and California’s common law duty to warn.
Monsanto relies primarily on PLIVA, a case concerning the
federal regulatory scheme governing generic drugs.
564 U.S. 604. But, as explained in PLIVA, “different federal
statutes and regulations may . . . lead to different pre-
emption results.” Id. at 626. Here, FIFRA’s regulatory
regime for pesticides differs meaningfully from the
regulatory scheme governing generic drugs in PLIVA and, as
a result, Monsanto’s implied preemption argument fails.

Under the regulatory scheme at issue in PLIVA, generic
drug manufacturers have an “ongoing federal duty of
sameness,” according to which they must use the same
labeling as the corresponding name-brand drug. Id. at 613
(internal quotation marks and citations omitted). Generic
drug manufacturers do not draft their products’ initial
labeling and do not have the power to revise labeling. See
id. As the Supreme Court explained, “[i]f [the generic drug
manufacturers] had [asked the FDA for help], and if the FDA
decided there was sufficient supporting information, and if
the FDA undertook negotiations with the brand-name
manufacturer, and if adequate label changes were decided on
and implemented, then the [generic drug] [m]anufacturers
26 HARDEMAN V. MONSANTO

would have started a Mouse Trap game that eventually led
to a better label.” Id. at 619. But, in PLIVA, the generic drug
manufacturer could not “independently satisfy . . . state
duties for pre-emption purposes” because it “cannot satisfy
its state duties without the Federal Government’s special
permission and assistance, which is dependent on the
exercise of judgment by a federal agency.” Id. at 623–24
(explaining that “[t]he only action the [generic drug]
[m]anufacturers could independently take” was “asking for
the FDA’s help”).

Unlike the FDCA and FDA regulatory scheme for
generic drug manufacturers, FIFRA and the EPA regulatory
scheme provide that pesticide manufacturers are responsible
for drafting their own product labels, § 136a(c)(1)(C), and
do not need to maintain the same labeling as another
manufacturer. Once a pesticide is registered, the
manufacturer has a “continuing obligation to adhere to
FIFRA’s labeling requirements.” Bates, 544 U.S. at 438
(citations omitted). When a label needs to be changed, the
manufacturer has the responsibility to change the label by
drafting and submitting the label to EPA for approval,
40 C.F.R. § 152.50(e), which EPA “shall” approve if it
determines the change will not violate FIFRA, § 136a(f)(1).
This is a far cry from the “special permission and assistance”
needed from the FDA in PLIVA to change a generic drug
label, a process constrained by a duty of sameness and the
added step of agency deliberations with name-brand
manufacturers. See 564 U.S. at 623–24.

Moreover, EPA permits pesticide manufacturers to make
certain changes to labels without prior approval. See id.
at 623. Specifically, manufacturers can make minor
modifications to labeling without prior EPA approval if EPA
is notified of the change. 40 C.F.R. § 152.46(a); EPA,
 HARDEMAN V. MONSANTO 27

Office of Pesticide Programs, Pesticide Registration Notice
98-10 (Oct. 22, 1998) (“PRN 98-10”). Thus, unlike the
generic drug manufacturers in PLIVA, pesticide
manufacturers “can act sufficiently independently under
federal law” when amending a label. See PLIVA, 564 U.S.
at 623.

Though Monsanto contends that “[a]dding a warning
about cancer would hardly qualify as a ‘minor
modification,’” EPA has repeatedly permitted pesticide
manufacturers to use the notification procedure to add
notices related to cancer to their products’ labels. 10
Nevertheless, Monsanto counters that there is no “single
example where EPA has allowed a registrant to use the
notification process” where EPA previously “found the
relevant chemical was not carcinogenic, much less where it
determined a cancer warning would render a label false and
misleading,” referring to the 2019 letter.

But neither EPA’s 2017 finding that glyphosate is not
carcinogenic nor the 2019 letter (which do not carry the force
of law) divert Monsanto to a different process for amending

10
For instance, pursuant to PRN 98-10, pesticide manufacturer
Bayer CropScience notified EPA “of a minor labeling amendment for
LARVIN Technical,” informing EPA that “[a]s required by California
Proposition 65, the following statement has been added to the label, ‘This
product contains a chemical known to the state of California to cause
cancer.’” Letter from Larry R. Hodges, Registration Manager, Bayer
CropScience, to EPA, Office of Pesticide Programs 4 (Nov. 29, 2012),
www3.epa.gov/pesticides/chem_search/ppls/000264-00343-20131217.
pdf. In response, EPA’s Registration Division “conducted a review of
this request for its applicability under PRN 98-10 and finds that the
action(s) requested fall within the scope of PRN 98-10.” Letter from
Jennifer Gaines, EPA, Office of Pesticide Programs, to Larry Hodges,
Bayer CropScience 2 (Dec. 17, 2012), www3.epa.gov/pesticides/chem_
search/ppls/000264-00343-20131217.pdf.
28 HARDEMAN V. MONSANTO

a label beyond those normally followed by pesticide
manufacturers under FIFRA and its regulations, as described
above. Considering the responsibility FIFRA places on
manufacturers to update pesticide labels and that EPA has
allowed pesticide manufacturers to add cancer warnings to
labels through the notification process without prior
approval, it is not impossible for Monsanto to add a cancer
warning to Roundup’s label. See PLIVA, 564 U.S. at 623;
see also Wyeth, 555 U.S. at 573 (explaining that
“[i]mpossibility pre-emption is a demanding defense”).

III

Whether the district court applied the correct legal
standard under Daubert is reviewed de novo, and the district
court’s decision to admit expert testimony is reviewed for
abuse of discretion. Estate of Barabin v. AstenJohnson, Inc.,
740 F.3d 457, 462 (9th Cir. 2014) (en banc), overruled on
other grounds by United States v. Bacon, 979 F.3d 766 (9th
Cir. 2020) (en banc). We hold that the district court
ultimately applied the correct legal standard under Daubert
and did not abuse its discretion by admitting Hardeman’s
general and specific causation expert testimony.

A

Under Federal Rule of Evidence 702, expert testimony
must be reliable to be admissible. Daubert, 509 U.S. at 589.
Scientific evidence is reliable when “the principles and
methodology used by an expert are grounded in the methods
of science.” Clausen v. M/V New Carissa, 339 F.3d 1049,
1056 (9th Cir. 2003). When determining reliability, district
court judges can consider the following non-exclusive
factors: (1) “whether the theory or technique employed by
the expert is generally accepted in the scientific
community;” (2) “whether it’s been subjected to peer review
 HARDEMAN V. MONSANTO 29

and publication;” (3) “whether it can be and has been tested;”
and (4) “whether the known or potential rate of error is
acceptable.” Daubert v. Merrell Dow Pharms., Inc., 43 F.3d
1311, 1316 (9th Cir. 1995) (citing Daubert, 509 U.S. at 593–
95). “Th[is] inquiry is ‘flexible,’” Wendell, 858 F.3d at 1232
(quoting Daubert, 509 U.S. at 594), and “should be applied
with a ‘liberal thrust’ favoring admission,” Messick v.
Novartis Pharms. Corp., 747 F.3d 1193, 1196 (9th Cir.
2014) (quoting Daubert, 509 U.S. at 588). 11

Monsanto contends that, by relying on a misguided
reading of Wendell and Messick, the district court
misinterpreted Daubert to be more forgiving of experts’
extrapolations than this circuit allows. But, in reaching its
conclusions, the district court followed this court’s
precedent and thus cannot be faulted for following binding
case law. Monsanto’s specific critiques are addressed
below.

First, according to Monsanto, the district court
erroneously stated there is “slightly more room for deference
to experts” in close cases, In re Roundup, 390 F. Supp. 3d
at 1113, and that courts in this circuit are “more tolerant of

11
This liberal thrust favoring admission is not without limits. “Just
as the district court cannot abdicate its role as gatekeeper, so too must it
avoid delegating that role to the jury.” Estate of Barabin, 740 F.3d at 464
(holding that district court erred by “pass[ing] its greatest concern about
[the expert’s] testimony to the jury to determine” and there was little
“indication that the district court assessed, or made findings regarding,
the scientific validity or methodology of [another expert’s] proposed
testimony”); see also United States v. Valencia-Lopez, 971 F.3d 891, 899
(9th Cir. 2020) (holding that district court erred in admitting expert
testimony without making a reliability determination by dismissing the
expert’s deficiencies as “going to the weight, not admissibility, of [the
expert’s] testimony” (quoting Nease v. Ford Motor Co., 848 F.3d 219,
230 (4th Cir. 2017))).
30 HARDEMAN V. MONSANTO

borderline expert opinions,” In re Roundup, 358 F. Supp. 3d
at 959.

As an initial matter, this court is not an outlier following
a more flexible Daubert approach than other circuits. The
cases on which the district court relied do not establish
otherwise. For instance, in the Fourth Circuit case relied on
by the district court, the expert failed to provide a proper
scientific basis for her differential diagnosis by “focus[ing]
almost exclusively on the fact that [plaintiff] took the drug
and later developed the disease, rather than explaining what
led her to believe that it was a substantial contributing factor
as compared to other possible causes.” In re Lipitor
(Atorvastatin Calcium) Mktg., Sales Pracs. & Prods. Liab.
Litig., 892 F.3d 624, 645 (4th Cir. 2018). But if we compare
the expert in Lipitor to, for instance, the expert in Messick,
the cases are readily distinguishable. Unlike the Lipitor
expert, the expert in Messick provided a scientific basis for
his conclusion by “refer[ing] to his own extensive clinical
experience as the basis for his differential diagnosis, as well
as his examination of [plaintiff’s] records, treatment, and
history.” 747 F.3d at 1198.

Similarly, in the Sixth Circuit case relied on by the
district court, the expert’s causation analysis was insufficient
because literature had only hypothesized but did not find a
link between the chemical and disease. Tamraz v. Lincoln
Elec. Co., 620 F.3d 665, 677–78 (6th Cir. 2010). The
Tamraz court explained, “the problem is not that [the expert]
failed to cite studies about [the chemical] causing [the
disease] . . . or could not quantify how much [of the
chemical] would lead to how much [of the disease]; the
problem is that he failed to cite any non-speculative evidence
for his conclusion.” Id. at 674. In contrast, the experts in
Wendell did not present that deficiency, as they “relied not
 HARDEMAN V. MONSANTO 31

just on . . . studies—which not only examined reported cases
but also used statistical analysis to come up with risk rates—
but also on their own wealth of experience and additional
literature.” 858 F.3d at 1236. Thus, the Fourth Circuit and
Sixth Circuit cases on which the district court relied are not
at odds with this court’s Daubert approach.

To the extent the district court relied on In re Zoloft
(Sertraline Hydrochloride) Products Liability Litigation,
858 F.3d 787, 800 (3d Cir. 2017), and McClain v. Metabolife
International, Inc., 401 F.3d 1233, 1244–45 (11th Cir.
2005), to show those courts adopted the any step
principles, 12 those cases do not reveal a more flexible
Daubert approach in this circuit. We have explained that
“expert evidence is inadmissible where the analysis is the
result of a faulty methodology or theory as opposed to
imperfect execution of laboratory techniques whose
theoretical foundation is sufficiently accepted in the
scientific community to pass muster under Daubert.” City
of Pomona v. SQM N. Am. Corp., 750 F.3d 1036, 1047–48
(9th Cir. 2014) (internal quotation marks and citation
omitted). Imperfect application of methodology may not
render expert testimony unreliable because “‘[a] minor flaw
in an expert’s reasoning or a slight modification of an
otherwise reliable method’ does not render expert testimony
inadmissible.” Id. at 1048 (quoting Amorgianos, 303 F.3d
at 267 (adopting the any step principles)) (alteration in
original). The reasoning guiding the any step principles is

12
“The Daubert ‘requirement that the expert testify to scientific
knowledge—conclusions supported by good grounds for each step in the
analysis—means that any step that renders the analysis unreliable under
the Daubert factors renders the expert’s testimony inadmissible.’”
Amorgianos v. Nat’l R.R. Passenger Corp., 303 F.3d 256, 267 (2d Cir.
2002) (quoting In re Paoli R.R. Yard PCB Litig., 35 F.3d 717, 745 (3d
Cir. 1994)); see also Zoloft, 858 F.3d at 797; McClain, 401 F.3d at 1245.
32 HARDEMAN V. MONSANTO

not dissimilar; namely, “[t]he judge should only exclude the
evidence if the flaw is large enough that the expert lacks
‘good grounds’ for his or her conclusions.” Amorgianos,
303 F.3d at 267 (quoting Paoli, 35 F.3d at 746).

Despite its incorrect assumption that this court is more
permissive than others in admitting Daubert testimony, the
district court still employed the correct legal standard for
reliability when it admitted Hardeman’s expert testimony.
For instance, the district court’s slight “deference to experts”
with “borderline . . . opinions” was proper under Daubert:
“[T]he interests of justice favor leaving difficult issues in the
hands of the jury and relying on the safeguards of the
adversary system . . . to ‘attack[] shaky but admissible
evidence.’” Wendell, 858 F.3d at 1237 (quoting Daubert,
509 U.S. at 596) (alteration in original). The Supreme Court
has not directed courts to follow a different rule since it first
decided Daubert almost 28 years ago.

Second, Monsanto takes issue with the district court’s
suggestion that courts in this circuit can admit opinions “that
lean strongly toward the ‘art’ side of the spectrum.” In re
Roundup, 358 F. Supp. 3d at 959 (citation omitted). Though
that may seem strange out of context, the district court was
only reiterating our precedent following Daubert. See
Messick, 747 F.3d at 1198 (“Medicine partakes of art as well
as science . . . .”). The district court did not suggest that
courts in this circuit allow “art” as a separate, standalone
category divorced from logic and science. Rather, in
referencing “art,” the district court followed Wendell and
Messick’s instructions that a testifying expert can rely on his
own extensive clinical experience under Daubert. See
Wendell, 858 F.3d at 1237 (“Where, as here, two doctors
who stand at or near the top of their field and have extensive
clinical experience with the rare disease or class of disease
 HARDEMAN V. MONSANTO 33

at issue, are prepared to give expert opinions supporting
causation, we conclude that Daubert poses no bar based on
their principles and methodology.”); Messick, 747 F.3d
at 1198 (allowing “extensive clinical experience” to form
basis of differential diagnosis opinion).

Monsanto attempts to distinguish Wendell by arguing
that it only allows experts to rely on clinical experience in
exceptional circumstances not present here, particularly
cases involving rare diseases with insufficient
epidemiological data. Considering that Wendell drew the
concept of “art” from Messick, a case which did not involve
a rare disease, we do not find that the application of art is
limited to exceptional circumstances.

The district court allowed experts to rely on clinical
experience, or “art,” only when conducting differential
diagnosis to render specific causation opinions. Allowing
experts to rely on clinical experience while conducting
differential diagnosis, as the district court did here, is
consistent with Messick. See 747 F.3d at 1198 (“[T]here is
nothing wrong with a doctor relying on extensive clinical
experience when making a differential diagnosis.”).

Monsanto further tries to distinguish Messick by
emphasizing that the expert there relied on clinical
experience as well as an examination of medical literature
and plaintiff’s records. But Hardeman’s experts did the
same thing here, if not more, by relying on epidemiological,
animal, and cell studies. Acknowledging this, Monsanto
counters that “there are numerous epidemiological studies
on the association between glyphosate and Hardeman’s
subtype of non-Hodgkin’s lymphoma that obviated the need
for any reliance on ‘art.’” But Monsanto contradicts its own
argument, asserting Wendell and Messick “state that
experience can supplement reliable scientific studies and
34 HARDEMAN V. MONSANTO

medical literature.” On this point, Monsanto is right:
Hardeman’s experts’ clinical experience could supplement
the epidemiological studies on which they relied.

Thus, the district court applied the correct legal standard
under Daubert by following our precedent and fulfilling its
“special obligation to determine the reliability of an expert’s
testimony.” Elsayed Mukhtar v. Cal. State Univ., Hayward,
299 F.3d 1053, 1063 (9th Cir. 2002) (internal quotation
marks and citation omitted), overruled on other grounds by
Estate of Barabin, 740 F.3d 457.

B

To establish general causation, Hardeman’s experts
needed to show that glyphosate can cause NHL at exposure
levels people realistically may have experienced. Here,
Hardeman’s general causation experts relied on three types
of studies: epidemiological, 13 animal, and cellular. Animal
studies are relevant evidence of causation where there is a
sound basis for extrapolating conclusions from those studies
to humans in real-world conditions. See Domingo ex rel.
Domingo v. T.K., 289 F.3d 600, 606 (9th Cir. 2002).
Similarly, cell studies can support more substantial evidence
of causation. Therefore, animal and cell studies can help
show causation so long as there is evidence of an association
between glyphosate and NHL in humans within the
epidemiological literature. This means that to be admissible
testimony, the experts must have reliably based their general
causation opinions on epidemiological evidence showing a
13
Epidemiology is “the field of public health and medicine that
studies the incidence, distribution, and etiology of disease in human
populations.” Michael D. Green et al., Reference Guide on
Epidemiology, in Reference Manual on Scientific Evidence 551, 551 (3d
ed. 2011) (“Reference Manual”).
 HARDEMAN V. MONSANTO 35

connection between glyphosate and cancer. As discussed
below, the district court did not abuse its discretion in
concluding that Hardeman’s experts satisfied this
requirement.

Monsanto maintains that the experts did not use the
epidemiological evidence reliably because they
(1) dismissed the Agricultural Health Study (“AHS”) and
(2) focused on case-control studies that did not sufficiently
account for confounding factors. These criticisms, however,
are not enough to render the expert opinions unreliable.

First, Monsanto criticizes Hardeman’s experts for
ignoring the AHS, which Monsanto considers to be the most
powerful evidence on the relationship between glyphosate
and NHL. That study was a cohort study conducted by the
National Institutes of Health that considered a range of
pesticide exposures on 57,000 participants over several
years. The AHS found no statistically significant association
between glyphosate and NHL and showed no dose-response
relationship, meaning “no evidence of higher rates of [NHL]
with more days of exposure.”

Nonetheless, Hardeman’s experts had a reasonable basis
for placing less weight on the AHS. For instance, an
epidemiologist employed by Monsanto wrote years before
the AHS results were announced that “the exposure
assessment in the AHS will be inaccurate” because the AHS
will have “spurious exposure-disease findings due to
exposure misclassification.” Similarly, Monsanto’s
toxicologist, Donna Farmer, recognized that “[m]any groups
have been highly critical of the study as being a flawed
study, in fact some have gone so far as to call it junk science
. . . . [T]he bottom line is scary . . . there will be associations
identified . . . just because of the way this study is designed.”
36 HARDEMAN V. MONSANTO

These criticisms from Monsanto employees resemble
those from Hardeman’s experts that the AHS is flawed and
unreliable. Though Monsanto changed its tune on the AHS
because the misclassification concerns were allegedly
addressed using “sensitivity analyses” as the study
progressed, the overlapping criticisms still show that
Hardeman’s expert opinions on the AHS are within “the
range where experts might reasonably differ.” Kumho Tire
Co. v. Carmichael, 526 U.S. 137, 153 (1999). Accordingly,
the district court did not abuse its discretion in concluding
that “the epidemiology evidence is open to different
interpretations” such that “an expert who places more weight
on the case-control studies than the AHS cannot be excluded
as categorically unreliable for doing so.” In re Roundup,
390 F. Supp. 3d at 1126.

Second, Monsanto criticizes Hardeman’s experts for
relying on three case-control studies: De Roos (2003),
McDuffie (2001), and Eriksson (2008), which allegedly
contain serious flaws. Although case-control studies are
“prone to recall bias,” 14 Hardeman’s experts gave the district
court valid reasons to discount this concern. For example,
the experts explained that epidemiology studies overall
found associations only between glyphosate and NHL, but
not between glyphosate and other cancers asked about in the
studies. The experts pointed out that, if participants were
pre-disposed to think glyphosate caused cancer and
consequently exhibited recall bias, the studies would have
reported associations for glyphosate and other cancers.

14
“[R]ecall bias[] occurs where people with a disease . . . are
differently able to recall past exposures than are people who never get
sick; generally, the assumption is that the cases will recall greater levels
of exposure, as those who become ill are more likely to ruminate about
the possible causes of their disease.” Reference Manual at 585–86.
 HARDEMAN V. MONSANTO 37

Hardeman’s experts also relied upon studies that sought to
validate self-reports of pesticide exposure and found similar
recall accuracy between controls and cases. Considering this
evidence, the district court did not abuse its discretion in
finding the “possible presence of recall bias” is “not
significant enough to require an expert categorically to
weight [the case-control studies] less heavily than the AHS.”
In re Roundup, 390 F. Supp. 3d at 1133.

Monsanto criticizes the De Roos study specifically for
“not properly account[ing] for [NHL’s] latency period”
because the study analyzed data collected between 1979 and
1986, but NHL takes “at least five to ten years to develop”
and Roundup was put on the market in 1974. As the district
court pointed out, a potential confounding variable 15 is an
important reason a study might show an association between
glyphosate and NHL shortly after glyphosate was put on the
market.

The De Roos study, however, reduced the risk of
confounding by adjusting for many other pesticides. While
Hardeman’s experts acknowledged that it is “always
possible” that the observed association was the result of
confounding not accounted for in De Roos, the adjustment
for many other pesticides in De Roos made it “significantly
less likely” that a pesticide other than glyphosate caused the
observed association. As a result, the district court properly
scrutinized the reliability of De Roos and did not abuse its
discretion in concluding that, “at least for the studies that
adjust for other pesticide exposures [i.e., De Roos], the
relatively short period between glyphosate exposure and
cancer development is not a concern so significant as to

15
Confounding variables are other factors that could explain an
observed association between a substance and the disease.
38 HARDEMAN V. MONSANTO

disqualify an expert who gives significant weight to the case-
control studies in rendering a causation opinion.” In re
Roundup, 390 F. Supp. 3d at 1123.

Nonetheless, Monsanto argues that Hardeman’s experts
still did not sufficiently consider confounding factors while
evaluating epidemiology. According to Monsanto,
“McDuffie did not account for the effect of exposure to
pesticides beyond glyphosate at all” and, “while Eriksson
did provide some results adjusted for the effect of other
pesticides, the adjusted results did not show a statistically
significant link between glyphosate and [NHL].”

But while the district court acknowledged that
“exclusive consideration of numbers unadjusted for other
pesticides, when adjusted numbers are available, would be
disqualifying,” In re Roundup, 390 F. Supp. 3d at 1140,
Hardeman’s experts did not do that here. For instance,
“Dr. Portier addressed the most significant concern—the
possibility that pesticides other than glyphosate caused the
observed cases of NHL—by focusing on data adjusted for
potential confounding by various other pesticides.” In re
Roundup, 390 F. Supp. 3d at 1133; see also id. at 1140–41,
1143 (discussing Dr. Ritz and Dr. Weisenburger). Further,
even where adjustment for other pesticides resulted in loss
of statistical significance, the results still showed a positive
association between glyphosate and NHL. 16 Thus, contrary

16
Monsanto criticizes Dr. Weisenburger for relying on a single
favorable odds ratio from the “earliest iteration” of the North American
Pooled Project. But such reliance is not enough to render
Dr. Weisenburger’s entire testimony unreliable. See, e.g., Wendell,
858 F.3d at 1233 (explaining that district courts should not look “too
narrowly at each individual consideration, without taking into account
the broader picture of the experts’ overall methodology”). The district
 HARDEMAN V. MONSANTO 39

to Monsanto’s criticisms, the general causation expert
opinions were sufficiently supported by reliable
epidemiological evidence, so admitting these experts’
testimony was not an abuse of discretion.

C

To establish specific causation, experts needed to show
that Hardeman’s NHL was caused by glyphosate, rather than
some other factor. To do so, Hardeman’s experts—
Dr. Weisenburger, Dr. Shustov, and Dr. Nabhan—used
“differential diagnosis,” which starts with ruling in “all
potential causes, then rul[ing] out the ones as to which there
is no plausible evidence of causation, and then determin[ing]
the most likely cause among those that cannot be excluded.”
Wendell, 858 F.3d at 1234; see also Clausen, 339 F.3d
at 1057. Here, Hardeman’s experts reliably used differential
diagnosis because they ruled in glyphosate based on the
epidemiological evidence supporting the general causation
opinions and ruled out alternative causes, such as idiopathy
and HCV.

1

Monsanto argues that Hardeman failed to adequately
rule out idiopathy, considering that 70% or more of NHL
cases have unknown causes. Monsanto acknowledges that
an expert can rule out idiopathy by reliably concluding that
the known factor (here, glyphosate) is a “substantial cause,”
which can be shown when a strong association exists
between the disease and that known risk factor. See Wendell,
858 F.3d at 1235, 1237 (even though expert “was not entirely

court properly considered this issue before concluding
Dr. Weisenburger’s testimony was sufficiently reliable.
40 HARDEMAN V. MONSANTO

able to rule” out idiopathy, he could conclude a “known risk
factor[]” was a substantial cause because “literature
show[ed] that patients exposed to” the drugs in question
were “at an increased risk for” the disease). But here,
Monsanto argues that Hardeman’s experts did not reliably
conclude that glyphosate was a substantial cause because no
strong association existed between glyphosate and NHL,
forcing the experts to rely on two flawed studies and their
own subjective judgment.

Specifically, Monsanto argues that Hardeman’s experts
did not rule in glyphosate as a substantial cause because,
unlike the experts in Wendell, they did not show a sharp
enough increased risk of cancer for those exposed to
glyphosate. Monsanto focuses on Hardeman’s experts’
inability to present a study with an adjusted odds ratio
above 2.0. But we have never suggested that a hardline
increase in a risk statistic, or even an adjusted odds ratio
above 2.0, is necessary for finding a strong association. See
id. at 1234. To the contrary, flexibility is warranted
considering the contextual nature of the Daubert inquiry.
Thus, it was not an abuse of discretion to admit expert
testimony—that glyphosate is a substantial cause—partly
based on the epidemiological studies from the general
causation opinions, where the general causation opinions
showed a “robust connection between glyphosate and NHL.”
In re Roundup, 358 F. Supp. 3d at 960.

Next, Monsanto criticizes Hardeman’s experts for
relying on “two flawed studies”—McDuffie and Eriksson—
linking glyphosate and NHL. Monsanto focuses on the
experts’ two attempted uses for those studies: (1) to assign a
quantified risk to Hardeman based on the studies’
“unadjusted numbers” and (2) to show that Hardeman’s risk
ratio must have exceeded 2.0 because he exceeded the
 HARDEMAN V. MONSANTO 41

exposure minimums from the two studies (i.e., two days per
year or ten lifetime days of exposure). But this focus is
misplaced. Though relying on McDuffie and Eriksson for
those propositions may have been problematic, that is not
what happened here. The district court explicitly considered
these issues and properly exercised its gatekeeping function
by precluding the experts from using the studies in those two
ways.

Instead, the district court allowed Hardeman’s experts to
rely on McDuffie and Eriksson to show a dose-response
relationship between glyphosate and NHL. And
Hardeman’s experts presented a sufficient basis for using
these studies (though unadjusted for other pesticides) to
show such a relationship. For instance, Dr. Weisenburger
clarified, if a chemical “shows a dose response, it’s very
likely an etiologic agent because it’s . . . unusual that a
chemical would cause a disease and not have a dose
response. So when you see a dose response, that gives you
some assurance that it really is causing the disease.”

Had the experts relied only on McDuffie and Eriksson to
show glyphosate is a substantial cause of NHL, their specific
causation opinions may have been unreliable. However,
Hardeman’s experts relied not only on McDuffie and
Eriksson but also other epidemiological evidence (like De
Roos) supporting a strong association, as well as their
clinical experience and review of plaintiff’s medical records.
Thus, as a whole, the evidence provided a sufficient basis for
reliably ruling out idiopathy by concluding glyphosate was
a substantial cause of Hardeman’s NHL. See Wendell,
858 F.3d at 1233–34 (ruling out idiopathy for disease with
70% idiopathy rate where expert relied on clinical
experience, literature, and medical records).
42 HARDEMAN V. MONSANTO

2

Monsanto also argues that Hardeman’s experts did not
reliably rule out HCV as an alternate cause. HCV is an
established cause of NHL. Even though Hardeman was
treated for HCV in 2005 and 2006, Monsanto claims that he
was vulnerable to cellular damage caused by the virus for
many years, including NHL. But, as Dr. Weisenburger
explained, to cause cancer, the virus must be active, and
there was no evidence that Hardeman’s HCV had been
active for the decade preceding his NHL diagnosis. And this
conclusion, as determined by the district court, had
significant support in the scientific literature.

Further, Dr. Weisenburger’s underlying methodology
for reaching this conclusion was sound. He relied on
Hardeman’s medical records and his clinical experience and
reviewed scientific literature (including seven studies) as the
basis for ruling out HCV. See Messick, 747 F.3d at 1199
(“[D]ifferential diagnosis grounded in significant clinical
experience and examination of medical records and
literature can certainly aid the trier of fact and cannot be
considered to be offering ‘junk science.’”). Thus,
Dr. Weisenburger reliably ruled out HCV as an alternate
cause of Hardeman’s NHL, and the district court did not
abuse its discretion in admitting Hardeman’s expert
testimony on specific causation.

IV

The district court’s decision to admit IARC’s
glyphosate classification as a “probable carcinogen” but
exclude contrary conclusions from other regulatory bodies is
reviewed for abuse of discretion. Estate of Barabin,
740 F.3d at 462. The district court made that decision to
 HARDEMAN V. MONSANTO 43

mitigate prejudice to Hardeman after granting Monsanto’s
request to bifurcate the trial.

Monsanto argues that admitting IARC’s classification
was an error because the classification’s minimal probative
value was outweighed by unfair prejudice and juror
confusion, which was allegedly exacerbated by the district
court’s exclusion of various foreign regulatory agencies’
rejections of IARC’s classification. We disagree.

Under Rule 403, the district court can “exclude relevant
evidence if its probative value is substantially outweighed by
a danger of one or more of the following: unfair prejudice,
confusing the issues, misleading the jury, undue delay,
wasting time, or needlessly presenting cumulative
evidence.” Fed. R. Evid. 403. “A district court’s Rule 403
determination is subject to great deference, because the
considerations arising under Rule 403 are susceptible only
to case-by-case determinations, requiring examination of the
surrounding facts, circumstances, and issues.” United States
v. Hinkson, 585 F.3d 1247, 1267 (9th Cir. 2009) (en banc)
(internal quotation marks and citation omitted).

According to Monsanto, IARC’s classification had
minimal probative value because it did not rely on new data
or gauge cancer risk from real-world glyphosate exposure.
But this misses the point: IARC’s classification was
admitted to mitigate prejudice to Hardeman from the trial’s
bifurcation. Monsanto had specifically requested
bifurcation to preclude evidence of its “attempting to
influence regulatory agencies and manipulate public opinion
regarding glyphosate.” Without IARC’s classification,
“jurors w[ould] be left wondering, during the causation
phase, how glyphosate could possibly be dangerous if it
ha[d] gone largely unregulated for decades.” Further, the
district court minimized the risk of prejudice to Monsanto by
44 HARDEMAN V. MONSANTO

only admitting IARC’s classification, not the underlying
details, and admitting the continued approval of glyphosate
from three other regulators—EPA, EFSA, and ECA.
Importantly, the district court instructed the jury to “not
defer” to the conclusions of any of these regulatory bodies
because they were not a substitute for the jurors’ “own
independent assessment of the evidence.” While other
regulatory agencies had also rejected IARC’s classification,
the district court did not err in concluding that evidence of
additional regulators’ post-IARC conclusions would have
been cumulative. Admitting all foreign regulatory
conclusions would have invited the jury to weigh competing
regulatory findings rather than independently assess the
scientific evidence.

Even if these evidentiary decisions were erroneous, any
error was harmless because it was “more probable than not
that the . . . admission of the evidence did not affect the
jury’s verdict.” United States v. Ramirez-Robles, 386 F.3d
1234, 1244 (9th Cir. 2004) (internal quotation marks and
citations omitted). Considering the strong limiting
instruction and the expert testimony linking glyphosate to
cancer, the jury would likely have reached the same
causation verdict even without evidence of IARC’s
classification or with more evidence of regulatory agency
rejections of that classification. Therefore, we affirm the
decision to admit the conclusions from IARC, EPA, EFSA,
and ECA, and to exclude evidence from additional
regulatory agencies.

V

Monsanto also challenges the district court’s causation
jury instruction. We review de novo whether that instruction
correctly states the law. Peralta v. Dillard, 744 F.3d 1076,
1082 (9th Cir. 2014) (en banc). We conclude that the district
 HARDEMAN V. MONSANTO 45

court’s causation jury instruction was inconsistent with the
Judicial Council of California Civil Jury Instructions
(“CACI”) and California case law. We conclude, however,
that any error was harmless. See Caballero v. City of
Concord, 956 F.2d 204, 206 (9th Cir. 1992).

The district court’s causation jury instruction included a
substantial factor and but-for causation instruction, drawing
from CACI 430, and a concurrent independent causes
instruction. 17 The first paragraph of this instruction (on
substantial factor and but-for causation) adopted the same

17
The jury was instructed as follows:

To prevail on the question of medical causation,
Mr. Hardeman must prove by a preponderance of the
evidence that Roundup was a substantial factor in
causing his non-Hodgkin’s lymphoma. A substantial
factor is a factor that a reasonable person would
consider to have contributed to the harm. It must be
more than a remote or trivial factor. It does not have
to be the only cause of the harm. Subject to the
additional instructions below, conduct is not a
substantial factor in causing harm if the same harm
would have occurred without that conduct.

The following additional instructions apply if you
believe that two or more NHL-causing factors
operated independently on Mr. Hardeman:

If you conclude that Mr. Hardeman has proven that his
exposure to Roundup was sufficient on its own to
cause his NHL, then you must find for Mr. Hardeman
even if you believe that other factors were also
sufficient on their own to cause his NHL. On the other
hand, if you conclude that Mr. Hardeman has not
proven that his exposure to Roundup was sufficient on
its own to cause his NHL, then you must find for
Monsanto.
46 HARDEMAN V. MONSANTO

language as CACI 430, the model “substantial factor”
instruction. The district court’s instruction included
CACI 430’s final optional sentence on but-for causation that
reads, “[c]onduct is not a substantial factor in causing harm
if the same harm would have occurred without that conduct.”
But CACI 430’s “Directions for Use” instruct courts to “not
include the [but-for instruction] in a case involving
concurrent independent causes,” which the district court did
here. As such, “the but-for test is inappropriate in cases
when two forces are actively operating and each is sufficient
to bring about the harm.” Lopez v. The Hillshire Brands Co.,
254 Cal. Rptr. 3d 377, 383–84 (Ct. App. 2019) (quoting
Major v. R.J. Reynolds Tobacco Co., 222 Cal. Rptr. 3d 563,
579 (Ct. App. 2017)). And this makes sense considering that
the two instructions tend to contradict each other when used
together.

Here, the district court’s causation jury instruction
erroneously incorporated the optional final sentence of
CACI 430. The concurrent independent causation
instruction was appropriate—otherwise, the jury might not
have found causation, even if it thought Roundup caused
Hardeman’s cancer, because HCV may have been an
additional cause. See Viner v. Sweet, 70 P.3d 1046, 1051
(Cal. 2003) (explaining there is an exception to but-for cause
for “multiple forces operating at the same time and
independently, each of which would have been sufficient by
itself to bring about the harm”). But because the concurrent
independent causation instruction inherently conflicted with
but-for causation, the district court’s jury instruction did not
state the law entirely correctly. See Peralta, 744 F.3d
at 1082. We recognize the district court tried to alleviate this
conflict by adding the introductory language of “[s]ubject to
the additional instructions below,” before providing the but-
 HARDEMAN V. MONSANTO 47

for causation instruction, but we still find that language
confusing, such that the instruction was “misleading.” Id.

An erroneous instruction does not require reversal,
however, when “the error is more probably than not
harmless.” Caballero, 956 F.2d at 206 (citation omitted).
That standard is “less stringent than review for harmless
error in a criminal case” and “more stringent than review for
sufficiency of the evidence.” Id. at 207. Because the
instruction given likely did not prejudice Monsanto, the
harmlessness standard is met. For instance, if the jury did
not view the but-for instruction as a bar to finding causation,
then it applied the appropriate causation standard. And even
if the jury interpreted the optional but-for sentence from
CACI 430 to mean Hardeman could only prevail if Roundup
was a but-for cause, then it would have also found legal
causation under the more flexible concurrent independent
causation standard. Thus, we affirm because the error in the
causation instruction was likely harmless.

VI

Monsanto argues it was entitled to judgment as a matter
of law on the failure-to-warn claims because it did not know
and could not have known that glyphosate caused cancer in
2012 (when Hardeman stopped using Roundup). But
reviewing de novo and “view[ing] the evidence in the light
most favorable to [Hardeman] . . . and draw[ing] all
reasonable inferences in h[is] favor,” Lakeside-Scott v.
Multnomah Cnty., 556 F.3d 797, 802 (9th Cir. 2009), we
conclude that sufficient scientific evidence was presented to
the jury to support that the association between glyphosate
and cancer was “knowable” by 2012.

To prevail on his failure-to-warn claim, Hardeman was
required to prove that the link between Roundup and cancer
48 HARDEMAN V. MONSANTO

was “known or knowable in light of the generally recognized
and prevailing best scientific and medical knowledge
available at the time of manufacture and distribution.”
Anderson v. Owens–Corning Fiberglas Corp., 810 P.2d 549,
558 (Cal. 1991). While the “scientific landscape” was “more
favorable” to Monsanto before 2012, there was sufficient
scientific evidence presented to the jury that the link between
glyphosate and cancer was “knowable.”

For instance, as early as 1985, EPA classified glyphosate
as a possible human carcinogen after reviewing a mouse
study finding that “[g]lyphosate was oncogenic in male
mice,” causing rare tumors. Even though EPA changed its
designation of glyphosate to non-carcinogenic in 1991,
several studies found an association between glyphosate and
cancer in the 1990s. In the late 1990s, Monsanto hired
Dr. Parry, a genotoxicologist, who found evidence that
glyphosate may be genotoxic and urged Monsanto to
conduct specific tests on Roundup’s genotoxicity. Though
Monsanto never conducted all the tests Dr. Parry
requested, 18 various independent scientific studies linking
glyphosate and cancer were released by 2012. Thus,
sufficient evidence was presented to the jury that the
association between glyphosate and cancer was, at
minimum, “knowable” by 2012, and Monsanto was
therefore not entitled to judgment as a matter of law.

VII

Finally, we address both parties’ challenges to the
punitive damages award. We review whether California law

Years later, in 2009, Monsanto toxicologist Donna Farmer said,
18

“you cannot say that Roundup does not cause cancer . . . [because] we
have not done carcinogenicity studies with ‘Roundup.’”
 HARDEMAN V. MONSANTO 49

permits a jury’s decision to award punitive damages for
substantial evidence. Kaffaga v. Estate of Steinbeck,
938 F.3d 1006, 1013 (9th Cir. 2019). We review de novo,
with an “[e]xacting appellate review,” the constitutionality
of a punitive damages award. State Farm Mut. Auto. Ins.
Co. v. Campbell, 538 U.S. 408, 418 (2003) (citing Cooper
Indus., Inc. v. Leatherman Tool Grp., Inc., 532 U.S. 424, 436
(2001)). “[W]e defer to the district court’s ‘findings of fact
unless they are clearly erroneous.’” Arizona v. ASARCO
LLC, 773 F.3d 1050, 1054 (9th Cir. 2014) (quoting Cooper
Indus., 532 U.S. at 440 n.14). Based on this review, we hold
that (1) California law permits a punitive damages award
because substantial evidence was presented to the jury that
Monsanto acted with malice, and (2) though the $75 million
punitive damages award was constitutionally excessive, the
reduced $20 million award comports with the outer limits of
the Due Process Clause.

A

Punitive damages were permissible under California law
because substantial evidence was presented that Monsanto
acted with malice by, among other things, ignoring
Roundup’s carcinogenic risks. See Kaffaga, 938 F.3d
at 1018.

Punitive damages are permissible under California law
when there is “clear and convincing evidence that the
defendant has been guilty of oppression, fraud, or malice.”
Cal. Civ. Code § 3294(a). As relevant here, “malice” means
“despicable conduct which is carried on by the defendant
with a willful and conscious disregard of the rights or safety
of others.” § 3294(c)(1).

That definition of malice requires that we examine what
constitutes “despicable conduct” and “conscious disregard.”
50 HARDEMAN V. MONSANTO

“Despicable conduct” is conduct “so vile, base,
contemptible, miserable, wretched or loathsome that it
would be looked down upon and despised by most ordinary
decent people.” Pac. Gas & Elec. Co. v. Super. Ct., 235 Cal.
Rptr. 3d 228, 236 (Ct. App. 2018) (internal quotation marks
and citation omitted). “Conscious disregard” requires that
the defendant “have actual knowledge of the risk of harm it
is creating and, in the face of that knowledge, fail to take
steps it knows will reduce or eliminate the risk of harm.” Id.
(internal quotation marks and citation omitted). But whether
a “defendant is aware of the probable dangerous
consequences of [its] conduct and [it] willfully fails to avoid
such consequences” can be “proved either expressly through
direct evidence or by implication through indirect evidence
from which the jury draws inferences.” Pfeifer v. John
Crane, Inc., 164 Cal. Rptr. 3d 112, 135 (Ct. App. 2013)
(quoting Angie M. v. Super. Ct., 44 Cal. Rptr. 2d 197, 204
(Ct. App. 1995)).

Substantial evidence of Monsanto’s malice was
presented to the jury, supporting punitive damages under
§ 3294(a). For example, internal emails were presented
supporting that Monsanto was consciously aware of the
potential health risks associated with Roundup. One email,
from Monsanto toxicologist Mark Martens, read, “I don’t
know for sure how suppliers would react—but if somebody
came to me and said they wanted to test Roundup I know
how I would react—with serious concern.” A second email,
from Monsanto toxicologist William Heydens, read,
“[g]lyphosate is OK but the formulated product (and thus the
surfactant) does the damage.” And a third email, from
Monsanto toxicologist Donna Farmer, read, “you cannot say
that Roundup is not a carcinogen . . . [because] we have not
done the necessary testing on the formulation to make that
statement.” These emails provide the substantial evidence
 HARDEMAN V. MONSANTO 51

necessary to support punitive damages based on Monsanto’s
awareness that Roundup posed a potential health risk.

There was also substantial evidence sufficient for a jury
to find that Monsanto “fail[ed] to take steps it kn[ew]
w[ould] reduce or eliminate the risk of harm.” Pac. Gas &
Elec. Co., 235 Cal. Rptr. 3d at 236 (internal quotation marks
and citation omitted). For instance, after its own hired
expert, Dr. Parry, found that glyphosate—alone and when
mixed with other chemicals in Roundup—had increased
genotoxic risks, evidence was sufficient to infer that
Monsanto largely failed to perform further studies. Instead,
Monsanto helped author an article downplaying
glyphosate’s health and safety concerns. Even though “it is
also possible to draw a contrary conclusion” that Monsanto
was ignorant or negligent (but not malicious), 19 the “jury’s
verdict must be upheld [because] it is supported by
substantial evidence” that Monsanto consciously
disregarded Roundup’s potential harm. See Pavao v. Pagay,
307 F.3d 915, 918 (9th Cir. 2002).

B

We next turn to the amount of punitive damages that
would still comport with the Due Process Clause. Hardeman
argues that the district court erred by reducing the jury’s
$75 million punitive damages award to $20 million. And
Monsanto contends that even the reduced punitive damages
award was unconstitutional under the Due Process Clause.

19
Monsanto also argues that it cannot be deemed to have acted with
malice because it complied with regulations. But “[a] defendant’s
compliance with, or actions consistent with, governmental regulations or
determinations about a product do not necessarily eviscerate a claim for
punitive damages.” Johnson & Johnson Talcum Powder Cases, 249 Cal.
Rptr. 3d 642, 678 (Ct. App. 2019).
52 HARDEMAN V. MONSANTO

Consistent with our “[e]xacting appellate review,” State
Farm, 538 U.S. at 418 (citing Cooper Indus., 532 U.S.
at 436), we lay out some fundamental principles underlying
the constitutionality of punitive damages awards.

“Compensatory damages and punitive damages serve
different purposes; compensatory damages redress concrete
loss caused by the defendant’s wrongful conduct, while
punitive damages are aimed at deterrence and retribution.”
Planned Parenthood of Columbia/Willamette Inc. v. Am.
Coal. of Life Activists, 422 F.3d 949, 953 (9th Cir. 2005)
(citing State Farm, 538 U.S. at 416; Cooper Indus., 532 U.S.
at 432). Further, “[t]he Supreme Court has instructed us to
go ‘no further’ if a ‘more modest punishment’ for the
‘reprehensible conduct’ at issue ‘could have satisfied the
State’s legitimate objectives’ of punishing and deterring
future misconduct.” Lompe v. Sunridge Partners, LLC,
818 F.3d 1041, 1065 (10th Cir. 2016) (quoting State Farm,
538 U.S. at 419–20). Ultimately, we are mindful that in
applying the Due Process Clause, it is “a constitution we are
expounding.” Tabares v. City of Huntington Beach,
988 F.3d 1119, 1122 (9th Cir. 2021) (quoting McCulloch v.
Maryland, 17 U.S. 316, 407 (1819)).

When punitive damages are “grossly excessive,” they
violate the Due Process Clause. State Farm, 538 U.S. at 416.
Whether punitive damages are “grossly excessive” depends
on three factors: “(1) the degree of reprehensibility of the
defendant’s misconduct; (2) the disparity between the actual
or potential harm suffered by the plaintiff and the punitive
damages award; and (3) the difference between the punitive
damages awarded by the jury and the civil penalties
authorized or imposed in comparable cases.” Id. at 418
(citing BMW of N. Am., Inc. v. Gore, 517 U.S. 559, 575
(1996)).
 HARDEMAN V. MONSANTO 53

1

The weightiest factor is “the degree of reprehensibility
of the defendant’s conduct.” Gore, 517 U.S. at 575. The
district court found “Monsanto’s approach to the safety of its
product was indeed reprehensible.” In re Roundup Prods.
Liab. Litig., 385 F. Supp. 3d 1042, 1047 (N.D. Cal. 2019).
The district court’s finding was reasonable and supported by
the facts presented to the jury. Thus, the question is to what
degree Monsanto’s actions were reprehensible. We
determine the reprehensibility of Monsanto’s conduct by
considering the following five factors: whether “[1] the harm
caused was physical as opposed to economic; [2] the tortious
conduct evinced an indifference to or a reckless disregard of
the health or safety of others; [3] the target of the conduct
had financial vulnerability; [4] the conduct involved
repeated actions or was an isolated incident; and [5] the harm
was the result of intentional malice, trickery, or deceit, or
mere accident.” State Farm, 538 U.S. at 419 (citing Gore,
517 U.S. at 576–77).

Several aggravating factors associated with
reprehensible conduct are present based on the evidence at
trial. First, the harm inflicted on Hardeman—cancer—was
physical, not purely economic. Hardeman has already been
well compensated for damages resulting from his physical
injury. Indeed, $5,066,667 of the compensatory damages—
about 96% of the jury’s $5,267,634.10 total compensatory
award—was based on noneconomic harm. See In re
Roundup, 385 F. Supp. 3d at 1045. The district court found
the $2 million in compensatory damages related to future
noneconomic harm were “borderline” high because
Hardeman’s cancer was in remission. Id. And while
remission is “no guarantee,” testimony showed his cancer is
unlikely to return. Id. But, as the district court explained,
54 HARDEMAN V. MONSANTO

this concern was “mitigate[d]” because “the jury likely
intended the future award to compensate a longer period of
suffering.” Id. These factual findings by the district court,
which are reasonable and not clearly erroneous, highlight the
reprehensibility of causing serious physical harm and the
need to deter future harm. While Hardeman was well
compensated for past and future harm, the serious nature of
the harm supports finding that Monsanto’s actions were
reprehensible.

Second, the district court’s factual conclusion that
Monsanto ignored safety risks is not clearly erroneous and
also supports reprehensibility. For example, the district
court found that “Monsanto’s behavior betrayed a lack of
concern about the risk that its product might be
carcinogenic.” Id. at 1047. In addition, it found that “the
evidence at trial painted the picture of a company focused on
attacking or undermining the people who raised concerns, to
the exclusion of being an objective arbiter of Roundup’s
safety.” Id. But the district court also found mitigating
evidence. Notwithstanding the jury’s verdict for Hardeman,
the district court explained that “the metaphorical jury is still
out on whether glyphosate causes NHL.” Id. Indeed, “there
is credible evidence on both sides of the scientific debate”
which “surely diminish[es]—to a degree—Monsanto’s
culpability.” Id. Moreover, “[t]he scientific landscape was
even more favorable to Monsanto during the time
Mr. Hardeman was using Roundup.” Id.

We also agree with the district court that no evidence was
presented that Monsanto knew Roundup in fact caused
cancer. Monsanto never conducted studies that may have
indicated (as its scientists suspected) that Roundup was
carcinogenic. And regulators, like EPA, have repeatedly
found glyphosate to not have carcinogenic risks. But, as the
 HARDEMAN V. MONSANTO 55

district court found, the evidence supports that Monsanto
knew Roundup might cause cancer, hence its concern and
reluctance to, for instance, conduct Dr. Parry’s
recommended studies. We have no quibble with any of the
district court’s findings of fact. Ultimately, evidence of
Monsanto’s conduct—downplaying concerns and failing to
fully assess Roundup’s safety after being alerted to possible
risks—supports that Monsanto acted with “indifference to or
a reckless disregard of the health or safety of others.” State
Farm, 538 U.S. at 419.

Third, State Farm asks us to look at Hardeman’s
financial vulnerability. Id. It goes without saying that this
is a case of a large corporation and an individual—not two
corporations on equal footing. Having said that, this factor
is not particularly relevant in a mostly noneconomic
damages case like this one. See Lompe, 818 F.3d at 1066
(“But as a practical matter, the financial vulnerability factor
does not have particular relevance . . . where the harm
[plaintiff] suffered was physical rather than a reprehensible
exploitation of financial vulnerability through fraud or other
financial misconduct.”). The district court below did not
analyze this factor. We do not find this factor helpful one
way or another to establish reprehensibility.

Fourth, the district court did not clearly err in finding that
Monsanto’s “conduct involved repeated actions” instead of
“an isolated incident.” See In re Roundup, 385 F. Supp. 3d
at 1047 (quoting State Farm, 538 U.S. at 419). Evidence
was presented that Monsanto repeatedly sold Roundup
without a warning label. Id. Thus, this factor supports
reprehensibility because “repeated misconduct is more
reprehensible than an individual instance of malfeasance.”
Gore, 517 U.S. at 577.
56 HARDEMAN V. MONSANTO

Fifth, the district court recognized Monsanto’s actions
exhibited malice but also made findings of fact that
mitigated this factor. The district court noted there was no
evidence “that Monsanto hid evidence from the EPA or,
alternatively, that it had managed to capture the EPA.” See
In re Roundup, 385 F. Supp. 3d at 1047. There was also no
evidence “that Monsanto was in fact aware that glyphosate
caused cancer but concealed it, thus distinguishing this case
from the many cases adjudicating the conduct of the tobacco
companies.” Id. Nonetheless, there was evidence of
Monsanto’s malice. As the district court found, “[d]espite
years of colorable claims in the scientific community that
Roundup causes NHL,” emails showed “Monsanto
employees crassly attempting to combat, undermine or
explain away challenges to Roundup’s safety.” Id. And “not
once was [the jury] shown an email suggesting that
Monsanto officials were actively committed to conducting
an objective assessment of its product.” Id. We do not find
the district court’s findings of fact clearly erroneous.

Based upon the district court’s findings, four of the five
factors support that Monsanto’s actions were reprehensible.
But in two of those factors, there were significant mitigating
considerations which suggest that Monsanto’s actions, while
reprehensible, were not “particularly egregious.” See Gore,
517 U.S. at 582.

2

We next examine the disparity between harm to
Hardeman and the punitive damages award by looking to the
Supreme Court’s guidelines on appropriate ratios. State
Farm, 538 U.S. at 424. The Supreme Court has explained
that “[s]ingle-digit multipliers are more likely to comport
with due process, while still achieving the State’s goals of
deterrence and retribution.” Id. at 425. “[A]n award of more
 HARDEMAN V. MONSANTO 57

than four times the amount of compensatory damages might
be close to the line of constitutional impropriety.” Id. (citing
Pac. Mut. Life Ins. Co. v. Haslip, 499 U.S. 1, 23–24 (1991)).
But there are “no rigid benchmarks that a punitive damages
award may not surpass” and greater ratios might “comport
with due process where ‘a particularly egregious act has
resulted in only a small amount of economic damages.’” Id.
(quoting Gore, 517 U.S. at 582).

Here, the jury awarded $5,267,634.10 in compensatory
damages and $75 million (approximately 14.2 times the
compensatory amount) in punitive damages. But
Monsanto’s conduct—though plausibly viewed as
reprehensible—was not “particularly egregious” as to
warrant a damages ratio above the single-digit range,
especially considering the absence of evidence showing a
known safety risk was intentionally concealed. See id. Thus,
we have little trouble holding that the jury’s 14.2 to 1 ratio
violated due process.

The $5,267,634.10 compensatory damages award was
substantial. See, e.g., Ramirez v. TransUnion LLC, 951 F.3d
1008, 1037 (9th Cir. 2020), cert. granted in part on other
grounds, No. 20-297, 2020 WL 7366280 (U.S. Dec. 16,
2020) (describing $8 million compensatory damages award
as “quite substantial”); Lompe, 818 F.3d at 1069
(“[C]ompensatory damages have often been considered
‘substantial’ when they are over $1,000,000.”). “When
compensatory damages are substantial, then a lesser ratio,
perhaps only equal to compensatory damages, can reach the
outermost limit of the due process guarantee.” State Farm,
538 U.S. at 425. But “State Farm’s 1:1 compensatory to
punitive damages ratio is not binding, no matter how
factually similar the cases may be.” Hangarter v. Provident
Life & Accident Ins. Co., 373 F.3d 998, 1014 (9th Cir. 2004).
58 HARDEMAN V. MONSANTO

Considering these precedents, we have held that “[i]n cases
where there are significant economic damages and punitive
damages are warranted but behavior is not particularly
egregious, a ratio of up to 4 to 1 serves as a good proxy for
the limits of constitutionality.” Planned Parenthood,
422 F.3d at 962.

Even though “substantial” compensatory damages were
awarded here, the evidence justifies a damages ratio higher
than 1 to 1. Monsanto intentionally downplayed and ignored
calls to test Roundup’s carcinogenic risks, and the jury
determined that Roundup caused Hardeman’s cancer.
Coupled with the physical damage—cancer—these factors
suggest a damages ratio up to 4 to 1 “serves as a good proxy
for the limits of constitutionality.” Id.; see State Farm,
538 U.S. at 425 (“The precise award in any case, of course,
must be based upon the facts and circumstances of the
defendant’s conduct and the harm to the plaintiff.”).

3

Third, the district court speculated that fines for failure
to warn of a product’s risk under FIFRA and the California
Health and Safety Code could potentially “over time[]
become quite high” because “both state and federal law
calculate penalties per violation.” In re Roundup, 385 F.
Supp. 3d at 1048; see also 7 U.S.C. § 136j(a)(1)(E);
40 C.F.R. § 19.4; Cal. Health & Safety Code § 25249.7. We
note the need to avoid speculation in analyzing this factor.
See State Farm, 538 U.S. at 428 (rejecting consideration of
speculative future penalties unrelated to plaintiffs’ harm).
The parties failed below, and again on appeal, to explain
what the relevant civil fines are, how they would be
calculated, and even whether they would be warranted. See
In re Roundup, 385 F. Supp. 3d at 1048. Monsanto points
out, however, that no civil or criminal fines have been
 HARDEMAN V. MONSANTO 59

imposed, apparently by any federal or any state agency,
including California.

Though California in 2017 categorized glyphosate as a
chemical known to the state to cause cancer, see Glyphosate
Proposition 65, it is also not clear that Monsanto would have
been subject to civil fines under California law in 2012.
Because neither party presents argument or evidence, we
agree with the district court that this guidepost is not
“particularly helpful here.” See id. at 1048 (“[A]bsent an
explanation from either party about how these penalties
would be calculated, it is difficult to use them as a
benchmark.”).

* * *

We hold that the jury’s $75 million punitive damages
award was “grossly excessive” given the mitigating factors
found by the district court. See State Farm, 538 U.S. at 416.
Considering the evidence of reprehensibility, however, we
hold that the district court’s reduced $20 million punitive
damages award (a 3.8 to 1 damages ratio), while at the outer
limits of constitutional propriety, ultimately comports with
due process. Planned Parenthood, 422 F.3d at 962; see also
Ramirez, 951 F.3d at 1037 (upholding 4 to 1 ratio where
$8 million compensatory damages awarded).

Though we uphold the district court’s $20 million
punitive damages award, we emphasize that the award is
“close to the line of constitutional impropriety.” See State
Farm, 538 U.S. at 425. Considering the number of cases
pending in this Roundup multidistrict litigation, we
recognize a smaller punitive damages award in other cases
may safely satisfy due process concerns by still imposing the
appropriate punishment and achieving the goals of
deterrence and retribution. Cf. Lompe, 818 F.3d at 1065
60 HARDEMAN V. MONSANTO

(“The Supreme Court has instructed us to go ‘no further’ if
a ‘more modest punishment’ for the ‘reprehensible conduct’
at issue ‘could have satisfied the State’s legitimate
objectives’ of punishing and deterring future misconduct.”
(quoting State Farm, 538 U.S. at 419–20)); see, e.g.,
Johnson v. Monsanto Co., 266 Cal. Rptr. 3d 111, 129 (Ct.
App. 2020), as modified on denial of reh’g (Aug. 18, 2020),
review denied (Oct. 21, 2020) (reducing punitive damages
award in a Roundup case to a 1 to 1 ratio with compensatory
damages of $10.3 million and where facts of Monsanto’s
reprehensibility were likely stronger than this case).

VIII

We are aware this appeal involves a bellwether trial with
potentially thousands of federal cases to follow. But many
of our holdings are fact-specific. Different Roundup cases
may present different considerations, leading to different
results. For example, were there evidence that EPA took
certain enforcement action against Monsanto after a cancer
warning was added to Roundup’s label, perhaps the
preemption analysis would lead to a different outcome. And
while our holding that expert testimony was admissible here
may be applicable to other Roundup cases, much of this
expert testimony was unique to Hardeman’s specific case.
Thus, it would not be unreasonable for the district court to
revisit the admissibility of expert testimony based upon the
facts raised in future cases. Similarly, despite the punitive
damages upheld here, a smaller punitive damages award in
future cases may better comport with due process.
Ultimately, we agree that the district court in this case either
reached the correct result or need not be reversed.

AFFIRMED.
 HARDEMAN V. MONSANTO 61

N.R. SMITH, Circuit Judge, dissenting to section VII.B.

After a mandated de novo review of the district court’s
punitive damages award, determining if the amount was
constitutionally excessive (not simply determining whether
the award was acceptable or reasonable), I must dissent. Let
me explain.

Punitive damages are “‘quasi-criminal,’ operat[ing] as
‘private fines’ intended to punish the defendant and to deter
future wrongdoing.” Cooper Indus., Inc. v. Leatherman
Tool Grp., Inc., 532 U.S. 424, 432 (2001) (quoting Pacific
Mut. Life Ins. Co. v. Haslip, 499 U.S. 1, 19 (1991)).
“Exacting appellate review ensures that an award of punitive
damages is based upon an ‘application of law, rather than a
decisionmaker’s caprice.’” State Farm Mut. Auto. Ins. Co.
v. Campbell, 538 U.S. 408, 418 (2003). As the majority has
stated, “the Supreme Court has instructed us to go ‘no
further’ if a ‘more modest punishment’ for the ‘reprehensible
conduct’ at issue ‘could have satisfied the State’s legitimate
objectives’ of punishing and deterring future misconduct.”
Lompe v. Sunridge Partners, LLC, 818 F.3d 1041, 1065
(10th Cir. 2016) (quoting State Farm, 538 U.S. at 419–20).
In order to determine de novo whether the punishment is
“grossly excessive,” the Supreme Court requires us “to
consider three guideposts: (1) the degree of reprehensibility
of the defendant’s misconduct; (2) the disparity between the
actual or potential harm suffered by the plaintiff and the
punitive damages award; and (3) the difference between the
punitive damages awarded by the jury and the civil penalties
authorized or imposed in comparable cases.” State Farm,
538 U.S. at 417–18. Of course, we always defer to the
district court’s findings of fact unless they are clearly
erroneous. Cooper Indus., 532 U.S. at 440 n.14.
62 HARDEMAN V. MONSANTO

The district court made the following findings of fact,
upon which one must make the analysis:

a. The jury found it was more likely than not
that Roundup (the glyphosate therein)
was a “substantial factor” in causing
Hardeman’s NHL. In order to evidence
that Roundup was a substantial factor in
this cause, Hardeman’s experts only
performed a differential diagnosis.
Differential diagnosis is a methodology
by which a physician “rules in” all
potential causes of a disease, “rules out”
those for “which there is no plausible
evidence of causation, and then
determines the most likely cause among
those that cannot be excluded.”

b. NHL is a cancer that affects white blood
cells in the immune system.
Approximately 70% or more of the NHL
cases are idiopathic, meaning they
develop for unknown reasons. However,
some causes of NHL, such as hepatitis C
(HCV), are well established. Hardeman
had HCV for 25 to 40 years before
developing NHL.

c. Hardeman was diagnosed with NHL in
early 2015. He started using Roundup in
the 1980s but ended his use in 2012.
During the time Hardeman was using
Roundup, the scientific landscape (of
whether it could cause cancer) was more
favorable to Monsanto than at the time of
trial. In 2012, EPA had little to no
 HARDEMAN V. MONSANTO 63

evidence that glyphosate was at all
carcinogenic in humans. Not until 2015
did the International Agency for Research
on Cancer (“IARC”) suggest that
glyphosate was probably carcinogenic to
humans.

d. Even today, there is credible evidence on
both sides with regard to whether
glyphosate causes NHL as documented
by the repeated approvals of glyphosate
by EPA, the European Chemicals
Agency, Health Canada, and other
worldwide regulatory agencies.

e. There is no evidence that Monsanto was
in fact aware that glyphosate caused
cancer; that Monsanto concealed it from
EPA; or that Monsanto somehow had
“captured” those in EPA, such that EPA
would not take a position contrary to
Monsanto.

f. The record at best shows that Monsanto
knew Roundup might cause cancer but
made minimal efforts to determine
whether the scientific evidence (finding
glyphosate may cause NHL) was
accurate.

g. However, Monsanto did attack or
undermine those who raised concerns for
Roundup’s safety.
64 HARDEMAN V. MONSANTO

h. Monsanto has sold Roundup without a
warning label. 1

i. The award of future noneconomic
damages was not based on physical pain
or impairment but was limited to
“anxiety, mental suffering, loss of
enjoyment of life, emotional distress, and
inconvenience.”

j. Hardeman’s NHL is now in remission,
his prognosis is “very good” and it is
“extremely unlikely” that his NHL will
return.

See In re Roundup Prod. Liab. Litig., 385 F. Supp. 3d 1042,
1047 (N.D. Cal. 2019).

1. The degree of reprehensibility of Monsanto’s
conduct.

Considering each of the three guideposts, the degree of
reprehensibility is “[t]he most important indicium of the
reasonableness of a punitive damages award.” See State
Farm, 538 U.S. at 419. The degree of reprehensibility is
determined by considering (1) “the harm caused was
physical as opposed to economic”; (2) “the tortious conduct
evinced an indifference to or a reckless disregard of the
health or safety of others”; (3) the target of the conduct had
financial vulnerability”; (4) “the conduct involved repeated
actions or was an isolated incident”; and (5) “the harm was

1
The district court noted that Monsanto continues to sell Roundup
without a warning label. However, “the conduct that harmed [plaintiff]
is the only conduct relevant to the reprehensibility analysis.” State Farm,
538 U.S. at 424.
 HARDEMAN V. MONSANTO 65

the result of intentional malice, trickery, or deceit, or mere
accident.” Id. Applying the facts (as determined by the
district court) de novo to these five considerations,
Monsanto’s conduct from the 1980s to 2012 did not
constitute the degree of “reprehensible conduct” that would
warrant an award of punitive damages at a 3.8:1 ratio. See
id. Reviewing these five considerations instead
demonstrates a low degree of reprehensibility.

First, while Hardeman suffered from physical harm
(NHL), he was well compensated for it by the jury.
Importantly, the physical harm suffered was not based on
acts or threats of violence, see Florez v. Delbovo, 939 F.
Supp. 1341, 1348 (N.D. Ill. 1996) (explaining that “acts of
violence or threats of bodily harm” are “the most
reprehensible” (citing BMW of N. Am., Inc. v. Gore, 517 U.S.
559, 575 (1996)), or “from some physical assault or trauma,”
see State Farm, 538 U.S. at 426. Further, (as demonstrated
below) Monsanto did not engage in deliberate conduct to
exploit Hardeman and expose him to a risk of cancer. Cf.
Bullock v. Philip Morris USA, Inc., 131 Cal. Rptr. 3d 382,
396 (Ct. App. 2011) (concluding “that in a case involving
physical harm, the physical or physiological vulnerability of
the target of the defendant’s conduct is an appropriate factor
to consider in determining the degree of reprehensibility,
particularly if the defendant deliberately exploited that
vulnerability”).

Second, one must determine whether Monsanto’s
conduct evinced “indifference to or a reckless disregard of
the health or safety of others.” State Farm, 538 U.S. at 419.
In California, Hardeman had to demonstrate that Monsanto
“had been guilty of oppression, fraud, or malice,” in order to
be awarded punitive damages from Monsanto. Cal. Civ.
Code § 3294(a). Section 3294(c)(1) outlines that
66 HARDEMAN V. MONSANTO

Monsanto’s conduct must have been undertaken “with a
willful and conscious disregard of the rights or safety of
others.” A conscious disregard “requires that the defendant
have actual knowledge of the risk of harm it is creating and,
in the face of that knowledge, fail[ed] to take steps it kn[ew
would] reduce or eliminate the risk of harm.” Ehrhardt v.
Brunswick, Inc., 231 Cal. Rptr. 60, 65 (Ct. App. 1986).
Given this standard, although ignoring evidence that
Roundup might cause cancer could be substantial evidence
to establish punitive damages against Monsanto, there was
and still exists “credible evidence on both sides of the
debate” about whether Roundup actually does cause cancer.
During the time that Hardeman used Roundup, the evidence
was scant that Roundup may cause cancer, but Monsanto did
disregard it. However, its conduct does not demonstrate (nor
did the court find) that Monsanto intentionally targeted
Hardeman.

Third, there is no evidence in the record that the target of
the conduct (Hardeman) had financial vulnerability. See
Clark v. Chrysler Corp., 436 F.3d 594, 604 (6th Cir. 2006)
(“The financial vulnerability of a target is particularly
relevant when the harm inflicted is economic in nature.”).
The wealth of Monsanto cannot justify an award of punitive
damages absent a connection of its “financial resources and
the physical injury suffered” by Hardeman. See id.

Fourth, Monsanto’s failure to place a warning on
Roundup’s label does not constitute “repeated actions.” 2 See
Gore, 517 U.S. at 577. “[E]vidence that a defendant has
repeatedly engaged in prohibited conduct while knowing or
suspecting that it was unlawful would provide relevant

2
As previously noted, the district court seems to rely on conduct
that occurred post 2012 in determining the reprehensibility.
 HARDEMAN V. MONSANTO 67

support for an argument that strong medicine is required to
cure the defendant’s disrespect for the law.” Id. at 576–77.
At the time Hardeman used the product, Monsanto was not
engaging in unlawful conduct. At that time, EPA had little
to no evidence that glyphosate was carcinogenic in humans.
In fact (again), there is credible evidence (to this day) on
both sides with regard to whether glyphosate causes NHL.
Notably, IARC did not decide to classify glyphosate as
“probably carcinogenic to humans” until 2015 (three years
after Hardeman stopped using Roundup). See In re Roundup
Prod. Liab. Litig., 385 F. Supp. 3d at 1047. Further, after
California’s passage of Proposition 65 (requiring a warning
label for glyphosate), Michael L. Goodis, EPA, Office of
Pesticide Programs, sent a letter to registrants (like
Monsanto) challenging Proposition 65 as contrary to “EPA’s
determination that glyphosate is ‘not likely to be
carcinogenic to humans.’” Proposed Interim Registration
Review at 11. The letter charged that the Proposition 65
warning was a “false and misleading statement” and violated
the Federal Insecticide, Fungicide, and Rodenticide Act
(“FIFRA”).

Fifth, Monsanto did not act with “intentional malice,
trickery, or deceit.” See State Farm, 538 U.S. at 419. As the
district court found, Monsanto acted with indifference, but
Monsanto did not engage in intentional acts, trickery, or
deceit. 3 See Williams v. First Advantage LNS Screening

3
Even the majority does not conclude that Monsanto acted with
“intentional malice, trickery, or deceit.” See Maj. Op. 53. Rather, it
describes Monsanto’s conduct as “malice.” Id. at 56. However,
“malice,” as found by the district court, means a “conscious disregard of
the rights or safety of others.” See In re Roundup Prod. Liab. Litig.,
385 F. Supp. 3d at 1046 (quoting Cal. Civ. Code § 3294(c)(1)). Thus,
there is no evidence that the harm suffered by Hardeman was the “result
of intentional malice”; Monsanto did not “intend[] to cause injury” to
68 HARDEMAN V. MONSANTO

Sols. Inc, 947 F.3d 735, 754 (11th Cir. 2020) (concluding
that “[a]t worst, Defendant acted recklessly, but without any
intent to harm Plaintiff”). In fact, Monsanto’s actions were
not contrary to “government regulations.” See Johnson &
Johnson Talcum Powder Cases, 249 Cal. Rptr. 3d 642, 678
(Ct. App. 2019), review denied (Oct. 23, 2019). Although
compliance with regulations cannot “eviscerate a claim for
punitive damages,” id., it does evidence that the harm “was
[not] the result of intentional malice, trickery, or deceit,”
State Farm, 538 U.S. at 419. As the district court found, the
association between Roundup and NHL “remains under
scientific investigation” and there was no evidence of
intentional acts on the part of Monsanto. Id.; cf. Satcher v.
Honda Motor Co., 52 F.3d 1311, 1316–17 (5th Cir. 1995)
(precluding punitive damages when there was a genuine
dispute in the scientific community).

Lastly, in reviewing these considerations, “some wrongs
are more blameworthy than others,” such as “violence,”
“trickery and deceit,” or “intentional malice” and are more
deserving of a higher punitive damages ratio. See Gore,
517 U.S. at 575. We have suggested that this “hierarchy of
reprehensibility” starts “with acts and threats of violence
. . . , followed by acts taken in reckless disregard for others’
health and safety, affirmative acts of trickery and deceit, and
finally, acts of omission and mere negligence.” Swinton v.
Potomac Corp., 270 F.3d 794, 818 (9th Cir. 2001) (citation

Hardeman. See Gober v. Ralphs Grocery Co., 40 Cal. Rptr. 3d 92, 106
(Ct. App. 2006). Nevertheless, despite a lack of evidence of this
“important criterion,” the majority still concludes that a punitive
damages award at outer constitutional boundaries for this case of
significant compensatory damages was appropriate. See Rhone-Poulenc
Agro, S.A. v. DeKalb Genetics Corp., 345 F.3d 1366, 1371 (Fed. Cir.
2003) (noting that this “factor has become an important criterion of what
the Constitution accepts as reprehensible conduct”).
 HARDEMAN V. MONSANTO 69

and quotation marks omitted). Thus, “[i]n order to justify a
substantial punitive damage award, a plaintiff ordinarily
must prove that the defendants’ conduct falls at the upper
end of the blameworthiness continuum, or, put another way,
that the conduct reflects a high level of culpability.”
Zimmerman v. Direct Fed. Credit Union, 262 F.3d 70, 82
(1st Cir. 2001).

No review of these considerations reflects “a high level
of culpability.” Id. Thus, Monsanto’s low degree of
reprehensibility cannot constitutionally justify the district
court’s substantial punitive damages award.

2. The disparity between harm suffered and
punitive damages award.

“The second and perhaps most commonly cited indicium
of an unreasonable or excessive punitive damages award is
its ratio to the actual harm inflicted on the plaintiff.” Gore,
517 U.S. at 580. In determining punitive damages for each
case, the Supreme Court has outlined that “the precise
award” of such damages “must be based upon the facts and
circumstances of the defendant’s conduct and the harm to the
plaintiff.” State Farm, 538 U.S. at 425. Although the Court
has not drawn a “bright-line ratio” for punitive damages, the
Court’s jurisprudence suggests “that, in practice, few awards
exceeding a single digit-ratio between punitive and
compensatory damages, to a significant degree, will satisfy
due process.” Id. “A higher ratio may . . . be justified in
cases” where (1) “the injury is hard to detect,” (2) “the
monetary value of noneconomic harm might have been
difficult to determine,” or (3) “a particularly egregious act
has resulted in only a small amount of [compensatory]
damages.” Gore, 517 U.S. at 582. The Court then clarified
the outer boundaries for such an award: “an award of more
than four times the amount of compensatory damages might
70 HARDEMAN V. MONSANTO

be close to the line of constitutional impropriety.” Id.
However, it also emphasized an outermost limit in making
such an award, stating “when compensatory damages are
substantial, then a lesser ratio [less than 4:1], perhaps only
equal to compensatory damages, can reach the outermost
limit of the due process guarantee.” Id. (emphasis added).

The compensatory damages in this case are substantial
($5,267,634.10) and the reasons to justify a higher ratio do
not exist. Thus, a punitive damages amount equal to
compensatory damages reaches the Supreme Court’s
outermost limit for punitive damages.

The California Supreme Court provides further
guidance, especially focusing on a case where there is a
relatively low reprehensibility. It said that “a ratio of one to
one might be the federal constitutional maximum in a case
involving . . . relatively low reprehensibility and a
substantial award of noneconomic damages: ‘When
compensatory damages are substantial, then a lesser ratio,
perhaps only equal to compensatory damages, can reach the
outermost limit of the due process guarantee.’” Roby v.
McKesson Corp., 219 P.3d 749, 769 (Cal. 2009) (quoting
State Farm, 538 U.S. at 425). In this case, the district court
reduced the jury’s 14.2:1 punitive damages award to nearly
a 4:1 ratio, which is generally reserved for a higher degree
of reprehensible conduct. Id. Monsanto’s conduct here did
not include (1) acts or threats of violence; or (2) acts of
trickery or deceit, evidencing a low degree of
reprehensibility.

Even in a case that involved conduct that was highly
reprehensible, the California Court of Appeal concluded that
“the permissible ratio of punitive to compensatory damages”
should be reduced when the noneconomic damages
“appear[ed] to include a punitive component.” See
 HARDEMAN V. MONSANTO 71

Bankhead v. ArvinMeritor, Inc., 139 Cal. Rptr. 3d 849, 866–
67 (Ct. App. 2012) (allowing punitive damages award at a
2.4:1 ratio). The Supreme Court agrees. See State Farm,
538 U.S. at 425–26, 429 (explaining that “in light of the
substantial compensatory damages awarded (a portion of
which contained a punitive element), a punitive damages
award at or near the amount of compensatory damages” was
justified).

The jury awarded substantial past and future
noneconomic damages totaling $5,066,667, which contain a
punitive element. The district court recognized this fact
when it noted that the $2,000,000 in future noneconomic
damages was “borderline,” because it was “somewhat
difficult to rationalize the conclusion that the suffering he
will face is, effectively, two-thirds of the suffering he has
already endured.” In re Roundup Prod. Liab. Litig. 385 F.
Supp. 3d. at 1045. Thus, a punitive damages award of 3.8:1
exceeded the constitutionally permissible limits. See State
Farm, 538 U.S. at 425; Roby, 219 P.3d at 769. The ratio of
punitive damages should be reduced to a 1:1 ratio. See id.

3. The difference between the punitive damages
awarded and the civil penalties authorized or
imposed in similar cases.

The third guidepost also supports punitive damages
equal to the compensatory damages award. Sanctions for
comparable misconduct can be determined by either the
“civil or criminal penalties that could be imposed for
comparable misconduct,” Gore, 517 U.S. at 583, or “the
existence of other civil awards against the defendant for the
same conduct,” Haslip, 499 U.S. at 22; see also Ismail v.
Cohen, 899 F.2d 183, 186 (2d Cir. 1990) (“Reference to
other awards in similar cases is proper.”).
72 HARDEMAN V. MONSANTO

One has difficulty comparing civil or criminal penalties
with this punitive damages award. During the time
Hardeman used Roundup, there were no federal or state
criminal or civil penalties for Monsanto’s conduct. Neither
the federal government nor the State of California had
imposed any penalties for the possibility that glyphosate may
cause cancer. 4 Although Monsanto’s conduct following the
harm can be considered in setting the punitive damages
award, see Johnson & Johnson Talcum Powder Cases,
249 Cal. Rptr. 3d at 678, California did not list glyphosate as
a chemical known to cause cancer until 2017. 5 It is similarly
difficult to determine how the federal government or
California would apply or calculate fines (which is probably
one of the reasons neither party really addressed this issue).

Comparing this case to the only other litigated case
against Monsanto regarding the sale of Roundup supports a
1:1 ratio. See Johnson v. Monsanto Co., 266 Cal. Rptr. 3d
111, 135 (Ct. App. 2020). In Johnson, the plaintiff
developed cancer in 2014 after using Roundup. Id. at 116–
17. Johnson sought damages, based on Monsanto’s
knowledge regarding Roundup’s carcinogenicity. Id. at 117.

4
Criminal and civil penalties may be imposed under federal and
state law. See 7 U.S.C. §§ 136j(a)(1)(E), 136l(a)(1); 40 C.F.R. § 19.4;
Cal. Health & Safety Code §§ 25249.6, 25249.7(a). Under federal law,
civil penalties may be assessed up to $5,000 for each offense. 7 U.S.C.
§ 136l(a)(1); see also 40 C.F.R. § 19.4 (2012). Criminal penalties may
result in either imprisonment of one year, a $50,000 fine, or both. Id.
§ 136l(b)(1). California law imposes a civil penalty up to $2,500 per day
for each violation. Cal. Health & Safety Code § 25249.7(a).
5
California Health & Safety Code section 25249.6 prohibits any
“person in the course of doing business [from] knowingly and
intentionally expos[ing] any individual to a chemical known to the state
to cause cancer or reproductive toxicity without first giving clear and
reasonable warning to such individual.”
 HARDEMAN V. MONSANTO 73

Reviewing the evidence, the district court reduced
compensatory damages to $10,253,309.32 and awarded
punitive damages at a 1:1 ratio. Id. at 129. The damages
awarded (for essentially “the same conduct”) in Johnson
provide a worthy comparison in assessing the
constitutionality of this punitive damages award. See
Haslip, 499 U.S. at 22; see also Restatement (Second) of
Torts § 908 cmt. e (1979) (noting that “[i]t seems appropriate
to take into consideration both the punitive damages that
have been awarded in prior suits and those that may be
granted in the future, with greater weight being given to the
prior awards”).

Finally, our sister circuits have come to similar
conclusions when dealing with substantial compensatory
damages (even when the conduct is highly reprehensible).
For example, in Boerner v. Brown & Williamson Tobacco
Co., the Eighth Circuit concluded that, despite American
Tobacco’s “highly reprehensible” conduct, the “punitive
damages award [of $15,000,000] [wa]s excessive when
measured against the substantial compensatory damages
award [of $4,025,000].” 394 F.3d 594, 603 (8th Cir. 2005).
Thus, it “conclude[d] that a ratio of approximately 1:1 would
comport with the requirements of due process.” Id.;
Saccameno v. U.S. Bank Nat’l Ass’n, 943 F.3d 1071, 1090
(7th Cir. 2019), cert. denied sub nom. Saccameno v. Ocwen
Loan Servicing, LLC, 140 S. Ct. 2674 (2020) (holding that
“a considerable compensatory award for the indifferent, not
malicious, mistreatment” and evidence that the “award
reflects emotional distress damages that ‘already contain [a]
punitive element’” “should not exceed 1:1”); Bridgeport
Music, Inc. v. Justin Combs Pub., 507 F.3d 470, 490 (6th
Cir. 2007) (“Given the large compensatory damages award
of $366,939, a substantial portion of which contained a
punitive element, and the low level of reprehensibility of
74 HARDEMAN V. MONSANTO

defendants’ conduct, a ratio of closer to 1:1 or 2:1 is all that
due process can tolerate in this case.”); see also Clark,
436 F.3d at 607 (holding that “because the compensatory
damage award here is not particularly large, a 1:1 ratio is
inappropriate. But due to the lack of several of
reprehensibility factors, any ratio higher than 2:1 is
unwarranted”).

4. Conclusion:

I start where I began. Because we are mandated to
review de novo the district court’s award of punitive
damages, one must undertake the review. In light of the
three guideposts, the district court’s $20,000,000 punitive
damages award exceeds the line of constitutionality. The
facts found by the district court do not support a 3.8:1 ratio
to compensatory damages. Most notably, Monsanto’s
conduct is not particularly reprehensible in light of the
ongoing scientific debate. The compensatory damages are
substantial; thus, punitive damages in an amount equal to
compensatory damages reaches the outermost limit of the
due process guarantee. Criminal and civil penalties and
punitive damages awarded in other cases do not suggest a
higher award. We then should go no further; this
punishment will satisfy the State’s legitimate objectives for
imposing such damages.