Genus Medical Technologies LLC v. FDA ( 2021 )

United States Court of Appeals
FOR THE DISTRICT OF COLUMBIA CIRCUIT
Argued December 10, 2020 Decided April 16, 2021
No. 20-5026
GENUS MEDICAL TECHNOLOGIES LLC,
APPELLEE
v.
UNITED STATES FOOD AND DRUG ADMINISTRATION,
APPELLANT
Appeal from the United States District Court
for the District of Columbia
(No. 1:19-cv-00544)
Daniel Winik, Attorney, U.S. Department of Justice,
argued the cause for appellant. With him on the briefs were
Jeffrey Bossert Clark, Acting Assistant Attorney General, Scott
R. McIntosh, Attorney, Robert P. Charrow, General Counsel,
U.S. Department of Health and Human Services, AnnaMarie
Kempic, Deputy Chief Counsel for Litigation.
Noam B. Fischman was on the brief for amicus curiae
Bracco Diagnostics Inc. in support of appellant.
James A. Boiani was on the brief for amicus curiae Giskit
B.V. in support of appellant.
2
Douglas B. Farquhar argued the cause and filed the brief
for appellee.
Before: HENDERSON, PILLARD and KATSAS, Circuit
Judges.
Opinion for the Court filed by Circuit Judge HENDERSON.
Opinion concurring in the judgment filed by Circuit Judge
PILLARD .
KAREN LECRAFT HENDERSON, Circuit Judge: The Federal
Food, Drug, and Cosmetic Act (FDCA or Act), 21 U.S.C.
§§ 301 et seq., sets forth separate and detailed regimes for the
regulation of medical products classified, inter alia, as drugs or
devices. The question before us is whether the U.S. Food and
Drug Administration (FDA) enjoys discretion to classify as a
“drug” a product that meets the statutory definition of a
“device.” The FDA claims that, if a medical product satisfies
the statutory definitions of both a “drug” and a “device,” the
Act’s overlapping definitions grant by implication the FDA
broad discretion to regulate the product under either regime.
Since 2017 the FDA has exercised its claimed discretion to
classify Genus Medical Technologies’ (Genus) “Vanilla SilQ”
line of diagnostic contrast agents as drugs, notwithstanding the
FDA’s recognition that the products “appear” to satisfy the
statutory definition for devices. Genus subsequently filed suit,
challenging the FDA’s classification decision as inconsistent
with the Administrative Procedure Act (APA), 5 U.S.C.
§ 706(2), and the FDCA. Finding that the FDCA
unambiguously forecloses the FDA’s interpretation, the district
court granted summary judgment in Genus’s favor and vacated
the FDA decision to classify Genus’s products as drugs. We
agree with the district court that the text, statutory structure and
legislative history of the Act make plain that the Congress did
3
not grant the FDA such sweeping discretion. Accordingly, we
affirm the district court’s grant of summary judgment.
I. BACKGROUND
A. Statutory & Regulatory Framework
The FDCA grants the FDA the authority to regulate certain
categories of medical products, including drugs, devices,
biologics and dietary supplements. Relevant here are the
statutory definitions for “drug” and “device.” The Act, in
relevant part, defines “drugs” to include:
articles intended for use in the diagnosis, cure,
mitigation, treatment, or prevention of disease
in man or other animals . . . .
21 U.S.C. § 321(g)(1)(B). “Devices” are defined to include:
an instrument, apparatus, implement, machine,
contrivance, implant, in vitro reagent, or other
similar or related article, including any
component, part, or accessory, which is . . .
intended for use in the diagnosis of disease or
other conditions, or in the cure, mitigation,
treatment, or prevention of disease, in man or
other animals, . . . and which does not achieve
its primary intended purposes through chemical
action within or on the body of man or other
animals and which is not dependent upon being
metabolized for the achievement of its primary
intended purposes.
4
Id. § 321(h)(1).1 Because the two definitions share a common
“intended-use clause”—that is, both definitions include articles
intended for use in the diagnosis, cure, mitigation, treatment or
prevention of disease—and because the drug definition
features no other relevant limitations, it is apparent that any
product that satisfies the “device” definition also satisfies the
definition of a “drug.” The converse, however, is not true.
Because a device must be “an instrument, apparatus,
implement, machine, contrivance, implant, in vitro reagent, or
other similar or related article,” and further, because it may
neither “achieve its primary intended purposes through
chemical action within or on the body of man” nor be
“dependent upon being metabolized for the achievement of its
primary intended purposes,” 2 the set of products that satisfy the
device definition is necessarily encompassed by, but narrower
than, the set of products that satisfy the drug definition.
Drugs and devices are subject to distinct regulatory
regimes. To begin, separate divisions of the FDA are primarily
responsible for each product category. Whereas drugs are
1
At the time of the FDA’s decision, the device definition was
located at 21 U.S.C. § 321(h). The Congress later relocated the
amendment to 21 U.S.C. § 321(h)(1). See Safeguarding
Therapeutics Act, Pub. L. No. 116-304, § 2(b), 134 Stat. 4915, 4916
(2011) (codified at 21 U.S.C. § 321(h)(1)).
2
Although FDA guidance refers to these “primary intended
purpose[]” limitations as the device definition’s “exclusionary
clause” or exclusionary clauses, Classification of Products as Drugs
and Devices & Additional Product Classification Issues: Guidance
for Industry and FDA Staff, U.S. Dep’t of Health and Human Servs.,
FDA, 6 & n.11 (Sept. 2017),
https://www.fda.gov/media/80384/download, we refer to them as the
“mode-of-action clauses” in order to distinguish them from the
“instrument clause,” which also has the effect of “excluding” certain
products that would otherwise satisfy the device definition. See
supra n.1.
5
generally regulated by the FDA’s Center for Drug Evaluation
and Research, devices are within the purview of the FDA’s
Center for Devices and Radiological Health.
The FDA holds new drugs to a high standard of pre-market
review and approval. To market a new prescription drug, the
sponsor (typically the manufacturer) must submit a new-drug
application and demonstrate through clinical trials that the drug
is safe and effective for its proposed use. 21 U.S.C. § 355(a)–
(b). Sponsors may, however, be able to take advantage of an
abbreviated new-drug application if their drug is sufficiently
similar to drugs that the FDA has previously approved. Id.§ 355(j).
The FDA’s pre-market review of devices is more varied.
Devices are assessed by the FDA and, with the assistance of
expert “classification panels,” classified into one of three
categories based on the risks they pose. Id. § 360c. First are
Class I devices, which are “subject only to minimal regulation
by ‘general controls’” because they “present no unreasonable
risk of illness or injury . . . .” Medtronic, Inc. v. Lohr, 518 U.S.
470, 476–77 (1996) (quoting 21 U.S.C. § 360c(a)(1)(A)).
Class II devices include “[d]evices that are potentially more
harmful” and, “although they may be marketed without
advance approval, manufacturers of such devices must comply
with federal performance regulations known as ‘special
controls.’” Id. at 477 (quoting 21 U.S.C. § 360c(a)(1)(B)).
Finally, devices that, inter alia, “‘presen[t] a potential
unreasonable risk of illness or injury,’ or which are ‘purported
or represented to be for a use in supporting or sustaining human
life or for a use which is of substantial importance in preventing
impairment of human health’” are generally classified as Class
III and, like drugs, subject to pre-market approval. Id.
(alteration in original) (quoting 21 U.S.C. § 360c(a)(1)(C)); 21
U.S.C. § 360e. To introduce a new Class III device into the
6
market, the sponsor must provide the FDA with “detailed
information regarding the safety and efficacy” of the device
and the FDA must have “‘reasonable assurance’ that the device
is both safe and effective.” Medtronic, 518 U.S. at 477(quoting 21 U.S.C. § 360e(d)(2)).
The regulatory differences do not end at the product
approval stage. Throughout the lifecycle of a medical product,
its treatment by the FDA depends upon its classification as
either a drug or a device. The FDCA sets forth separate rules
for, inter alia, annual manufacturer registration, compare 21
U.S.C. § 360(b)(1) (registration requirements for drug
manufacturers), with id. § 360(b)(2) (registration requirements
for device manufacturers); routine manufacturer inspections,
compare 21 U.S.C. § 360(h)(3) (risk-based inspection
schedules for drug manufacturers), with id. § 360(h)(2) (risk-
based inspection schedules for device manufacturers); routine
product reporting, see id. § 356i (reporting of marketing status
for drugs only); and adverse-event reporting, compare id.
§ 355b (reporting of adverse drug events), with id. § 360i
(records and reports on devices, including reporting of adverse
device events).
The result is that, on average, it is more costly for a sponsor
to develop and market a product as a drug than it would be to
develop and market an otherwise identical product as a device.
Genus maintains that its cost would be approximately $60,000
to seek device clearance for Vanilla SilQ—the product line in
question here. Genus Med. Techs., LLC v. FDA, 427 F. Supp.
3d 74, 78 (D.D.C. 2019). If, however, the same product line
were classified as drugs, Genus estimates that it would cost
them more than $500,000 to obtain pre-market approval in
addition to a recurring cost of more than $186,000 per year to
continue marketing their products as drugs. Id.7
Fortunately for sponsors, the FDCA contemplates at least
a limited role for sponsor input in the course of the product
classification process. Specifically, if the classification of a
product is unclear, a product sponsor may file a request for
designation (RFD) to obtain a formal, binding determination
from the FDA as to the “classification of the product . . . or . . .
the component of the [FDA] that will regulate the product.” 21
U.S.C. § 360bbb-2(a). A sponsor submits its RFD—including
a recommended classification—to the FDA’s Office of
Combination Products (OCP) and the OCP must respond
thereto no later than 60 days after the RFD’s filing. Id.
§ 360bbb-2(b). If the OCP fails to respond, the sponsor’s
recommended classification becomes final. Id. § 360bbb-2(c).
A classification made through the RFD process cannot be
changed “except with the written consent of the [sponsor], or
for public health reasons based on scientific evidence.” Id.
§ 360bbb-2(b)–(c).
B. Factual and Procedural History
Genus has manufactured its Vanilla SilQ product line
since 2015. Compl. ¶ 25. Vanilla SilQ belongs to a category
of products known as contrast agents. Contrast agents are used
in medical imaging to improve the visualization of tissues,
organs and physiological processes. According to Genus,
Vanilla SilQ is an oral solution used in combination with X-ray
examinations or other radiologic procedures to enhance the
visualization of the gastrointestinal tract for diagnostic
purposes. The product’s key ingredient is an inert metal salt
known as barium sulfate. When swallowed, the barium sulfate
coats the inside of the individual’s gastrointestinal tract and
facilitates the absorption of X-rays. Subsequently, the X-ray
examination will appear lighter for areas coated with barium
sulfate and darker for the surrounding tissues that are not
coated. Although some contrast agents cannot be classified as
8
devices because they achieve their primary intended purpose
through metabolization or chemical action within or on the
body of man, the FDA agrees that Genus’s Vanilla SilQ
product line “appear[s] to meet the definition of ‘device’”
insofar as it does not achieve its primary intended purposes
through either of the excluded modes.3 Appellant’s Br. 12–13.
Genus avers “that before and after it started producing
Vanilla SilQ, it sought FDA clearance to distribute its
products” as either devices or grandfathered drugs (which,
unlike new drugs, do not require pre-market approval). Genus,
427 F. Supp. 3d at 79. In June 2016, however, the FDA
conducted a three-day inspection of Genus’s distribution
facility. Id. The result of the inspection was a warning letter,
issued on May 2, 2017, notifying Genus that its products
constituted “drugs” within the meaning of the FDCA. Id.
Genus, responding to the FDA in a letter dated May 19, 2017,
asserted that its products are devices and that the FDA could
not regulate them as drugs because they do not “achieve [their]
3
We note that it is not immediately obvious to us how a contrast
agent satisfies the device definition’s requirement that the regulated
product be “an instrument, apparatus, implement, machine,
contrivance, implant, in vitro reagent, or other similar or related
article, including any component, part, or accessory . . . .” 21 U.S.C.
§ 321(h)(1). Nor is it altogether settled that Vanilla SilQ satisfies the
device definition’s mode-of-action clauses. Compare, e.g., Amicus
Bracco Br. 5–6 (arguing that Vanilla SilQ may not be regulated as a
device because it achieves its primary intended purpose through
chemical action), with Appellee’s Br. 52 (arguing that Vanilla SilQ
does not achieve its primary intended purpose through chemical
action). Because neither question is part of the administrative
decision now under review—the FDA found only that Genus’s
products “appear to meet” the device definition, see Joint Appendix
(J.A.) 122, 152, and both parties continue to agree that they do—we
reserve the question whether Vanilla SilQ satisfies the device
definition’s instrument and mode-of-action clauses.
9
primary intended purposes through chemical action within or
on the body” or through “metaboliz[ation].” J.A. 157–61
(quoting 21 U.S.C. § 321(h)(1)). On September 6, 2018, the
FDA responded that, “[a]lthough [the Vanilla SilQ products]
appear to meet the definition of ‘device’ . . . they also meet the
definition of ‘drug’ [under the FDCA] because they are articles
intended for use in the diagnosis of disease.” Id. at 152. The
FDA stated that “[w]hile [it] generally regulates products that
meet the definition of a device under the device authorities,
there are certain exceptions” and “[b]ecause not all contrast
agents meet the definition of a device, but all of them do meet
the definition of a drug, [it] has for many years regulated these
products as drugs in order to regulate them consistently under
the same authority . . . .” Id.Having failed to convince the FDA through its
correspondence, Genus next submitted an RFD, in which it
formally requested that the OCP classify its Vanilla SilQ
products as devices under the FDCA. Genus, 427 F. Supp. 3d
at 79–80. The OCP responded with an official Designation
Letter in which it echoed the FDA’s previous reasoning that,
although the Vanilla SilQ products appeared to meet the
definitions for both a device and a drug, it was nonetheless
appropriate to regulate uniformly all contrast agents as drugs.
Id. at 80.
On February 28, 2019 Genus filed suit in district court. In
addition to certain claims not relevant here, Genus claimed that
the FDA’s decision to regulate Vanilla SilQ as a drug rather
than as a device was arbitrary and capricious and in excess of
statutory authority under the FDCA and the APA. In a decision
filed December 6, 2019, the district court granted summary
judgment to Genus, concluding that the plain language of the
FDCA unambiguously requires that “a product that meets the
device definition must be regulated as such” and that the court
10
must therefore “end[] its analysis at Chevron step one.” Genus,
427 F. Supp. 3d at 84. The district court vacated the FDA’s
classification of Vanilla SilQ as a drug and remanded the
matter to the FDA for further proceedings. Id. at 87.
II. ANALYSIS
Our review of a summary judgment grant is de novo,
affirming only if “there is no genuine issue as to any material
fact [and] the moving party is entitled to judgment as a matter
of law.” Mylan Labs., Inc. v. Thompson, 389 F.3d 1272, 1278–
79 (D.C. Cir. 2004) (alteration in original) (quoting Trans
Union LLC v. Fed. Trade Comm’n, 295 F.3d 42, 48 (D.C. Cir.
2002)). In a case like this one, in which the district court
reviewed an agency action under the APA, “[w]e review the
administrative record and give no particular deference to the
District Court’s views.” Eagle Pharms., Inc. v. Azar, 952 F.3d
323, 329–30 (D.C. Cir. 2020) (quoting Am. Bankers Ass’n v.
Nat’l Credit Union Admin., 934 F.3d 649, 662 (D.C. Cir.
2019)). We review the FDA decision to classify Genus’s
products, then, under the familiar standards of the
Administrative Procedure Act, which require that we uphold
the FDA decision unless it is “arbitrary, capricious, an abuse of
discretion, or otherwise not in accordance with law” or “in
excess of statutory jurisdiction, authority, or limitations . . . .”
5 U.S.C. § 706(2). We defer to the FDA’s interpretation of the
FDCA “so long as the Congress has not unambiguously
forbidden it and it is otherwise permissible.” Cal. Metro
Mobile Commc’ns, Inc. v. FCC, 365 F.3d 38, 43 (D.C. Cir.
2004) (citing Chevron U.S.A., Inc. v. Nat. Res. Def. Council,
Inc., 467 U.S. 837, 842–43 (1984)); see also Teva Pharms.
USA, Inc. v. Sebelius, 595 F.3d 1303, 1315 (D.C. Cir. 2010)
(applying Chevron framework to FDA interpretations of
FDCA contained in letter rulings); Mylan Labs., 389 F.3d at
1279–80 (same). Our task requires that “[w]e examine the
11
statute’s text, structure, purpose, and legislative history to
determine if the Congress has expressed its intent
unambiguously.” Eagle Pharms., 952 F.3d at 330 (alteration
omitted) (quoting U.S. Sugar Corp. v. EPA, 830 F.3d 579, 605
(D.C. Cir. 2016) (per curiam)).
A. FDCA’s Text
The question before us is a purely legal one: whether the
FDCA grants the FDA discretion to classify as a “drug” a
product that satisfies the statutory definitions of both a “drug”
and a “device.” In answering the question, “[w]e begin ‘where
all such inquiries must begin: with the language of the statute
itself.’” Caraco Pharm. Labs., Ltd. v. Novo Nordisk A/S, 566
U.S. 399, 412 (2012) (quoting United States v. Ron Pair
Enters., Inc., 489 U.S. 235, 241 (1989)). We are mindful,
however, that if the text alone is insufficient to end the inquiry,
we may turn to other “customary statutory interpretation tools,”
including “‘structure, purpose, and legislative history.’” Cal.
Metro Mobile, 365 F.3d at 44–45 (quoting Consumer Elecs.
Ass’n v. FCC, 347 F.3d 291, 297 (D.C. Cir. 2003)); see also
Chevron, 467 U.S. at 843 n.9 (“If a court, employing traditional
tools of statutory construction, ascertains that Congress had an
intention on the precise question at issue, that intention is the
law and must be given effect.”) (emphasis added). We
conclude that the FDCA’s text unambiguously forecloses the
FDA’s interpretation.
The parties’ dispute is purely legal. Genus contends that
when a product satisfies both the drug and device definitions of
the FDCA, the product is a device. Conversely, the FDA
argues that it can choose whether to treat products that satisfy
both definitions as drugs or devices. Because the FDA’s legal
theory did not require it to do so, it made no factual findings
about whether the Vanilla SilQ products satisfied the particular
12
requirements of the FDCA’s device definition. Instead, it
found only that the products fell within the drug definition and
remarked that they “appear” to also satisfy the device
definition.
Beginning with the statute’s text, the FDA argues that,
because it is possible for a product to simultaneously satisfy the
linguistic demands of both the drug and device definitions, the
Congress must have granted the FDA discretion in such
instance to choose a classification. Simply put, any product
meeting the device definition may be classified as a device, any
product meeting the drug definition may be classified as a drug
and, according to the FDA’s reading, any product meeting both
definitions may be classified as either. To the extent the FDCA
is silent on how to treat products that meet both definitions, the
FDA argues that we should read it as a sign of statutory
ambiguity and defer to the FDA’s purportedly reasonable
interpretation.
Whereas the FDA draws our attention to the definitional
overlap, Genus urges us to focus on the elements of the device
definition that set it apart, including, most prominently, its
mode-of-action clauses. Genus argues that, because the drug
and device definitions are broadly similar except for the device
definition’s mode-of-action clauses—excluding products that
achieve their primary intended purposes through “chemical
action within or on the body of man” or “metaboliz[ation],” 21
U.S.C. § 321(h)(1)—products that do not achieve their primary
intended purposes through either excluded mode (and that
otherwise satisfy both definitions) must be regulated as devices
and devices alone. According to Genus, any other
interpretation would “effectively read[] the Mode of Action
Clause[s] out of the statute.” Appellee’s Br. 23.
13
Genus also urges us to rely on two traditional canons of
statutory construction. First, the “old and familiar rule” that
“the specific governs the general.” RadLAX Gateway Hotel,
LLC v. Amalgamated Bank, 566 U.S. 639, 645–46 (2012)
(quoting United States v. Chase, 135 U.S. 255, 260 (1890);
Morales v. Trans World Airlines, Inc., 504 U.S. 374, 384
(1992)). And second (and relatedly), the basic interpretive
canon that a “statute should be construed [to give effect] to all
its provisions, so that no part will be inoperative or superfluous,
void or insignificant.” Corley v. United States, 556 U.S. 303,
314 (2009) (quoting Hibbs v. Winn, 542 U.S. 88, 101 (2004)).
Applying these canons, Genus argues that the FDA’s
interpretation would render the device definition’s mode-of-
action clauses inoperative and allow the device definition to be
“swallowed by the more general drug definition.” Appellee’s
Br. 30–31 (quoting Genus, 427 F. Supp. 3d at 83).
Although we are unpersuaded that the FDA’s
interpretation would render the mode-of-action clauses
completely inoperative—under the FDA’s theory, the mode-
of-action clauses would still be necessary for differentiating
products that may be regulated as devices from those that may
not—we nonetheless agree with Genus that this is a case where
the specific must govern the general. The FDA does not
dispute that the FDCA’s definition of a “device” is drawn more
narrowly than its definition of a “drug.” Indeed, as we
discussed above, supra Section I.A, the set of products that
satisfy the device definition is necessarily encompassed by, but
narrower than, the set of products that satisfy the drug
definition.4 Moreover, the general-specific canon is
4
The concurring opinion contends that recognizing complete
overlap in the definitions would render the instrument clause
surplusage. Concurring Op. 7–8. But the instrument clause, like the
mode-of-action clauses, necessarily restricts which medical products
are devices. It does clear work in determining which medical
14
particularly appropriate where, as here, the provisions at issue
are “interrelated and closely positioned” as “parts of the same
statutory scheme.” RadLAX Gateway Hotel, 566 U.S. at 645
(alteration adopted) (quoting HCSC-Laundry v. United States,
450 U.S. 1, 6 (1981) (per curiam)). Thus, the device
definition’s instrument and mode-of-action clauses make it a
classic candidate for application of the canon that the specific
governs the general, and to the extent the drug and device
definitions conflict, it is the narrower definition—the device
definition—to which we must give effect. See D. Ginsberg &
Sons, Inc. v. Popkin, 285 U.S. 204, 208 (1932) (“Specific terms
prevail over the general in the same or another statute which
otherwise might be controlling.”).
The only question, then, is whether the two definitions are
truly in conflict. The FDA claims they are not. More
specifically, according to the FDA, the general-specific canon
is inapplicable here because it is “most frequently applied to
statutes in which a general permission or prohibition is
contradicted by a specific prohibition or permission” or where
“a general authorization and a more limited, specific
authorization exist side-by-side.” RadLAX Gateway Hotel, 566
U.S. at 645. The FDA argues that there is no such contradiction
here because the provisions in question are definitions as
opposed to authorizations or prohibitions and both definitions
can be given simultaneous effect. There is “no reason,”
according to the FDA’s opening brief, that “the statute must be
read so that a given product qualifies as either a ‘drug’ or a
‘device,’ but not both.” Appellant’s Br. 24 (emphasis in
original).
On this point the FDA is mistaken. In theory, it may be
possible for a product to satisfy both definitions at once. What
products, from among those that satisfy the broader drug definition,
also satisfy the narrower device definition.
15
the FDA omits, however, is that the FDCA’s statutory
definitions are meaningful only insofar as they carry concrete
regulatory consequences. As discussed, the FDCA elaborates
distinct regulatory regimes for drugs and devices. And each
scheme is mandatory: The FDCA prohibits the sale of “any
new drug” not approved under the regime for drug approvals.
21 U.S.C. § 355(a) (emphasis added). Similarly, all new Class
III devices are “required” to satisfy the pre-market review
regime for devices, id. § 360e(a), and Class I and Class II
devices must meet other distinct requirements, see id. § 360c.
Nor can the Secretary circumvent these requirements. Id.
§ 355(c)(1) (Secretary “shall” either approve new-drug
application pursuant to drug regime or deny application);
§ 360c(b)(1) (Secretary “shall” classify “all” new devices
intended for human use into three device classes). In short, it
is not textually possible to say that an item is a drug (or device)
but need not be regulated as such. And no one suggests that
the FDCA requires products meeting both definitions to be
regulated both as drugs and devices, which would create a
breathtaking example of statutory redundancy. The statute,
then, is clear: a product may be regulated as a drug or a device,
but not both, and while a single product may simultaneously
satisfy the linguistic elements of two definitions, it is not
possible for the FDA to give simultaneous effect to both. Thus,
this is precisely the sort of setting in which we must give effect
to the specific over the general. To do otherwise would be in
violation of the “settled” principle that “[h]owever inclusive
may be the general language of a statute, it will not be held to
apply to a matter specifically dealt with in another part of the
same enactment.” Fourco Glass Co. v. Transmirra Prods.
Corp., 353 U.S. 222, 228 (1957) (quoting Clifford F. MacEvoy
Co. v. U.S. ex rel. Calvin Tomkins Co., 322 U.S. 102, 107
(1944)).
16
Before proceeding to the parties’ structural claims, we
briefly dispatch with the FDA’s argument that we should be
guided by a 1990 amendment to the FDCA’s drug definition.
Specifically, the FDA argues that interpreting the drug and
device definitions as mutually exclusive would be to
“effectively read[] back into the statute” an old version of the
drug definition the Congress affirmatively abandoned when it
adopted the Safe Medical Devices Act of 1990 (SMDA), Pub.
L. No. 101-629, 104 Stat. 4511. Appellant’s Br. 21–22. Before
1990, the FDCA definition of a drug specifically excluded
“devices or their components, parts, or accessories.” FDCA,
Pub. L. No. 75-717, § 201(g), 52 Stat. 1040, 1041 (1938)
(codified as amended at 21 U.S.C. § 321(g)(1)). The 1990
SMDA struck this exclusionary language, thereby making it
possible for a single product to satisfy—simultaneously—the
terms of both definitions. SMDA § 16(b)(1), 104 Stat. at 4526.
The FDA argues that, by eliminating the drug definition’s
exclusionary language, the Congress granted it authority to
regulate certain products as either drugs or devices.
This argument presumes that the Congress dramatically
expanded the FDA’s authority by deleting a phrase from a
statutory definition. As the Supreme Court has counseled,
“[f]undamental changes in the scope of a statute are not
typically accomplished with so subtle a move.” Kellogg Brown
& Root Servs., Inc. v. U.S. ex rel. Carter, 575 U.S. 650, 661
(2015). Instead, we conclude that “the removal of the
. . . provision was more plausibly driven by” a narrower
concern. Id. The change occurred in a section of the statute
authorizing the FDA to “regulate products that constitute a
combination of a drug, device, or biological product,”
depending on “the primary mode of action of the combination
product.” SMDA § 16(a), 104 Stat. at 4526 (codified as
amended at 21 U.S.C. § 353(g)) (authorizing the regulation of
“combination products”). These new provisions thus created a
17
distinct regulatory regime that gave the Secretary flexibility to
determine the standards for pre-market review of these
combination products. See, e.g., 21 U.S.C.
§ 353(g)(2)(A)(ii)(I) (entitling sponsors of combination
products to meet with the Secretary to “address the standards
and requirements for market approval or clearance of the
combination product”); id. § 353(g)(7) (“Nothing in this
subsection shall prevent the Secretary from using any agency
resources of the [FDA] necessary to ensure adequate review of
the safety, effectiveness, or substantial equivalence of an
article.”). The definitional change helped to implement the
scheme by removing the previously categorical prohibition on
ever treating a drug as a device and vice versa. See Miller v.
Mylan Inc., 741 F.3d 674, 677 (6th Cir. 2014) (“The deletion
reflected the replacement of the binary scheme with a tripartite
scheme[ that included combination products].”). But the
amended definition provides no affirmative support for the
proposition that the FDA may treat drugs as devices—and vice
versa—even absent any combination. As explained above, the
FDCA’s basic textual architecture forecloses such reading of
the statute.
Legislative history confirms that the amendments seek
only to facilitate the FDA’s regulation of the new category of
“combination products.”5 See S. Rep. No. 101-513, at 43
(1990) (“Section 19 [of the SMDA] alters the drug and device
definitions in [21 U.S.C. § 321]. Language is removed from
the drug definition that will permit an approval of a
drug/device combination.”) (emphasis added); id. at 30 (“By
deleting this language, a product whose primary mode of action
5
We note that our analysis of the FDA’s argument regarding
the 1990 SMDA depends upon the FDCA’s legislative history, which
we further discuss in the following section, infra Section II.B.
Because the FDA’s SMDA argument is primarily textual, we address
it here.
18
is attributable to a drug, but has a device component, may be
reviewed under this Act’s drug authority.”). Thus, we read the
SMDA to facilitate the regulation of combination products, not
to grant the FDA near-limitless discretion to categorize as
drugs any product meeting the device definition.6
B. FDCA’s Structure, Purpose and Legislative History
We turn next to the FDCA’s structure, purpose and
legislative history. See, e.g., Pharm. Rsch. & Mfrs. of Am. v.
Thompson, 251 F.3d 219, 224 (D.C. Cir. 2001); see also
Roberts v. Sea-Land Servs., Inc., 566 U.S. 93, 101 (2012)
(“[Because s]tatutory language . . . ‘cannot be construed in a
vacuum . . . [i]t is a fundamental canon of statutory construction
that the words of a statute must be read in their context and with
a view to their place in the overall statutory scheme.’”)
(quoting Davis v. Mich. Dep’t of Treasury, 489 U.S. 803, 809
(1989)). All of these considerations reinforce our conclusion.
As set out above, supra Section I.A, the FDCA establishes
two distinct regulatory tracks, one for drugs and one for
devices. Although certain aspects of the regulatory regimes are
common, see, e.g., 21 U.S.C. § 352 (defining a single standard
by which “[a] drug or device shall be deemed to be
misbranded”), several vital aspects are not. Especially salient
here are the FDCA’s dual regimes for pre-market review and
approval. Subject to limited exceptions, new drugs require pre-
6
The concurring opinion reads the amendment as “just the kind
of ‘textual indication’ that may override” the general-specific canon.
See Concurring Op. 6–7 (alteration adopted) (quoting RadLAX
Gateway Hotel, 566 U.S. at 646). But deleting the express exclusion
from the drug definition only raised the general-specific question
without answering it. And as discussed, we decline to embrace such
a large grant of authority from a negative inference in statutory
history.
19
market approval based upon clinical showings of safety and
efficacy, see 21 U.S.C. § 355(a)–(b), while devices are subject
to varying levels of pre-market review depending upon the risk
they pose. Only Class III devices—so classified because they
“present[] a potential unreasonable risk of illness or injury” or
because they are “purported or represented to be for a use in
supporting or sustaining human life or for a use which is of
substantial importance in preventing impairment of human
health,” id. § 360c(a)(1)(C)—generally require pre-market
approval. Id. § 360e. Nor is the FDA’s authority over the
classification of devices entirely unfettered. The FDA is
required to convene expert panels to provide recommendations
on its device classification decisions, id. § 360c(b), and device
sponsors are entitled to participate in the classification process.
Id. § 360c(b)(6). It would make little sense, then, for the
Congress to have constructed such elaborate regulatory
regimes—carefully calibrated to products’ relative risk
levels—only for the FDA to possess the authority to upend the
statutory scheme by reclassifying any device as a drug, no
matter its relative risk level.
The legislative history underscores our analysis. As
discussed above, see supra Section I.A, setting aside products
that are not “an instrument, apparatus, implement, machine,
contrivance, implant, in vitro reagent, or other similar or related
article,” 21 U.S.C. § 321(h)(1), what distinguishes a drug from
a device under the FDCA is that a device excludes a product
that achieves its primary intended purposes through either
chemical action or metabolization. This, however, was not
always the case. The two mode-of-action clauses were added
to the FDCA’s device definition by the Medical Device
Amendments of 1976 (MDA), Pub. L. No. 94-295, 90 Stat.
539, and the legislative history strongly suggests that the
Congress’s aim, at least in part, was to formalize a
distinction—apparently already “administratively developed”
20
by the FDA—between drugs and devices based upon their
modes of action and the relative risk levels created by those
modes of action. See S. Rep. No. 94-33, at 6 (1975) (explaining
that the “FDA has administratively developed a distinction
between drug and device, which favors classifying a product as
a drug if its intended action is chemical, or based on highly
complex technology potential hazards of which may be
reduced through new drug controls”). Granted legislative
history is hardly dispositive, but we nonetheless see in it
additional evidence that the Congress established separate
regulatory tracks for drugs and devices and that the device
definition’s mode-of-action clauses were critical to
effectuating this bifurcated scheme.
The FDA offers its own arguments regarding the FDCA’s
statutory structure but none is compelling. Specifically, the
FDA directs our attention to two provisions of the FDA
Reauthorization Act of 2017 (FDARA), Pub. L. No. 115-52,
131 Stat. 1005, both of which relate to the marketing of a
medical imaging device intended for use with a prior-approved
contrast agent but the contrast agent’s new intended use is
different from the prior-approved intended use. The first
provision grants the FDA authority to approve certain “medical
imaging device[s]” notwithstanding they “involve[] the use of
a contrast agent” in a manner different from that described in
the agent’s approved labeling. Id. § 706(a), 131 Stat. at 1058–
59 (codified at 21 U.S.C. § 360j(p)). The FDA latches onto
language in this provision defining a “contrast agent,” in
relevant part, as “a drug that . . . is intended for use in
conjunction with an applicable medical imaging device . . . .”
Id. § 706(a), 131 Stat. at 1059 (emphasis added) (codified at 21
U.S.C. § 360j(p)(4)(B)). The second provision grants the FDA
authority to approve a contrast agent’s new intended use based
upon the submission of a supplement to the sponsor’s original
new drug application. Id. § 706(b), 131 Stat. at 1059–60
21
(codified at 21 U.S.C. § 355(y)). The FDA argues that both
FDARA provisions evince a congressional intent to ratify the
FDA’s practice of uniformly regulating all contrast agents as
drugs.
Despite its superficial appeal, the FDA’s argument is
unavailing. First, the FDARA’s definitional language—
defining a contrast agent as “a drug”—was explicitly provided
“[f]or purposes of this subsection . . . .” Id. § 706(a), 131 Stat.
at 1059 (emphasis added) (codified at 21 U.S.C. § 360j(p)(4)).
Subsection 360j(p) authorizes treating as devices the
combination of imaging devices paired with contrasting agents
previously approved as drugs. 21 U.S.C. § 360j(p). But some
contrast agents work through chemical action within the body
and thus plainly are drugs. See id. § 321(h)(1). The special
rules provided in this context thus hardly suggest that the FDA
may generally treat all contrast agents as drugs, much less
constitute “express congressional approval” for such a rule.
Gen. Am. Transp. Corp. v. ICC, 872 F.2d 1048, 1053 (D.C. Cir.
1989) (quoting AFL-CIO v. Brock, 835 F.2d 912, 915 (D.C.
Cir. 1987)). The second provision cited by the FDA, 21 U.S.C.
§ 355(y), likewise governs contrast agents previously approved
as drugs, which does not suggest that all contrast agents must
(or even may be) so classified. Even more fundamentally, we
strongly doubt that the Congress would have chosen to hide
such a major grant of regulatory discretion in so narrow an
amendment. Had it intended to endorse the view that the FDA
may regulate all contrast agents as drugs rather than devices, it
would have used more explicit language to do so. Here as
elsewhere we “must be guided to a degree by common sense as
to the manner in which Congress is likely to delegate a policy
decision of such economic and political magnitude to an
administrative agency” and we are skeptical that the Congress
would grant the FDA such vast authority “in so cryptic a
22
fashion.” FDA v. Brown & Williamson Tobacco Corp., 529
U.S. 120, 133, 160 (2000).
The FDCA’s structure, purpose and legislative history
confirm our reading of the text. See Teva Pharms., 595 F.3d at
1315 (analyzing statutory structure and setting aside FDA’s
interpretation of FDCA where text alone “hardly rules out
alternative readings that, absent consideration of statutory
structure, also appear plausible”). They make plain that the
Congress did not grant the FDA near-limitless discretion to
classify any device as a drug. Rather, the Congress has
elaborated separate regulatory tracks for drugs and devices and,
to the extent that the FDA possesses the discretion to choose
one track or the other, such discretion must be exercised in a
manner consistent with the statutory “drug” and “device”
definitions.
We note that we are especially troubled by the FDA’s
inability to articulate a limiting principle with which to cabin
its asserted discretion. The FDA offers only that, like all other
agency actions, its classification decisions are subject to the
APA’s arbitrary and capricious standard. But the arbitrary and
capricious standard is necessarily narrow, see Motor Vehicle
Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463
U.S. 29, 43 (1983), and seldom is that more true than in the
context of highly technical matters like the FDA’s medical
product classifications, see, e.g., Kennecott Greens Creek Min.
Co. v. Mine Safety & Health Admin., 476 F.3d 946, 954–55
(D.C. Cir. 2007) (agency is entitled to “extreme degree of
deference” when “evaluating scientific data within its technical
expertise”) (quoting Hüls Am. Inc. v. Browner, 83 F.3d 445,
452 (D.C. Cir. 1996)). Thus, what the FDA attempts to claim
for itself is the near-limitless authority to classify any device as
a drug, subject only to a highly deferential standard of judicial
review. We cannot reasonably infer such broad discretion
23
without a clearer statement. See Whitman v. Am. Trucking
Ass’ns, Inc., 531 U.S. 457, 467–68 (2001) (“Congress, we have
held, does not alter the fundamental details of a regulatory
scheme in vague terms or ancillary provisions—it does not, one
might say, hide elephants in mouseholes.”).7
The concurring opinion stresses the need for the FDA to
bring its expertise to bear on close questions regarding whether
the instrument clause covers particular kinds of medical
products. Concurring Op. 11–14. We do not deny that when
the instrument clause (and, for that matter, the mode-of-action
clauses) are ambiguous in their application to some kind of
medical product, reviewing courts should respectfully consider
the expert views of the FDA, which may even qualify for
Chevron deference. But here, the FDA did not invoke its
expertise to contend that Vanilla SilQ does not satisfy the
device definition and so should be regulated as a drug. Instead,
it assumed that Vanilla SilQ meets the definition of a device
but nevertheless undertook to regulate it as a drug. In rejecting
that position, we do nothing to restrict the agency’s discretion
to determine, in close cases, whether a particular product
satisfies the device definition. Indeed, we expressly reserve
that question in this case. See supra n.3.
We emphasize the purely legal nature of the question
before us. Because its interpretation of the FDCA did not
require it to do so, the FDA made no factual findings with
respect to Vanilla SilQ except that it is an article intended for
use in the diagnosis, cure, mitigation, treatment or prevention
7
We also disagree that the FDA’s restrained use of discretion in the
past blunts this concern or otherwise provides a basis to limit any
purported discretion. See Carlson v. Postal Reg. Comm’n, 938 F.3d
337, 349 (D.C. Cir. 2019) (“[N]o amount of historical consistency
can transmute an unreasoned statutory interpretation into a reasoned
one.” (internal quotations omitted)).
24
of disease—that is, the minimum findings necessary for
classification as a drug. We do not and cannot consider
whether the FDA’s assumption that Vanilla SilQ satisfies the
device definition is a valid one because it was not the basis for
the FDA’s decision. See SEC v. Chenery Corp., 318 U.S. 80,
93–94 (1943). Instead, we necessarily address only the FDA’s
conclusion that the FDCA grants it discretion to classify as a
“drug” any product that meets the statutory definition of a
“device.” We hold that it does not. Excepting combination
products, see 21 U.S.C. § 353(g), devices must be regulated as
devices and drugs—if they do not also satisfy the device
definition—must be regulated as drugs.8 Thus, the FDA’s
decision must be set aside because it was based on an erroneous
interpretation of law. See 5 U.S.C. § 706(2).
For the foregoing reasons, the judgment of the district
court is affirmed.
So ordered.
8
Our reading does not, as the concurring opinion suggests, limit the
drug definition “to products with the modes of action specified in the
device definition’s exclusions.” Concurring Op. 14. A product that
satisfies the drug definition and the mode-of-action clauses in the
device definition would still be a drug if it could not satisfy the
instrument clause.
PILLARD, Circuit Judge, concurring in the judgment: Our
role on review is limited to determining whether Congress has
unambiguously foreclosed the FDA’s statutory interpretation,
and, if not, whether the agency’s decision is nonarbitrary and
reasonably explained. In regulating Vanilla SilQ as a drug, the
FDA asserted that the FDCA’s drug definition fully subsumes
the device definition so grants the FDA authority to regulate
any device as a drug. See J.A. 121-22; Appellant’s Br. 6-7, 15-
17. Vanilla SilQ is a contrast agent used in radiologic
procedures including X-rays. The mode of action of some
other contrast agents prevents their regulation as devices.
Because the agency deemed it administratively convenient to
regulate Vanilla SilQ together with other contrast agents,
whatever their mode of action, it decided to regulate Vanilla
SilQ under the drug regime. See J.A. 122 & n.5.
The majority acknowledges that the text of the drug and
device definitions describes some overlap. See Maj. Op. 4, 13.
But, cued by the FDA’s overbroad claim of discretion, my
colleagues view the textual overlap as conferring implausibly
“sweeping,” “near-limitless” power on the FDA to regulate any
device as a drug, contrary to the congressional scheme. Id. at
3, 18, 22. Cognizant that Congress defined devices separately
from drugs for important reasons, the majority resorts to
statutory structure and history to conclude that the drug and
device definitions overlap not at all, so are entirely mutually
exclusive. Id. at 11, 14-15, 17-18, 22, 24.
I agree with my colleagues that the FDA misread the
statute in concluding that the drug definition fully subsumes
devices. But I join the judgment only, because the majority
overshoots in the other direction by insisting the statute
unambiguously eliminates all overlap of the drug and device
definitions. There is overlap, but it is only partial.
Both the FDA’s and my colleagues’ readings overtly
disregard textual specificity within the device definition that
2
assigns particular types of products to the regulatory pathway
for devices, not drugs. Whereas the drug definition refers to
“articles” intended for medical uses, 21 U.S.C. § 321(g)(1)(B),
the device definition singles out as devices any “instrument,
apparatus, implement, machine, contrivance, implant, in vitro
reagent, or other similar or related article, including any
component, part, or accessory,” intended for those same uses,
id. § 321(h)(1). That more specific list—the “instrument
clause,” for short—communicates Congress’s intent not to
grant the FDA the broad discretion it claims. The FDA is
wrong to view the device definition as describing a fully nested
subset of the drug definition; it clearly prevents assignment of
many devices—from bathroom scales and band-aids to
respirators and ultrasound machines—to the regulatory
pathway for drugs.
The FDA has yet to grapple with the product
characteristics Congress deemed relevant. The agency simply
assumed that Vanilla SilQ meets the device definition. It
treated the instrument clause as a nullity and proceeded as if
the mode-of-action exclusions did not apply to Vanilla SilQ.
See Appellant’s Br. 21; J.A. 121-22, 152; see also Oral Arg.
Tr. 5:10-6:12. It neither identified in what respect Vanilla SilQ
might or might not be considered an “instrument,” nor verified
how it avoids the device definition’s mode-of-action
exclusions. The agency accordingly failed to explain its
decision to regulate Vanilla SilQ as stringently as a drug in any
way that accounts for the factors Congress deemed relevant to
its design of distinct drug and device regimes. The explanation
it did provide, turning on little more than administrative
convenience, falls short, so requires remand to the FDA. That
is as far as we need to go to decide this case.
3
I.
Our review of the FDA’s classification decision pursuant
to the FDCA turns on “whether Congress has unambiguously
foreclosed the agency’s statutory interpretation.” Vill. of
Barrington v. Surface Transp. Bd., 636 F.3d 650, 659 (D.C.
Cir. 2011) (internal quotation marks and citation omitted). “If
the agency’s interpretation is not unambiguously foreclosed by
the statute, we defer to its interpretation so long as it is
reasonable.” Sorenson Commc’ns, LLC v. FCC, 897 F.3d 214,
224 (D.C. Cir. 2018) (internal quotation marks and citation
omitted); see also Serono Lab’ys, Inc. v. Shalala, 158 F.3d
1313, 1319-22 (D.C. Cir. 1998) (applying that framework to
the FDA’s interpretation of the FDCA). We cannot defer to an
agency decision that rests on an erroneous statutory
interpretation.
Our deference to the agency depends on its having
engaged in reasoned decisionmaking, explaining how it
accounted for all the factors relevant to the exercise of the
authority Congress has given it. See Judulang v. Holder, 565
U.S. 42, 53 (2011); 5 U.S.C. § 706(2). We cannot sustain a
ruling that fails to consider and explain an important aspect of
the issue at hand. Motor Vehicle Mfrs. Ass’n of U.S., Inc. v.
State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983); see
also Pharm. Mfg. Rsch. Servs., Inc. v. FDA, 957 F.3d 254, 262
(D.C. Cir. 2020) (applying that standard to FDA action taken
pursuant to the FDCA).
The bulk of my colleagues’ analysis trains on the
functional separation and mandatory character of the drug and
device regulatory pathways. But nobody questions that, when
faced with a product with the intended use common to both the
drug and device definitions—i.e., a product intended for use in
the diagnosis, cure, mitigation, treatment, or prevention of
4
disease—the agency must regulate it as either a drug or a
device, not both. See Appellant’s Br. 16-17, 25-26; Maj. Op.
14-15. And all agree that the respective drug and device
regimes are mandatory on the agency and the regulated parties.
See Oral Arg. Tr. 9:18-10:9; Maj. Op. 15. The dispute here is
whether the statute does all the work of sorting drugs from
devices or whether it affords the agency some discretion in
doing so and, if the latter, how much and with what guideposts.
The FDA claims leeway well beyond what the statute
affords. My colleagues overlook definitional overlap the
statute allows. I start with the overlap the majority denies, and
then discuss how the FDA’s discretion is bounded in ways the
agency overlooks. I defend Congress’s decision to allow for
some overlap in the drug and device definitions in view of the
complexity and heterogeneity of medical products and the
challenge of assessing their efficacy and safety.
A.
The Supreme Court has long acknowledged “obvious
areas of overlap in definition” of drugs and devices under the
FDCA, even under an earlier version of the statute that
manifested exclusivity in a way that it does not today. United
States v. Article of Drug, Bacto-Unidisk, 394 U.S. 784, 799
(1969). Congress codified a particular definitional overlap in
1990, when it cut from the drug definition its prior, express
exclusion of “devices or their components, parts, or
accessories.” Safe Medical Devices Act of 1990 (SMDA),
Pub. L. No. 101-629, § 16(b)(1), 104 Stat. 4511, 4526. It did
so to clearly permit the FDA to regulate as drugs some products
also meeting the device definition. The products foremost on
the congressional agenda at that time were “combination
products”—such as surgical mesh with anesthetic coating and
drug-eluting cardiovascular stents that are “instruments” as
5
referenced in the device definition but that also function in part
through chemical action or metabolization that would define
them as drugs. Capsular Decisions – Products Assigned to
CDRH, FDA (Feb. 16, 2018),
https://www.fda.gov/combination-products/rfd-jurisdictional-
decisions/capsular-decisions-products-assigned-cdrh. But the
SMDA did not limit the amendment to those products, so we
have no business doing so.
The majority insists Congress removed the drug category’s
exclusion of devices “only to facilitate the FDA’s regulation of
the new category of ‘combination products.’” Maj. Op. 17.
But it is the enacted text “rather than the principal concerns of
our legislators by which we are governed.” Bostock v. Clayton
County, 140 S. Ct. 1731, 1749 (2020) (quoting Oncale v.
Sundowner Offshore Servs., Inc., 523 U.S. 75, 79 (1998)).
“The fact that Congress may not have foreseen all of the
consequences of a statutory enactment is not a sufficient reason
for refusing to give effect to its plain meaning.” Union Bank v.
Wolas, 502 U.S. 151, 158 (1991); see Antonin Scalia & Bryan
A. Garner, Reading Law: The Interpretation of Legal Texts 56
(2012) (“[T]he purpose must be derived from the text, not from
extrinsic sources such as legislative history or an assumption
about the legal drafter’s desires.”). Congress easily could have
limited the scope of the amendment as the majority says it
meant to do by writing it as an exception within the new
provision for combination products, 21 U.S.C. § 353(g),
relieving only combination products’ device components from
the drug definition’s exclusion. Instead, Congress removed the
device exclusion altogether.1 It thus did not enact the mutual
1
Even in referencing combination products, the legislative
history does not describe the amendment as confined to them. See
S. Rep. No. 101-513, at 43 (1990) (explaining that Congress removed
the device exclusion from the drug definition to “permit”
combination products); id. at 30 (explaining that the amendment
6
exclusivity of the statute’s drug and device definitions that my
colleagues detect.
The majority sees the device definition’s mode-of-action
exclusions as “critical to effectuating” non-overlapping
statutory drug and device definitions. See Maj. Op. 20; see also
Oral Arg. Tr. 29:2-17 (Genus arguing that the mode-of-action
clauses are what distinguish drugs from devices). But those
exclusions do not unambiguously remove definitional overlap.
They prevent regulation of a product as a “device” insofar as it
“achieve[s] its primary intended purposes through chemical
action within or on the body of man or other animals” or is
“dependent upon being metabolized for the achievement of its
primary intended purposes.” 21 U.S.C. § 321(h)(1). But by
their terms those exclusions operate only in one direction: The
text nowhere provides that the drug definition is confined to
products reliant on the modes of action that the device
definition excludes. Definitional overlap remains for products
that do not achieve their primary intended purposes by
chemical or metabolic action—a category that the FDA here
assumed without deciding includes Vanilla SilQ.
The general-governs-the-specific canon on which the
majority’s reading depends is not to the contrary. See Maj. Op.
13-15. That canon is “a strong indication of statutory meaning”
but “not an absolute rule.” RadLAX Gateway Hotel, LLC v.
Amalgamated Bank, 566 U.S. 639, 646 (2012). It cannot
resolve all cases when the more specific provision is as
qualitative as the device definition’s instrument clause. (More
on that below.) Plus, the partial overlap wrought by the 1990
Amendment is just the kind of “textual indication[]” that may
ensured that drugs with device components “may be reviewed” as
drugs).
7
override the canon in a case in which the agency provides
sufficient explanation. Id.Neither the mode-of-action exclusions nor the 1990
Amendment requires that Vanilla SilQ be regulated as a device.
Nor do they together effectuate mutual exclusivity of the drug
and device definitions.
B.
At the same time, acknowledgment of the statutory drug
and device definitions’ overlap—and the FDA’s factual
assumption that Vanilla SilQ falls within it—does not suffice
to allow the FDA to classify the product as a drug. The agency
claims that, apart from products the mode-of-action clauses
identify as drugs, any product with the requisite intended use
may be regulated as either a device or a drug. But the FDA is
wrong that the mode-of-action exclusions are the only relevant
constraint. See Appellant’s Br. 6-7; see also Oral Arg. Tr.
5:10-6:12, 48:20-23. The device definition also has
inclusionary language, defining a product as a device if it is an
“instrument, apparatus, implement, machine, contrivance,
implant, in vitro reagent, or other similar or related article,
including any component, part, or accessory.” 21 U.S.C.
§ 321(h)(1).
The FDA sees “device” as a fully nested subset of “drug”
by noting that the general reference to “articles” in the drug
definition, id. § 321(g)(1), is broad enough to subsume every
“instrument, apparatus, . . . or other similar or related article”
listed in the device definition’s instrument clause, id.
§ 321(h)(1). See Appellant’s Br. 6-7; Oral Arg. Tr. 8:12-20.
Genus agrees, at least when it comes to the instrument clause.
See Oral Arg. Tr. 29:2-17 (noting that both definitions use the
term “articles,” and that the statute distinguishes them with the
device definition’s mode-of-action clauses). That reading is
8
impermissible. “A statute should be construed so that effect is
given to all its provisions, so that no part will be inoperative or
superfluous, void or insignificant.” Corley v. United States,
556 U.S. 303, 314 (2009) (citation omitted and formatting
modified). The FDA violates the canon against superfluity by
failing to give meaning to a big piece of the device definition.
The only argument the FDA puts forward is that its reading
creates no superfluity because the two definitions remain
distinct: “Products that satisfy the ‘device’ definition may be
regulated either as drugs or as devices, whereas those that
satisfy the ‘drug’ definition but not the ‘device’ definition may
be regulated only as drugs.” Appellant’s Br. 16. But that
misses the point. It does nothing to explain why Congress used
twenty words in the device definition’s instrument clause if it
meant nothing more specific than is expressed by the word
“articles” alone in the drug definition.
Notwithstanding that the FDA treats it as a nullity here, the
instrument clause has all along done substantial work in
assigning products to the device rather than the drug pathway.
For starters, the FDA accounts for that clause in classifying as
devices all manner of medical products, such as crutches, X-
ray machines, and other “things that go clank.” Oral Arg. Tr.
6:3; see, e.g., 21 C.F.R. § 890.3150 (classifying a crutch as a
device); id. § 892.1680 (classifying a stationary X-ray system
as a device). With the instrument clause in view, it is obvious
that the statute does not afford the agency the “near-limitless
discretion,” Maj. Op. 18, 22, that the FDA says it does and that
animates the majority’s too-restrictive response, see id. at 16
(voicing concern that the 1990 Amendment “dramatically
expanded” the FDA’s authority). The instrument and mode-
of-action clauses make clear the elephant that the majority sees
is only a mouse. See id. at 23 (citing Whitman v. Am. Trucking
Ass’ns, Inc., 531 U.S. 457, 467-68 (2001)).
9
In the decades since Congress in 1990 removed the device
exclusion from the FDCA’s drug definition to create some
textual overlap, there is little evidence that the FDA has treated
as drugs what should be regulated as devices. See Oral Arg.
Tr. 11:21-23 (FDA referencing the lack of cases); cf. Bracco
Diagnostics, Inc. v. Shalala, 963 F. Supp. 20, 28 (D.D.C. 1997)
(saying of contrast agents that “all likely meet both . . .
definition[s] . . . and the FDA therefore has discretion in
determining how to treat them,” though it cannot “permit two
sets of similar products to run down separate tracks . . . for no
apparent reason”). The only other instance the parties identify
of a product that meets the device definition but that the FDA
nonetheless regulates as a drug is sunblock. As with contrast
agents, the FDA uniformly regulates the entire group of
sunscreen products as drugs even though some, “such as those
commonly marketed for use with infants and small children,”
are also eligible to be regulated as devices insofar as they
operate by providing a physical barrier (e.g., zinc) against solar
rays, rather than through chemical action that would subject
them to the device definition’s mode-of-action exclusion.
Appellant’s Br. 31-32; see also Appellee’s Br. 47. The dearth
of litigated cases, or even illustrative examples, of products
arguably meeting the device definition but being regulated as
drugs owes much to the instrument clause.
The FDA here treats the instrument clause as superfluous,
but its own guidance and regulations recognize the clause’s
robust role informing the FDA’s product classifications. The
FDA’s 2017 Guidance points out that, “[i]n some cases,”
products that are not themselves instruments, apparatuses, or
so forth “are appropriately considered ‘similar or related
articles’” under the instrument clause “and may be classified as
devices.” J.A. 332. It notes how, for example, “gels or
powders put on the skin” come within the instrument clause
when used “as a barrier,” “gases” satisfy the clause when “used
10
as space fillers,” and certain “liquids” qualify when “used to
clean either surgical instruments or contact lenses.” Id.; see,
e.g., 21 C.F.R. § 886.5928 (classifying “contact lens care
products” for soft contact lenses, including solutions, as
devices); see also Topical Drug Products for Over-the-Counter
Human Use; Products for the Prevention of Swimmer’s Ear and
for the Drying of Water-Clogged Ears; Final Rule, 60 Fed. Reg.
8916, 8917 (Feb. 15, 1995) (explaining that the products at
issue were not devices because they did not satisfy the
instrument clause, and that they were drugs even though they
worked through physical, not chemical or metabolic, means);
Capsular Decisions – Products Assigned to CDER, FDA (Feb.
16, 2018), https://www.fda.gov/combination-products/rfd-
jurisdictional-decisions/capsular-decisions-products-assigned-
cder (listing “[d]ye mouthrinse to examine oral tissue” as a
drug).
The FDA is bound by “the core administrative-law
principle that an agency may not rewrite clear statutory terms
to suit its own sense of how the statute should operate.” Util.
Air Regul. Grp. v. EPA, 573 U.S. 302, 328 (2014). Because it
incorrectly treated the instrument clause as a nullity and
assumed without deciding that the mode-of-action clauses do
not apply to Vanilla SilQ, see Appellant’s Br. 21 (citing J.A.
122), the agency concluded the statute was silent as to whether
Vanilla SilQ should be regulated as a drug or device. With
none of the device definition’s clauses placing any restriction
on its action here, the FDA says, it had free rein to choose how
to regulate Vanilla SilQ, and it chose the drug pathway. See id.
at 15-18.
The FDA is not entirely wrong that the drug and device
definitions overlap—they do, in part. But it fell short in neither
acknowledging the detailed instrument clause nor providing a
lawful and nonarbitrary explanation of whether and how
11
regulating Vanilla SilQ as a drug accords with both that clause
and the mode-of-action exclusions in the device definition. It
is no answer that the FDA’s classification decisions are subject
to APA review, nor that its thousands of duly promulgated
product classifications constrain its decisions as a practical
matter. See Oral Arg. Tr. 16:12-17:7. The statutory framework
governing the FDA’s exercise of its discretion is what provides
traction for those procedural safeguards—and the instrument
clause that the agency ignored here is a crucial part of that
framework.
We have said, in relation to the FDCA, that a “statutory
phrase must be read in the context of the kind of drug at issue.”
Serono Lab’ys, 158 F.3d at 1319. The FDA’s interpretation of
the drug and device definitions and their application to Vanilla
SilQ failed even “to wrestle with the relevant statutory
provisions,” including the instrument clause, and “we cannot
do [the agency’s] work for it.” Hosp. of Barstow, Inc. v. NLRB,
820 F.3d 440, 445 (D.C. Cir. 2016) (quoting Children’s Hosp.
& Rsch. Ctr. of Oakland, Inc. v. NLRB, 793 F.3d 56, 59 (D.C.
Cir. 2015)). We are therefore “left wondering how the
[agency] in these circumstances interprets the statute”—that is,
how it would account for the instrument clause as a limit on the
definitional overlap and a constraint on its discretion. Id.(internal quotation marks omitted). Because of that
uncertainty, I would follow “[o]ur general practice in these
sorts of situations” and remand for the agency to interpret the
statute in the first instance, including the instrument clause that
it simply did not apply. Id.C.
The nub of my disagreement with the majority concerns
whether, as between dueling statutory definitions, the correct
choice might legitimately turn in part on the agency’s expert
12
determination of predicate facts and its sound discretion as to
which definition best applies. It seems unremarkable to me that
Congress conferred such authority on the FDA.
The majority disagrees. It concludes that, because the
FDCA “elaborates distinct regulatory regimes” in which “each
scheme is mandatory” for the agency and regulated parties,
Maj. Op. 15, there can be no definitional overlap of the drug
and device categories. Where the text does not clearly
eliminate overlap, the majority says that interpretive canons
make the definitions’ mutual exclusivity unmistakable. But the
premise that the statute makes sense only if there is no
definitional overlap is wrong.
There is no escaping some classificatory judgment by the
FDA. To see why, it is helpful to identify two layers of
judgment that can be required to apply the right definition to a
given product. My colleagues take no issue with the first, but
hold that the second is inimical to the statutory structure and
function. But the two are not materially different. Both call
for expert determinations and judgments by the agency—
choices that the statutory scheme as written does not obviate
and that, no matter how much detail Congress might add, could
not be wholly eliminated.
First, even accounting for textual specificity the FDA
ignores, gray areas remain. With the help of the instrument and
mode-of-action clauses, applied in relation to a product’s
intended use, most products can readily be identified as either
a drug or a device. But classification is not always obvious.
Nor could it be. The drug and device definitions—in relevant
part, each a single sentence—apply to many thousands of
widely heterogeneous products. Those products present varied
and often complex questions of safety and efficacy.
Determining the appropriate scrutiny—as drug or device—can
13
be complicated by the detailed knowledge and experience and
the high stakes involved in assessing certain products’ efficacy
and safety for use in the diagnosis, cure, mitigation, treatment,
or prevention of disease. Congress knew it could not codify all
that distinguishes drugs from devices. It was content to sketch
the basics and leave it to the FDA to bring to bear the distinct
expertise of the Center for Drug Evaluation and Research and
the Center for Devices and Radiological Health, informed by
the detailed input of many expert panels and advisory
committees. See id. at 4-5; see also Advisory Committees,
FDA, https://www.fda.gov/advisory-committees (last visited
Apr. 14, 2021).
As already discussed, the instrument clause contains
material specificity. But, at the end of the day, its terms are
unavoidably qualitative and imprecise. For example, deciding
whether a product is a “contrivance” or “apparatus”—let alone
“similar or related” to one—requires some judgment about the
product’s character. That judgment is not unguided but, per the
FDCA and APA, informed by whether treating it as a drug or
device best aligns with what Congress meant to achieve in
articulating separate definitions keyed to distinct regulatory
approaches. Also relevant is how the product compares to
others already classified, not least because the agency acts
arbitrarily if it regulates similar products differently. But there
is no avoiding the exercise of judgment at the margins.
Provided the agency makes nonarbitrary determinations based
on substantial evidence, we must defer. My colleagues express
no disagreement with this. See Maj. Op. 8 n.3 (noting that it is
“not immediately obvious” how a contrast agent satisfies the
device definition’s instrument clause, “[n]or is it altogether
settled that Vanilla SilQ satisfies the device definition’s mode-
of-action clauses”); id. at 23 (acknowledging that, when the
instrument and mode-of-action clauses “are ambiguous in their
14
application to some kind of medical product, reviewing courts
should respectfully consider the expert views of the FDA”).
A second layer of classification judgment, called into play
by definitional overlap, is what the majority reads the statute to
unambiguously eliminate. If a product is neither clearly within
or outside the instrument clause, for example, and not excluded
by the mode-of-action clauses, as a definitional matter it is
plausibly both a drug and device. Even though it ultimately
can only be regulated as one or the other, the statute does not
alone determine which one it is. Assume, for example, that
nanotechnology ingested or injected for the purpose of
delivering light to cancer cells is not excluded from regulation
as a device because it relies not on chemical action, but on some
purely physical process. On my colleagues’ reading, that
product could be regulated only as a device. Provided that it
might plausibly be thought to fall within the instrument clause,
they would hold that the device definition’s mode-of-action
clauses—“critical to effectuating” mutual exclusivity, id. at
20—necessarily sort it into the device category. On my
reading, the statute authorizes the FDA to make an informed
judgment whether to regulate that type of product—rare as it
may be—as a drug or a device. Congress did not
unambiguously leave that call to us rather than to the FDA.
The majority’s reading has another type of anomalous
effect. Without the overlap Congress in 1990 built into the
definitions, an “article” that does not meet the device
definition’s more specific “instrument” clause, and that also
does not rely on chemical or metabolic action to achieve its
primary intended purposes, would be neither drug nor device.
By insisting, through negative implication, that the drug
category is confined to products with the modes of action
specified in the device definition’s exclusions, the majority
eliminates coverage for any product not described as a device
15
by the instrument clause but also not deemed a drug by the
mode-of-action clauses. For example, if Vanilla SilQ’s
presumed mode of action were confirmed as non-chemical and
non-metabolic, and if it were proposed to be used not for its
arguably device-like function of blocking X-rays with its
molecular density (and thus, e.g., an “accessory” to an X-ray
machine) but instead to be consumed to treat disease in its own
right (and thus an uneasy fit for “instrument”), the majority’s
reading would exclude it from both the drug and device
categories. Interpreting the statute to exclude a category of
products intended for use in the diagnosis, cure, mitigation,
treatment, or prevention of disease from both the drug and
device definitions squarely conflicts with Congress’s design in
a manner that acknowledging some overlap does not.
In my view, the statutory overlap effected by the 1990
Amendment would give the FDA a choice how to classify such
products—provided its justification addressed the product’s
material characteristics and Congress’s definitions of the
distinct regulatory pathways. Whether the product was
nanotechnology or Vanilla SilQ, the agency would have to
consider whether its device-like character and function as it is
intended to be used suggest it should be grouped with devices.
It would have to analyze whether its mode of action, even if not
chemical or metabolic, presents questions of efficacy or risk
that warrant regulating it on the same pathway with otherwise
similar products that have different modes of action. And it
would have to explain how its judgments on those points
accorded with the text, nature, and purpose of the distinct
definitions to advance the statute’s objectives.
These two types of classificatory judgments—at the
margins of definitions, and in the overlap between them—are
not all that different. Both depend on legal and factual
expertise and judgment calls informed by scientific and
16
regulatory experience. The majority accepts the first kind yet
rejects the second. But Congress did not unmistakably
eliminate the type of judgment calls that definitional overlap
would require of the FDA. And nothing about that is
anomalous.
Similar judgment calls are built into multiple provisions of
the FDCA. For example, we have observed that, “although the
consequences of classification as a ‘drug claim’ or a ‘health
claim’ are quite substantial”—if the former, a product would
have to be approved as a drug to be marketed in that way—
“Congress has given definitions that at least partially overlap”
with “little guidance as to how the FDA should sort out claims
that seem to fit both definitions.” Whitaker v. Thompson, 353
F.3d 947, 949 (D.C. Cir. 2004). Faced with that statutory
overlap, we deferred to the agency’s reasonable interpretation.
Id. at 951-52.
In Serono Laboratories, Inc. v. Shalala, we likewise
deferred to the FDA’s interpretation of an FDCA provision
requiring the agency to approve a generic drug with the “same”
active ingredients as a listed drug, in part because the statute
did not foreclose treating a generic as clinically the same even
if not completely chemically identical to the pioneer version.
158 F.3d at 1318-20. In so holding, we emphasized the
significant judgment that Congress entrusted to the agency in
the FDCA: “The FDA’s determination of what is required to
establish ‘sameness’ for purposes of the Act rests on the
agency’s evaluations of scientific data within its area of
expertise, and hence is entitled to a high level of deference.”
Id. at 1320 (internal quotation marks and citations omitted). In
sum, the majority’s statutory analysis is unpersuasive, and the
concern that animates it illusory.
17
But because the FDA has neither recognized the statutory
constraints that bear on its classification of Vanilla SilQ nor
made findings on the relevant facts, I agree we must remand to
the agency to do so.
II.
I credit the majority for recognizing a lack of clarity at the
margins, specifically “reserv[ing] the question whether Vanilla
SilQ satisfies the device definition’s instrument and mode-of-
action clauses” and thus not prejudging the outcome. Maj. Op.
8 n.3; accord id. at 23. The court’s decision leaves the agency
with several options on remand. For instance, the FDA might
determine that Vanilla SilQ—an oral solution consumed by
patients to improve imaging of their gastrointestinal tracts with
X-ray machines and CT scanners—is sufficiently “related” to
those devices or an “accessory” to them, 21 U.S.C. § 321(h)(1),
such that it, too, is a device within the meaning of the FDCA
and must be regulated as such. Or the FDA might make a
contrary determination, in which case Vanilla SilQ remains
amenable to classification and regulation as a drug. And if,
after further study, the FDA were to determine that Vanilla
SilQ in fact achieves its primary intended purpose through
chemical action, as an amicus urges, see Amicus Bracco Br. 5-
6, that conclusion would call for its regulation as a drug. (At
this juncture, no one suggests that Vanilla SilQ achieves its
primary intended purpose through metabolization, see id. at 7,
but that possibility likewise remains open and would require
regulation as a drug.)
That is all to say that the agency must make the requisite
factual determinations and attend to all relevant statutory
provisions to decide whether to classify Vanilla SilQ as either
a drug or device in the first instance, and that it may “deal with
the problem afresh” on remand. Dep’t of Homeland Sec. v.
18
Regents of the Univ. of Cal., 140 S. Ct. 1891, 1908 (2020)
(quoting SEC v. Chenery Corp., 332 U.S. 194, 201 (1947)).
Once the FDA does the work of applying the explicit
constraints of the device definition’s instrument and mode-of-
action clauses, it might determine that Vanilla SilQ is not a
device, but only a drug. In that event, there would have been
no reason to address the existence or not of definitional overlap
between drugs and devices.
The majority goes further than required to resolve this
appeal when it interprets the device and drug definitions as
mutually exclusive. I have explained why the majority errs as
a matter of statutory interpretation. More fundamentally, we
need not—so should not—decide that issue because it is not yet
apparent whether the existence or not of definitional overlap
matters in the case of Vanilla SilQ. If the FDA on remand were
to confirm that it believes that Vanilla SilQ could be either a
device or drug and make an assignment with the proper
reasoning in support, we would then be faced with the question
whether statutory overlap permits it to do so, and whether its
assignment comports with the APA. For now, it suffices to
identify the flaws in the FDA’s decision and the inadequacy of
the unelaborated administrative convenience rationale it gave
for subjecting contrast agents with materially varying modes of
action to the same degree of regulatory rigor.
***
Because the FDCA does not give the FDA the discretion
that it claims to regulate any device as a drug, and the agency
has failed to explain its choice in a manner that grapples with
the applicable statutory terms, I concur in the court’s judgment.