The Judge Rotenberg Educational Center, Inc. v. FDA

United States Court of Appeals
FOR THE DISTRICT OF COLUMBIA CIRCUIT

Argued April 23, 2021 Decided July 6, 2021

No. 20-1087

THE JUDGE ROTENBERG EDUCATIONAL CENTER, INC.,
PETITIONER

v.

UNITED STATES FOOD AND DRUG ADMINISTRATION, ET AL.,
RESPONDENTS

Consolidated with 20-1088

On Petitions for Review of an Order
of the Food & Drug Administration

Max D. Stern argued the cause for petitioners Luis Aponte,
et al. With him on the briefs were Joseph M. Cacace, and
Alexandra H. Deal.

Michael P. Flammia argued the cause for petitioner The
Judge Rotenberg Educational Center, Inc. With him on the
 2
briefs were Christian B. W. Stephens, Matthew D. Rodgers,
Edward J. Longosz, II, and Jeffrey N. Gibbs.

Richard A. Samp was on the brief for amicus curiae The
New Civil Liberties Alliance in support of petitioners.

Daniel Aguilar, Attorney, U.S. Department of Justice,
argued the cause for respondents. With him on the brief were
Sarah E. Harrington, Deputy Assistant Attorney General, and
Scott R. Mcintosh, Attorney.

Felicia H. Ellsworth was on the brief for amici curiae
American Academy of Pediatrics, et al. in support of
respondents.

Before: SRINIVASAN, Chief Judge, KATSAS, Circuit Judge,
and SENTELLE, Senior Circuit Judge.

Opinion for the Court filed by Senior Circuit Judge
SENTELLE.

Dissenting opinion filed by Chief Judge SRINIVASAN.

SENTELLE, Senior Circuit Judge: The Judge Rotenberg
Educational Center and the parents and guardians of its patients
both petition for review of a Food and Drug Administration
(FDA) rule banning electrical stimulation devices used to treat
aggressive or self-injurious behavior. In its rule, the FDA
determined that the devices present an unreasonable and
substantial risk of illness or injury, but only when used to treat
aggressive or self-injurious behaviors. The petitioners contend
that banning a medical device for a particular purpose regulates
the practice of medicine in violation of 21 U.S.C. § 396. We
agree, grant the petitions for review, and vacate the FDA’s rule.
 3

I. Background

A. Factual background

The Judge Rotenberg Educational Center is a facility in
Massachusetts that treats patients with severe mental
disabilities. The Center admits patients that other facilities
could not successfully treat. According to the Center, some of
its patients suffer from severe self-injurious and aggressive
behaviors that are difficult or impossible to treat using
conventional behavioral and pharmacological techniques. The
most common self-injurious behaviors include head-banging
and self-biting. The behaviors of some patients are extreme
enough that they have suffered self-inflicted brain trauma,
broken and protruding bones, and blindness.

Before the ban at issue in this case, the Center treated some
of its patients exhibiting severe self-injurious or aggressive
behavior with an electrical stimulation device. The device,
called a graduated electronic decelerator, briefly shocks
patients causing them to reduce or cease their self-injurious
behaviors. Banned Devices; Electrical Stimulation Devices for
Self-Injurious or Aggressive Behavior, 85 Fed. Reg. 13,312,
13,314 (March 6, 2020). The Center is the only facility in the
country that uses electric shock therapy to treat individuals who
severely self-injure or are aggressive. Other health care
practitioners not affiliated with the Center, however,
administer electrical stimulation devices to treat a wide variety
of other conditions, including tobacco, alcohol, and drug
addictions, as well as inappropriate sexual behaviors following
traumatic brain injuries. Id. at 13,317. The Center manufactures
its own devices. The Center treats approximately 20% of its
patients with this treatment at any given time.
 4

The devices are subject to extensive federal and state
regulation. The FDA regulates aversive conditioning devices,
including ones that use electrical shocks, as Class II devices.
21 C.F.R. § 882.5235. That classification includes all medical
devices that the FDA determines are reasonably safe and
effective when subject to special controls like postmarket
surveillance and patient registries. 21 U.S.C. § 360c(a)(1)(B).
In addition to the federal regulation, Massachusetts requires
several entities to approve electrical shock treatment. See
Judge Rotenberg Educ. Ctr. v. Comm’r of the Dep’t of Dev.
Servs., Dkt. No. 86E-0018-GI, at 2–8 (Bristol, Mass. Prob. &
Fam. Ct., June 20, 2018). Before the Center treats a patient with
the devices, Massachusetts requires multiple health care
practitioners to certify that no other treatments were effective
or that the shock treatment is not contraindicated. It further
requires that peer review and that human rights committees
ratify the treatment. Further, a state court must determine that
the treatment was appropriate. Id. The intricate system of state
regulation arose as a combination of state statutes, regulations,
and a consent decree that the Center and Massachusetts entered
in 1987. Id.

B. Procedural background

In April 2016, the FDA proposed banning electrical
stimulation devices for self-injurious or aggressive behavior.
See Banned Devices; Proposal to Ban Electrical Stimulation
Devices Used to Treat Self-Injurious or Aggressive Behavior,
81 Fed. Reg. 24,386 (Apr. 25, 2016). The notice of proposed
rulemaking stated that the therapy presented several
“psychological and physical risks: Depression, fear, escape and
avoidance behaviors, panic, aggression, substitution of other
behaviors (e.g., freezing and catatonic sit-down), worsening of
underlying symptoms (e.g., increased frequency or bursts of
 5
self-injury), pain, burns, tissue damage, and errant shocks from
device misapplication or failure.” Id. at 24,387. Literature
addressing other electrical devices that shock patients further
suggested treatment with such devices could result in
posttraumatic stress disorder. Id.

The FDA also reviewed the evidence of the devices’
effectiveness and concluded that the evidence was weak.
According to the FDA, some studies showed that the devices
immediately interrupt the targeted behavior, but that the
evidence was inconclusive as to whether the devices
“achieve[d] durable long-term reduction of [self-injurious or
aggressive behaviors].” Id. at 24,387. In reaching those
conclusions, the FDA reviewed the medical literature at large
and data from the Center itself. Id. Based on the evidence of
harm to patients, and what it regarded as weak evidence of
durable effectiveness, the FDA determined that the devices
presented a substantial and unreasonable risk to self-injurious
and aggressive patients, justifying banning the devices for that
purpose. In 2020, the FDA promulgated its final rule. See 85
Fed. Reg. 13,312. The final rule adopted the conclusions set
forth above on the risks and efficacy of electrical stimulation
devices to treat self-injury and aggression. Id. at 13,315. The
FDA, in reviewing comments, also concluded that it had the
legal authority to ban a device for a particular purpose. Id. at
13,345.

Both the Center and parents and guardians of patients who
receive or seek to receive treatment using an electrical
stimulation device now petition this court to review the FDA’s
ban raising several issues. We determine that a single issue is
determinative of the case. That issue is: Does the FDA have
legal authority to ban an otherwise legal device from a
particular use? The other arguments will not require separate
analysis.
 6

II. Analysis

The answer to the controlling issue is determined by the
application of two statutes: 21 U.S.C. § 360f, which authorizes
the FDA to ban medical devices, and 21 U.S.C. § 396, which
prohibits the FDA from regulating the practice of medicine. We
begin by setting forth the relevant portion of each statute.

Section 360f, which Congress passed in the Medical
Device Amendments of 1976 to the Food, Drug, and Cosmetic
Act, grants the FDA authority to ban medical devices. The
section provides: “Whenever the Secretary finds . . . that a
device intended for human use presents substantial deception
or an unreasonable and substantial risk of illness or injury,” and
that risk cannot be “corrected or eliminated by labeling,” the
Secretary “may initiate a proceeding to promulgate a regulation
to make such device a banned device.” 21 U.S.C. § 360f(a).
“Device” is a defined term within the Food, Drug, and
Cosmetic Act, meaning “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.” Id. § 321(h)(1). A banned device is
considered adulterated, authorizing the FDA to seize the device
if it has been or may be introduced into interstate commerce
and making it a crime to introduce the device into interstate
commerce or manufacture it. Id. §§ 331(a), (g); 333; 334(a)(1);
351(g).

Section 396 constrains the FDA’s authority by prohibiting
it from regulating the practice of medicine. In the Food and
Drug Modernization Act of 1997, Congress provided:
 7
Nothing in this chapter shall be construed to
limit or interfere with the authority of a health
care practitioner to prescribe or administer any
legally marketed device to a patient for any
condition or disease within a legitimate health
care practitioner-patient relationship. This
section shall not limit any existing authority of
the Secretary to establish and enforce
restrictions on the sale or distribution, or in the
labeling, of a device that are part of a
determination of substantial equivalence,
established as a condition of approval, or
promulgated through regulations.
21 U.S.C. § 396. Section 396 ensures that once the FDA
permits a device to be marketed for one use, health care
practitioners have the flexibility to draw on their expertise to
prescribe or administer the device for any condition or disease,
not just the use the FDA approved—in short, to practice
medicine. See Buckman Co. v. Plaintiffs’ Legal Comm., 531
U.S. 341, 349–50 (2001). Such “‘off-label’ usage of medical
devices . . . is an accepted and necessary corollary of the
FDA’s mission,” and indeed, duty, “to regulate in this area
without directly interfering with the practice of medicine.” Id.
at 350. Section 396 protects the liberty of doctors and patients
to use approved devices in any manner they wish. See
Caplinger v. Medtronic, Inc., 784 F.3d 1335, 1344 (10th Cir.
2015) (Gorsuch, J.); see also H.R. Rep. No. 105-399, at 97
(1997) (“[T]he off-label use of a medical device by a physician
using his or her best medical judgment in determining how and
when to use the medical product for the care of a particular
patient is not the province of the FDA.”).

The FDA argues that section 396 does not restrict its
authority under section 360f at all. It argues that section 396
only prohibits the FDA from limiting the authority of a
 8
practitioner to prescribe or administer a legally marketed
device, and a device is not legally marketed if it is banned. The
FDA also points to the Act’s definition of a “device” with
reference to its intended use, such that it would be appropriate
for the FDA to ban devices with reference to particular uses.
Finally, the FDA also argues that it would be a peculiar
construction of the statute if the statute authorizes it to ban a
device completely, or not approve it in the first instance, but
prohibits the FDA from using its expertise to narrowly tailor a
ban to those circumstances in which a device presents a
uniquely substantial risk.

The petitioners contend not only that section 360f does not
authorize use-specific banning, but also contend that the plain
text of section 396 prohibits the FDA from banning a medical
device for a particular purpose. They further argue that the
section reserves the “tailoring” that the FDA suggests is
appropriate to medical practitioners, not the FDA. Finally, they
note that the FDA’s construction is not appropriate because it
interferes with states’ traditional authority to regulate the
practice of medicine.

Ordinarily, we evaluate an agency’s interpretation of a
statute it administers under Chevron USA Inc. v. Natural
Resources Defense Council, Inc., 467 U.S. 837, 842–43 (1984).
Under the framework, we first consider “whether Congress has
directly spoken to the precise question at issue.” Id. at 842. If
so, “the court, as well as the agency, must give effect to the
unambiguously expressed intent of Congress.” Id. at 842–43.
If Congress has not spoken clearly, we defer to an agency’s
interpretation if it is “based on a permissible construction of the
statute.” Id. at 843. In this case, the FDA did not invoke
Chevron deference or even cite the case in its briefing. Perhaps
this is because the agency concluded that the relevant statutes
are unambiguous. We agree that the statutes are unambiguous,
 9
although this does not mandate the FDA’s conclusion that the
statute authorizes it to take this action.

A. Statutory analysis

We begin our analysis of the scope of the FDA’s banning
authority with section 360f, the statute that authorizes banning
in the first instance. The statute states that the FDA may make
“such device a banned device,” and the natural reading of that
language suggests a device either is banned or it is not. It
speaks of no authority to place a device in an intermediate state
of “banned in some uses.” But the statute also requires the FDA
to determine whether the risks a device presents are reasonable,
presumably in light of the benefits that the device provides to
individuals that use it. If a device has multiple possible uses,
each use will present a different benefit-risk profile. The risks
a device presents could therefore be reasonable for some uses
but not for others. Focusing on “reasonable” in the statute
echoes the FDA’s reasoning that it should be permitted to tailor
a ban to those circumstances in which it applies its expertise to
find an unreasonable risk.

However, section 396 expressly denies the FDA authority
to construe any part of the Food, Drug, and Cosmetic Act,
including its authority to ban devices under section 360f, to
permit the FDA to “limit[] or interfere[]” with practitioners’
authority to prescribe or administer “legally marketed
device[s]” to patients. The questions for us then, are whether a
ban “limits or interferes,” and whether a device that the FDA
has attempted to ban for a particular purpose is “legally
marketed.”

A use-specific ban limits or interferes with a practitioner’s
authority by restricting the available range of devices through
regulatory action. Rather than being a peculiar interpretation,
 10
as the FDA argues, this understanding is consistent with both
the ordinary meaning of the terms “limit” and “interfere,” as
well as Supreme Court precedent. To limit is “to restrict the
bounds or limits of,” or “to curtail or reduce in quantity or
extent”; to interfere, in turn, is “to interpose in a way that
hinders or impedes: come into collision or be in opposition.”
Interfere, Limit, Merriam-Webster.com; see also Buckman,
531 U.S. at 350. Preventing further manufacture of a device
and seizing existing devices both “limits” and “interferes” with
a physician’s ability to prescribe or administer them.

As to the statutory reference to “legally marketed,” a
device is legally marketed if it is lawful for a manufacturer to
sell the device or a practitioner to prescribe or administer it.
The statute does not suggest, not should we read into it, a
limitation that the device must be marketed for the particular
use for which the practitioner wants to utilize the device.
Indeed, that would eviscerate the statute’s protection of off-
label use. See Buckman, 531 U.S. at 350. Any device that the
FDA attempts to ban for one but not all uses will, accordingly,
still be legally marketed. In this case, practitioners can still
prescribe or administer electrical stimulation devices for other
conditions, like smoking. 85 Fed. Reg. at 13,317. Electrical
stimulation devices are therefore legally marketed, and as
discussed previously, banning them for a particular use limits
or interferes with a practitioner’s ability to administer or
prescribe them as the practitioner sees fit. The plain meaning
of the first sentence of section 396 demonstrates that the FDA
does not have the authority to limit practitioners’ use of a
device for a particular purpose.

The FDA’s alternative interpretation of “legally marketed
device” is unpersuasive. The agency argues that because the
Act requires a device to have a use, the appropriate
construction of “device” is a pairing of a particular instrument
 11
with a particular use. It would follow that an electrical
stimulation device for self-injurious and aggressive behavior is
not a “legally marketed device” once the FDA bans that pairing
of an instrument and use. Congress potentially had such an
interpretation in mind, as legislative history contains references
to the FDA being permitted to treat multiple “use[s] . . . as
constituting a different device for purposes of classification or
other regulation.” H.R. Rep. No. 94-853, at 14–15 (1976). That
is not an appropriate interpretation.

First, that construction would allow the FDA to escape the
constraints of section 396 whenever it bans a device. If
Congress wished to have section 396 apply to everything
except the FDA’s banning authority, it could have done so.
Instead, the statute begins “Nothing in this chapter shall be
construed,” mandating that this section constrain the FDA’s
banning authority. Next, interpreting the definition of
“device” as requiring a pairing of an instrument with a
particular use is inconsistent with the definition of “device”
itself. Section 321 defines “device” by reference to “an
instrument, apparatus, implement, machine, contrivance,
implant, in vitro reagent, or other similar or related article,”
focusing on the physical item itself rather than any particular
intended use. It then uses the terms “diagnosis of disease or
other conditions,” and “cure, mitigation, treatment, or
prevention of disease,” when defining a device, not “a
condition,” “a disease,” or “a particular disease.” The plurality
and lack of an article in the statute render its ordinary meaning
to be that a “device” must be intended to diagnose, cure,
mitigate, treat, or prevent some number of conditions or
diseases, not necessarily a particular one. Contact lenses, for
example, are one medical device even though they treat both
nearsightedness and farsightedness.
 12
The second sentence of section 396, which authorizes the
Secretary to “establish and enforce restrictions on the sale or
distribution . . . of a device that are . . . promulgated through
regulations,” does not rescue the FDA. Although one could
colloquially refer to a ban as a restriction, there is significant
space between the definitions of the two terms. A “restriction”
is a “limitation or qualification,” while a “ban” is a “legal or
otherwise official prohibition.” Black’s Law Dictionary (11th
ed. 2019). More important, however, is the fact that the Act
structurally segregates bans and restrictions. As usual, we
presume that the same words, used in the same act of Congress,
have the same meaning. Atlantic Cleaners & Dyers, Inc. v.
United States, 286 U.S. 427, 433 (1932). Congress, in a
subsection titled “Restricted Devices,” authorized the FDA to
“require that a device be restricted to sale, distribution, or use
only upon the written or oral authorization of a practitioner
licensed by law to administer or use such device, or upon such
other conditions as the Secretary may prescribe in such
regulation.” 21 U.S.C. § 360j(e).

The restrictions that the FDA promulgates through section
360j(e) include, for example, requiring the special controls
appropriate for Class II devices. See id. § 360c(a)(1)(B). By
contrast, section 360f, authorizes the FDA to ban a medical
device. The two separate sections would be sufficient to
demonstrate that restrictions and bans are distinct. Looking
more closely at the statute reinforces the point. Reading
“restriction” in section 396 to include bans is inconsistent with
section 360j(e), which makes clear that a restriction is a
condition upon which a device may still be sold. If we
somehow overlook the plain language of section 360j(e) and
permit a ban as a restriction under section 360j(e), then section
360f itself would be surplusage as the FDA could find all the
banning authority it wanted elsewhere.
 13
As we suggested above, we are not persuaded that because
the FDA possesses the “greater” power to completely ban a
medical device, it must have the “lesser” power to tailor a ban
to only certain uses. Courts regularly recognize that a greater
power does not imply the existence of a lesser power,
especially when the exercise of that claimed lesser power
uniquely offends some external constraint. States may hold
elections for state judges or they may not, but the First
Amendment prohibits them from “conduct[ing] elections under
conditions of state-imposed voter ignorance.” Republican
Party of Minn. v. White, 536 U.S. 765, 788 (2002) (quoting
Renne v. Geary, 501 U.S. 312, 349 (1991) (Marshall, J.,
dissenting)). States can regulate dentists or allow them to
participate in an unregulated market, but they cannot permit
dentists to anticompetitively self-regulate in defiance of the
Sherman Act. N.C. State Bd. of Dental Exam’rs v. FTC, 574
U.S. 494, 505–06 (2015). Congress may provide Medicaid
funds to states or it may not, but once states have relied upon
those sizeable funds, it cannot condition their receipt on an
expansion of Medicaid without running afoul of the Tenth
Amendment. See Nat’l Fed’n of Indep. Bus. v. Sebelius, 567
U.S. 519, 581–83 (2012) (opinion of Roberts, C.J.). As a final
example, and most relevant to this case, the Attorney General
may ban all uses of certain controlled substances, but he cannot
prohibit specific uses that states regard as legitimate medical
practice. See Gonzales v. Oregon, 546 U.S. 243, 272–73
(2006). In this case too, no one disputes that section 360f
permits the FDA to ban a device completely. The FDA could
even decline to approve a device in the first instance. The
problem is that once the FDA approves a device and then tries
to ban it for specific uses, it defies the limitation that section
396 imposes. Just as in other contexts, section 396 operates as
an external constraint—preserving the ability of physicians to
make professional judgments about off-label uses—that
 14
prevents the FDA from exercising a lesser power merely
because it possesses a greater one.

B. Federalism concerns

While our analysis thus far has focused on the text of
section 396, the statute’s role in preserving the balance of
powers between the federal government and the states provides
further support for our conclusion. “The Constitution created a
Federal Government of limited powers.” Gregory v. Ashcroft,
501 U.S. 452, 457 (1991). “[T]he general government is not to
be charged with the whole power of making and administering
laws. Its jurisdiction is limited to certain enumerated objects,
which concern all the members of the republic, but which are
not to be attained by the separate provisions of any.” THE
FEDERALIST NO. 14 (James Madison). Courts have recognized
this bedrock principle since the earliest years of the republic.
See Marbury v. Madison, 5 U.S. (1 Cranch) 137, 176 (1803)
(“The powers of the legislature are defined, and limited; and
that those limits may not be mistaken, or forgotten, the
constitution is written.”) (Marshall, C.J.).

The Tenth Amendment instructs us that the powers which
the Constitution does not delegate to the federal government
belong to the states. U.S. CONST. amend. X. The very structure
of the Constitution underscores the balance of powers between
the federal and state governments “[w]ith its careful
enumeration of federal powers and explicit statement that all
powers not granted to the Federal Government are reserved.”
United States v. Morrison, 529 U.S. 598, 618 n.8 (2000).
Congress may legislate on naturalization, bankruptcy, patents,
and copyrights, for example, U.S. CONST. Art. I, § 8, but cannot
generally mandate individuals purchase particular goods or
regulate the practice of law. See NFIB, 567 U.S. at 557–58
(opinion of Roberts, C.J.); Am. Bar Ass’n v. FTC, 430 F.3d 457,
 15
471–72 (D.C. Cir. 2005). As a result, before we will construe a
statute to permit federal action in an area that is traditionally
the province of state law, we require Congress to make its
intention to “alter the usual constitutional balance between the
States and the Federal Government . . . unmistakably clear in
the language of the statute.” Will v. Michigan Dep’t of State
Police, 491 U.S. 58, 65 (1989) (internal quotations omitted).

“This principle applies with equal force to the so-called
modern administrative state.” Michigan v. EPA, 268 F.3d 1075,
1081 (D.C. Cir. 2001). Federal agencies are creatures of statute.
They possess only those powers that Congress confers upon
them. Bowen v. Georgetown Univ. Hosp., 488 U.S. 204, 208
(1988). If no statute confers authority to a federal agency, it has
none. If Congress has forbidden an agency from taking an
action, the agency cannot so act. See FDA v. Brown &
Williamson Tobacco Co., 529 U.S. 120, 132–33, 155–56
(2000). Accordingly, federal agencies like the FDA are doubly
restricted: they may only exercise powers Congress has
delegated to them, and that delegation itself must be a lawful
exercise of Congress’s constitutional authority. Just as we
require a clear statement when interpreting a statute in the first
instance, we require an explicit authorization from Congress
before we will permit an agency to regulate in an area that alters
the balance of powers between states and the federal
government. See Am. Bar Ass’n, 430 F.3d at 471–72. In
American Bar Association, a federal agency, in that case the
Federal Trade Commission, attempted to regulate the practice
of law. We held that it could not do so absent explicit
authorization by Congress. See id. In this case, the Food and
Drug Administration attempts to regulate the practice of
medicine, not only without explicit authorization from
Congress, but in the face of an explicit congressional command
not to do so.
 16
States, not the federal government, traditionally have
regulated the practice of medicine. See Gonzales, 546 U.S. at
275. Choosing what treatments are or are not appropriate for a
particular condition is at the heart of the practice of medicine.
See State v. Miller, 542 N.W.2d 241, 246 (Iowa 1995); State v.
Smith, 135 S.W. 465, 469 (Mo. 1911). Indeed, Massachusetts
has taken a very active role in regulating the Center’s use of
electrical stimulation devices. See Judge Rotenberg Educ. Ctr.
v. Comm’r of the Dep’t of Dev. Servs., Dkt. No. 86E-0018-GI,
at 2–8 (Bristol, Mass. Prob. & Fam. Ct., June 20, 2018).
Therefore, before we would permit the FDA to dictate whether
practitioners may administer electrical stimulation therapy to
self-injuring and aggressive patients, we would require an
explicit statement from Congress to that effect. Will, 491 U.S.
at 65; Am. Bar Ass’n, 430 F.3d at 471–72. When Congress
chooses to authorize the FDA to explicitly list what conditions
a physician can use a drug or device to treat, it does so. E.g.,
21 U.S.C. § 333(e) (requiring human growth hormone to be
approved for a particular purpose). In this case, we have quite
the opposite of an explicit authorization—an explicit statement
from Congress that the FDA cannot act. Section 396 explicitly
limits the power of the FDA. The FDA has no authority to
choose what medical devices a practitioner should prescribe or
administer or for which conditions.

CONCLUSION

In the end, despite the length of our discussion, the
resolution of the controlling issue is quite easily expressed.
When Congress has spoken in a statute, we assume that it says
what it means and that the statute means what it says. In this
case, the statute says that the FDA is not to construe its statute
so as to interfere with the practice of medicine. That means that
the FDA may not enact the regulation at issue before us.
Because we conclude that the FDA lacks the statutory authority
 17
to ban a medical device for a particular use, we do not address
petitioners’ other arguments, including whether the ban was
arbitrary and capricious or whether substantial evidence
supported the FDA’s factual determinations. We grant the
petitions for review and vacate the FDA’s rule banning
electrical stimulation devices for self-injurious and aggressive
behavior.
So ordered.
 SRINIVASAN, Chief Judge, dissenting: The Federal Food,
Drug, and Cosmetic Act, 21 U.S.C. § 301 et seq., grants the
FDA various types of regulatory authority over medical
devices. The agency, for instance, may classify devices as
Class I, II, or III devices, depending on the degree of risk a
device poses. See id. § 360c(a)(1). The riskier the device, the
greater the FDA’s power to regulate it. See id. The Act also
vests the FDA with authority to go further: the agency can
outright ban a device upon determining that it presents “an
unreasonable or substantial risk of illness or injury.” Id.
§ 360f(a)(1). A banned device cannot be manufactured,
introduced, or received in interstate commerce. Id. §§ 331(a),
(c), (g), 351(g). The FDA has exercised the banning power
very sparingly in the 45 years of its existence. Medical Device
Amendments of 1976, Pub. L. No. 94-295, 90 Stat. 539
(codified at 21 U.S.C. § 360c et seq.). The agency has banned
a device on just three occasions, including, most recently, the
ban in question in this case. See 85 Fed. Reg. 13,312 (Mar. 6,
2020) (Rule); 81 Fed. Reg. 91,722 (Dec. 19, 2016); 48 Fed.
Reg. 25,126 (June 3, 1983).

There is no dispute that the FDA has power under the
statute to ban a medical device altogether—i.e., across all its
potential uses. The issue in this case is whether the FDA can
exercise its banning authority in a more tailored fashion: rather
than ban a device across the board, can the agency ban the
device for a particular intended use while allowing it for other
uses? One can readily envision why the FDA might wish to do
so. Suppose a device has been approved to treat two distinct
conditions, meaning it has two intended uses. When treating
the first condition, the device proves to be of limited
effectiveness and poses an acute risk of injury. But when
treating the second condition, the device proves highly
effective and presents only a negligible risk of injury. The
FDA presumably would want to ban the device for the first
intended use but permit it for the second. And it is hard to
perceive why Congress could want to deny the agency that
 2
middle-ground option. Why force the agency to make an all-
or-nothing choice—either abolish a highly beneficial use so it
can stamp out a highly risky one, or stomach the highly risky
use so it can preserve the highly beneficial one?

This case is illustrative. The FDA conducted a years-long
examination of whether to ban the use of electrical stimulation
devices intended to treat self-injurious or aggressive behaviors.
81 Fed. Reg. 24,386, 24,392–93 (Apr. 25, 2016) (Proposed
Rule). Persons who engage in those behaviors often suffer
from intellectual or developmental disabilities. Id. at 24,389.
Electrical stimulation devices deliver a powerful and painful
electric shock to the wearer’s skin, in an effort to punish and
thereby discourage self-injurious and aggressive behaviors
when they manifest. Id. That type of treatment has fallen into
disuse over the past three decades: petitioner The Judge
Rotenberg Center is the sole facility in the United States that
still uses electrical stimulation devices to treat self-injurious or
aggressive behavior in disabled persons. Id. at 24,391, 24,409.

The FDA found that use of electrical stimulation devices
to treat those behaviors poses a number of health and safety
risks—from physical injuries such as severe pain, skin burns,
and tissue damage, to psychological injuries such as panic,
anxiety, and post-traumatic stress disorder. See, e.g., Rule, 85
Fed. Reg. at 13,323–26. The agency further concluded that the
devices are of dubious efficacy in treating self-injurious or
aggressive behaviors, and that alternative treatments (not
involving the infliction of pain) have proven more effective and
less risky. Id. at 13,333. The FDA thus decided to impose a
ban on electrical stimulation devices intended to treat those
behaviors. See id. at 13,315.

The agency, though, did not extend its ban to encompass
electrical stimulation devices intended to treat other conditions.
 3
The devices, for instance, can be used to treat smoking
addiction. When used for that purpose, the agency found, the
devices present a different, and acceptable, benefit-risk profile.
Id. at 13,317. That is in part because a smoker typically
controls the device and can stop its use when it causes undue
pain, and she can also communicate any harmful symptoms to
a healthcare provider. Id. By contrast, persons with
intellectual or developmental disabilities receiving treatment
for self-injurious or aggressive behaviors often do not control
the electrical stimulation devices they wear (rather, the devices
are controlled by a third party or automatically trigger upon
detecting certain movements). Id.; Proposed Rule, 81 Fed.
Reg. at 24,394, 24,396. Those persons also may lack the ability
to discern a causal connection between the devices and
physical or psychological harm, or may be unable to
communicate the harmful symptoms to others. Proposed Rule,
81 Fed. Reg. at 24,395. Because electrical stimulation devices
have varying benefit-risk profiles depending on their intended
use, the FDA confined its ban to the intended use determined
to pose an unreasonable danger—the treatment of self-
injurious or aggressive behaviors. Rule, 85 Fed. Reg. at
13,317.

Again, no one doubts the FDA’s statutory authority to
impose a blanket ban on electrical stimulation devices covering
all their potential uses. The question here is whether the agency
could adopt a less sweeping, more tailored approach: banning
the devices for treatment of self-injurious or aggressive
behaviors while allowing the devices for treatment of other
conditions. Petitioners contend, and my colleagues agree, that
the agency possesses only an all-or-nothing banning power: it
must either ban a device across all its potential uses or refrain
from banning it at all. Respectfully, I do not read the statute to
compel that counterintuitive result.
 4
Petitioners, joined by my colleagues, ground their all-or-
nothing understanding in 21 U.S.C. § 396. That provision
states that “[n]othing in this chapter shall be construed to limit
or interfere with the authority of a health care practitioner to
prescribe or administer any legally marketed device to a patient
for any condition or disease within a legitimate health care
practitioner-patient relationship.” 21 U.S.C. § 396. In
petitioners’ view, because banning a device generally makes it
unavailable, a ban naturally “limit[s] or interfere[s] with the
authority of a health care practitioner to prescribe or
administer” the device within the meaning of § 396. Id. (For
present purposes, I will assume arguendo the correctness of that
understanding.)

The key to petitioners’ all-or-nothing conception of the
FDA’s banning authority lies in their interpretation of the
phrase “legally marketed device” in § 396. That provision
protects a physician’s ability to “prescribe or administer” a
device—but only if it is a “legally marketed device.” Id.
According to petitioners, when the FDA bans a device across
the board, the device cannot be “legally marketed” at all,
rendering § 396 inapplicable. But if the FDA attempts to ban
a device only for a particular use, petitioners reason, the device
can still be “legally marketed” for other intended uses. And
because the device, on petitioners’ reading, then counts as a
“legally marketed device,” § 396 applies, such that the single-
use ban impermissibly “limit[s] or interfere[s] with the
authority of a health care practitioner to prescribe or
administer” the device for the banned use. The upshot of that
reading is that the agency must either ban a device for all its
uses or not ban it at all.

The FDA construes § 396 differently. The agency
specifically addressed (and rejected) petitioners’ understanding
of § 396 when promulgating the Rule. Rule, 85 Fed. Reg. at
 5
13,345–46. Whereas petitioners believe that a device counts as
a “legally marketed device” as long as it can be legally
marketed for any purpose, the FDA understands that a device
can be a “legally marketed device” for some purposes but not
others. In the FDA’s view, if the agency bans a device for a
specific purpose, the device cannot be “legally marketed” for
that purpose. And because the device then does not count as a
“legally marketed device” in connection with that purpose,
§ 396 is inapplicable to the ban. Consequently, the FDA
explained in the Rule, § 396 “does not . . . in any way limit
FDA’s ability to determine which devices can be legally
marketed and the uses for which they can be legally marketed.”
Id. at 13,346. It follows that electrical stimulation devices
“manufactured and used at [The Judge Rotenberg Center]” to
treat self-injurious and aggressive behaviors “are not legally
marketable devices” due to the ban, and that the “FDA’s
issuing of this rule in no way conflicts with section [396].” Id.

The agency’s interpretation of § 396 is judged under
Chevron’s two-step framework. See Chevron, U.S.A., Inc. v.
Nat. Res. Def. Council, Inc., 467 U.S. 837 (1984);
SoundExchange, Inc. v. Copyright Royalty Bd., 904 F.3d 41, 54
(D.C. Cir. 2018). The question at Chevron’s first step is
whether the statute unambiguously forecloses the agency’s
interpretation of its banning power, under which it can tailor a
ban to a device’s most problematic uses. See Catawba Cnty. v.
EPA, 571 F.3d 20, 35 (D.C. Cir. 2009). Or put alternatively,
does the statute unambiguously compel petitioners’ all-or-
nothing understanding of the FDA’s banning authority? My
colleagues believe it does. Maj. Op. 9. In their view,
petitioners’ interpretation of “legally marketed device” in
§ 396 is the only possible reading of the statute. In my view,
however, while petitioners’ reading is certainly a permissible
one, the FDA’s competing interpretation is permissible as well.
 6
And that is enough to get the agency past Chevron’s first step.
See Catawba Cnty., 571 F.3d at 35.

The statute does not expressly define what “legally
marketed device” means in § 396, or otherwise directly address
whether the term accommodates the FDA’s use-specific
reading. But outside the context of a ban, whether a device can
be “legally marketed”—whether it will be approved, what kind
of approval process it must undergo, what sorts of labeling
requirements it must satisfy, and whether it must meet
additional restrictions before sale, see 21 U.S.C.
§§ 360c(a)(1)(A)–(B), 360e(c)(1), 360j(e)(1)(A)—can vary
based on the device’s intended use. Indeed, when Congress
first granted the FDA authority to regulate medical devices, a
House Committee Report expressly recognized that “there may
be instances in which a particular device is intended to be used
for more than one purpose,” and anticipated that “each use
may . . . be treated as constituting a different device for
purposes of classification or other regulation.” H.R. Rep. No.
94–853, at 14–15 (1976). Accordingly, the FDA has long held
the position that it can “regulate[] . . . devices . . . based on the
intended uses [of] the products.” 59 Fed. Reg. 59,820, 59,821
(Nov. 18, 1994).

As an example, the FDA has assigned the same device
differing classifications (Class I, II, or III) depending on its
intended use. Consider, in that regard, cranial electrotherapy
stimulators. They are Class II devices when intended to treat
insomnia or anxiety but are Class III devices when intended to
treat depression. See, e.g., 21 C.F.R. §§ 882.5800(b)(1), (b)(2).
As a result, whether a cranial electrotherapy device can be
“legally marketed” will turn at least in part on its intended use.
If the device is intended to treat depression, it can be marketed
only after undergoing the rigorous premarket approval process
applicable to Class III devices. See id. § 882.5800(c); 21
 7
U.S.C. § 360c(a)(1)(C). But if the device is intended to treat
insomnia or anxiety, it can be marketed without regard to that
process. See 21 U.S.C. § 360c(a)(1)(B). To know whether a
device may be “legally marketed,” then, it may be necessary to
consider the device’s intended use.

Under the FDA’s interpretation of § 396, the same is true
of banned devices: when a device is banned for one intended
use, it is not a “legally marketed device” in connection with
that use. Nothing in the terms of § 396, or in the broader
statutory context, unambiguously forecloses that
understanding. To the contrary, in light of the use-specific
operation of the FDA’s regulatory authority over devices more
generally, it stands to reason that the agency’s banning power
can be understood to function in the same way.

My colleagues suggest that the FDA’s understanding of
the term “legally marketed devices” in § 396 would eviscerate
the provision’s basic object. See Maj. Op. 10. Section 396, all
agree, aims centrally to protect a physician’s ability to
administer devices for “off-label” uses—i.e., uses beyond those
for which FDA approval has already been obtained. See
Chaney v. Heckler, 718 F.2d 1174, 1180 (D.C. Cir. 1983),
rev’d on other grounds, 470 U.S. 821 (1985). But in explaining
its interpretation of § 396 in the Rule itself, the FDA
specifically recognized—as it has for decades—that the
provision “makes clear . . . that a doctor may prescribe an
approved device for a use different from those for which it has
been approved.” Rule, 85 Fed. Reg. at 13,346; see also 59 Fed.
Reg. at 59,821. Far from undercutting § 396’s protection of a
physician’s ability to prescribe off-label uses, then, the Rule’s
reading of § 396 reaffirms that very protection.

My colleagues also invoke the federalism-based interest in
construing federal statutes to avoid unduly impinging on
 8
traditional state prerogatives—here, regulating the practice of
medicine. See Maj. Op. 14–16. But again, all agree that the
agency possesses statutory authority to impose a blanket ban
on a device covering all its intended uses. It is hard to see how
allowing the agency to fashion a less intrusive ban would give
rise to a more significant federalism-based concern. In fact,
recognizing the FDA’s ability to tailor a ban to a device’s most
problematic uses will enable the agency to avoid affecting state
regulation of the practice of medicine more than is necessary.

For those reasons, § 396 does not unambiguously
foreclose the FDA’s reading of the statute at Chevron’s first
step. We then move on to Chevron’s second step, under which
we defer to the FDA’s interpretation as long as it is “reasonable
and consistent with the statute’s purpose.” UC Health v.
NLRB, 803 F.3d 669, 675 (D.C. Cir. 2015). I believe it is.

The primary purpose of the Federal Food, Drug, and
Cosmetics Act is to “protect consumers from dangerous
products.” United States v. Sullivan, 332 U.S. 689, 696 (1948).
To that end, the Act undisputedly grants the FDA power to
impose a blanket ban on an unsafe device covering all its uses.
Viewed in that light, it is eminently reasonable—and entirely
consistent with the statute’s purposes—to conclude that the
FDA may impose a more targeted ban focused solely on a
device’s unreasonably dangerous intended uses. To be sure, a
use-specific ban may be seen to constrain a physician’s ability
to acquire the device for the banned purpose. But her ability to
acquire the device would be equally (if not more) constrained
in the case of an across-the-board ban, which all agree the
agency can impose. And the FDA’s ability to take myriad other
actions that can also prevent a device from reaching physicians
is well-established: no one disputes that the agency may
decline to approve a device in the first place, for instance, or
 9
withdraw its prior approval of a device. See 21 U.S.C.
§§ 360e(d)(2), 360e(e)(1).

Against that backdrop, the agency’s authority to fashion a
partial ban on a device is unexceptionable. I would therefore
sustain the FDA’s understanding of § 396 as a permissible
exercise of the agency’s interpretive authority. I respectfully
dissent from my colleagues’ contrary conclusion.