United States ex rel. Deborah Sheldon v. Allergan Sales, LLC

24 F.4th 340

• Court: Court of Appeals for the Fourth Circuit
• Decided: January 25, 2022
• Docket: 20-2330
• Status: Published

Opinion

PUBLISHED

UNITED STATES COURT OF APPEALS FOR THE FOURTH CIRCUIT

No. 20-2330

UNITED STATES EX REL. DEBORAH SHELDON, Executrix of the Estate of Troy Sheldon, United States of America, ex rel.,

Plaintiff – Appellant,

v.

ALLERGAN SALES, LLC,

Defendant – Appellee.

------------------------------

UNITED STATES OF AMERICA; TAXPAYERS AGAINST FRAUD EDUCATION FUND,

Amici Supporting Appellant.

WASHINGTON LEGAL FOUNDATION; CHAMBER OF COMMERCE OF THE UNITED STATES OF AMERICA; PHARMACEUTICAL RESEARCH AND MANUFACTURERS OF AMERICA,

Amici Supporting Appellee.

Appeal from the United States District Court for the District of Maryland, at Baltimore. Ellen L. Hollander, Senior District Judge. (1:14-cv-02535-ELH)

Argued: October 28, 2021 Decided: January 25, 2022

Before WILKINSON, WYNN, and RICHARDSON, Circuit Judges. Affirmed by published opinion. Judge Wilkinson wrote the opinion, in which Judge Richardson joined. Judge Wynn wrote a dissenting opinion.

ARGUED: Joshua Yrion Dos Santos, UNITED STATES DEPARTMENT OF JUSTICE, Washington, D.C.; Joseph M. Callow, Jr., KEATING, MUETHING & KLEKAMP PLL, Cincinnati, Ohio, for Appellant. John Patrick Elwood, ARNOLD & PORTER KAYE SCHOLER LLP, Washington, D.C., for Appellee. ON BRIEF: Gregory M. Utter, Paul R. Kerridge, Collin L. Ryan, KEATING MUETHING & KLEKAMP PLL, Cincinnati, Ohio; Joel D. Hesch, THE HESCH FIRM, LLC, Lynchburg, Virginia, for Appellant. Michael A. Rogoff, Paula R. Ramer, New York, New York, Jeffrey L. Handwerker, Christian D. Sheehan, ARNOLD & PORTER KAYE SCHOLER LLP, Washington, D.C., for Appellee. Brian M. Boynton, Acting Assistant Attorney General, Michael S. Raab, Charles W. Scarborough, Civil Division, UNITED STATES DEPARTMENT OF JUSTICE, Washington, D.C., for Amicus United States of America. Jacklyn De Mar, TAXPAYERS AGAINST FRAUD EDUCATION FUND, Washington, D.C.; John W. Black, Samuel J. Buffone, Jr., BLACK & BUFFONE PLLC, Washington, D.C., for Amicus Taxpayers Against Fraud Education Fund. John M. Masslon II, Cory L. Andrews, WASHINGTON LEGAL FOUNDATION, Washington, D.C., for Amicus Washington Legal Foundation. James C. Stansel, Melissa B. Kimmel, PHARMACEUTICAL RESEARCH AND MANUFACTURERS OF AMERICA, Washington, D.C., for Amicus Pharmaceutical Research and Manufacturers of America. Tara S. Morrissey, Andrew R. Varcoe, UNITED STATES CHAMBER LITIGATION CENTER, Washington, D.C., for Amicus Chamber of Commerce of the United States of America. John C. O’Quinn, Matthew S. Owen, Matthew D. Rowen, Andrea R. Butler, KIRKLAND & ELLIS LLP, Washington, D.C., for Amici Pharmaceutical Research and Manufacturers of America and Chamber of Commerce of the United States of America.

2 WILKINSON, Circuit Judge:

Plaintiff Troy Sheldon filed a False Claims Act qui tam suit against his employer,

Forest Laboratories, LLC. He alleged that Forest engaged in a fraudulent price reporting

scheme under the Medicaid Drug Rebate Statute, 42 U.S.C. § 1396r-8, by failing to

aggregate discounts given to separate customers for purposes of reporting “Best Price.”

Because Forest’s reading of the Rebate Statute was at the very least objectively reasonable

and because it was not warned away from that reading by authoritative guidance, it did not

act “knowingly” under the False Claims Act. As a result, we affirm the district court’s

dismissal of Sheldon’s complaint.

We thank our friend for his thoughtful dissent. We do of course agree with him that

“[t]he False Claims Act is the government’s primary litigative tool for the recovery of

losses sustained as the result of fraud against the government.” Dissenting Op. at 32

(quoting Avco Corp. v. U.S. Dep’t of Just., 884 F.2d 621, 622 (D.C. Cir. 1989)).

Regrettably, despite all protestations, the dissent nullifies the whole concept of scienter

about which the Supreme Court has shown an especial solicitude. The FCA unquestionably

has a punitive aspect, and the kinship between civil scienter and criminal mens rea in this

case is closer than Sheldon or the dissent is willing to acknowledge.

Sheldon’s position takes the FCA a very long step toward a strict liability statute. It

conflates factual fraud and legal fraud, thereby facilitating steep liability for those whose

factual representations are not alleged to be either false or duplicitous and those whose

legal position is not only arguable but correct. Sheldon does not so much as allege reckless

disregard or deliberate indifference or nefarious knowledge here with respect to, in the

3 operative word of the statute, the “information.” 31 U.S.C. § 3729(b)(1)(A). Yet the

relator’s position instead makes sinister actors out of parties who have followed the law in

every respect and sought administrative guidance where none was ever provided. Given

the veritable thicket of Medicaid regulations, it is not too much to expect something more

in the way of clarity and direction than was ever offered here. To reward the state with

treble damages for this treatment of parties in the private sector is something no court

should do.

Sheldon would disregard Judge Hollander’s sound counsel that the Rebate Statute’s

“plain and natural reading” did not require aggregating discounts, along with her sensible

conclusion that there was not “a single example where CMS explicitly state[d] that

manufacturers must aggregate discounts to different customers along the supply chain in a

given sale.” United States ex rel. Sheldon v. Forest Laboratories, LLC, 499 F. Supp. 3d

184, 209, 211 (D. Md. 2020). Sheldon in addition recommends we ignore all our sister

circuits which have followed the framework that the Supreme Court has set forth in Safeco

Insurance Co. of America v. Burr, 551 U.S. 47 (2007), thus opening wide a stark circuit

split. See United States ex rel. Schutte v. SuperValu Inc., 9 F.4th 455, 459 (7th Cir. 2021);

United States ex rel. Streck v. Allergan, Inc., 746 F. App’x 101, 106 (3d Cir. 2018); United

States ex rel. McGrath v. Microsemi Corp., 690 F. App’x 551, 552 (9th Cir. 2017); United

States ex rel. Donegan v. Anesthesia Assocs. of Kansas City, PC, 833 F.3d 874, 879–80

(8th Cir. 2016); United States ex rel. Purcell v. MWI Corp., 807 F.3d 281, 290–91 (D.C.

Cir. 2015). Moreover, Sheldon proposes to disregard the Supreme Court’s insistence that

the concept of scienter be given “rigorous” application, Universal Health Servs., Inc. v.

4 United States ex rel. Escobar, 136 S. Ct. 1989, 2002 (2016), and the dissent dismisses as

“dictum” Supreme Court guidance which it finds inconvenient, Dissenting Op. at 31. All

this—at all three levels of the judicial system—Sheldon and the dissent would overturn, in

deference to a view that is not sustainable under law or under any notion of notice and due

process with which we are familiar.

I.

A.

Medicaid offers federal financial assistance to states that reimburse certain medical

expenses for eligible individuals. Pharm. Rsch. & Mfrs. of Am. v. Walsh, 538 U.S. 644,

650 (2003). One of those expenses is prescription drugs. 42 U.S.C. § 1396d(a)(12). To

make sure that Medicaid programs receive “the benefit of the best price for which a

manufacturer sells a prescription drug to any public or private purchaser,” H.R. Rep. No.

101-881, at 96 (1990), Congress enacted the Medicaid Drug Rebate Statute in 1990, see 42

U.S.C. § 1396r-8.

Under the Rebate Statute, manufacturers seeking to have their drugs covered by

Medicaid must enter into Rebate Agreements with the Secretary of Health and Human

Services and provide quarterly rebates to states on Medicaid sales of covered drugs. Id.

§ 1396r-8(a)(1), (c)(1)(A). The manufacturer reports the “Average Manufacturer Price”

and the “Best Price” for its covered drugs to the Centers for Medicare & Medicaid Services

(CMS); CMS then calculates the rebate amount that the manufacturer must pay to the states

for each drug. See id. § 1396r-8(b)(3)(A). For covered drugs, the rebate amount is the

greater of two numbers: (1) the statutory minimum rebate percentage, or (2) the difference

5 between the Average Manufacturer Price and the Best Price. Id. § 1396r-8(c)(1)(A).

Federal payments to each state are reduced by the rebates that the state receives from

manufacturers. Id. § 1396r-8(b)(1)(B).

The Rebate Statute defines Best Price as “the lowest price available from the

manufacturer during the rebate period to any wholesaler, retailer, provider, health

maintenance organization, nonprofit entity, or governmental entity,” which “shall be

inclusive of cash discounts, free goods that are contingent on any purchase requirement,

volume discounts, and rebates.” Id. § 1396r-8(c)(1)(C)(i), (ii)(I). CMS regulations likewise

define Best Price as “the lowest price available from the manufacturer during the rebate

period to any entity in the United States,” including “all sales and associated rebates,

discounts and other price concessions provided by the manufacturer to any entity.” 42

C.F.R. § 447.505(a) (2007). Best Price “shall be net of cash discounts . . . and any other

discounts or price reductions and rebates . . . which reduce the price available from the

manufacturer.” Id. § 447.505(e)(1) (2007). And the Rebate Agreement defines Best Price

as “the lowest price at which the manufacturer sells the [covered drug] to any purchaser in

the United States,” which “shall be adjusted by the manufacturer if cumulative discounts,

rebates or other arrangements subsequently adjust the prices actually realized.” J.A. 213;

see 56 Fed. Reg. 7049, 7050 (Feb. 21, 1991).

Acknowledging Medicaid’s complexity, the Rebate Agreement provides that “[i]n

the absence of specific guidance,” manufacturers should “make reasonable assumptions in

[their] calculations of . . . Best Price, consistent with the requirements and intent of [the

Rebate Statute], Federal regulations and the terms of this agreement.” J.A. 217. In

6 subsequent rulemaking, CMS has reaffirmed the need for manufacturers to make such

reasonable assumptions. See, e.g., Medicaid Program; Prescription Drugs, 72 Fed. Reg.

39,142, 39,164 (July 17, 2007).

Because Medicaid involves submitting claims to the government, it implicates the

False Claims Act (FCA). Relevant here, the FCA imposes liability if a person “knowingly

makes, uses, or causes to be made or used, a false record or statement material to an

obligation . . . to the Government” or “knowingly conceals or knowingly and improperly

avoids or decreases an obligation . . . to the Government.” 31 U.S.C. § 3729(a)(1)(G). The

FCA defines “knowingly” to mean that a person “(i) has actual knowledge of the

information; (ii) acts in deliberate ignorance of the truth or falsity of the information; or

(iii) acts in reckless disregard of the truth or falsity of the information.” Id. § 3729(b)(1)(A).

It “require[s] no proof of specific intent to defraud.” Id. § 3729(b)(1)(B).

The FCA allows private individuals known as relators to bring qui tam actions “for

the person and for the United States Government.” Id. § 3730(b)(1). The United States can

choose to intervene in the relator’s action if it wishes. Id. § 3730(b)(2), (4). When, as here,

the government declines to intervene, the relator generally receives 25–30% of any

proceeds of the action, plus attorney’s fees and costs. Id. § 3730(d)(2). If an FCA action

succeeds, defendants are liable for treble damages as well as a civil penalty of up to $10,000

per claim. Id. § 3729(a).

7 B.

Relator Troy Sheldon filed this FCA suit against his employer Forest Laboratories,

LLC in 2014. 1 In essence, Sheldon alleged that Forest gave discounts to separate customers

along distribution chains but failed to account for the combined amount of all discounts in

calculating Best Price, which led to the submission of false pricing reports to the

government. This allegedly reduced the rebates that Forest paid to participating states and

resulted in the federal government paying at least $680 million more than it would have if

Forest had accurately reported Best Price.

To give an example: on one covered drug, Sheldon alleged that in FY2013 Forest

gave a 20% discount to a patient’s insurance company and a 10% discount to the same

patient’s pharmacy—two different entities on the distribution chain. See J.A. 98. Sheldon

alleged that Forest was required to aggregate these discounts, report a Best Price of 70%,

and give Medicaid a 30% rebate. Instead, Forest did not aggregate these discounts because

they were given to different entities, reported a Best Price of 80% (based on the highest

discount given to a single entity), and gave Medicaid a 23.1% rebate (the statutory

minimum rebate percentage for that year, see 42 U.S.C. § 1396r-8(c)(1)(B)(i)(VI)).

1 Troy Sheldon died after filing this action and Deborah Sheldon, his wife, was substituted as plaintiff. And in 2018, Forest merged into Allergan Sales, LLC. For clarity, we refer to Troy Sheldon rather than Deborah and to Forest rather than Allergan.

Sheldon sued on behalf of the United States. The suit was initially filed under seal. See 31 U.S.C. § 3730(b)(2). After a five-year investigation and every opportunity to intervene, the government declined to do so, and the suit was unsealed in October 2019.

8 Sheldon alleges that this led to the federal government paying 6.9% more for this drug than

it would have if Forest had accurately reported Best Price.

Forest moved to dismiss Sheldon’s complaint, and the district court in a thoughtful

opinion granted Forest’s motion. 499 F. Supp. 3d 184. The district court found that Sheldon

had failed to plead both that the claims at issue were false and that Forest had made them

knowingly. 2 Relevant here, it held that Forest had offered “a plausible and objectively

reasonable interpretation” of the Rebate Statute. Id. at 209. Beginning with the statutory

text, the district court found that its “plain and natural reading” did not require aggregating

discounts. Id. And looking at the regulatory language and history, the district court did not

find “a single example where CMS explicitly state[d] that manufacturers must aggregate

discounts to different customers along the supply chain in a given sale.” Id. at 211. The

district court then concluded that CMS guidance “was not so clear as to warn Forest away

from its interpretation,” especially considering the complexity of the statutory scheme. Id.

at 212. So it held that Forest did not act with the requisite scienter when submitting Best

Price reports to the government.

II.

We review de novo the dismissal of a relator’s complaint under Rule 12(b)(6).

United States ex rel. Rostholder v. Omnicare, Inc., 745 F.3d 694, 700 (4th Cir. 2014). To

plead his FCA claim, Sheldon must plausibly allege that Forest (1) made a false statement;

2 Because we hold that Forest did not act knowingly under the FCA, we have no occasion to address the district court’s holding as to falsity.

9 (2) with the requisite scienter (“knowingly”); (3) that was material; and (4) that caused the

government to pay out money. Id.; see also 31 U.S.C. § 3729(a)(1)(G). Here, we interpret

the second element, scienter, in line with the Supreme Court’s guidance in Safeco.

Applying that analysis, we hold that Forest did not act knowingly under the FCA.

1.

We are tasked with “strict enforcement” of the FCA’s “rigorous” scienter

requirement. Escobar, 136 S. Ct. at 2002. As noted, the FCA defines “knowingly” to mean

that a person “(i) has actual knowledge of the information; (ii) acts in deliberate ignorance

of the truth or falsity of the information; or (iii) acts in reckless disregard of the truth or

falsity of the information.” 31 U.S.C. § 3729(b)(1)(A). Yet it does not further define these

terms or signify how they apply in situations where it is unclear if a defendant complied

with the law.

Fortunately, we are not without guidance in this area. In Safeco, the Supreme Court

interpreted the Fair Credit Reporting Act’s analogous scienter provision. Like every other

circuit to consider the issue, we hold that Safeco applies with equal force to the FCA’s

scienter requirement. See Schutte, 9 F.4th at 459; Streck, 746 F. App’x at 106; McGrath,

690 F. App’x at 552; Donegan, 833 F.3d at 879–80; Purcell, 807 F.3d at 290–91.

Safeco interpreted the scienter requirement of the Fair Credit Reporting Act

(FCRA), which required defendants to act “willfully.” See 15 U.S.C. § 1681n(a). Because

the FCRA did not define this common law term, the Court looked to its common law

meaning. Safeco, 551 U.S. at 58. It interpreted the FCRA’s “willfulness” requirement to

10 cover both knowing and reckless violations of the statute. Id. at 57. Then it defined

recklessness as “conduct violating an objective standard: action entailing ‘an unjustifiably

high risk of harm that is either known or so obvious that it should be known.’” Id. at 68

(quoting Farmer v. Brennan, 511 U.S. 825, 836 (1994)). Accordingly, it found a

defendant’s subjective intent irrelevant: “To the extent that [plaintiffs] argue that evidence

of subjective bad faith can support a willfulness finding even when the company’s reading

of the statute is objectively reasonable, their argument is unsound.” Id. at 70 n.20.

The Safeco Court set forth a two-step analysis as to reckless disregard, first asking

whether defendant’s interpretation was objectively reasonable and then determining

whether authoritative guidance might have warned defendant away from that reading. Id.

at 69–70. Because defendant’s reading “was not objectively unreasonable” and “ha[d] a

foundation in the statutory text,” it did not act recklessly—even though its reading was

ultimately “erroneous.” Id. And defendant had no guidance from the courts of appeals or

the implementing agency that “might have warned it away from the view it took.” Id. at

70. “Given this dearth of guidance and the less-than-pellucid statutory text, [defendant’s]

reading was not objectively unreasonable, and so falls well short of raising the

‘unjustifiably high risk’ of violating the statute necessary for reckless liability.” Id. Failure

to meet this recklessness standard precluded a finding of knowledge as well: “Where, as

here, the statutory text and relevant court and agency guidance allow for more than one

reasonable interpretation, it would defy history and current thinking to treat a defendant

who merely adopts one such interpretation as a knowing or reckless violator.” Id. at 70

n.20.

11 As noted above, several of our sister circuits have applied Safeco’s scienter analysis

to the FCA. And with good reason. The FCA defines “knowingly” as including actual

knowledge, deliberate ignorance, and reckless disregard. 31 U.S.C. § 3729(b)(1)(A).

Safeco interpreted “willfully” to include both knowledge and recklessness. 551 U.S. at 57,

68. Given this parallel, we hold that Safeco’s reasoning applies to the FCA’s scienter

requirement. Under the FCA, a defendant cannot act “knowingly” if it bases its actions on

an objectively reasonable interpretation of the relevant statute when it has not been warned

away from that interpretation by authoritative guidance. This objective standard precludes

inquiry into a defendant’s subjective intent.

In adopting this standard, we join each and every circuit that has considered Safeco’s

applicability to the FCA. For example, the Seventh Circuit reasoned that Safeco “defined

a similar common law term . . . which the Court interpreted as encompassing the same

common law scienter terms used in the FCA.” 9 F.4th at 465. It rightly concluded that

Safeco “announced a standard inquiry for reckless disregard” and found “no reason why

the scienter standard established in Safeco (for violations committed knowingly or with

reckless disregard) should not apply to the same common law terms used in the FCA.” Id.

After all, the Supreme Court has held that the FCA “does employ the common law

meaning” for other common law terms like false and fraudulent, so long as there are no

textual indicia to the contrary. Id. (citing Escobar, 136 S. Ct. at 1999 & n.2). Finding none

here, there was “no barrier to importing the Safeco standard to the FCA.” Id.

Sheldon claims that Safeco should not apply, alluding to Halo Electronics, Inc. v.

Pulse Electronics, Inc., 136 S. Ct. 1923 (2016). But that case does not suggest a different

12 result. See Schutte, 9 F.4th at 466–67 (finding Safeco more analogous to FCA than Halo

Electronics). Halo Electronics interpreted § 284 of the Patent Act, which allowed for treble

damages in certain infringement cases but did not specify scienter. 136 S. Ct. at 1928; see

35 U.S.C. § 284 (“[T]he court may increase the damages up to three times the amount

found or assessed.”). The Court found that such damages “are generally reserved for

egregious cases of culpable behavior” and clarified that a showing of objective recklessness

was not necessary in a context of “such deliberate wrongdoing.” Id. at 1932. It also

emphasized the district court’s discretion and the lack of textual limitations on that

discretion. Id. at 1931–32. The Court acknowledged Safeco’s standard but did not apply it

in the context of the Patent Act because its “precedents [made] clear that ‘bad-faith

infringement’ is an independent basis for enhancing patent damages.” Id. at 1933 n.*. In

this situation, a test of objective recklessness “impermissibly encumber[ed] the statutory

grant of discretion to district courts.” Id. at 1932 (quoting Octane Fitness, LLC v. ICON

Health & Fitness, Inc., 572 U.S. 545, 553 (2014)).

Context matters, and here two differences stand out. First, § 284 did not include a

scienter requirement, while the FCA clearly limits liability to claims that are made

“knowingly.” And the Supreme Court has instructed that this “rigorous” requirement ought

to find “strict enforcement” in the courts. Escobar, 136 S. Ct. at 2002. Second, while § 284

concerned whether district courts could issue a particular amount of damages after finding

liability, the relevant provision here concerns whether liability exists at all. Taking these

differences into account, the gap between the FCA and the Patent Act is much wider than

13 that between the FCA and the FCRA—both of which include an explicit scienter standard

(covering both knowledge and recklessness) that speaks to liability rather than damages.

Sheldon also argues that Safeco improperly collapses the FCA’s statutory

definitions. But applying Safeco does not sap the FCA’s three scienter definitions of

independent meaning. Safeco itself recognized that recklessness and knowledge were

separate subcategories of willfulness. 551 U.S. at 60. Yet it still held that its standard served

as the starting point for both, refusing to treat a defendant who adopted a reasonable

interpretation “as a knowing or reckless violator.” Id. at 70 n.20 (emphasis added). The

same is true here. That actual knowledge, deliberate ignorance, and reckless disregard are

distinct—which we do not dispute—does not preclude them from sharing a threshold

requirement. See Schutte, 9 F.4th at 468. Nor does it preclude them from functioning as a

hierarchy, as is commonly understood. Reckless disregard has been called the “most

capacious,” United States ex rel. Watson v. King-Vassel, 728 F.3d 707, 712 (7th Cir. 2013),

the “loosest,” Purcell, 807 F.3d at 288, and the “baseline,” Schutte, 9 F.4th at 465, of the

FCA’s scienter standards. So if a defendant has not acted with reckless disregard in its view

of the statute, “it follows a fortiori” that it has not acted with deliberate ignorance or actual

knowledge, which “plainly demand[] even more culpability.” Urquilla-Diaz v. Kaplan

Univ., 780 F.3d 1039, 1058 n.15 (11th Cir. 2015).

2.

Safeco does not apply to all FCA suits. There are two general categories of false

claims under the FCA: those that are factually false and those that are legally false. See

United States ex rel. Wilkins v. United Health Grp., Inc., 659 F.3d 295, 305 (3d Cir. 2011).

14 The paradigmatic FCA action targets factually false claims—those in which someone “has

submitted an incorrect description of goods or services provided or a request for

reimbursement for goods or services never provided.” United States ex rel. Polukoff v. St.

Mark’s Hosp., 895 F.3d 730, 741 (10th Cir. 2018) (citation omitted); see, e.g., United States

ex rel. Citynet, LLC v. Gianato, 962 F.3d 154, 157 (4th Cir. 2020) (complaint alleged that

defendant billed the federal government for “material and labor it did not provide, and for

[projects] that were not constructed”); Affinity Living Grp., LLC v. StarStone Specialty Ins.

Co., 959 F.3d 634, 636 (4th Cir. 2020) (complaint alleged that defendant “submitted

reimbursement claims for resident services that were never provided”). Of a different

vintage are legally false claims, which “generally require knowingly false certification of

compliance with a regulation or contractual provision as a condition of payment.” Polukoff,

895 F.3d at 741.

Safeco simply does not reach factually false claims, where the law is clear. Instead,

it is narrowly cabined to legally false claims—like the one here—which involve contested

statutory and regulatory requirements. As we have recognized, “establishing even the

loosest standard of knowledge, i.e., acting in reckless disregard of the truth of falsity of the

information, is difficult when falsity turns on a disputed interpretive question.” United

States ex rel. Complin v. N.C. Baptist Hosp., 818 F. App’x 179, 184 (4th Cir. 2020)

(quoting Purcell, 807 F.3d at 288). After all, “[a] defendant might suspect, believe, or

intend to file a false claim, but it cannot know that its claim is false if the requirements for

that claim are unknown.” Schutte, 9 F.4th at 468.

15 Nor does Safeco write defendants a blank check. To start, Safeco’s first step requires

an objectively reasonable reading of the statute. If a defendant bases its actions on an

unreasonable view of the law, it runs a considerable litigation risk. Knowing an FCA claim

is waiting in the wings, it takes a serious chance that a court will find liability if it attempts

to concoct strained justifications for its actions. Much better to steer clear of danger than

to risk it all defending a questionable interpretation in court.

And not every objectively reasonable reading will suffice. Safeco’s second step

allows the government to issue authoritative guidance that clarifies its interpretation of the

law and so warns defendants away from otherwise reasonable interpretations. The test thus

“does not shield bad faith defendants that turn a blind eye to guidance indicating that their

practices are likely wrong.” Id. But it does put the burden where it belongs. If the

government wants to hold people liable for violating labyrinthine reporting requirements,

it at least needs to indicate a way through the maze. See, e.g., Gates & Fox Co. v. OSHRC,

790 F.2d 154, 156 (D.C. Cir. 1986) (Scalia, J.) (citation omitted) (“If a violation of a

regulation subjects private parties to criminal or civil sanctions, a regulation cannot be

construed to mean what an agency intended but did not adequately express.”).

Safeco’s standard duly ensures that defendants must be put on notice before facing

liability for allegedly failing to comply with complex legal requirements. Without such

notice, defendants are not likely to receive due process. “A fundamental principle in our

legal system is that laws which regulate persons or entities must give fair notice of conduct

that is forbidden or required.” FCC v. Fox Television Stations, Inc., 567 U.S. 239, 253

(2012). Such “clarity in regulation is essential to the protections provided by the Due

16 Process Clause of the Fifth Amendment,” id., especially when, as here, defendants are

faced with “damages that are essentially punitive in nature,” Vt. Agency of Nat. Res. v.

United States ex rel. Stevens, 529 U.S. 765, 784 (2000) (describing FCA); see also Tex.

Indus. v. Radcliff Materials, Inc., 451 U.S. 630, 639 (1981) (“The very idea of treble

damages reveals an intent to punish past, and to deter future, unlawful conduct, not to

ameliorate the liability of wrongdoers.”).

It is profoundly troubling to impose such massive liability on individuals or

companies without any proper notice as to what is required. Safeco avoids this trouble by

making the government “provide a reasonably clear standard of culpability to circumscribe

the discretion of the enforcing authority and its agents.” United States v. Hoechst Celanese

Corp., 128 F.3d 216, 224 (4th Cir. 1997) (citation omitted). Rightly so. As the Supreme

Court has made clear, “concerns about fair notice and open-ended liability can be

effectively addressed through strict enforcement” of the FCA’s “rigorous” scienter

requirement. Escobar, 136 S. Ct. at 2002 (citation omitted). Safeco’s careful analysis is

just the right means to further this end. See, e.g., Purcell, 807 F.3d at 287 (“Strict

enforcement of the FCA’s knowledge requirement helps to . . . avoid[] the potential due

process problems posed by penalizing a private party for violating a rule without first

providing adequate notice of the substance of the rule.”) (citation omitted). We therefore

decline Sheldon’s invitation to make our circuit an outlier.

B.

Applying Safeco’s test to Forest’s conduct, we conclude that Forest did not act

“knowingly” under the False Claims Act. Forest’s reading of the Rebate Statute was not

17 only objectively reasonable but also the most natural. And Forest was not warned away

from its reading by authoritative guidance from CMS. As a result, Sheldon failed to plead

scienter as required by the FCA. 3

We must first determine whether Forest’s reading was objectively reasonable by

examining the text of the statute. Safeco, 551 U.S. at 69–70. The Rebate Statute defines

Best Price as “the lowest price available from the manufacturer during the rebate period to

any wholesaler, retailer, provider, health maintenance organization, nonprofit entity, or

governmental entity.” 42 U.S.C. § 1396r-8(c)(1)(C)(i). The plain language here indicates

that Best Price is one offered to a single entity.

Notably, both “price” and all of the entities listed are singular, joined by the

disjunctive “or.” And “any” usually means a single member in a class if used with singular

nouns. Any, Oxford English Dictionary (3d ed. 2021). This linguistic construction (singular

3 Sheldon argues that it was improper for the district court to decide the scienter question on a motion to dismiss. Yet the Supreme Court has generally urged us to resolve cases on a motion to dismiss when a claim is not “plausible on its face.” Ashcroft v. Iqbal, 556 U.S. 662, 678 (2009) (quoting Bell Atl. Corp v. Twombly, 550 U.S. 544, 570 (2007)). This plausibility standard “asks for more than a sheer possibility that a defendant has acted unlawfully.” Id. And that standard bars Sheldon’s claim, which does not allege any plausible theory of recovery. In addition, we have specifically held that a “district court did not err in deciding the issue of [FCA] scienter at the Rule 12(b)(6) motion-to-dismiss stage,” Complin, 818 F. App’x at 183 n.5 (citing Rostholder, 745 F.3d at 703)—even when the case involved the question of whether a defendant was warned away from its interpretation, see id. at 184 n.6. Other circuits have similarly conducted the Safeco analysis in the FCA context of a motion to dismiss. See, e.g., Streck, 746 F. App’x 101; United States ex rel. Hixson v. Health Mgmt. Sys., Inc., 613 F.3d 1186 (8th Cir. 2010). This is especially appropriate when, as here, the question of whether a defendant has been warned away depends upon the interpretation of legal materials.

18 nouns plus the disjunctive) strongly advises against aggregating discounts to multiple

entities. Change some nouns to see why. If, when striking a deal for baseball equipment,

the thrifty Kansas City Royals asked for “the lowest price available from the manufacturer

to any wholesaler, retailer, professional baseball team, minor-league organization, or

collegiate program,” no one would think that the equipment company needs to aggregate

prices. The Royals are just asking for the best deal that any one of the other entities

received. Or imagine you ask a friend about “the lowest apple price available to any

wholesaler, grocery store, or restaurant.” You would not expect your friend to aggregate

prices between grocery stores and restaurants, but instead report to you the single lowest

price at which someone can readily purchase apples.

Finally, “available” means “suitable or ready for use,” “at hand,” or “readily

obtainable.” The Random House Dictionary of the English Language 142 (2d ed. 1987).

The statute is thus talking about an actual price, not something that is purely hypothetical.

A price is not “available” to an entity if the manufacturer must first aggregate other prices.

Overall, this plain language conveys that Forest was not required to aggregate

discounts given to separate customers. Yet this does not give Forest a free ride. The Rebate

Statute most naturally reads as requiring drug manufacturers to give Medicaid the lowest

price that was provided to any single purchaser. This includes aggregating discounts to a

single entity even if given at different points in time. But the statute cannot be stretched

beyond this singular point.

Other provisions in the Rebate Statute confirm this reading. For example, the Rebate

Statute defines Average Manufacturer Price as “the average price paid to the manufacturer

19 for the drug” by “wholesalers” and “retail community pharmacies.” Id. § 1396r-

8(k)(1)(A)(i)–(ii). An “average,” by definition, requires some sort of combination. And

something “paid to the manufacturer” might incorporate discounts to different entities. Yet

Average Manufacturer Price is also limited to a narrower class of entities than is Best Price,

making the reporting problem less onerous. We refuse to ignore such distinctions in the

statutory scheme. Congress chose dissimilar language for the two terms, and these

linguistic differences must be given legal effect. See, e.g., Soliman v. Gonzales, 419 F.3d

276, 283 (4th Cir. 2005) (“Where Congress has utilized distinct terms within the same

statute, . . . we endeavor to give different meanings to those different terms.”).

Beyond faithfulness to the statutory text, this reading also accords with practical

realities. Well has it been said that Medicaid statutes and regulations “are among the most

completely impenetrable texts within human experience.” Rehab. Ass’n of Va., Inc. v.

Kozlowski, 42 F.3d 1444, 1450 (4th Cir. 1994). And discount aggregation in particular

raises some of the thorniest issues in government price reporting. See, e.g., Astra USA, Inc.

v. Santa Clara Cnty., 563 U.S. 110, 115 (2011) (“Calculation of a manufacturer’s ‘average’

and ‘best’ prices . . . is a complex enterprise.”). Numerous entities—including state

Medicaid agencies, Pharmacy Benefit Managers, manufacturers, wholesalers, and

pharmacies—are involved in increasingly complicated customer relationships. See, e.g.,

Rachel Dolan & Marina Tian, Pricing and Payment for Medicaid Prescription Drugs,

Kaiser Family Foundation (Jan. 23, 2020), https://www.kff.org/medicaid/issue-

brief/pricing-and-payment-for-medicaid-prescription-drugs/ (depicting “complex drug

supply and payment chain” for prescription drugs covered by Medicaid). Because of these

20 complex sales practices, “manufacturers may find it difficult to determine how to treat

certain sales practices when calculating prices.” U.S. Department of Health & Human

Services, Office of Inspector General, Reasonable Assumptions in Manufacturer Reporting

of AMPs and Best Prices 3–4 (Sept. 2019), https://oig.hhs.gov/oei/reports/oei-12-17-

00130.pdf (OIG Report). Given this considerable difficulty, it makes good sense to think

that manufacturers are expected to report a price actually given to a purchaser, rather than

cobbling together bits and pieces to fashion a price never “available” to any actual entity.

We turn next to the CMS regulations. Of course, courts, not agencies, are the

ultimate interpreters of statutes. See, e.g., Chevron U.S.A., Inc. v. Nat. Res. Def. Council,

Inc., 467 U.S. 837, 842–43 (1984) (limiting deference in statutory interpretation to

situations where the law is ambiguous and the agency interpretation is reasonable). And to

the extent that CMS regulations are relevant, here they simply mirror the statutory

language. CMS defines Best Price as “the lowest price available from the manufacturer

during any rebate period to any entity in the United States.” 42 C.F.R. § 447.505(a) (2007).

Again, each term is singular, most naturally referring to the lowest price given to a single

entity. Likewise, the Rebate Agreement (also promulgated by CMS regulation) defines

Best Price as “the lowest price at which the manufacturer sells the [covered drug] to any

purchaser in the United States.” J.A. 213; see 56 Fed. Reg. at 7050. This straightforward

language—“any purchaser,” again singular—counsels in favor of Forest’s interpretation.

And while Sheldon makes much of three other words in the Rebate Agreement (“prices

actually realized”) to argue that discounts must be aggregated, these words cannot be

wrenched out of context or used to subvert the Rebate Statute’s natural meaning. Read

21 consistently with the governing statute (to which it is subordinate), the Rebate Agreement’s

“prices actually realized” simply means prices the manufacturer receives on sales to each

individual customer.

Clearly, Forest’s reading “has a foundation in the statutory text.” Safeco, 551 U.S.

at 69–70. Not only that; it is the best reading of that text. We agree with the district court

that the “plain and natural reading” of the Rebate Statute means that Best Price entails “the

lowest price available by the manufacturer, including all price concessions, to any one of

the listed entities, but not to multiple entities.” Sheldon, 499 F. Supp. 3d at 209. There is

nothing in the statute to suggest that Best Price requires aggregating discounts given to

separate entities. Thus, we hold that Forest has offered, at minimum, an objectively

reasonable reading of the Rebate Statute. It in turn becomes more difficult to conclude that

a party “knowingly” presented a false claim, 31 U.S.C. § 3729(b)(1)(A), when that claim

is premised on such a textually sound view.

Next we ask whether authoritative guidance warned Forest away from its

interpretation. See Safeco, 551 U.S. at 70. To function as a warning, authoritative guidance

requires both the right source and sufficient specificity. When it comes to source, either

circuit court precedent or guidance from the relevant agency is required. See id.; Schutte,

9 F.4th at 471 (limiting authoritative guidance to these two sources); Purcell, 807 F.3d at

289 (considering only these two sources); Streck, 746 F. App’x at 106, 108 (considering

only these two sources). And the guidance must “canvass the issue” with sufficient

specificity to be able to function as a warning. Safeco, 551 U.S. at 70 n. 19. It does not

22 suffice for agency guidance merely to be related to the question at hand; instead,

“authoritative guidance must have a high level of specificity to control an issue.” Schutte,

9 F. 4th at 471; see also Safeco, 551 U.S. at 70 n.20 (agency guidance did not warn away

when it “allow[ed]” defendant’s interpretation). Because CMS never clearly stated that

discount aggregation to different entities was required, it did not act with the specificity

necessary to warn Forest away from its interpretation.

CMS knew as early as 2006 that manufacturers were not aggregating discounts

given to different entities along supply chains. After CMS submitted its proposed rule on

Medicaid drug pricing, several manufacturers, including Forest, offered comments. These

comments expressed a uniform view that Best Price “has always been interpreted to mean

the single lowest price to a particular customer.” J.A. 239; accord J.A. 271 (“[Best Price]

is the single lowest price at which the manufacturer sells the product to a single

customer.”); J.A. 285 (“We therefore request that CMS confirm that best price will

continue to be the lowest price at which a drug is actually sold.”); J.A. 305 (“Best price is

not calculated as a price derived by aggregating price concessions to different customers.”).

And the manufacturers asked CMS to “clarify” or “confirm” that it would continue to be

so. J.A. 239, 271, 285.

CMS nonetheless failed to clarify and thereby maintained strategic ambiguity. But

in all material respects, the final rule adopted the proposed rule’s Best Price definition. 72

Fed. Reg. at 39,242–43. As we have seen, that language simply reflected the Rebate

Statute, which most naturally supports Forest’s interpretation.

23 Sheldon points to two CMS responses to comments that, he says, should have

warned Forest away. While both were related to the broad issue here—Best Price reporting

and discounts—neither spoke directly to whether manufacturers were required to aggregate

discounts given to separate entities on the supply chain. As a result, they were not sufficient

to warn Forest away from its objectively reasonable interpretation.

The first scenario involved Pharmacy Benefit Managers (PBMs), which the

proposed rule had initially included in Best Price. See Medicaid Program; Prescription

Drugs, 71 Fed. Reg. 77,174, 77,197 (Dec. 22, 2006) (proposing that 42 C.F.R

§ 447.505(c)(2) include PBM rebates). After receiving public comments, CMS agreed to

generally remove PBM rebates from Best Price calculation in its final rule but noted one

situation where PBM rebates might be included. See 72 Fed. Reg. at 39,198, 39,242; 42

C.F.R § 447.505(d)(13) (2007) (“Best price excludes PBM rebates, discounts, or other

price concessions except . . . where such rebates, discounts, or other price concessions are

designed to adjust prices at the retail or provider level.”). As Sheldon conceded below, this

example has nothing to do with whether discounts should be aggregated in calculating Best

Price; instead, “CMS’s comments involving PBMs simply addressed how rebates to an

excluded entity might nevertheless fall within Best Price.” D. Ct. Docket 79 at 22. It thus

does not provide sufficient specificity to warn Forest away from its position on aggregating

discounts to included entities.

The second scenario proves similarly lacking, as it concerned two discounts

administered through a single entity. One commenter asked if Best Price calculations

required aggregating prompt pay discounts to wholesalers and wholesaler chargeback

24 agreements, and CMS confirmed that they did. 72 Fed. Reg. at 39,199. Yet as the district

court noted, “the different price concessions . . . both actually function as price concessions

to [a] single entity—the wholesaler.” Sheldon, 499 F. Supp. 3d at 211. The prompt pay

discount lowers the wholesaler’s price at the time of sale. And the chargeback agreement

means that “the wholesaler delivers the product to the favored purchaser at the discounted

price and then ‘charges back’ the manufacturer for the difference.” In re Brand Name

Prescription Drugs Antitrust Litig., No. 94-cv-897, 1996 WL 167350, at *2 (N.D. Ill. Apr.

4, 1996). It thus functions as a lagged price concession to the wholesaler and is properly

included in a Best Price calculation because it affects the price available to a single entity.

The Rebate Statute, after all, does require aggregating discounts if they are given to a single

entity. But as the district court noted, CMS’s comments here “did not actually clarify

whether there is a requirement to aggregate concessions from multiple entities in separate

arrangements.” Sheldon, 499 F. Supp. 3d at 211. So they were not precise enough to warn

Forest away.

Sheldon’s other examples fare no better. 4 Mostly, they involve language about

“prices actually realized” or stay at high levels of generality. This is simply insufficient.

All told, Sheldon has not, in the words of the district court, “pointed to a single example

where CMS explicitly state[d] that manufacturers must aggregate discounts to different

4 While Sheldon twice alleged that Forest’s conduct continued “to the present,” J.A. 106, 107, his complaint contains no factual allegations concerning Forest’s conduct after 2014 (when Forest terminated Sheldon). Two conclusory references about continuing conduct are simply insufficient to meet Rule 12(b)(6)’s standard, which requires some level of “factual content.” Iqbal, 556 U.S. at 678.

25 customers along the supply chain in a given sale.” Id. It thus did not warn Forest away from

its well-grounded interpretation. 5

Instead of a warning, CMS issued manufacturers like Forest a permission slip.

CMS’s Rebate Agreement provides that “in the absence of specific guidance,”

manufacturers should “make reasonable assumptions in their calculations of . . . Best Price,

consistent with the requirements and intent of [the Rebate Statute], Federal regulations and

the terms of this agreement.” J.A. 217. In the very rulemaking that Sheldon highlights,

CMS reaffirmed the need to make reasonable assumptions—not once, not twice, but nine

times. See 72 Fed. Reg. at 39,164, 39,166, 39,167, 39,171, 39,191, 39,211. Combine this

exhortation with the complex statutory scheme and it is no wonder that reliance on

reasonable assumptions is widespread. See OIG Report at 24.

In fact, a 2019 HHS Inspector General report found that eighty percent of

manufacturers reported making reasonable assumptions about the precise issue here:

whether discounts given to separate entities must be aggregated. Id. at 9. And this issue is

far from unique. More than fifty percent of responding manufacturers reported making

reasonable assumptions in fourteen different areas identified by the Inspector General. Id.

at 9–10. Importantly, it is not the case that manufacturers are taking advantage of CMS’s

silence; almost two thirds reported a desire for additional guidance on these very issues.

Id. at 11. Facing these requests, CMS demurs. Indeed, “CMS specifically instructs

5 Because Safeco focuses on objective reasonableness and forecloses inquiry into subjective beliefs, see 551 U.S. at 70 n.20, Sheldon’s allegations regarding Forest’s motivation for undertaking a data audit are simply irrelevant.

26 manufacturers not to submit their assumptions to the agency, and states that if a

manufacturer does so, CMS will not review the assumptions.” Id. at 20.

What CMS once gave with one hand it now wants to take away with the other.

Having told manufacturers to rely on reasonable assumptions, the government cannot

receive damages when Forest has done exactly that. Moreover, it cannot do so when CMS

has refused to respond to manufacturer requests for clarification. What a troubling result:

companies ask for explanation and at first are told to do their best but then are subjected to

potentially ruinous liability for following those instructions. How can this—which looks

more like Calvinball than the rule of law—possibly qualify as a sufficient warning? See

Bill Watterson, The Calvin & Hobbes Tenth Anniversary Book 129 (1995) (“People have

asked how to play Calvinball. It's pretty simple: you make up the rules as you go.”).

Of course, CMS may not wish to specify its position on the issue. From its vantage

point, that might be understandable. Clear regulations constrain regulatory power and limit

future flexibility, which is why an agency might find them undesirable. See, e.g., Kisor v.

Wilkie, 139 S. Ct. 2400, 2440–41 (Gorsuch, J., concurring in the judgment) (“Whether

purposeful or not, the agency’s failure to write a clear regulation winds up increasing its

power.”). To be sure, there are plenty of reasons why agencies might prefer ambiguity. But

such reasons are not necessarily permissible. Retaining ambiguity in order to expand

potential liability for regulated entities cannot pass muster. In a world where the

administrative state “wields vast power and touches almost every aspect of daily life,” Free

Enter. Fund v. Pub. Co. Acct. Oversight Bd., 561 U.S. 477, 499 (2010), allowing agencies

to take advantage of companies like this would not be right.

27 CMS did not warn Forest away from its objectively reasonable reading. None of its

guidance dealt with aggregating discounts to different entities, and it even invited Forest

to make reasonable assumptions. So the district court correctly dismissed Sheldon’s

complaint for failure to allege scienter.

III.

Safeco’s two prongs are interrelated; though separate, they are not totally divorced.

Looking at both the statute’s text and the agency’s guidance, a coherent picture emerges.

Forest made eminently reasonable assumptions based on the statutory text, and CMS

invited assumptions precisely of this sort. The False Claims Act does not assess liability

through ambush. Companies must instead knowingly submit a false claim to be liable. And

Forest simply did not do so here.

We cannot accept the idea that a defendant acts “knowingly” when its reading of a

statute is both objectively reasonable and in fact the best interpretation; when the agency’s

regulation mirrors, rather than repudiates, that interpretation; when the agency resists

attempts to get it to clarify its view; and when the agency explicitly invites regulated parties

to make reasonable assumptions. It is not plausible to accuse Forest of acting “knowingly”

in these circumstances.

All that said, the government is not without recourse. Should Congress so wish, it

can alter the Rebate Statute to require the aggregate reporting of discounts to separate

entities. But the burden is on the government to be clear. As the district court recognized,

this case presents no sound rationale for the immense consequences the relator would have

this court impose.

28 The judgment of the district court is hereby affirmed.

AFFIRMED

29 WYNN, Circuit Judge, dissenting:

Those who believe that some judicial decisions usurp the power of elected

legislatures by making the law rather than merely interpreting it can add another tally to

their ledgers. Today, with the stroke of a pen, my thoughtful friends in the majority opinion

effectively neuter the False Claims Act—the Government’s primary tool for fighting

fraud—by eliminating two of its three scienter standards (actual knowledge and deliberate

ignorance) and replacing the remaining standard with a test (objective recklessness) that

only the dimmest of fraudsters could fail to take advantage of.

Over thirty years ago, Congress grew concerned that years of restrictive court

interpretations had artificially narrowed the False Claims Act’s scienter requirement. To

remedy this problem, Congress crafted three distinct and expansive scienter standards.

Today’s majority opinion undoes that work by making a new law that reads two of those

three scienter standards right out of existence. In their place, the majority opinion erects its

own threshold scienter test that allows fraudsters to escape any liability so long as they can

come up with a post hoc legal rationale that passes the smell test.

But the majority opinion’s legal hand-waving cannot cover the stench here. Troy

Sheldon plausibly alleges that for years, pharmaceutical giant Forest Laboratories, LLC

failed to include stacked rebates when reporting its best drug prices to the Government.

When alerted that its scheme was unlawful, Forest hired a data-scrubbing firm to identify

and eliminate rebate stacking for many of its customers. However, it continued to pay out

stacked rebates to its preferred customers, rebates that it then failed to report in its best

30 price calculations for years to come. That fraudulent scheme bilked the federal Government

out of $680 million.

Yet, the majority opinion finds it unnecessary to even address these facts due to its

wholesale revision of the False Claims Act’s scienter standard. But what, you might ask,

empowers judges to trade in their judicial robes for congressional pins, rewrite the statute,

and ignore the factual record? The underwhelming answer: a dictum single footnote buried

at the end of a Supreme Court opinion on credit reporting.

Tellingly, the majority opinion spends 4/5 of its introduction cavalierly dismissing

the recognition of its judicial overreach as mere “protestations.” Majority Op. at 3. But the

fact that it found the need to say so with a first breath pontification—without providing any

context for the reader—says otherwise.

At any rate, that first breath does nothing to dispel the substantive concerns

identified in this dissenting opinion: it does not, for example, tangle with the damning facts

of this case, explain why importing mismatched common law into the False Claims Act is

a good idea, or, most importantly, defend its decision to write two of the Act’s three scienter

standards out of existence. Instead, it accuses the dissenting opinion—which seeks to

maintain the statutory status quo by keeping the three scienter standards created by

Congress—of somehow taking a “very long step toward a strict liability statute.” Id. And

without any sense of irony, it protests that the dissenting opinion “nullif[ies] the whole

concept of scienter” for the False Claims Act. Id. But as explained below, that is precisely

what the majority opinion accomplishes by rewriting the Act’s scienter standard to suit its

own policy ends.

31 Because I cannot join in this judicial overhaul of the False Claims Act—an overhaul

that will require further congressional correction—I dissent.

“The False Claims Act is the government’s primary litigative tool for the recovery

of losses sustained as the result of fraud against the government.” Avco Corp. v. U.S. Dep’t

of Just., 884 F.2d 621, 622 (D.C. Cir. 1989) (citing S. Rep. No. 99-345, at 1 (1986),

reprinted in 1986 U.S.C.C.A.N. 5266, 5266). However, the Act only reaches “knowingly”

false conduct. 31 U.S.C. § 3729(a)(1)(A)–(B).

Individuals act “knowingly” if they (1) have “actual knowledge of the [falsity of

the] information”; (2) act “in deliberate ignorance of the truth or falsity of the information”;

or (3) act “in reckless disregard of the truth or falsity of the information.” Id.

§ 3729(b)(1)(A). Thus, though the Act does “not punish honest mistakes or incorrect claims

submitted through mere negligence,” United States ex rel. Owens v. First Kuwaiti Gen.

Trading & Contracting Co., 612 F.3d 724, 728 (4th Cir. 2010) (citation omitted), it does

require “those doing business with the Government . . . to make a limited inquiry to ensure

the claims they submit are accurate,” United States ex rel. Phalp v. Lincare Holdings, Inc.,

857 F.3d 1148, 1155–56 (11th Cir. 2017) (quoting S. Rep. No. 99-345, at 7, 1986

U.S.C.C.A.N. at 5272).

A careful review of the full record here reveals no “honest mistakes,” “negligence,”

or adequate inquiry. In fact, the record shows a deliberate plan to frustrate the requirements

of the Medicaid Rebate Act and bilk the federal Government out of $680 million. Though

the majority opinion dismisses these inconvenient facts—and the record itself—as “simply

32 irrelevant” to its allegedly purely legal inquiry, Majority Op. at 26 n.5, it is worth

describing the facts it skimmed over in detail. With this context in mind, I then turn to the

majority’s ill-fated application of Safeco Insurance Co. of America v. Burr, 551 U.S. 47

(2007), to the fraud context. Finally, I conclude that even if Safeco applied, the majority

erred by finding that Forest wasn’t “warned away” from its stacked-rebate scheme.

When ruling on a Rule 12(b)(6) motion to dismiss, “a judge must accept as true all

of the factual allegations contained in the complaint” and must “draw all reasonable

inferences in favor of the plaintiff.” E.I. du Pont de Nemours & Co. v. Kolon Indus., 637

F.3d 435, 440 (4th Cir. 2011) (citation omitted). The following facts are largely taken from

Sheldon’s amended complaint.

Under the Medicaid Drug Rebate program, drug manufacturers that wish to sell their

drugs to state Medicaid agencies must first enter into rebate agreements with the Secretary

of Health and Human Services. 42 U.S.C. § 1396r-8(a). These agreements require the

manufacturers to provide states with rebates on drugs purchased for Medicaid

beneficiaries. Id. § 1396r-8(b). These rebates are then passed along to the federal

Government by offsetting them against federal Medicaid assistance provided to the states.

Id. § 1396r-8(b)(1)(B).

Calculating these rebates “is a complex enterprise requiring recourse to detailed

information about the company’s sales and pricing.” Astra USA, Inc. v. Santa Clara Cnty.,

563 U.S. 110, 115 (2011). For most drugs, the rebate amount is equal to the greater of two

numbers: (1) the statutory minimum rebate percentage of the “average manufacturer price”

33 (currently 23.1%) and (2) “the difference between the average manufacturer price and the

best price.” 42 U.S.C. § 1396r-8(c)(1)(A), (c)(1)(B)(i)(VI). The “average manufacturer

price” means “the average price paid to the manufacturer for the drug in the United States

by . . . wholesalers . . . [and] retail community pharmacies.” Id. § 1396r-8(k)(1)(A). The

“best price” is “the lowest price available from the manufacturer . . . to any wholesaler,

retailer, provider, health maintenance organization, nonprofit entity, or governmental

entity.” Id. § 1396r-8(c)(1)(C)(i).

Allergan Sales, LLC and its predecessors Forest Laboratories, LLC and Forest

Pharmaceuticals (collectively “Forest”) is a leading pharmaceutical-drug manufacturer. In

2014, Forest’s expected annual revenues topped $15 billion. A significant portion of this

business is supported by drug reimbursements from state Medicaid programs.

From the 1990s until 2014, relator Sheldon worked at Forest. Sheldon served in

several managerial roles and was responsible for billions of dollars in revenue streams.

Sheldon was also directly involved in the sale of Forest’s drugs, including the negotiation

of discounts, rebates, and other incentives. As a result, he had “direct, personal knowledge

of the drug rebates and other discounts given to Forest customers that impact[ed] the

reported Best Price for each drug.” J.A. 63.

In 2005, Sheldon discovered that Forest was failing to account for rebates provided

to two separate customers on the same dispensed drug unit. Specifically, Forest was

providing one rebate to private insurance companies and another to pharmacy providers or

group purchasing organizations (“GPOs”). Because some of the patients treated by these

pharmacies or GPOs were also covered by these private insurers, Sheldon believed that

34 Forest was benefiting from double rebates but illegally reporting only one rebate as the

basis of its “[b]est [p]rice.” J.A. 67.

Shortly after Sheldon’s discovery, the Centers for Medicare & Medicaid Services

(“CMS”) proposed a rule that would codify and clarify the definition of “best price,” among

other things. Medicaid Program; Prescription Drugs, 71 Fed. Reg. 77,174 (proposed Dec.

22, 2006). That proposed rule defined best price as “the lowest price available from the

manufacturer during the rebate period to any entity in the United States in any pricing

structure,” including “all sales and associated discounts and other price concessions

provided by the manufacturer to any entity unless . . . specifically excluded by statute or

regulation.” Id. at 77,197 (emphases added). It further clarified that best price “shall be

[the] net of cash discounts . . . and any other discounts or price reductions and

rebates . . . which reduce the price available from the manufacturer,” and required

manufacturers to “adjust the best price for a rebate period if cumulative discounts, rebates,

or other arrangements subsequently adjust the prices available from the manufacturer.” Id.

at 77,198 (emphases added). In the preamble, CMS noted that “any price adjustment which

ultimately affects those prices which are actually realized by the manufacturer . . . should

be included in the calculation of best price.” Id. at 77,182 (emphasis added).

Forest submitted written comments on the rulemaking, noting that “the proposed

rule suggests that CMS views best price as the net amount realized by the manufacturer on

a sale rather than the lowest price to a particular customer.” J.A. 239 (emphases added).

It urged CMS to clarify that “only discounts and price concessions to the same entity to

which a drug is sold should be included in the computation of best price to that entity.”

35 J.A. 239 (emphasis added). It believed the “statutory definition of best price has always

been interpreted to mean the single lowest price to a particular customer,” and that “prices

to unrelated entities in the chain of distribution should not be aggregated . . . even if they

concern the same unit of a drug.” J.A. 239–40 (emphasis added). Several other drug

manufacturers submitted similar comments.

Nearly a year later, CMS published its final rule. Medicaid Program; Prescription

Drugs, 72 Fed. Reg. 39,142 (July 17, 2007) (codified at 42 C.F.R. pt. 447). CMS declined

to change the offending language identified by Forest or the other drug manufacturers,

reiterating that the “best price represents the lowest price available from the manufacturer

to any entity . . . [and] any price concession associated with that sale should be netted out

of the price received by the manufacturer in calculating best price and best price should be

adjusted by the manufacturer if other arrangements subsequently adjust the prices actually

realized.” Id. at 39,150 (emphases added).

CMS also took the opportunity to clear up confusion regarding a stacked-rebate

situation involving pharmacy benefit managers (“PBMs”). These entities serve as

middlemen between drug manufacturers, pharmacies, health insurance companies, and end

users. Linda L. Ujifusa & J. Mark Ryan, Pharmacy Benefit Managers: The Mystery

Bureaucrats Managing your Prescription Drugs, Uprise RI (Aug. 25, 2021),

https://upriseri.com/pharmacy-benefit-managers/. Originally, CMS proposed including

rebates paid to PBMs when determining best price. 71 Fed. Reg. at 77,182–83. Some

“industry analysts” believed that this proposal obligated manufacturers “to add concessions

paid to PBMs to the concessions paid to customers of the PBMs in calculating best price.”

36 72 Fed. Reg. at 39,198. Multiple commentators objected, arguing that “if Congress had

intended anything other than a customer-by-customer analysis of separate prices, the statute

would have combined each customer with the word ‘and’ instead of the disjunctive ‘or.’”

Id. (emphasis added). In conclusion, they asked that “CMS reaffirm that best price is the

lowest price available from the manufacturers” to a single customer. Id.

In no uncertain terms, CMS replied that “[w]e do not agree with the commenters.”

Id. It noted that although the final rule had largely removed any requirement that rebates

paid to PBMs be included in best price, the rule reiterated that best price must “reflect the

lowest price available from the manufacturer to any purchaser, inclusive of rebates,

discounts, or price concessions that adjust the price realized.” Id. (emphasis added). 1

In response, top-level managers at Forest prepared reports and held a series of

meetings that examined the stacked-rebate issue. As the result of these meetings, Forest

decided to hire a data-audit firm to identify stacked rebates claimed by its commercial

customers—mostly private insurance companies—“for the same dispensed drug units to

the same patient.” J.A. 69. After claims involving double rebates were identified, Forest

1 The majority argues that this example is irrelevant because, “[a]s Sheldon conceded below, this example has nothing to do with whether discounts should be aggregated in calculating Best Price; instead, ‘CMS’s comments involving PBMs simply addressed how rebates to an excluded entity might nevertheless fall within Best Price.’” Majority Op. at 24 (quoting Res. in Opp’n to Def.’s Mot. to Dismiss Am. Compl. at 22, United States ex rel. Sheldon v. Forest Lab’ys, (D. Md. 2020), ECF No. 79). However, Sheldon did not concede anything of the sort, see Res. in Opp’n to Def.’s Mot. to Dismiss Am. Compl., supra, at 4 (arguing that this example showed that “CMS explicitly rejected Forest’s interpretation”), and the majority opinion offers no explanation for CMS’s express repudiation of the commentators’ single-customer approach.

37 paid the first entity that claimed a rebate but refused to pay the second. Forest was able to

do this because its sales contracts at the time—for these customers, at least—included a

“clause providing that Forest would only pay one company when there are two entities

claiming a rebate for the same drug to a single patient.” J.A. 35–36. The purpose of only

allowing a single rebate to be claimed was to ensure that stacked “discounts on the same

pill would [not] have to be added together” when reporting best prices to CMS. J.A. 69.

But Forest took a different tack with its preferred customers: pharmacy providers,

GPOs, and certain private insurance companies. To “avoid negatively impacting its

relationships” with these entities, Forest declined to audit their rebates or add a first-come-

first-serve rebate clause to their sales contracts. Instead, it continued to pay these entities

stacked rebates on the same drug unit “quarter after quarter,” while only reporting one of

those rebates as the basis of its best price. J.A. 70. By Sheldon’s calculation, this led to

Forest underpaying its rebates to state Medicaid programs—and by extension, the federal

Government—by over $680 million between 2005 and 2014.

Although these damning facts strongly suggest that Forest was actually aware it was

submitting false best-price reports, the majority finds said facts “simply irrelevant” due to

the Supreme Court’s decision in Safeco. Majority Op. at 26 n.5. That decision interpreted

the scienter requirement for the Fair Credit Reporting Act. The majority claims that if we

import Safeco’s common-law definition of reckless disregard from the Fair Credit

Reporting Act into the False Claims Act, then any defendant who “bases its actions on an

objectively reasonable interpretation of the relevant statute when it has not been warned

38 away from that interpretation” “cannot act ‘knowingly.’” Id. at 12; see also id. at 11

(“Failure to meet this [objective] recklessness standard preclude[s] a finding of knowledge

as well.” (emphasis added)). In other words, the actual-knowledge and deliberate-

ignorance standards are mere surplusage; a purely legal “threshold” recklessness test is

now the alpha and the omega of False Claims Act scienter. Id. at 14.

But Safeco itself and the Supreme Court’s subsequent decision in Halo Electronics,

Inc. v. Pulse Electronics, Inc., 579 U.S. 93 (2016), counsel against importing Safeco

wholesale into a vastly different statutory context. And even if we did, neither Safeco nor

the majority opinion’s sketchy logic justifies finding that Safeco’s objective-recklessness

test allows us to scrap two of the False Claims Act’s three scienter standards.

Safeco concerned a narrow issue: the proper interpretation of the Fair Credit

Reporting Act’s scienter requirement. Safeco, 551 U.S. at 52. While the Fair Credit

Reporting Act requires “willful[]” violations, it does not further define this term. Id. at 56–

57 (quoting 15 U.S.C. § 1681n(a) (2007)). As a result, the Court looked to the common

law and held that “willfulness” includes both “knowing and reckless disregard of the

law”—but not before exhaustively examining whether “Congress had something different

in mind.” Id. at 59, 69 (emphases added).

To start, the Court pored over the drafting history of the Fair Credit Reporting Act,

finding some support for the notion that “liability was supposed to attach only to knowing

violations,” but dismissing such evidence as “shaky, and certainly no match for the

following clue in the text as finally adopted.” Id. at 58–59. Specifically, the Court noted

39 that the Fair Credit Reporting Act imposed heightened liability for “knowing[]” violations.

Id. at 59. But if “willfully” only meant “knowingly,” then this heightened liability standard

would be both “superfluous and incongruous.” Id. Since the Court’s primary directive was

to “[g]ive effect, if possible, to every clause and word of a statute,” the Fair Credit

Reporting Act’s scienter term had to encompass both knowing and reckless violations. Id.

at 60 (quoting United States v. Menasche, 348 U.S. 528, 538–39 (1955)).

Next, the Court reasoned that since the Fair Credit Reporting Act did not define

recklessness, it made sense to invoke the common law once more—but not before again

assessing whether “Congress had something different in mind.” Id. at 69 (emphasis added).

After concluding it did not, the Court held that “a company subject to [the Fair Credit

Reporting Act] does not act in reckless disregard of [that statute],” id. (emphasis added),

unless it runs an “unjustifiably high risk” of violating the law “that is either known or so

obvious that it should be known,” id. at 68 (quoting Farmer v. Brennan, 511 U.S. 825, 836

(1994)).

Ultimately, the Supreme Court recognized that a scienter term’s “construction is

often dependent on the context in which it appears.” Id. at 57 (quoting Bryan v. United

States, 524 U.S. 184, 191 (1998)). Thus, the Court carefully parsed through the Fair Credit

Reporting Act’s legislative history, considered appropriate statutory context, and adopted

a common-law definition that gave effect to “every clause and word of [the] statute.” Id. at

60 (quoting Menasche, 348 U.S. at 538). In simple terms, the Supreme Court took the time

and effort to truly understand whether any evidence “point[ed] to something different in

40 [the Fair Credit Reporting Act]” that would require a “deviat[ion] from the common law.”

Id. at 58, 69.

The same cannot be said of today’s majority opinion. Its “analysis” of whether it

makes sense to import the Safeco Court’s common-law definition of recklessness into the

False Claims Act spans all of three sentences: “The [False Claims Act] defines ‘knowingly’

as including actual knowledge, deliberate ignorance, and reckless disregard. Safeco

interpreted ‘willfully’ to include both knowledge and recklessness. Given this parallel, we

hold that Safeco’s reasoning applies to the [False Claims Act]’s scienter requirement.”

Majority Op. at 12 (citations omitted). But what the majority opinion passes off as

reasoning is no more than say-so. That should not be sufficient to upend the law of frauds

in our Circuit.

Instead, it is necessary to take the time—as the Safeco Court said we must—to ask

whether “Congress had something different in mind” with the False Claims Act. By doing

so, it becomes evident that we should not import the Fair Credit Reporting Act’s objective

recklessness standard, for a few reasons.

To start, the Fair Credit Reporting Act’s and False Claims Act’s vastly different

contexts make them a poor match for common-law cross-pollination. The Fair Credit

Reporting Act is a primarily prescriptive statute intended “to ensure fair and accurate credit

reporting, promote efficiency in the banking system, and protect consumer privacy.”

Safeco, 551 U.S. at 52. The False Claims Act is an entirely proscriptive statute intended to

prevent fraud. Universal Health Servs. v. United States ex rel. Escobar, 579 U.S. 176, 181–

41 82 (2016). And fraud often revolves around a defendant’s subjective state of mind. See

Restatement (Second) of Torts § 526 cmts. c, e (Am. L. Inst. 1977) (noting scienter for

fraud can be established when a defendant has actual “knowledge of falsity,” “believes the

representation to be false,” or makes a false representation with “careless [disregard] of

whether it is true or false”); see also United States ex rel. Drakeford v. Tuomey, 792 F.3d

364, 384 (4th Cir. 2015) (holding that the “subjective inquiry” of whether a defendant

“knew that its claims were in violation of the [law is] covered under the [False Claims

Act’s] knowledge element” (emphasis added)).

Therefore, it makes little sense to import the Fair Credit Reporting Act’s objective

recklessness test into the False Claims Act—especially when this “threshold” test

effectively becomes a be-all-and-end-all scienter requirement. See Halo, 579 U.S. at 104,

106 n.* (declining to import Safeco’s recklessness test into the patent context because “bad

faith” was relevant in that context and a “threshold [objective recklessness] requirement

excludes from discretionary punishment many of the most culpable offenders”).

The majority opinion’s wholesale adoption of this Fair Credit Reporting Act test

makes even less sense when one considers the sources of common law underlying it. In

Safeco, those sources were the Restatement (Second) of Torts § 500 (Am. L. Inst. 1963–

1964) and the Court’s previous decision in Farmer v. Brennan. But the Restatement

(Second) § 500 pertains not to the common law of fraud, but rather to the common law of

physical safety. See § 500 (stating that conduct “must involve an easily perceptible danger

of death or substantial physical harm” to qualify as reckless). Likewise, Farmer’s “civil-

law recklessness” definition—which drew on § 500’s physical-safety standard—also relied

42 on the common law of physical injury. 511 U.S. at 837; id. at 836 (finding its recklessness

standard equivalent to “deliberate indifference to a substantial risk of serious harm to a

prisoner” (emphasis added)). Both sources, therefore, are inapposite in the fraud context.

In fact, the Restatement (Second) has another section that deals specifically with the

scienter requirement for common-law fraud. See Restatement (Second) of Torts § 526. And

though this directly relevant body of common law surely has bearing on the meaning of

reckless disregard in fraud, the majority opinion ignores it.

The majority opinion counters that “every other circuit to consider the issue” has

“h[e]ld that Safeco applies with equal force to the [False Claims Act]’s scienter

requirement.” Majority Op. at 10. Not so. In United States ex rel. Phalp v. Lincare

Holdings, Inc., the Eleventh Circuit received extensive briefing on the recklessness

standard recognized in Safeco and declined to import it into the False Claims Act. 2 See 857

F.3d at 1155 (rejecting the conclusion—recognized in Safeco—“that a finding of scienter

can be precluded by a defendant’s identification of a reasonable interpretation of an

ambiguous regulation”).

To be sure, other courts have gone the other way, but most of these cases are either

unpublished or easily distinguishable. See United States ex rel. Streck v. Allergan, Inc., 746

F. App’x 101, 106 (3d Cir. 2018) (unpublished); United States ex rel. McGrath v.

Microsemi Corp., 690 F. App’x 551, 552 (9th Cir. 2017) (unpublished); United States ex

2 Though the Phalp opinion did not explicitly cite to Safeco, it squarely rejected the very holding the majority claims is commanded by Safeco.

43 rel. Donegan v. Anesthesia Assocs. of Kansas City, PC, 833 F.3d 874, 879–80 (8th Cir.

2016) (citing Safeco only to explain that the plaintiff had not created a material issue of

fact regarding whether the defendant was warned away from its reasonable interpretation);

United States ex rel. Purcell v. MWI Corp., 807 F.3d 281, 290–91 (D.C. Cir. 2015) (citing

Safeco in holding that a reasonable interpretation of a contract precluded False Claims Act

liability). And all but one either predates or fails to distinguish the Supreme Court’s

decision in Halo. But see United States ex rel. Schutte v. SuperValu Inc., 9 F.4th 455, 467

(7th Cir. 2021) (foreshadowing the majority opinion’s flawed attempt to distinguish Halo).

Because Halo explicitly declined to import Safeco’s objective recklessness test into

an analogous context, it deserves further explanation. In Halo, the Supreme Court

interpreted the scienter requirement for enhanced damages under § 284 of the Patent Act.

579 U.S. at 97. Though the Patent Act does not include a specific scienter standard for

these damages, for “nearly two centuries” the Supreme Court and the courts of appeal had

“[c]onsistent[ly]” interpreted the statute to require “willful misconduct.” Id. at 106. But in

2007, the Federal Circuit created a test for “willful” infringement that wholly relied on

Safeco’s definition of objective recklessness. See In re Seagate Tech., LLC, 497 F.3d 1360,

1371 (Fed. Cir. 2007), abrogated by Halo, 579 U.S. 93 (2016). Like the standard crafted

by the majority opinion, the Federal Circuit’s objective-recklessness test was a “threshold

requirement” for liability. Halo, 579 U.S. at 104; see also id. (“Under Seagate, a district

court may not even consider enhanced damages for [a willful] pirate, unless the court first

determines that his infringement was ‘objectively’ reckless.”); see Majority Op. at 14.

44 The Halo Court squarely rejected the Federal Circuit’s Safeco test. Though the Fair

Credit Reporting Act and § 284 share the same scienter requirement—willfulness—the

Halo Court noted that “‘willfully’ is a word of many meanings whose construction is often

dependent on the context in which it appears.” Halo, 579 U.S. at 106 n.* (quoting Safeco,

551 U.S. at 57). And because the “subjective willfulness of a patent infringer, intentional

or knowing, may warrant enhanced damages, without regard to whether his infringement

was objectively reckless,” the Federal Circuit erred by crafting a threshold objective test

for § 284. Id. at 105 (emphasis added); see also id. at 106 n.* (rejecting the respondents’

argument that Safeco’s footnote required the Court to find that “bad faith was not relevant

absent a showing of objective recklessness” because “‘bad-faith infringement’ is an

independent basis for enhancing patent damages”). Safeco’s common-law definition of

“willfulness” simply did not apply. Id. at 104–106; cf. Farmer, 511 U.S. at 840 (declining

to adopt an objective-recklessness test for Eighth Amendment violations based on textual

and contextual clues).

It’s hard to see much daylight between Halo and the present case. Both address the

use of a “threshold” Safeco test that precludes inquiry into “deliberate wrongdoing.” Halo,

579 U.S. at 104. Both concern the application of said test to statutes that revolve around

the “subjective willfulness” or subjective knowledge of the statutory violator—unlike the

Fair Credit Reporting Act—and punish transgressions with up to treble damages. Id. at

105, 109. And in both cases, as explained in more detail below, an unthinking application

of Safeco’s test would “mak[e] dispositive the ability of the [statutory violator] to muster

a reasonable (even though unsuccessful) defense at . . . trial.” Id. at 105. Because of these

45 serious contextual concerns, the Halo Court declined to import Safeco’s test into § 284.

See id. at 105–07. We should too.

The majority opinion struggles to explain why Halo should not control. In the end,

it lands on two weak distinctions between the False Claims Act and § 284: (1) “§ 284 d[oes]

not include a scienter requirement, while the FCA clearly limits liability to claims that are

made ‘knowingly,’” and (2) “while § 284 concerned whether district courts could issue a

particular amount of damages after finding liability, the relevant provision here concerns

whether liability exists at all.” Majority Op. at 13. Neither distinction holds any water.

To start, while § 284 might not include an explicit scienter requirement, for almost

two centuries courts have interpreted the Patent Act to require “willful” violations for

enhanced damages. Halo, 579 U.S. at 106. When Congress enacted § 284 in 1952, it

legislated “against this backdrop.” Id. at 100. Thus, whether the courts, as ratified by

Congress, or Congress itself created § 284’s “willful” standard, its standard remains the

same as the Fair Credit Reporting Act’s. If anything, § 284 is an even closer analog to the

Fair Credit Reporting Act than the False Claims Act; while § 284 and the Fair Credit

Reporting Act have the exact same scienter standard—willfulness—the False Claims Act

requires only knowing violations. See 31 U.S.C. § 3729(a)(1)(A)–(B). Instead of

acknowledging this potentially critical difference, the majority opinion simply ignores it.

See Majority Op. at 12 (noting simply that the Fair Credit Reporting Act and False Claims

Act have “parallel” scienter requirements).

The majority opinion’s second distinction is even weaker. While it attempts to draw

a hard line between scienter terms for “damages after [a] finding [of] liability” and those

46 for “liability” alone, it does not, and perhaps cannot, explain why this distinction is

important. Id. at 13 (simply noting that these “differences” create a “gap” between the False

Claims Act and the Patent Act). In fact, neither statute suggests that this difference is

meaningful at all: a patent infringer is only “liab[le] for enhanced damages” if they acted

willfully, Halo, 579 U.S. at 104, just as a fraudster is only “liable” for treble damages if

they acted knowingly, 31 U.S.C. § 3729(a).

Nonetheless, the majority opinion doubles down, arguing that the False Claims Act

and the Fair Credit Reporting Act are analogs because both “speak[] to liability rather than

damages.” Majority Op. at 14. But even if this was a relevant point of analysis, it simply

isn’t so. The relevant section of the Fair Credit Reporting Act plainly states that “[a]ny

person who willfully fails to comply” with the statute “with respect to any consumer is

liable to that consumer . . . [for] any actual damages[;] . . . punitive damages as the court

may allow; and . . . reasonable attorney’s fees as determined by the court.” 15 U.S.C. §

1681n(a) (emphases added). So if the discussion of § 284 in Halo is irrelevant for our

purposes in understanding the False Claims Act because § 284 “concern[s] whether district

courts [can] issue a particular amount of damages after finding liability” whereas the False

Claims Act “concerns whether liability exists at all,” Majority Op. at 13, then the statute

upon which the majority hangs its hat—the Fair Credit Reporting Act, as understood in

Safeco—is irrelevant for precisely the same reason. In other words, the very statute the

majority opinion claims to be analogizing to elides the very distinction it attempts to make.

At the end of the day, though the majority opinion ironically spends more time

distinguishing the False Claims Act from § 284 than analogizing the False Claims Act to

47 the Fair Credit Reporting Act, its facile analysis still fails. Nor does it provide any answer

for the most troubling concern identified by the Halo Court—that adopting Safeco’s

objective recklessness test makes “deliberate wrongdoing” completely irrelevant, despite

Congress’s clear intention to impose liability in such circumstances. Halo, 579 U.S. at 104

(“In the context of such deliberate wrongdoing . . . it is not clear why an independent

showing of objective recklessness . . . should be a prerequisite” to recovery.).

3.

It would seem to be enough to point out that the majority treads on thin ice by

copying and pasting mismatched common law into the False Claims Act. But instead of

retreating after hearing the cracking beneath its feet, it takes yet another step and plunges

into the depths below.

That next step occurs when the majority opinion holds that if we adopt Safeco’s

objective-recklessness test for False Claims Act allegations, then a “[f]ailure to meet this

recklessness standard preclude[s] a finding of knowledge as well.” Majority Op. at 11

(emphasis added). The majority opinion claims this result is commanded by Safeco and

logic. Failing that, it makes undisguised appeals to notions of public policy. Neither

argument withstands even the slightest scrutiny.

The majority opinion’s Safeco argument can be traced to a single footnote at the

very end of that opinion. That footnote proclaims that “[w]here, as here, the statutory text

and relevant court and agency guidance allow for more than one reasonable interpretation,

it would defy history and current thinking to treat a defendant who merely adopts one such

48 interpretation as a knowing or reckless violator.” Safeco, 551 U.S. at 70 n.20 (emphasis

added). According to the majority opinion, this single footnote gives it permission to strike

the “actual knowledge” and “deliberate ignorance” standards from the text of the False

Claims Act, at least with regard to “legally false claims.” Majority Op. at 14–15.

But nothing suggests that the Supreme Court intended to upend the law of frauds in

a terse footnote in an opinion on credit-reporting requirements. In fact, the Court

clarified—in the very same footnote—that it was focused on whether “subjective bad faith

must be taken into account in determining whether a company acted knowingly or

recklessly for purposes of § 1681n(a)” of the Fair Credit Reporting Act. Safeco, 551 U.S.

at 70 n.20 (emphasis added). So, the Court’s seemingly broad references to “a defendant,”

“knowing or reckless violator[s],” and “subjective bad faith,” see id., are limited to the Fair

Credit Reporting Act context—as the Court itself plainly noted in Halo, 579 U.S. at 106

n.* (rejecting an analogy to Safeco’s footnote because a “showing of bad faith was not

relevant absent a showing of objective recklessness” under the Fair Credit Reporting Act,

while “‘bad-faith infringement’ is an independent basis for enhancing patent damages”).

Even if we ignored this critical context—which we should not—Safeco’s

conclusory conflation of knowing and reckless violations would be dictum. Safeco did not

involve any knowing violation of the Fair Credit Reporting Act; the plaintiffs’ entire action

rested on allegedly reckless failures. Safeco, 551 U.S. at 52–58. Therefore, the Supreme

Court’s discursion on “knowing” violations is a classic example of a “peripheral” statement

that “may not have received the full and careful consideration of the court that uttered it”

and “that could have been deleted without seriously impairing the analytical foundations

49 of the holding.” Payne v. Taslimi, 998 F.3d 648, 654–55 (4th Cir. 2021) (quoting Pittston

Co. v. United States, 199 F.3d 694, 703 (4th Cir. 1999)). And while we give “great weight

to Supreme Court dicta,” NLRB v. Bluefield Hosp. Co., 821 F.3d 534, 541 n.6 (4th Cir.

2016), dicta “cannot serve as a source of binding authority in American jurisprudence,”

United States v. Pasquantino, 336 F.3d 321, 329 (4th Cir. 2003) (en banc), aff’d, 544 U.S.

349 (2005).

Undeterred, the majority opinion insists that even if Safeco’s footnote is not

controlling, when “a defendant has not acted with reckless disregard in its view of the

statute, ‘it follows a fortiori’ that it has not acted with deliberate ignorance or actual

knowledge, which ‘plainly demand[] even more culpability.’” Majority Op. at 14 (quoting

Urquilla-Diaz v. Kaplan Univ., 780 F.3d 1039, 1058 n.15 (11th Cir. 2015)).

As support, it offers a syllogism with a major premise stating that reckless disregard

is the “most capacious,” “loosest,” or “baseline” scienter standard, and a deeply flawed

minor premise stating that actual knowledge and deliberate ignorance necessarily fall

within the “capacious” reckless-disregard standard. Id. at 14 (citations omitted). That minor

premise is foreclosed by Safeco itself, which said that “action falling within the knowing

subcategory does not simultaneously fall within the reckless alternative.” Safeco, 551 U.S.

at 60 (emphasis added); see also Halo, 579 U.S. at 105 (“The subjective willfulness of a

patent infringer, intentional or knowing, may warrant enhanced damages, without regard

to whether his infringement was objectively reckless.” (emphasis added)).

And there are even stronger reasons to reject the majority opinion’s overall result.

At the outset, it is a “cardinal rule of statutory construction that we are ‘obliged to give

50 effect, if possible, to every word Congress used.’” Taylor v. Grubbs, 930 F.3d 611, 617

(4th Cir. 2019) (quoting Nat’l Ass’n of Mfrs. v. Dep’t of Def., 138 S. Ct. 617, 632 (2018));

see also Safeco, 551 U.S. at 60 (recognizing its obligation to “[g]ive effect, if possible, to

every clause and word of a statute” (citation omitted)). But the majority opinion’s test

creates a “threshold requirement” that renders the statutory text’s “actual knowledge” and

“deliberate ignorance” standards totally superfluous. Majority Op. at 14. Taking the

majority opinion at its word: the objective-recklessness standard is a threshold inquiry.

That means that if one can satisfy the majority’s objective-recklessness standard, there is

no need to assess actual knowledge or deliberate ignorance, since liability has already been

established. If one cannot satisfy the majority’s objective-recklessness standard, then we

are precluded from assessing these other scienter standards at all. Id. at 12. There is no

escaping this result. Yet, the majority opinion claims that “applying Safeco does not sap

the FCA’s three scienter definitions of independent meaning.” Id. at 14. But claiming it to

be so does not make it so.

That’s because reading two of the three scienter standards out of the statute is not

only inconsistent with a cardinal rule of statutory construction but also inconsistent with

Safeco itself. That decision teaches us that “a common law term in a statute comes with a

common law meaning” unless “Congress had something different in mind.” Safeco, 551

U.S. at 58, 69 (emphasis added). The fact that Congress crafted three distinct scienter

standards—not one threshold objective-recklessness test—compels the conclusion that it

did have something different in mind.

51 In case there was any doubt about this, the drafting history of the False Claims

Reform Act confirms it. Over thirty years ago, Congress grew concerned that overly

“restrictive court interpretations” of the False Claims Act were “thwart[ing] the

effectiveness of the statute.” S. Rep. No. 99-345, at 4, 1986 U.S.C.C.A.N. at 5269. In

particular, “inappropriate” narrowing of the Act’s scienter requirement was hamstringing

the Government’s ability to fight “rampant fraud.” Id. at 7, 13, 1986 U.S.C.C.A.N. at 5272,

5278. To remedy this problem, Congress crafted three distinct and expansive scienter

standards and eliminated any requirement to show bad faith. 31 U.S.C. § 3729(b)(1). The

clear intent of these amendments was to adopt a broad, “remedial” scienter standard that

would allow the Government to “hold responsible those corporate officers who insulate

themselves from knowledge of false claims submitted by lower-level subordinates.” S.

Rep. No. 99-345, at 7, 1986 U.S.C.C.A.N. at 5272. In other words, Congress was trying to

capture more fraud, not less.

Yet rather than turning to this history, the majority opinion instead repeats the

mistakes made by courts before Congress amended the False Claims Act in 1986 by

adopting its own overly “restrictive” interpretation of the Act. Id. at 4, 1986 U.S.C.C.A.N.

at 5269. Thusly, it reads two of the three scienter standards right out of existence: the

actual-knowledge and deliberate-ignorance standards that concern “deliberate

wrongdoing.” Halo, 579 U.S. at 104 (emphasis added). By striking these two standards

from the statute, the majority effectively “insulat[es] some of the worst [scammers] from

any liability” whatsoever. Id. (emphasis added). The majority opinion’s new law thereby

frustrates the clear intent of Congress—as evidenced by both the text and legislative

52 history—to expand False Claims Act liability to cover situations precisely like that alleged

by Sheldon today.

Perhaps sensing the weight of authority against it, the majority opinion claims that

its redlined version of the Act will “not apply to all [False Claims Act] suits.” Majority Op.

at 14. Rather, it contends the opinion “is narrowly cabined to legally false claims—like the

one here—which involve contested statutory and regulatory requirements.” Id. at 15

(emphases added). But this is not a minor universe of cases. It might take a lifetime just to

list all of the contested statutory and regulatory requirements out there. Even if we only

consider what requirements might conceivably be contested for a single program like

Medicaid, the mind reels. After all, as the majority itself acknowledges, “Medicaid statutes

and regulations ‘are among the most completely impenetrable texts within human

experience,’” id. at 20 (quoting Rehab. Ass’n of Va., Inc. v. Kozlowski, 42 F.3d 1444, 1450

(4th Cir. 1994)), involving “complex” and “labyrinthine reporting requirements” that

“raise[] some of the thorniest issues in government price reporting,” id. at 16, 20. If this is

true, then what qualifies as a contested Medicaid requirement is only limited by the

“ingenuity” of defense attorneys. Halo, 579 U.S. at 105.

In other words, the majority opinion’s “narrow[]” holding is actually as broad as

defendants want it to be. Majority Op. at 15. So long as a legal fraudster can “muster a

reasonable (even though unsuccessful) defense” at trial—which should not be much of a

lift, especially for complex programs like Medicaid—they can “escape any comeuppance.”

Halo, 579 U.S. at 105. This creates a truly perverse incentive; the more that defendants

53 steal via fraud, the easier it is for them to hire high-priced attorneys who can dream up

reasonable explanations to justify said fraud after the fact.

Post hoc rationalizations like these are only possible because under the majority

opinion’s test, a defendant does not need to have “act[ed] on the basis of the defense” or

“even [be] aware of it” at the time the fraud was committed. Id. They just need to advance

an “objectively reasonable” interpretation that “ha[s] a foundation in the statutory text,”

even if that reading is ultimately “erroneous.” Majority Op. at 11 (quoting Safeco, 551 U.S.

at 69–70). Whether the defendant was actually operating under this interpretation when it

committed the alleged fraud is both unnecessary and impossible to discern under the

majority’s test because any “inquiry into a defendant’s subjective intent” or “subjective

beliefs” is completely precluded. Id. at 12, 26 n.5. Forbidding such an inquiry, however,

violates another cardinal principle: that “culpability is generally measured against the

knowledge of the actor at the time of the challenged conduct.” Halo, 579 U.S. at 105. It

also allows the “most culpable offenders”—those who commit fraud with actual

knowledge and “without any reason to suppose [their] conduct is arguably defensible”—

to craft their own get-out-of-jail-free cards whenever they like. Id. at 104–105.

The majority opinion counters that any concerns about deliberate fraudsters

escaping liability are blunted by Safeco’s second step. That step asks “whether authoritative

guidance might have warned [the] defendant away from [their objectively reasonable]

reading.” Majority Op. at 11. According to the majority opinion, a defendant cannot truly

know that they are filing a false claim until they obtain authoritative guidance from either

the courts of appeal or the relevant agency that “clarifies [their] interpretation of the law

54 and so warns defendants away from otherwise reasonable interpretations.” Id. at 16. Before

this point, a “defendant might suspect, believe, or intend to file a false claim, but it cannot

know that its claim is false.” Id. at 15 (quoting Schutte, 9 F.4th at 468).

That borders on the nonsensical. It is self-aggrandizing to suppose that the biggest

pharmaceutical companies on the planet, with some of the highest-paid experts in health

care law, are incapable of reading a statute or regulation and “knowing” they are breaking

the law until a court or CMS spells it out for them. And even if a lack of authoritative

guidance precludes “actual knowledge”—which it shouldn’t—it certainly could not

preclude a finding of “deliberate ignorance.” 31 U.S.C. § 3729(b)(1)(A) (emphases

added). After all, this standard is intended to reach the “‘ostrich’ type situation where an

individual has ‘buried his head in the sand’ and failed to make simple inquiries which

would alert him that false claims are being submitted.” S. Rep. No. 99-345, at 21, 1986

U.S.C.C.A.N. at 5286. In other words, the False Claims Act’s deliberate-ignorance

standard is designed to capture the very conduct the majority says cannot be captured under

Safeco’s second step: situations where a defendant “suspect[s]” or “believe[s]” they are

committing fraud but avoids making inquiries that would confirm their suspicions.

Majority Op. at 15 (quoting Schutte, 9 F.4th at 468).

Applying Safeco’s second step here also leads to absurd results. For example, under

the majority opinion’s test, a defendant could know they are committing fraud, be told by

a court that they are doing so, and nevertheless escape liability because (1) they advance a

post hoc explanation that, while wrong, is still “reasonable,” and (2) neither the

Government nor the court had said anything “authoritative” at the time of the fraud.

55 It also has the effect of basically freezing judicial interpretation of the statute at

issue. Cf. Camreta v. Greene, 563 U.S. 692, 706 (2011) (noting that the doctrine of

qualified immunity, as applied to claims under 42 U.S.C. § 1983, “may frustrate the

development of constitutional precedent” because courts need not reach the merits of the

constitutional claim where qualified immunity applies (internal quotation marks omitted)).

As noted above, under Safeco’s first step, a defendant need only advance an objectively

reasonable statutory interpretation. When analyzing this claim, a court does not have to

decide what the statute actually says; it only has to determine if the defendant’s reading is

“reasonable.” This is precisely what happened below, and precisely what the majority does

today. Majority Op. at 22 (“[W]e hold that Forest has offered, at minimum, an objectively

reasonable reading of the Rebate Statute.”). The problem is that by doing so, the court

necessarily forgoes the opportunity to provide “authoritative guidance,” which is needed

at Safeco’s second step to warn the defendant away from their fraudulent scheme. With

judicial interpretation stalled, the defendant is free to continue committing knowing fraud

as long as they desire unless CMS steps in with new guidance.

And even that might not be enough. For example, though CMS issued new guidance

in 2007 that clearly warned Forest away from most of its rebate stacking—as evidenced by

its high-level meetings, data scrubbing, sales contracts, and its “first come, first served”

56 rebate policy—the majority opinion decides, as a matter of law and without considering

these facts, that this warning-away could not possibly have occurred. 3 Id. at 23.

Its legal arguments exhausted, the majority opinion next turns to naked

considerations of public policy. It accuses CMS—without any basis in the record—of

deliberately “maintain[ing] strategic ambiguity” in its Medicaid regulations “in order to

expand potential liability for regulated entities.” Majority Op. at 23, 27; see also id. at 27

(“Clear regulations constrain regulatory power and limit future flexibility, which is why an

agency might find them undesirable.”). In other words, the majority opinion baldly accuses

the executive branch of regulating in bad faith in order to saddle innocent companies with

“potentially ruinous liability.” 4 Id. at 27. Incredibly, the majority opinion then doubles

down, alleging that CMS is simply “mak[ing] up the rules as [it] go[es]” along, id. (quoting

Bill Watterson, The Calvin & Hobbes Tenth Anniversary Book 129 (1995)), and trying “to

take advantage of companies like [Forest]” “through ambush,” id. at 27–28.

3 In 2016, CMS issued a new rulemaking stating that “[i]f a manufacturer offers multiple price concessions to two entities for the same drug transaction . . . all discounts related to that transaction which adjust the price available from the manufacturer should be considered in the manufacturer’s final price of that drug when determining the best price to be reported for the drug.” Medicaid Program; Covered Outpatient Drugs, 81 Fed. Reg. 5170, 5253 (Feb. 1, 2016) (codified at 42 C.F.R. pt. 447). CMS believed this understanding was consistent with the regulation promulgated in 2007. Id. But the majority opinion fails to even mention this rulemaking. 4 This is likely an overstatement. As explained above, Forest’s annual revenues top $15 billion per year. Therefore, the majority opinion’s teeth-gnashing over the “potentially ruinous liability” for pharmaceutical companies like Forest is sorely misplaced. Majority Op. at 27.

57 Finally, the majority opinion circles back to the False Claims Act, finding it

“profoundly troubling” that the Act could be used to impose “massive liability on

individuals or companies without any proper notice as to what is required.” Id. at 17. The

majority opinion then states that since the Act imposes “damages that are essentially

punitive in nature,” a lack of appropriate notice means that “defendants are not likely to

receive due process.” Id. at 16–17 (citations omitted). However, it says adopting Safeco

allows us to “avoid[] this trouble” because it forces courts to “strict[ly] enforce[]” the False

Claims Act’s “rigorous” scienter requirement. Id. at 17 (quoting Escobar, 579 U.S. at 192).

Having set up this artificial construct, the majority concludes that Safeco’s standard

provides “just the right means to further [the majority’s] end”: preventing the ever-

expanding “administrative state” from “tak[ing] advantage of companies” like Forest. Id.

at 17, 27.

But “[t]he seriousness of [the majority opinion’s] policy concerns cannot justify

imposing an artificial construct such as the [Safeco] test on the” False Claims Act. Halo,

579 U.S. at 109. This is especially true when imposing such a construct obviates the clear

commands of Congress. Ironically, while it is the majority opinion that accuses CMS of

“mak[ing] up the rules as [it] go[es]” along, it is the majority opinion that ends up playing

its own version of “Calvinball” by using Safeco to shred two of the Act’s scienter standards.

Majority Op. at 27 (quoting Watterson, supra, at 129). The majority opinion claims that

this outcome is justified by the Supreme Court’s command to “strict[ly] enforce[]” the

Act’s “rigorous” scienter requirement. Id. at 17 (quoting Escobar, 579 U.S. at 192). But

there is a big difference between strictly enforcing all three scienter standards created by

58 Congress and deleting two of them altogether. And because nothing even suggests that the

False Claims Act, as currently written, violates due process, we must give effect to all three

standards—not rewrite them based on our own notions of a better public policy.

C.

For the reasons explained above, Safeco should not be imported into the False

Claims Act. But even if it is, Sheldon has plausibly alleged a claim under the Safeco

framework. Under Safeco’s first step, we assess whether Forest’s reading of the Rebate

Statute is objectively reasonable. While I agree that its reading would be reasonable if we

were interpreting on a blank slate, we aren’t. Even if Forest survives Safeco’s first step, the

majority errs by finding—at Safeco’s second step—that Forest was not warned away from

its fraudulent scheme.

Though the majority opinion barely mentions it, our interpretation of the Rebate

Statute is governed by the familiar framework articulated in Chevron U.S.A. v. Natural

Resources Defense Council, Inc., 467 U.S. 837 (1984). Under Chevron, courts first

examine “whether Congress has directly spoken to the precise question at issue.” Id. at 842.

If it has, “that is the end of the matter.” Id. But “[i]f the statute is ambiguous, courts then

‘move to Chevron’s second step and defer to the agency’s interpretation so long as it is

based on a permissible construction of the statute.’” Sierra Club v. U.S. Army Corps of

Eng’rs, 909 F.3d 635, 643 (4th Cir. 2018) (cleaned up) (quoting King v. Burwell, 759 F.3d

358, 367 (4th Cir. 2014), aff’d, 576 U.S. 473 (2015)).

59 i.

The Rebate Statute’s definition of “best price” is certainly ambiguous. Best price

means “the lowest price available from the manufacturer . . . to any wholesaler, retailer,

provider, health maintenance organization, nonprofit entity, or governmental

entity . . . inclusive of cash discounts, free goods that are contingent on any purchase

requirement, volume discounts, and rebates.” 42 U.S.C. § 1396r-8(c)(1)(C) (emphases

added). In general, Congress’s “use of the word ‘any’ suggests an intent to use that term

expansive[ly].” Smith v. Berryhill, 139 S. Ct. 1765, 1774 (2019) (quoting Ali v. Fed.

Bureau of Prisons, 552 U.S. 214, 218–19 (2008)). “Any” can mean “one, some, or all,”

depending on context. Any, Merriam Webster Dictionary, https://www.merriam-

webster.com/dictionary/any (last visited Dec. 19, 2021). And the context here is the

statute’s broadly remedial purpose: ensuring that Medicaid programs pay the same rate as

private entities for prescription drugs. H.R. Rep. No. 101-881, at 96 (1990), reprinted in

1990 U.S.C.C.A.N. 2017, 2108. Therefore, it seems reasonable to read “any” to refer to

one or more of the entities listed—especially since Congress did not say “any single” or

“any particular” entity, for example. After all, if spreading rebates for the same drug unit

around to different entities in the supply chain was not captured in the “best price,” it would

not make much sense to call it that.

As the majority opinion notes, two context clues suggest that “any” here means

“one” and not “some” or “all.” Majority Op. at 18–19. But neither can bear the weight the

majority opinion would place upon them in its bid to render the text unambiguous. First,

each of the entities in the statute is listed in the singular form. And “when ‘any’ is used in

60 context of the singular noun,” it ordinarily refers to a “single” item. United States v.

Dunford, 148 F.3d 385, 389–90 (4th Cir. 1998) (nonetheless rejecting this reading). But

neither Forest nor the majority opinion account for the Dictionary Act—“which supplie[s]

rules of construction for all legislation,” Ngiraingas v. Sanchez, 495 U.S. 182, 190 (1990)

(citation omitted)—which says that “words importing the singular include and apply to

several persons, parties, or things.” 1 U.S.C. § 1. Second, the statute includes the

disjunctive “or,” which also suggests that each entity must be considered apart from the

other. But this is not determinative. “Unsurprisingly, statutory context can overcome the

ordinary, disjunctive meaning of ‘or.’” Encino Motorcars, LLC v. Navarro, 138 S. Ct.

1134, 1141 (2018); see also Confederated Tribes & Bands of Yakama Nation v. Yakima

Cnty., 963 F.3d 982, 990 (9th Cir. 2020) (“[C]ourts are often compelled to construe ‘or’ as

meaning ‘and,’ and again ‘and’ as meaning ‘or.’” (quoting United States v. Fisk, 70 U.S.

445, 447 (1865))). And again, the context here is Congress’s broad intent to stop “pay[ing]

overly inflated prices for prescription drugs.” 136 Cong. Rec. S12,954 (daily ed. Sept. 12,

1990) (statement of Sen. David Pryor).

The majority opinion makes several other arguments, but none clear up the issue.

To start, it provides a few simplistic examples using baseballs and apples to suggest

“aggregating discounts to multiple entities” cannot be required by the Rebate Statute.

Majority Op. at 19. But by their very nature, these everyday examples lack the critical

legislative context animating the best-price provision. They do not, for example, assume

that the “thrifty Kansas City Royals” or your “friend” have been repeatedly swindled and

forced to pay exorbitant amounts for the same goods purchased by everyone else at a much

61 lower price. Id. Nor do they account for the overlapping nature of the supply chain for drug

manufacturing and delivery.

Next, the majority opinion suggests that since the statute says “the lowest price

available from the manufacturer” and “‘available’ means ‘suitable or ready for use,’” the

statute must be “talking about an actual price, not something that is purely hypothetical.”

Id. at 19 (emphasis added). But the majority opinion itself recognizes that “available” is a

more elastic word than this argument suggests. For example, it notes that “wholesaler

chargeback agreements”—discounts that the wholesaler delivers to its customers and later

“charges back” to the manufacturer—can be included in best price, even though they are

not “at hand” or immediately “available” from the manufacturer and in fact operate as a

“lagged price concession.” Id. at 19, 24–25.

The majority opinion also points out differences between the definitions of best

price and “[a]verage [m]anufacturer [p]rice.” Id. at 19–20. Specifically, the former refers

to “the lowest price available from the manufacturer” while the latter refers to the “the

average price paid to the manufacturer.” 42 U.S.C. § 1396r-8(c)(1)(C)(i), (k)(1)(A)

(emphases added). The majority opinion vaguely notes that “something ‘paid to the

manufacturer’ might incorporate discounts to different entities” but fails to explain why

this is true, or how “paid” and “from” create a meaningful “distinction[] in the statutory

scheme.” Majority Op. at 20.

It also seems odd to interpret these standards in dramatically different ways since

the difference between the two is what determines the manufacturer’s rebate payment. See

42 U.S.C. § 1396r-8(c)(1)(A)(ii). Mathematically, it usually only makes sense to subtract

62 like terms from each other. Addition and Subtraction of Algebraic Expressions, Cuemath,

https://www.cuemath.com/algebra/addition-and-subtraction-of-algebraic-expressions/

(last visited Dec. 19, 2021) (“Unlike terms cannot be combined by adding or subtracting.”).

But if average manufacturer price could incorporate stacked rebates but best price could

not, then drug manufacturers would be stuck subtracting apples from oranges. It also would

lead to bizarre results: normally, we would expect the best price to be lower than the

average price. But if average price could include rebates from multiple entities but best

price cannot, the difference between the two would diminish or even disappear. Such a

result would be out of step with Congress’s intent, which was to “achieve significant

Medicaid savings” by getting the “same discounts” that private entities enjoy. H.R. Rep.

No. 101-881, at 96, 98, 1990 U.S.C.C.A.N. at 2108, 2110.

The majority opinion counters that aggregating prices to different entities is

difficult, so it makes sense to read the statute to not require manufacturers to do so. Majority

Op. at 21. But that’s not a canon of construction—whether compliance with the law is

taxing has no bearing on what the law itself requires.

In sum, the Rebate Statute is ambiguous which means Chevron’s second step is

implicated.

Addressing Chevron’s second step, it is worth pointing out from the outset that no

one debates that CMS has the authority to make rules interpreting the Rebate Statute with

the “force of law.” See United States v. Mead Corp., 533 U.S. 218, 226–27 (2001) (limiting

Chevron deference to interpretations made by agencies acting with the “force of law”

63 pursuant to that authority). The real question is whether there is any reasonable agency

interpretation to defer to in the first place. See Fogo De Chao (Holdings) Inc. v. U.S. Dep’t

of Homeland Sec., 769 F.3d 1127, 1135 (D.C. Cir. 2014) (“[W]here ‘the underlying

regulation does little more than restate the terms of the statute itself[,]’ the agency has left

the statute as it found it, adding nothing material to Congress’s language and providing

nothing of its own in which to ground an interpretation to which a court might defer.”

(quoting Gonzales v. Oregon, 546 U.S. 243, 257 (2006))). The majority opinion finds that

CMS’s regulations “simply mirror the statutory language,” so no deference is appropriate.

Majority Op. at 21. Not so.

CMS has issued three distinct notice-and-comment rulemakings on best price. 5 In

1991, CMS promulgated the Rebate Agreement, which copied the statutory language on

“best price” but added that the “best price for a quarter shall be adjusted by the

Manufacturer if cumulative discounts, rebates or other arrangements subsequently adjust

the prices actually realized.” Medicaid Program; Drug Rebate Agreement, 56 Fed. Reg.

7049, 7050 (Feb. 21, 1991). In 2007, CMS promulgated a regulation defining “best price”

as “the lowest price available from the manufacturer during the rebate period to any entity

in the United States in any pricing structure.” 72 Fed. Reg. at 39,242 (emphases added). It

further clarified that best price “shall be [the] net of cash discounts . . . which reduce the

price available from the manufacturer,” and required manufacturers to “adjust the best price

5 “When an agency’s interpretation derives from notice-and-comment rulemaking, it will almost inevitably receive Chevron deference.” Sierra Club, 909 F.3d at 644 (citation and internal quotation marks omitted).

64 for a rebate period if cumulative discounts, rebates, or other arrangements subsequently

adjust the prices available from the manufacturer.” Id. at 39,242–43 (emphases added).

Finally, in 2016, CMS promulgated another regulation that best price must include “all

prices, including applicable discounts, rebates, or other transactions that adjust prices either

directly or indirectly to the best price-eligible entities” listed in the statutory definition. 81

Fed. Reg. at 5351 (emphases added). 6

These rulemakings’ broad references to “any entity,” “any pricing structure,” “net”

cash discounts, “prices actually realized,” and “other arrangements subsequently

adjust[ing] prices” strongly suggest that CMS is focused on the “net” result—the price the

manufacturer actually realizes for the sale of a single drug unit. In fact, it is this “net” result

language that prompted Forest and other pharmaceutical companies to suggest that “CMS

views best price as the net amount realized by the manufacturer on a sale rather than the

lowest price to a particular customer.” J.A. 239 (emphases added). I agree with the drug

companies that this is precisely what CMS intended. I also find that this interpretation is

reasonable for the reasons explained above, as well as the fact that it best comports with

our obligation to interpret a statute “in light of its object and policy.” United States v.

Turpin, 65 F.3d 1207, 1210 (4th Cir. 1995).

6 Forest claims this regulation is irrelevant because Sheldon did not include any particularized factual allegations concerning the company’s conduct after 2014. Response Br. at 27–28. Even if this is true, the regulation still shows that CMS has consistently interpreted the Rebate Act to require stacked rebates be included in best price. 81 Fed. Reg. at 5253 (noting that the 2016 regulation is consistent with the 2007 regulation).

65 Ultimately, despite the majority opinion’s protestations, we must defer to the

reasonable interpretation of CMS. If we do, then we must find that Forest acted under an

objectively unreasonable reading of the Rebate Statute. 7

Even if we conclude that Forest’s reading was reasonable, it still falters at Safeco’s

second step because it was warned away from that reading. But before I get to that, I must

first address the majority opinion’s flawed warned-away standard.

According to the majority opinion, a defendant may only be warned away from an

erroneous statutory reading by two “authoritative” sources: “circuit court precedent or

guidance from the relevant agency.” Majority Op. at 22. As support, it cites Safeco and

several out-of-circuit cases. Id. However, Safeco did not expressly limit the warned-away

exception to just these two sources. See 551 U.S. at 70 (addressing “guidance from the

courts of appeals or the [relevant agency]” but not expressly limiting the inquiry to these

sources only). And in fact, we have already held that the warned-away exception extends

beyond these two sources.

In United States ex rel. Lutz v. Mallory, 988 F.3d 730 (4th Cir. 2021), we considered

whether a blood-testing lab knowingly violated the Anti-Kickback Statute and thus ran

afoul of the False Claims Act. Id. at 735–36. At trial, the Government offered evidence that

7 A final note on the interpretation of the Rebate Statute. Though the majority opinion’s statutory analysis is couched in absolute terms, its holding is much more modest: it only concludes “that Forest has offered, at minimum, an objectively reasonable reading of the Rebate Statute.” Majority Op. at 22. Therefore, the majority opinion’s reading of the statute is not binding on this or any other court.

66 the defendants’ own attorneys warned them their scheme might violate the statute. Id. at

736. In addition, the Government “offered evidence that outside lawyers warned all three

[of the] [d]efendants about the illegality of the[ir kickbacks].” Id. at 736–37. The jury found

the defendants had knowingly violated the Anti-Kickback Statute, and we declined to

reverse as a matter of law. Id. at 735–36.

The defendants argued that “because the Anti-Kickback Statute is ambiguous, they

could have reasonably concluded that the statute did not prohibit [their scheme], and so

they cannot have knowingly violated the False Claims Act.” Id. at 737. We disagreed,

noting that “[the d]efendants were repeatedly ‘warned away from [their] interpretation’ of

purportedly ambiguous terms, including by legal practitioners.” Id. (emphasis added)

(quoting Purcell, 807 F.3d at 288). Because the Mallory Court expressly held that guidance

from legal practitioners can satisfy the “warned-away” exception, the majority opinion’s

attempt to limit the same exception to appellate precedent and agency guidance must fail.

United States v. Spinks, 770 F.3d 285, 290 (4th Cir. 2014) (explaining that “if two circuit

precedents conflict, the earlier one . . . controls over the later”).

The majority opinion’s failure to heed our precedent leads it to make yet another

error by holding that Forest was not warned away as a matter of law. Majority Op. at 22–

28. To wit, because the majority opinion erroneously considers only appellate precedent or

agency guidance relevant, it finds it can resolve the entire warned-away issue by

interpreting these “legal materials” on its own. Id. at 18 n.3. However, Mallory forecloses

this view. 988 F.3d at 737 (recognizing the fact-intensive nature of the warned-away

exception). And other courts, including the D.C. Circuit in United States ex rel. Purcell v.

67 MWI Corp.—a case the majority opinion repeatedly relies on—have consistently held that

whether an entity was warned away “cannot readily be labeled as a ‘purely legal’ question.”

See, e.g., Purcell, 807 F.3d at 288; see also id. at 289 (“[T]he factual question remains

whether there was sufficient evidence that [the defendant] was warned away from its

interpretation.”); United States ex rel. Brown v. Celgene Corp., 226 F. Supp. 3d 1032, 1051

(C.D. Cal. 2016) (“Whether [the defendant] was warned away from the view it took is a

question of fact.”); United States ex rel. Streck v. Bristol-Myers Squibb Co., 370 F. Supp.

3d 491, 497 (E.D. Pa. 2019) (noting “a factual determination remains whether [the

defendant] had been warned away from its interpretation by CMS[]”).

This makes sense when you take the time to think about what being “warned away”

means. A full warned-away inquiry might require determining what legal guidance existed,

what it said, who said it, how authoritative it was, when the defendant knew or should have

known about it, how the defendant responded, and what other advice the defendant might

have received from its own or outside attorneys. See, e.g., Mallory, 988 F.3d 736–37

(reviewing a timeline of legal memos, board meetings, emails, agency commentary,

judicial opinions, and legal opinions authored by outside lawyers to assess whether the

defendants were warned away). At most, this is a mixed question of law and fact.

Therefore, the majority opinion errs by finding that Forest could not have been warned

away as a pure matter of law.

With the proper framework in mind, there is no doubt that Sheldon plausibly alleged

Forest was warned away. As explained at length above, Forest explicitly asked CMS to

remove language from the 2007 regulation it believed would require rebate stacking. CMS

68 refused, and expressly rejected a “customer-by-customer” approach to best price. 8 72 Fed.

Reg. at 39,198. In response, Forest held a series of high-level meetings and instituted a data

audit to eliminate rebate stacking. It also introduced language prohibiting its customers

from claiming stacked rebates and instituted a “first come first serve” policy for rebates on

the same drug units to avoid having to report double rebates to CMS. Thus, Forest was not

only “warned away” by CMS, but also clearly took that warning to heart—at least for its

non-preferred customers. Unfortunately, under the majority opinion’s purely legal—and

purely impermissible—warned-away test, the jury will never get to consider these facts

and make its own assessment of Forest’s liability under the False Claims Act.

If the majority opinion wants to consider the impact this decision has on policy, then

here are some facts from which we can take judicial notice.

Every year, between $100 and $360 billion are lost to health care fraud. See National

Health Care Anti-Fraud Association, The Challenge of Health Care Fraud,

8 The majority opinion counters that this same rulemaking repeatedly urged manufacturers like Forest “to make reasonable assumptions” when calculating best price. Majority Op. at 26. But as the majority acknowledges, a manufacturer may only make such assumptions “[i]n the absence of specific guidance,” and such assumptions must be “consistent with the general requirements and the intent of the [Rebate Statute], [and] Federal regulations.” 72 Fed. Reg. at 39,164. For the reasons explained above, the majority opinion errs by finding the 2007 guidance was not specific; after all, it was specific enough to trigger Forest to conduct a data audit, alter its sales-contract language, and refuse to make stacked-rebate payments for most of its customers. Similarly, the majority opinion cannot explain how Forest’s neat trick—directly paying out rebates to different customers instead of paying one rebate to its wholesaler to avoid reporting double rebates to CMS— is consistent with the intent of the Rebate Statute.

69 https://www.nhcaa.org/tools-insights/about-health-care-fraud/the-challenge-of-health-

care-fraud/ (last visited Dec. 19, 2021). And these numbers are only growing. See, e.g.,

Mike Stankiewicz, Medicaid Wasted $37B on Improper Payments in 2017, CMS Shrugs

Off GAO Advice, Fierce Healthcare (Apr. 13, 2018), https://www.fiercehealthcare.com/

payer/medicaid-wasted-37b-improper-payments-gao (noting fraud has “spiked in recent

years”).

In this swelling sea of fraud, the Government is bailing out with an ever-shrinking

teaspoon. In fiscal year 2020, the Government recovered only $1.8 billion in settlements

and judgments for health care fraud using the False Claims Act. Press Release, U.S. Dep’t

of Justice, Justice Department Recovers Over $2.2 Billion from False Claims Act Cases in

Fiscal Year 2020 (Jan. 14, 2021) (noting over 80% of the total fraud recovery in 2020

related to the health care industry). This was almost a 30% decline from the amount

recovered in 2019, and over a 40% decline from the $3.1 billion high-water mark in 2012.

Id. Thus, it is not only the “sad truth . . . that [fraud] against the Government

often does pay,” S. Rep. No. 99-345, at 3, 1986 U.S.C.C.A.N. at 5268, but getting away

with it is also getting easier.

Unfortunately, today’s majority opinion only worsens this trend. In doing so, the

majority opinion joins a long and ignominious line of cases that have “thwart[ed] the

effectiveness of the [Act]” by adopting overly “restrictive” scienter standards. Id. at 4, 1986

U.S.C.C.A.N. at 5269.

70 Thirty years ago, Congress stepped in to correct the worst of these judicial abuses.

If the majority decision stands, Congress will be forced—unnecessarily—to do the same

again. With respect for my colleagues in the majority, I dissent. 9

9 The majority opinion finds it unnecessary to address the district court’s falsity finding because it concludes that Sheldon did not plausibly allege scienter. Majority Op. at 9 n.2. But the falsity finding was plainly inconsistent with the text of the False Claims Act and our precedent.

The district court found that the False Claims Act only punishes “objective falsehoods,” United States ex rel. Sheldon v. Forest Lab’ys, 499 F. Supp. 3d 184, 212 (D. Md. 2020)—those “expressions of fact” that are capable of “empirical verification” and, thus, can be shown to be empirically false, United States ex rel. Wilson v. Kellogg Brown & Root, Inc., 525 F.3d 370, 377–78 (4th Cir. 2008) (quoting Harrison v. Westinghouse Savannah River Co., 176 F.3d 776, 792 (4th Cir. 1999)). Since Forest acted under an objectively reasonable interpretation of the statute, the district court concluded, its best- price reports were not verifiably and objectively “false” for False Claims Act purposes. Sheldon, 499 F. Supp. 3d at 212.

There are three major problems with this analysis. First, on its face, the False Claims Act is not limited to “objective falsehoods”—it merely requires “a false or fraudulent claim” or “statement.” 31 U.S.C. § 3729(a)(1)(A)–(B). And at common law, “false or fraudulent claims” include “more than just claims containing express [or empirical] falsehoods.” Escobar, 579 U.S. at 187; see also id. at 188 (noting that even statements that are technically true can be “actionable misrepresentations”).

Second, injecting “objectivity” at this stage impermissibly conflates scienter with falsity. See Mallory, 988 F.3d at 737 (holding that whether a defendant failed to comply with an “ambiguous” statutory term “go[es] to whether the government proved knowledge” (quoting Purcell, 807 F.3d at 287)); United States ex rel. Oliver v. Parsons Co., 195 F.3d 457, 463 (9th Cir. 1999) (“[W]hile the reasonableness of [a defendant’s] interpretation of the applicable [statute] may be relevant to whether it knowingly submitted a false claim, the question of ‘falsity’ itself is determined by whether [a defendant’s] representations were accurate in light of applicable law.”). Forest “either complied with” the Rebate Statute “or [it] didn’t”; its allegedly “reasonable” reading of the statute plays no part in the falsity inquiry. Drakeford, 792 F.3d at 383–84.

(Continued) 71 Third, even if we conclude that the False Claims Act requires an “objective falsehood,” the district court erred by concluding that compliance with the law in this case is not empirically verifiable. This Court has held that whether an entity complied with the law is an “objective inquiry.” Id. at 384 (emphasis added). Again, Forest’s statements “either complied with” the Rebate Statute “or [they] didn’t.” Id. at 383–84. And, if they did not, then they would be objectively false. The district court never determined whether that was the case here.