United States Ex Rel. May v. Purdue Pharma L.P.

Vacated and remanded by published opinion.

Chief Judge TRAXLER wrote the opinion, in which Judge DIAZ and Judge GROH joined.

Appellants Steven May and Angela Radcliffe brought this action under the False Claims Act, 31 U.S.C. §§ 3729-33 (the “FCA”), against Purdue Pharma L.P. and Purdue Pharma, Inc. (together, “Purdue”). Giving preclusive effect to this court’s decision in United States ex rel. Radcliffe v. Purdue Pharma L.P., 600 F.3d 319 (4th Cir.2010), the district court dismissed the action on res judicata grounds. Because we agree with the appellants that this action is not barred by res judicata, we vacate the decision of the district court and remand for further proceedings.

I.

Mark Radcliffe, the husband of appellant Angela Radcliffe, was a district sales manager for Purdue. Radcliffe was laid off as part of a reduction in force in *911 June 2005, and he subsequently executed a general release (the “Release”) of all claims against Purdue in order to receive an enhanced severance package. Radcliffe thereafter filed an FCA action against Purdue (“Qui Tam I ”) 1 in which he alleged that Purdue falsely marketed its narcotic pain medication OxyContin to physicians as being twice as potent as MS Contin (a cheaper, off-patent drug also manufactured by Purdue), thus making it appear that OxyContin was cheaper per dose than MS Contin. The government investigated Radcliffe’s allegations and declined to intervene in his action.

The district court eventually dismissed Qui Tam I with prejudice, concluding that Radcliffe’s amended complaint did not satisfy the heightened pleading requirements of Rule 9. See Fed.R.Civ.P. 9(b) (“In alleging fraud or mistake, a party must state ■with particularity the circumstances constituting fraud or mistake.... ”). On appeal, we affirmed the with-prejudice dismissal on alternate grounds, concluding that the Release barred Radcliffe’s FCA claims. See Radcliffe, 600 F.3d at 333.

After we issued our opinion in Radcliffe, Steven May and Angela Radcliffe (the “Relators”) commenced this FCA action against Purdue (“Qui Tam II”) setting forth allegations nearly identical to those advanced by Mark Radcliffe in Qui Tam I. As noted, Angela Radcliffe is Mark Radcliffe’s wife; Steven May was formerly a sales representative for Purdue under Mark Radcliffe’s supervision.

Purdue moved to dismiss the Relators’ complaint on res judicata grounds, arguing that our decision in Radcliffe barred the Relators from proceeding with Qui Tam II. See, e.g., Martin v. Am. Bancorporation Retirement Plan, 407 F.3d 643, 650 (4th Cir.2005) (“Res judicata ... precludes the assertion of a claim after a judgment on the merits in a prior suit by the parties or their privies based on the same cause of action.”). Purdue also argued that the FCA’s public-disclosure bar, see 31 U.S.C. § 3730(e)(4), divested the district court of jurisdiction over the action and that the complaint did not allege fraud with the particularity required by Rule 9.

As to the res judicata question, Purdue contended that Radcliffe was a judgment on the merits because it. affirmed a with-prejudice dismissal; that the claims asserted in Qui Tam I and Qui Tam II were identical; and that the parties were identical because Qui Tam I was “brought on behalf of the United States as the real party in interest,” such that the government “and any other relators seeking to allege identical claims are bound by its judgment.” J.A. 83. The Relators argued that Radcliffe was not a decision on the merits for res judicata purposes, but they did not directly dispute Purdue’s contention that the parties were identical.

Citing Adkins v. Allstate Insurance Co., 729 F.2d 974 (4th Cir.1984), the district court held that Radcliffe was necessarily a decision on the merits because it affirmed the grant of a summary-judgment motion. See Adkins, 729 F.2d at 976 n. 3 (“For purposes of res judicata, a summary judgment has always been considered a final disposition on the merits.”). And because the Relators did not challenge the other res-judicata requirements, the district court held without further analysis that “the instant case is barred by the doctrine of res judicata.” J.A. 225. The district court therefore dismissed the action with *912 out considering the other issues raised by Purdue. This appeal followed.

II.

The Relators argue on appeal that the district court erred by giving preclusive effect to Radcliffe and dismissing their action on res judicata grounds. The preclusive effect of a judgment issued by a federal court is a legal question governed by federal common law and subject to de novo review. See Taylor v. Sturgell, 553 U.S. 880, 891, 128 S.Ct. 2161, 171 L.Ed.2d 155 (2008) (federal common law determines preclusive effect of federal-court judgment); Clodfelter v. Republic of Sudan, 720 F.3d 199, 210 (4th Cir.2013) (district court’s application of res judicata reviewed de novo).

Generally speaking, whether res judicata precludes a subsequent action “turns on the existence of three factors: (1) a final judgment on the merits in a prior suit; (2) an identity of the cause of action in both the earlier and the later suit; and (3) an identity of parties or their privies in the two suits.” Clodfelter, 720 F.3d at 210 (4th Cir.2013) (internal quotation marks omitted).

A.

The Relators contend that Radcliffe was not a “judgment on the merits” because the decision was premised on a determination that Mark Radcliffe lacked standing to pursue the FCA claims. Because Article III standing requirements are jurisdictional, see, e.g., United States v. Day, 700 F.3d 713, 721 (4th Cir.2012), cert. denied, — U.S.-, 133 S.Ct. 2038, 185 L.Ed.2d 887 (2013), and jurisdictional dismissals are not “judgment[s] on the merits for purposes of res judicata,” Goldsmith v. Mayor of Balt., 987 F.2d 1064, 1069 (4th Cir.1993), 2 the Relators argue that Radcliffe is not entitled to preclusive effect.

We disagree with the Relators’ reading of our decision in Radcliffe. Standing principles require the plaintiff to have suffered an “injury in fact.” Lujan v. Defenders of Wildlife, 504 U.S. 555, 560, 112 S.Ct 2130, 119 L.Ed.2d 351 (1992) (internal quotation marks omitted). In the context of the FCA, however, it is the government, not the private-citizen relator, that has been injured by the defendant’s fraud. FCA relators nonetheless have standing to bring an FCA action because the FCA “effect[s] a partial assignment of the Government’s damages claim” and thus statutorily vests private citizens with standing. Vt. Agency of Natural Res. v. United States ex rel. Stevens, 529 U.S. 765, 773, 120 S.Ct. 1858, 146 L.Ed.2d 836 (2000).

In Radcliffe, we discussed FCA standing principles in the course of rejecting one of Radcliffe’s arguments against enforcement of the Release. As we explained, “Radcliffe had a statutory [FCA] claim, and the necessary legal standing as partial assign-ee” once the government suffered an injury and Radcliffe became aware of the fraud. Radcliffe, 600 F.3d at 329 (emphasis added). We did not conclude that Radcliffe lost standing when he executed the Release, but instead simply held that his execution of the Release effected a waiver of his right to sue Purdue. See id. at 329 (explaining that Mark Radcliffe “had the right” to bring an FCA action before he signed the Release, “a right he waived under the terms of the Release”).

B.

Although we reject the Relators’ assertion that Radcliffe was a jurisdiction *913 al dismissal, we nonetheless agree with their bottom-line position that the district court erred by giving Radcliffe preclusive effect.

As the government notes in its amicus brief, the traditional res-judicata inquiry is modified in cases where the earlier action was dismissed in accordance with a release or other settlement agreement. See Keith v. Aldridge, 900 F.2d 736, 740-41 (4th Cir.1990). A judgment entered “based upon the parties’ stipulation, unlike a judgment imposed at the end of an adversarial proceeding, receives its legitimating force from the fact that the parties consented to it.” Norfolk S. Corp. v. Chevron, U.S.A., Inc., 371 F.3d 1285, 1288 (11th Cir.2004). Thus, where a dismissal is “based on a settlement agreement, ... the principles of res judicata apply (in a somewhat modified form) to the matters specified in the settlement agreement, rather than the original complaint.” Id. That is, given the contractual nature of consent decrees and settlement agreements, the preclusive effect of a judgment based on such an agreement can be no greater than the preclusive effect of the agreement itself. 3 See Keith, 900 F.2d at 740 (“When a consent judgment entered upon settlement by the parties of an earlier suit is invoked by a defendant as preclusive of a later action, the preclusive effect of the earlier judgment is determined by the intent of the parties.”); 18A Charles A. Wright, Arthur R. Miller & Edward H. Cooper, Federal Practice & Procedure § 4427 (“Judgments that-rest on stipulations, admissions in pleadings, or consent to the very'judgment itself should be given effect according to the intention of the parties.... ”); see also Ohio Valley Envtl. Coal. v. Aracoma Coal Co., 556 F.3d 177, 211 (4th Cir.2009) (“Settlement agreements operate on contract principles, and thus the preclusive effect of a settlement agreement should be measured by the intent of the parties.” (internal quotation marks omitted)). 4

The Release éxecuted by Mark Radcliffe in Qui Tam I was personal to him and addressed only his rights and the claims that he might assert against Purdue. Neither the Relators nor the government were parties to or intended beneficiaries of the Release. See Restatement (Second) of Contracts § 302; see also United States ex rel. Ubl v. IIF Data Solutions, 650 F.3d 445, 451 (4th Cir.2011) (explaining that'the effect of an agreement settling FCA claims is a question of federal common law as to which the Restatement (Second) of Contracts provides guidance). The Release itself, therefore, could not serve as a defense to any claims that the Relators (or other non-signatories) might assert against Purdue. Indeed, we made this very point in Radcliffe when we noted that the Release “did not prohibit the government or another relator from pursuing similar *914 claims against Purdue.” Radcliffe, 600 F.3d at 329 n. 8. Our decision in Raddiffe enforcing the Release did not (and could not) broaden the scope. of the Release. Accordingly, because the Release does not bar non-signatories from proceeding against Purdue, the judgment enforcing the Release cannot bar such claims.

Purdue’s arguments to the contrary are not persuasive. Our dismissal in Raddiffe may well have been a dismissal “on the merits” under Rule 41. See Fed.R.Civ.P. 41 (“Unless the dismissal order states otherwise, a dismissal under this subdivision (b) and any dismissal not under this rule— except one for lack of jurisdiction, improper venue, or failure to join a party under Rule 19 — operates as an adjudication on the merits.”); Shoup v. Bell & Howell Co., 872 F.2d 1178, 1181 (4th Cir.1989) (“[F]or purposes of res judicata, a summary judgment has always been considered a final disposition on the merits.” (internal quotation marks omitted)). As the Supreme Court has explained, however, “it is no longer true that a judgment ‘on the merits’ [for purposes of Rule 41] is necessarily a judgment entitled to claim-preclusive effect.” Semtek Int’l, Inc. v. Lockheed Martin Corp., 531 U.S. 497, 503, 121 S.Ct. 1021, 149 L.Ed.2d 32 (2001) (emphasis added). As discussed above, the preclusive effect of a judgment enforcing a settlement agreement is determined by the intent of the parties as reflected by the terms of that agreement, and the Release did not bar anyone other than Mark Radcliffe from bringing suit against Purdue. Regardless of the procedural vehicle through which our decision enforcing the Release was entered, our decision simply did not broaden-the scope of the Release. See Am. Cyanamid Co. v. Capuano, 381 F.3d 6, 17 (1st Cir.2004) (“[A] dismissal with prejudice contained in a consent decree is not a ruling on the merits that applies to others under the law of claim preclusion.” (internal quotation marks and alterations omitted)). Accordingly, the district court erred by dismissing Qui Tam II as barred by principles of res judicata.

III.

We turn now to the contention urged by Purdue and the government that the district court’s dismissal can be affirmed because the action is prohibited by 31 U.S.C. § 3730(e)(4), the FCA’s “public disclosure” bar. Addressing that argument requires us to first determine which version of the statute applies to this case.

A.

The complaint focuses on conduct occurring between 1996 and 2005. At that time, the public-disclosure bar provided:

No court shall have jurisdiction over an action under this section based upon the public disclosure of allegations or transactions in a criminal, civil, or administrative hearing, in a congressional, administrative, or Government Accounting Office report, hearing, audit, or investigation, or from the news media, unless the action is brought by the Attorney General or the person bringing the action is an original source of the information.

31 U.S.C. § 3730(e)(4)(A) (2005) (emphasis added).

Section 3730(e)(4), however, was amended on March 23, 2010 — after the occurrence of the conduct alleged in the complaint, but before the commencement of this action. See Patient Protection & Affordable Care Act, Pub.L. 111-148, § 10104(j)(2), 124 Stat. 119, 901-02. The statute as amended provides that:

The court shall dismiss an action or claim under this section, unless opposed by the Government, if substantially the *915 same allegations or transactions as alleged in the action or claim were publicly disclosed—

(i) in a Federal criminal, civil, or administrative hearing in which the Government or its agent is a party;

(ii) in a congressional, Government Accountability Office, or other Federal report, hearing, audit, or investigation; or

(iii) from the news media,

unless the action is brought by the Attorney General or the person bringing the action is an original source of the information.

31 U.S.C. § 3730(e)(4)(A) (2010) (emphasis added). Purdue argues that the amended version of the statute applies, while the Relators argue that the prior version of the statute applies.

“[T]he principle that the legal effect of conduct should ordinarily be assessed under the law that existed when the conduct took place has timeless and universal appeal.” Landgraf v. USI Film Prods., 511 U.S. 244, 265, 114 S.Ct. 1483, 128 L.Ed.2d 229 (1994) (internal quotation marks omitted). Accordingly, a “presumption against retroactive legislation-is deeply rooted in our jurisprudence,” id., and that “time-honored presumption” must apply “unless Congress has clearly manifested its intent to the contrary,” Hughes Aircraft Co. v. United States ex rel. Schumer, 520 U.S. 939, 946, 117 S.Ct. 1871, 138 L.Ed.2d 135 (1997). The presumption against retroactivity, however, is limited to statutes “that would have genuinely ‘retroactive’ effect.” Landgraf, 511 U.S. at 277, 114 S.Ct. 1483. A statute has retroactive effect if it “takes away or impairs vested rights acquired under existing laws, or creates a new obligation, imposes a new duty, or attaches a new disability, in respect to transactions or considerations already past.” Id. at 269, 114 S.Ct. 1483 (internal quotation marks omitted).

Applying these principles, the Supreme Court has twice held that the 2010 FCA amendments may not be applied to cases arising before the effective date of the amendments. See Graham Cnty. Soil & Water Conservation Dist. v. United States ex rel. Wilson, 559 U.S. 280, 283 n. 1, 130 S.Ct. 1396, 176 L.Ed.2d 225 (2010) (“The legislation makes no mention of retroactivity, which would be necessary for its application to pending cases given that it eliminates petitioners’ claimed defense to a qui tarn suit.”); see also Schindler Elevator Corp. v. United States ex rel. Kirk, — U.S. -, 131 S.Ct. 1885, 1889 n. 1, 179 L.Ed.2d 825 (2011) (citing Graham County and stating that the 2010 amendments “are not applicable to pending .cases”). The circuit courts .considering the issue have likewise applied the pre-2010 version of the statute. See United States ex rel. Zizic v. Q2Administrators, LLC, 728 F.3d 228, 232 n. 3 (3d Cir.2013); United States ex rel. Goldberg v. Rush Univ. Med. Ctr., 680 F.3d 933, 934 (7th Cir.2012); United States ex rel. Jamison v. McKesson Corp., 649 F.3d 322, 326 n. 6 (5th Cir.2011); United States ex rel. Poteet v. Bahler Med., Inc., 619 F.3d 104, 107 n. 2 (1st Cir.2010); United States ex rel. Hixson v. Health Mgmt. Sys., Inc., 613 F.3d 1186, 1188 n. 3 (8th Cir.2010).

Purdue suggests the analysis should be different in this case, however, because Graham County and Schindler, unlike this ease, involved complaints that were filed before the statute was amended. We disagree. The retroactivity inquiry looks to when the underlying conduct occurred, not when the complaint was filed. See Landgraf, 511 U.S. at 265, 114 S.Ct. 1483 (“[T]he legal effect of conduct should ordinarily be assessed under the law that existed when the conduct took place.... ” *916 (emphasis added)). While changes in jurisdictional and procedural rules are often applied to pending cases, that is not because the date of filing controls, see Hughes Aircraft, 520 U.S. at 946, 117 S.Ct. 1871 (refusing to apply 1986 FCA amendments to action that was commenced after the effective date of the amendments), but because application of those new rules often does not have an impermissible retroactive effect. See Landgraf, 511 U.S. at 274, 114 S.Ct. 1483 (“Application of a new jurisdictional rule usually takes away no substantive right but simply changes the tribunal that is to hear the casé.” (internal quotation marks omitted)); id. at 275, 114 S.Ct. 1488 (“Because rules of procedure regulate secondary rather than primary conduct, the fact that a new procedural rule was instituted after the conduct giving rise to the shit does not make application of the rule at trial retroactive.”).

The Supreme Court determined in Graham County and Schindler that application of the 2010 amendments would have retroactive effect if applied in those cases, and we conclude that the amendments likewise would have retroactive effect if applied in this case. See Baldwin v. City of Greensboro, 714 F.3d 828, 836 (4th Cir. 2013) (retroactivity inquiry looks ' to “whether the new statute would have retroactive effect as applied to the particular case” (internal quotation marks omitted)); Gordon v. Pete’s Auto Serv. of Denbigh, Inc., 637 F.3d 454, 459 (4th Cir.2011) (“Th[e retroactivity] inquiry is narrow, for it asks not whether the statute may possibly have an impermissible retroactive effect in any case, but specifically whether applying the statute to the person objecting would have a retroactive consequence in the disfavored sense.” (internal quotation marks and citation omitted)).

Under the prior version of the statute, § 3730(e)(4) operated as a jurisdictional limitation — the public-disclosure bar, if applicable, divested the district court of subject-matter jurisdiction over the action. See 31 U.S.C. § 3730(e)(4) (2005) (“No court shall have jurisdiction over an action under this section based upon the public disclosure of allegations.... ” (emphasis added)); Rockwell Int’l Corp. v. United States, 549 U.S. 457, 468-69, 127 S.Ct. 1397, 167 L.Ed.2d 190 (2007) (explaining that § 3730(e)(4) is a “jurisdiction-removing provision”). It is apparent, however, that the public-disclosure bar is no longer jurisdictional.' The amended statute does not mention jurisdiction but instead states that in cases where the bar is applicable, the court “shall dismiss” the action “unless opposed by the Government.” 31 U.S.C. § 3730(e)(4) (2010). The 2010 amendments thus deleted the unambiguous jurisdiction-removing language previously contained in § 3730(e)(4) and replaced it with a generic, not-obviously-jurisdictional phrase (“shall dismiss”), while at the same time retaining jurisdiction-removing language in §§ 3730(e)(1) and (e)(2). 5 In our view, these changes make it clear that the public-disclosure bar is no longer a jurisdiction-removing provision. See, e.g., Brewster v. Gage, 280 U.S. 327, 337, 50 S.Ct. 115, 74 L.Ed. 457 (1930) (“The deliberate selection of language so differing from that used in the earlier acts indicates that a change of law was intended.”); Pirie v. Chi. Title & Trust Co., 182 U.S. 438, 448, 21 S.Ct. 906, 45 L.Ed. 1171 (1901) (“When the purpose of a prior law is *917 continued, usually its words are,’ and an omission of the words implies an omission of the purpose.”); Chertkof v. United States, 676 F.2d 984, 987 (4th Cir.1982) (“[T]he deletion of language, having so distinct a meaning, almost compels the opposite result when words of such plain meaning are excised.”). Indeed, it is difficult to understand how the amended public-disclosure bar could be jurisdictional when the government has the ability to veto a dismissal under that section. See Gonzalez v. Thaler, — U.S.-, 132 S.Ct. 641, 648, 181 L.Ed.2d 619 (2012) (“Subject-matter jurisdiction can never be waived or forfeited.”); Brickwood Contractors, Inc. v. Datanet Eng’g, Inc., 369 F.3d 385, 390 (4th Cir.2004) (en banc) (“Subject-matter jurisdiction cannot be conferred by the parties, nor can a defect in subject-matter jurisdiction be waived by the parties.”). And even if the changes somehow did not establish Congress’ intent to convert the public-disclosure bar into a non-jurisdictional basis for dismissal, the omission of the jurisdictional language would nonetheless require us to treat the amended public-disclosure bar as such. See Sebelius v. Auburn Reg’l Med. Ctr., — U.S. -, 133 S.Ct. 817, 824, 184 L.Ed.2d 627 (2013) (Unless “Congress has clearly stated that the [statutory limitation] is jurisdictional ..., courts should treat the restriction as nonjurisdic-tional in character.” .(internal quotation marks and alteration omitted)),

Moreover, the 2010 amendments significantly changed the scope of the public-disclosure bar. Under the prior version of the statute, disclosures in federal and state trials and hearings qualify as public disclosures, see, e.g., McElmurray v. Consol. Gov’t of Augusta-Richmond Cnty., 501 F.3d 1244, 1252 (11th Cir.2007), and disclosures in federal and state reports, audits, or investigations likewise constitute public disclosures, see Graham Cnty., 559 U.S. at 301, 130 S.Ct. 1396. After the amendments, however, only disclosures in federal trials and hearings and in federal reports and investigations qualify as public disclosures. See 31 U.S.C. §§ 3730(e)(4)(A)® & (ii) (2010). The 2010 amendments thus substantially narrowed the class of disclosures that can trigger the public-disclosure bar. By the same token, the amendments expand the number of private plaintiffs entitled to bring qui tam actions by including plaintiffs who learn of the underlying fraud through disclosures in state proceedings or reports.

And as we will discuss in more detail in the next section, the 2010 amendments also changed the required connection between the plaintiffs claims and the qualifying public disclosure. Under the pre-amendment version of the statute, an action is barred if the action is “based upon” a qualifying public disclosure, see 31 U.S.C. § 3730(e)(4)(A) (2009), a standard we have interpreted to mean that the plaintiff must have “actually derived” his knowledge of the fraud from the public disclosure. United States ex rel. Siller v. Becton Dickinson & Co., 21 F.3d 1339, 1348 (4th Cir.1994). As amended, however, the public-disclosure bar no longer requires actual knowledge of the public disclosure, but instead applies “if substantially the same allegations or transactions were publicly disclosed.” 31 U.S.C.. § 3730(e)(4)(A) (2010). Because the Relators allege that they did not derive their knowledge of Purdue’s fraud from any public disclosure, their claims are viable under the pre-amendment version of the FCA, but not under the amended version, which focuses on the similarity of the allegations of fraud rather than the derivation of the knowledge of fraud.

We believe that these significant revisions to the statute “change® the substance of the existing cause of action,” *918 Hughes Aircraft, 520 U.S. at 948,117 S.Ct. 1871, such that the amended statute would have retroactive effect if applied in this case. The 2010.amendments deprive Purdue of the previously available jurisdictional defense and replace it with a non-jurisdictional defense that is triggered by a substantially narrower range of public disclosures and is,, even then, subject to veto by the government. See id. (1986 FCA amendment had retroactive effect because it “eliminate[d] a defense to a qui tam suit ... and therefore change[d] the substance of the existing cause of action for qui tam defendants” (internal quotation marks and alteration omitted)); id. at 948-49, 117 S.Ct. 1871 (1986 amendment “create[d] a new cause of action” by “exten[ding] ... an FCA cause of action to private parties in circumstances where the action was previously foreclosed” (internal quotation marks omitted)). The 2010 amendments similarly imperil the Relators’ right to assert their claims against Purdue, a right they possessed and could have acted upon up until the moment that the amendments took effect. See Landgraf, 511 U.S. at 269, 114 S.Ct. 1483 (statute has retroactive effect if it “takes away or impairs vested rights acquired under' existing laws” (internal quotation marks omitted)); cf. Brown v. Angelone, 150 F.3d 370, 373 (4th Cir.1998) (“When application of a new limitation period would wholly eliminate claims for substantive rights or remedial actions considered timely under the old law, the application is impermissibly retroactive. The legislature cannot extinguish an existing cause of action by enacting a new limitation period without first providing a reasonable time after the effective date of the new limitation period in which to initiate the action.” (citations and internal quotation marks omitted)). Accordingly, because the 2010 amendments have retroactive effect and the legislation is silent as to retroactivity, the 2010 version of the public-disclosure bar cannot be applied in this case, notwithstanding the fact that the complaint was filed after the effective date of the amendments. See Hughes Aircraft, 520 U.S. at 946, 117 S.Ct. 1871 (declining to apply 1986 FCA amendments to action alleging pre-amendment fraud that was commenced after the effective date of the amendments).

B.

Having concluded that the pre2010 version of § 3730(e)(4) applies, we turn to the question of whether the public-disclosure bar requires dismissal of this action.

As previously noted, the preamendment version of the public-disclosure bar provides that:

No court shall have jurisdiction over an action under this section based upon the public disclosure of allegations or transactions in a criminal, civil, or administrative hearing, in a congressional, administrative, or Government Accounting Office report, hearing, audit, or investigation, or from the news media, unless the action is brought by the Attorney General or the person bringing the action is an original source of the information.

31 U.S.C. § 3730(e)(4)(A) (2005) (emphasis added). Although most circuits have interpreted the “based upon” language to bar actions where the allegations of fraud were “supported by” or “substantially similar” to.fraud that had been publicly disclosed, see, e.g., United States ex rel. Mistick PBT v. Housing Auth., 186 F.3d 376, 386 (3d Cir.1999) (collecting cases), this circuit has interpreted the clause as barring only those actions where the relator’s knowledge of the fraud alleged was actually derived from the public disclosure itself. See Siller, 21 F.3d at 1348 (“[A] relator’s *919 action is ‘based upon’ a public disclosure of allegations only where the relator has actually derived from that disclosure the allegations upon which his qui tam action is based.” (emphasis added)). The public-disclosure bar applies and requires dismissal if the action is “even partly” derived from prior public disclosures. See United States ex rel. Vuyyuru v. Jadhav, 555 F.3d 337, 351 (4th Cir.2009).

Whether a relator derived his knowledge of the fraud from a public disclosure is a jurisdictional fact to be resolved by the district court. See id. at 348, 350; Siller, 21 F.3d at 1349. Although the district court dismissed this action on res judicata grounds without addressing the public-disclosure bar, Purdue contends that the record nonetheless establishes that the allegations in this action were at least partly derived from the publicly disclosed allegations contained in the Qui Tam I complaint. Purdue points out that the allegations of the complaints in Qui Tam I and Qui Tam II are nearly identical, and that many of the allegations in Qui Tam II are verbatim copies of Qui Tam I allegations. In Purdue’s view, “[t]he verbatim overlap of the complaints forecloses any argument that the complaint in this action was not at least partly based on the ... [c]omplaint in Qui Tam I.” Br. of Resp’t at 31. We disagree.

Under Siller, the question is not whether the allegations set out in the relator’s complaint are .similar to publicly disclosed allegations of fraud; the question is whether the relator’s knowledge of the fraud was actually derived from the public disclosure — that is, whether the relator learned about the fraud from the public disclosure. See Siller, 21 F.3d at. 1347, 1348 (“[T]he only fair construction” of § 3730(e)(4) is that “a qui tam action is only ‘based upon’ a public disclosure where the relator has actually derived from that disclosure the knowledge of the facts underlying his action.” (emphasis added)); see also id. at 1348 (explaining that an FCA action could “include[] allegations that happen to be similar (even identical) to those already publicly disclosed, but were not actually derived from those public disclosures”). Indeed, the standard urged by Purdue is the standard adopted by other circuits but rejected by Siller. See id. (“We are aware ... that other circuits have not embraced this interpretation of the phrase, assuming instead that an action is based upon a public disclosure of allegations if its allegations are identical or similar to those already publicly disclosed.”).

The Relators both submitted affidavits to the district court asserting that then-knowledge of Purdue’s fraud was not derived -from the Qui .Tam I complaint or any other qualifying public disclosure, but from conversations with Mark Radcliffe and, in Steven May’s case, from his own experiences as a Purdue sales representative. The similarity between the allegations in each complaint could provide a basis for disbelieving the Relators’ assertions, see Vuyyuru, 555 F.3d at 350-51, but that is an issue for the district court as fact-finder, not this court. Because the district court has not made the factual findings necessary to determine whether the public-disclosure bar precludes this action, we must remand this case to the district court for discovery and other proceedings as necessary to resolve the issues related to the applicability of the public-disclosure bar. See United States ex rel. Carter v. Halliburton Co., 710 F.3d 171, 184 (4th Cir.2013), petition for cert. filed, 82 U.S.L.W. 3010 (June 24, 2013) (“Be: cause the district court should have the opportunity in the first instance to address the facts relevant to public disclosure, we remand this issue to the district court.”); Siller, 21 F.3d at 1349 (remanding for district court to determine whether allega *920 tions were “actually derived” from prior suit). If the district court determines that the Relators’ knowledge of the fraud alleged here was actually derived, even in part, from a qualifying public disclosure and that the Relators are not original sources of the information, then the district court must dismiss this action for lack of subject-matter jurisdiction. See Vuyyuru, 555 F.3d at 355.

IV.

Purdue makes two additional arguments for sustaining the district court’s dismissal of this action that do not require extended discussion.

First, Purdue contends that dismissal. was proper because the Relators’ complaint fails to allege fraud with the specificity required by Rule 9 of the Rules of Civil Procedure. We disagree. Assuming without deciding that the complaint does not allege the fraudulent conduct with the specificity required by Rule 9, see U.S. ex rel. Nathan v. Takeda Pharm. N. Am., Inc., 707 F.3d 451, 456-57 (4th Cir. 2013), 'petition for cert. filed, 81 U.S.L.W. 3650 (May 10, 2013), 6 the Relators have yet to amend their complaint, and they requested an opportunity to amend if the court believed the allegations deficient. Leave to amend a complaint should generally be freely granted, and there is at present no basis in the record for this court to conclude that any efforts to amend would be futile or otherwise improper. See, e.g., Mayfield v. NASCAR, Inc., 674 F.3d 369, 379 (4th Cir.2012) (“[A] request to amend should only be denied if one of three facts is present: the amendment would be prejudicial to the opposing party, there has been bad faith on the part of the moving party, or amendment would be futile.” (internal quotation marks omitted)). Because the Relators have not had the opportunity to amend their complaint, we believe it would be improper to rely on any Rule 9 deficiencies to affirm the district court’s dismissal of the action with prejudice. The district court on remand is free to consider Purdue’s Rule 9 argument in the first instance.

Second, Purdue argues that we can affirm the district court’s order because dismissal is required by the FCA’s “first to file” bar. See 31 U.S.C. 3730(b)(5). Section 3730(b)(5) provides that “[w]hen a person brings an action under this subsection, no person other than the Government may intervene or bring a related action based on the facts underlying the pending action.” Although this action is clearly based on the facts underlying Qui Tam I, we recently held that the first-to-file bar applies only if the first-filed action was still pending when the subsequent action was commenced. See Carter, 710 F.3d at 182-83. By the time this action was commenced, Qui Tam I had been dismissed by the district court, the dismissal had been affirmed by this court in Radclijfe, and certiorari had been denied by the Supreme Court. Qui Tam I, therefore, was no longer pending at the time this action was commenced, thus making the first-to-file bar inapplicable. See Carter, 710 F.3d at 183 (“[0]nce a case is no longer pending the first-to-file bar does not stop a relator from filing a related case.”).

V.

Accordingly, for the foregoing reasons, we vacate the district court’s order dismissing this action on res judicata grounds *921 and remand for further proceedings consistent with this opinion.

VACATED AND REMANDED