Notice: This opinion is subject to formal revision before publication in the
Federal Reporter or U.S.App.D.C. Reports. Users are requested to notify
the Clerk of any formal errors in order that corrections may be made
before the bound volumes go to press.




       United States Court of Appeals
                  FOR THE DISTRICT OF COLUMBIA CIRCUIT




Argued November 25, 2003                    Decided January 20, 2004

                               No. 02-5410

                PUREPAC PHARMACEUTICAL COMPANY,
                           APPELLEE

                                     v.

                     TOMMY G. THOMPSON,
         SECRETARY OF HEALTH AND HUMAN SERVICES, AND
        LESTER M. CRAWFORD JR., DEPUTY COMMISSIONER OF
                       FOOD AND DRUGS,
                          APPELLEES

                  TORPHARM, INC. AND APOTEX, INC.,
                            APPELLANTS



                           Consolidated with
                               03-5121



         Appeals from the United States District Court
                  for the District of Columbia
                        (No. 02cv01657)
                         (No. 03cv00254)
                           –————
 Bills of costs must be filed within 14 days after entry of judgment.
The court looks with disfavor upon motions to file bills of costs out
of time.
                              2

  William A. Rakoczy argued the cause for appellants. With
him on the briefs were Hugh S. Balsam and Arthur Y. Tsien.
Hugh L. Moore, Jacqueline H. Eagle, and Matthew O. Brady
entered appearances.
  Andrew E. Clark, Attorney, U.S. Department of Justice,
argued the cause for the federal appellees. With him on the
brief were Peter D. Keisler, Assistant Attorney General;
Eugene M. Thirolf, Director; Alex M. Azar II, General
Counsel, U.S. Department of Health & Human Services;
Daniel E. Troy, Chief Counsel; and Karen E. Schifter,
Associate Chief Counsel. Christine N. Kohl, Attorney, U.S.
Department of Justice; Douglas N. Letter, Litigation Coun-
sel; and Howard S. Scher, Attorney, entered appearances.
  Charles J. Raubicheck argued the cause for appellee Pure-
pac Pharmaceutical Company. With him on the brief was
Steven M. Amundson. James M. Webster, Mark C. Hansen,
and Richard H. Stern entered appearances.

  Before: GINSBURG, Chief Judge, and EDWARDS and TATEL,
Circuit Judges.
  Opinion for the Court filed by Circuit Judge TATEL.
   TATEL, Circuit Judge: To encourage the marketing of low-
cost generic drugs, the 1984 Hatch-Waxman amendments to
the Food, Drug, and Cosmetic Act grant companies that
successfully challenge drug patents the right to sell their
generic drugs without competition for 180 days. In this case,
two companies, each seeking to market a generic drug, com-
peted for the right to exclusivity. The Food and Drug
Administration ruled that neither company could earn exclu-
sivity by challenging the first of two patents, but it awarded
exclusivity to one of the companies based on that company’s
challenge to the second patent. The district court rejected
challenges to these two decisions, and the company denied
exclusivity now appeals. Finding no error in the district
court’s two thorough and well-reasoned opinions, we affirm in
all respects.
                                3

                                I.
  The Food, Drug, and Cosmetic Act, 21 U.S.C. §§ 301–99
(FDCA), provides that a company wishing to market a new
brand-name drug must submit a New Drug Application,
known as an NDA, to the Food and Drug Administration.
See 

(2000 & Supp. III 2003). Usually quite
lengthy, NDAs must include, among other things, evidence of
the drugs’ safety and effectiveness, as well as information
about patents that cover or might cover the drugs. 

Congress passed the ‘‘Hatch-Waxman’’ amend-
ments to the FDCA. See The Drug Price Competition and
Patent Term Restoration Act of 1984, Pub. L. No. 98–417, 98
Stat. 1585 (1984) (codified in scattered sections of titles 21, 35,
and 42 U.S.C.). Enacted to expedite the process by which
companies gain approval to sell generic versions of already-
approved brand-name drugs, the amendments allow compa-
nies seeking such approval to submit Abbreviated New Drug
Applications, known as ANDAs, that ‘‘piggyback’’ on the
safety-and-effectiveness information that the brand-name
manufacturers submitted in their NDAs. See 21 U.S.C.
§ 355(j)(2)(A); 21 C.F.R. § 314.94(a)(3) (2003). ‘‘The result
[is] to make practical the manufacture of generic copies which
theretofore had been uneconomical.’’ Mead Johnson Pharm.
Group v. Bowen, 

, 1333 (D.C. Cir. 1988).
  Like NDAs, ANDAs must address patents that cover or
might cover the relevant drugs. For each patent, companies
can satisfy this requirement by including in their ANDAs one
of several ‘‘certifications’’ that explain why the FDA should
approve the application despite the patent’s claim on the
drug. 21 U.S.C. § 355(j)(2)(A)(vii). The certification at issue
in this case—a ‘‘paragraph IV certification,’’ named for the
subsection of the law that describes it—states ‘‘that such
patent is invalid or will not be infringed by the manufacture,
use, or sale of the new drug.’’ 

In
                               4

essence, applicants use paragraph IV certifications to chal-
lenge the validity of brand-name manufacturers’ patents.
  An applicant that includes one or more paragraph IV
certifications in its ANDA must inform both the patent holder
and the company that submitted the NDA on which the
ANDA ‘‘piggybacks.’’ 

Once an appli-
cant gives notice, the FDA must wait forty-five days before
approving the ANDA, thereby giving the patent holder that
much time to file a patent-infringement suit.                

21 C.F.R. § 314.107(f)(2). If the patent
holder sues, the FDA must wait thirty months from the
notice date before approving the ANDA unless the applicant
wins the suit sooner or the court hearing the suit shortens the
thirty-month period. 21 U.S.C. § 355(j)(5)(B)(iii).
   In order to encourage paragraph IV challenges, thereby
increasing the availability of low-cost generic drugs, the
FDCA provides that the first company to win FDA approval
of an ANDA containing a paragraph IV certification has the
right to sell its drug without competition for 180 days. 

The statute and the implementing regula-
tion create this exclusivity period by prohibiting the FDA
from approving any other ANDA that contains a paragraph
IV challenge to the same patent until 180 days after the first
company markets its drug or 180 days after the first company
wins a patent-infringement suit involving that patent, which-
ever comes first. Id.; 21 C.F.R. § 314.107(c)(1).
   Paragraph IV certifications are not the only way for ANDA
applicants to satisfy their obligation to address all relevant
patents. Applicants can instead submit one or more ‘‘section
viii statements,’’ named, again, after the relevant FDCA
subsection—section 505(j)(2)(A)(viii). A section viii statement
indicates that a patent poses no bar to approval of an ANDA
because the applicant seeks to market the drug for a use
other than the one encompassed by the patent. See 21 U.S.C.
§ 355(j)(2)(A)(viii). For example, if a brand-name manufac-
turer’s patent covers a drug’s use for treating depression, and
the ANDA applicant seeks approval to use the drug to treat
any other condition, then a section viii statement would be
                               5

appropriate. Thus, whereas applicants use paragraph IV
certifications to challenge the validity of admittedly applicable
patents, they use section viii statements to assert that patents
do not apply. The FDA has long required that for every
patent ANDA applicants use either a paragraph IV certifica-
tion or a section viii statement—they may not use both. As
the FDA puts it, ‘‘either the applicant is seeking approval for
the use claimed in the patent, or it is not.’’ TorPharm, Inc. v.
Thompson, 

, 77 (D.D.C. 2003) (quoting the
record in that case) (internal quotation marks omitted).
   Paragraph IV certifications and section viii statements have
quite different consequences. Applicants submitting section
viii statements have no obligation to provide notice, nor must
they wait thirty months for FDA approval. As the district
court explained, ‘‘the FDA may [thus] approve a section viii
application immediately, making it an attractive route for
generic manufacturers, even though a section viii statement
does not entitle a successful applicant to the 180–day period
of exclusivity bestowed on paragraph IV applicants.’’ Pure-
pac Pharm. Co. v. Thompson, 

, 195
(D.D.C. 2002).
   In order to determine what patents cover existing brand-
name drugs and hence whether any paragraph IV certifica-
tions or section viii statements are needed, applicants look in
the ‘‘Orange Book,’’ an FDA publication that includes all
patent information that companies have submitted to the
agency. ‘‘Method-of-use patents’’—which cover specific uses
for drugs—can be included in the Orange Book only if they
cover drug uses that the FDA has approved. 21 C.F.R.
§ 314.53(b). In other words, companies cannot use the
Orange Book to claim protection for uses that the FDA has
not approved. The FDA, however, does not evaluate patent
information that companies submit to it; it just passively
publishes information it receives. This means that if one
company submits patent information that a second company
believes is false or violates the FDA’s regulation prohibiting
the listing of unapproved-use patents, the second company—
after having the FDA verify the information’s accuracy with
the first company—must go to court to resolve the dispute.
                               6

Meanwhile, the second company must, if it submits an ANDA,
treat the disputed patent as valid. In other words, it must
include in its ANDA either a paragraph IV certification or a
section viii statement.
   This case involves gabapentin, a drug sold by Pfizer, Inc.
that the FDA has approved for treating epilepsy. In 1998,
appellee Purepac Pharmaceutical Company submitted an
ANDA seeking permission to sell a generic version of Pfizer’s
brand-name drug, Neurontin. Purepac’s ANDA listed three
patents that Warner-Lambert Company, which later assigned
Pfizer the rights to the drug, had included in its gabapentin
NDA. One of the three—the only one relevant to this case—
was No. 5,084,479 (the ’479 patent), a method-of-use patent.
Seeking to sell its drug as a treatment for epilepsy, Purepac
submitted a section viii statement about this patent because
as it read the Orange Book, the patent covered gabapentin’s
unapproved use for treating neurodegenerative diseases.
That is, Purepac asserted that the ’479 patent posed no bar to
FDA approval of its ANDA because the patent covered a use
other than the one for which Purepac sought permission.
Given that under FDA regulations, only approved-use meth-
od-of-use patents may be listed in the Orange Book, Pure-
pac’s interpretation of the ’479 patent would, if true, have
meant that the patent’s inclusion in the Orange Book was
improper.
   About a month after Purepac submitted its ANDA, appel-
lant TorPharm, Inc. filed its own ANDA, similarly seeking
permission to market a generic version of Neurontin. Unlike
Purepac, TorPharm submitted both a paragraph IV certifica-
tion and a section viii statement regarding the ’479 patent.
By doing so, TorPharm effectively hedged its bets on whether
the information that Warner-Lambert had submitted to the
FDA showed that the patent covered gabapentin’s use for
treating epilepsy or for treating neurodegenerative diseases.
TorPharm’s notice to Warner-Lambert, however, indicated
that TorPharm agreed with Purepac’s interpretation of what
use the ’479 patent covered: the notice stated that Tor-
Pharm’s generic drug posed no danger of infringement be-
cause ‘‘[a]ll of the claims of the ’479 patent are directed to a
method of using gabapentin and its derivatives in the treat-
                               7

ment of neurodegenerative diseases.’’ Warner-Lambert Co.
v. Apotex Co., No. 98 C 4293, 

, at *2 (N.D.
Ill. Sept. 14, 2001) (quoting the record in that case) (internal
quotation marks omitted).
  The FDA initially informed Purepac that because Warner-
Lambert claimed that the ’479 patent covered gabapentin’s
use for treating epilepsy, Purepac should have filed a para-
graph IV certification. Purepac protested that this repre-
sented a reversal by the FDA, pointing out that when the
agency first listed the patent in the Orange Book, the con-
comitant ‘‘use code’’—the code the FDA assigns to identify
the use that a patent covers—bore the title ‘‘treatment of
neurodegenerative diseases.’’ By contrast, the use code that
the FDA gave to another of the patents that Warner-
Lambert had mentioned in its NDA was entitled ‘‘epilepsy.’’
Since the FDA acknowledges that ‘‘in assigning use codes it
relies exclusively on the NDA holder’s statements regarding
a patent’s coverage,’’ 

n.10,
Purepac argued that the FDA’s choice of use code indicated
that the agency had previously decided that according to
Warner-Lambert, the ’479 patent covered gabapentin’s use
for treating neurodegenerative diseases rather than epilepsy.
According to Purepac, it therefore needed to file a section viii
statement.
   Unconvinced, the FDA informed Purepac that the company
would have to add a paragraph IV certification before its
ANDA could be approved. Purepac did not follow this in-
struction. Instead, perhaps recognizing that complying with
the FDA’s demand would mean that TorPharm would receive
the 180–day exclusivity as the first company to submit an
approvable ANDA containing all necessary paragraph IV
certifications, Purepac sued the FDA in the United States
District Court for the District of Columbia, arguing that the
FDA’s rejection of its section viii statement and the agency’s
insistence that the company submit a paragraph IV certifica-
tion were arbitrary and capricious.
  The district court, through Judge Huvelle, ruled for Pure-
pac, holding that the FDA should have concluded on the basis
                               8

of the evidence before it that the ’479 patent covered gaba-
pentin’s use for treating neurodegenerative diseases. Be-
cause Purepac sought approval for a different use, the district
court directed the FDA to accept the company’s section viii
statement. See 

. The court’s
ruling did not entitle Purepac to exclusivity, however, as only
paragraph IV certifications can earn exclusivity. It simply
required the FDA to accept Purepac’s ANDA as of the date
the company originally submitted it, which occurred before
TorPharm filed its ANDA. That was important because both
Purepac’s and TorPharm’s ANDAs contained paragraph IV
challenges to other patents, and whichever company first
submitted its ANDA would be eligible for exclusivity based on
those other challenges.
   Although the district court required the FDA to accept
Purepac’s section viii statement, it declined to decide what the
agency should do about TorPharm’s paragraph IV certifica-
tion. Noting that the FDA had long insisted that paragraph
IV certifications and section viii statements were mutually
exclusive, the court observed that the agency ‘‘has not taken a
definitive position as to whether equitable considerations
might ultimately persuade it to TTT approve TorPharm’s
application with a paragraph IV certification to the ’479
patent even if the Court were to direct the agency to accept
Purepac’s application with a section viii statement.’’ 

This question was critical, for if the FDA allowed
TorPharm to submit a paragraph IV certification, then Tor-
Pharm would be entitled to exclusivity as the first applicant
to win FDA approval for an ANDA containing a paragraph
IV challenge to the ’479 patent. The district court decided to
allow the FDA to address that issue on remand. The FDA
did not appeal.
   On remand, the FDA stood by its position that paragraph
IV certifications and section viii statements are mutually
exclusive, ruling TorPharm’s paragraph IV certification as to
the ’479 patent improper. At approximately the same time,
the FDA asked Pfizer to consent to the removal of the patent
from the Orange Book, pointing out that the patent claimed
protection for a use that the agency had yet to approve.
                               9

When Pfizer consented, the FDA removed the patent from
the Orange Book. As a consequence, neither TorPharm nor
any other company that had submitted a gabapentin-related
application had any obligation to address the ’479 patent in its
ANDA. In fact, the FDA instructed all applicants to amend
their ANDAs by removing any section viii statements or
paragraph IV certifications about the patent. This meant
that no company would enjoy the 180–day exclusivity period
based on the ’479 patent because no company could submit a
proper paragraph IV certification regarding it.
   TorPharm then sued the FDA, challenging the removal of
the patent from the Orange Book and the agency’s refusal to
accept the company’s paragraph IV certification. Again
speaking through Judge Huvelle, the district court rejected
TorPharm’s claim, concluding ‘‘that the FDA acted reason-
ably in not departing from its well-settled rule that a section
viii statement and paragraph IV certification cannot be filed
as to the same patent, [and that] the agency’s corresponding
conclusion that no applicant was entitled to exclusivity on
the ’479 patent must be upheld.’’ 

.
   In its decision, the district court also addressed the FDA’s
ruling about the other gabapentin-related patent at issue in
this case, No. 6,054,482 (the ’482 patent). Whereas the ’479
patent is a method-of-use patent, the ’482 patent is a ‘‘drug-
product’’ patent, covering the drug’s overall composition and
formulation. Warner-Lambert first submitted information
about the ’482 patent to the FDA after TorPharm and
Purepac filed their ANDAs, at which point both companies
amended their ANDAs by adding paragraph IV challenges to
the ’482 patent. Although the FDCA requires applicants to
provide notice ‘‘when’’ they file their amended ANDAs, 21
U.S.C. § 355(j)(2)(B)(iii), and although the implementing reg-
ulation requires notice ‘‘at the same time’’ as filing, 21 C.F.R.
§ 314.95(d), Purepac did not send notice to Warner-Lambert
about its paragraph IV certification until June 13, 2000, some
two-and-a-half weeks after the FDA received, and deemed
filed, Purepac’s amended ANDA on May 26. TorPharm also
sent its notice to Warner-Lambert on June 13, the same day
                                10

it mailed its amended ANDA to the FDA. The FDA, howev-
er, deemed TorPharm’s ANDA to have been filed on June 16,
the day it received the ANDA. This meant that Purepac,
despite the lag between its filing with the FDA on May 26
and its notice to Warner-Lambert on June 13, completed both
tasks first. The FDA thus awarded the 180–day exclusivity
period to Purepac.
   In the district court, TorPharm argued that the FDA’s
award of exclusivity to Purepac effectively eliminated the
statutory and regulatory requirements of simultaneous notice.
The district court disagreed, concluding that neither the
statute nor the regulation specify the penalty for failing to
notify immediately and ruling that the FDA had reasonably
filled that gap by stating that if a company waits to provide
notice, it runs the risk that another company will both file and
provide notice first, thereby winning exclusivity. 

–81. The district court also rejected
TorPharm’s challenge to the FDA’s decision to use a date-of-
receipt rule rather than a mailbox rule, seeing nothing arbi-
trary or capricious in the agency’s choice. See 

   Appealing both district court decisions, TorPharm argues
that the court erred by (1) reversing the FDA’s determination
about the claimed scope of the ’479 patent (Purepac), (2)
upholding the FDA’s decision to ‘‘delist’’ the ’479 patent and
deny exclusivity based on that patent (TorPharm), and (3)
sustaining the FDA’s award of exclusivity to Purepac based
on Purepac’s challenge to the ’482 patent (also TorPharm).
‘‘Because the district court entered a summary judgment, we
review its decision de novo and therefore, in effect, review
directly the decision of the [agency].’’ Lozowski v. Mineta,

, 845 (D.C. Cir. 2002). We will set aside an FDA
decision only if it is ‘‘arbitrary, capricious, an abuse of discre-
tion, or otherwise not in accordance with law.’’ 5 U.S.C.
§ 706(2)(A) (2000). FDA interpretations of the FDCA re-
ceive deference, Serono Labs, Inc. v. Shalala, 

,
1319 (D.C. Cir. 1998), as do its interpretations of its own
regulations, 

                              11

                               II.
   We start our analysis with the district court’s Purepac
decision. In that case the court had to answer the following
question: According to the information that Warner-Lambert
submitted to the FDA, what use of gabapentin did the ’479
patent cover? In other words, setting aside the question of
what use the patent actually covered—a question the FDA
leaves to the courts—what use did Warner-Lambert say the
patent covered? As noted, the FDA told Purepac that ac-
cording to Warner-Lambert the patent covered gabapentin’s
use for treating epilepsy. This meant that Purepac, like
TorPharm, had to include a paragraph IV challenge to the
patent in its ANDA. It also meant that TorPharm, as the
first to have submitted a paragraph IV challenge to the
patent, would be eligible to earn exclusivity.
   We agree with the district court that the FDA’s conclusion
about the claimed coverage of the ’479 patent was arbitrary
and capricious. In the Orange Book, the FDA assigned the
patent a use code corresponding to neurodegenerative dis-
eases. Having thus concluded that Warner-Lambert asserted
that the patent covered gabapentin’s use for treating such
diseases, the FDA, when evaluating Purepac’s and Tor-
Pharm’s ANDAs, could not change course and decide that
Warner-Lambert claimed that the patent covered the drug’s
use for treating epilepsy. Yet in rejecting Purepac’s section
viii statement, the FDA did just that. Because this unex-
plained reversal represents the height of arbitrary and capri-
cious decision making, the district court rightly disallowed it.
As we have said, ‘‘[w]hile the scope of review under the
arbitrary and capricious standard is narrow and a court is not
to substitute its judgment for that of the agency, neither may
a court sanction agency action when the agency TTT fails to
justify seeming inconsistencies in its approach.’’ Prof’l Pilots
Fed’n v. FAA, 

, 771 (D.C. Cir. 1997) (internal
quotation marks omitted).
  TorPharm argues that the FDA’s choice of use code has no
relevance because ‘‘use codes are not required by statute TTT
[or] by the controlling regulation.’’ Appellant’s Reply Br. at 5
                               12

(internal quotation marks omitted). This misses the point.
That use codes are not required by statute has nothing
whatsoever to do with the fact that the FDA has given no
reason for making one decision for purposes of selecting a use
code and a different decision for purposes of determining
whether ANDA applicants had to submit paragraph IV certi-
fications or section viii statements.
   In fact, the FDA’s action was doubly flawed. Not only did
the agency make inconsistent decisions, but evidence before
the FDA did not support its conclusion that Warner-Lambert
claimed that the ’479 patent covered gabapentin’s use for
treating epilepsy. As the district court pointed out, Warner-
Lambert repeatedly told the FDA that the patent ‘‘covers a
method for treating neurodegenerative diseases with gaba-
pentin.’’ At no time did Warner-Lambert tell the FDA that
the patent covered a method of treating epilepsy. As the
district court aptly put it, ‘‘[t]he agency TTT tried to construct
a legal fiction about the scope of the ’479 patent and to use
that construct to ignore crucial facts (i.e. what Warner-
Lambert actually said) about that patent’s reach, facts that
reveal the ultimate falsity of the agency’s fiction.’’ 

.
   TorPharm points out that in each of the documents in
which Warner-Lambert stated that the patent covered neuro-
degenerative diseases, the company also said that the patent
covered ‘‘the use’’ of Neurontin. According to TorPharm,
because Neurontin has only one approved use—the treatment
of epilepsy—and because FDA regulations permit listing
approved uses only, the phrase ‘‘the use’’ must have referred
to the approved use, epilepsy. But as the district court
explained, the fact that an FDA regulation allows the listing
of only approved-use method-of-use patents does not prove
that every method-of-use patent in the Orange Book actually
covers an approved use. The reason is obvious: the FDA
does not evaluate information that applicants submit about
patents, but simply accepts that information passively. Thus,
we have no reason to believe that because applicants are
supposed to submit information about approved uses only,
they in fact do so. Such a benign view, the district court
concluded, ‘‘represents the triumph of hope over reality.’’ 

moreover, Warner-Lambert’s repeated claims that the
patent covered neurodegenerative diseases gave the FDA
ample reason to suspect that the company had chosen to
ignore the regulations. The FDA’s ‘‘decision [wa]s thus
factually unsupportable and irreconcilable with the language
and intent of the FDCA. For these reasons, it violate[d] the
APA.’’ 

   TorPharm also calls our attention to a letter that Pfizer
sent to the FDA ‘‘confirm[ing] that the ’479 patent was
properly listed in the Orange Book.’’ Appellant’s Reply at 4
(citing the record). But Pfizer did not say that the patent
belonged in the Orange Book because it covered the only
approved use of gabapentin. Rather, Pfizer stated that the
patent belonged in the Orange Book because in the company’s
opinion, the FDA regulation limiting the Orange Book to
approved uses violated the FDCA. Undermining TorPharm’s
position, moreover, the letter includes the following state-
ment: ‘‘Pfizer agrees that the ’479 patent does not claim
methods of use for which Neurontin has been approved.
Pfizer reconfirms that neither Pfizer nor Warner-Lambert
ever represented to FDA that the ’479 patent claimed an
approved use.’’
   In essence, then, TorPharm argues that the FDA rightly
ignored Warner-Lambert’s repeated explicit assertions about
neurodegenerative diseases in favor of oblique references to
‘‘the use’’ and other unhelpful statements. Perhaps recogniz-
ing the weakness of this position, TorPharm asserts that
Warner-Lambert’s express statements ‘‘were not required by
statute or any FDA regulation and therefore had no regulato-
ry significance.’’ Appellant’s Br. at 25. This argument is no
more convincing than the company’s similar claim regarding
use codes. Obligated like any agency to base its decisions on
the entire record, see Achernar Broad. Co. v. FCC, 

, 1446 (D.C. Cir. 1995) (‘‘Failure to weigh the entire
record would constitute reversible errorTTTT’’), the FDA may
not ignore some evidence before it just because an entity
submitted that evidence despite the absence of a legal re-
quirement to do so.
                              14

   In sum, we agree with the district court that ‘‘the FDA’s
determination that the ’479 patent claims the use of treating
epilepsy ‘runs counter to the evidence before the agency,’ and
is thus arbitrary and capricious.’’ 

(quoting Sinclair Broad. Group, Inc. v. FCC, 

, 159 (D.C. Cir. 2002)). According to TorPharm, this
conclusion contravenes the FDA’s longstanding policy under
which it refuses to determine independently what use a
patent covers and instead accepts at face value the use
claimed by the patent holder. But the district court did not
decide that the FDA should have scrutinized the claimed use,
nor did it address what the patent actually covered. The
court simply deemed the FDA’s conclusion regarding what
Warner-Lambert claimed about the patent to be unjustified.
Under the district court’s analysis, what use the patent
actually covered had no relevance, and properly so.
   Having considered TorPharm’s other challenges to Purepac
and finding them without merit, we will affirm the judgment
of the district court.

                             III.
   We next consider the district court’s rulings about the ’479
patent in TorPharm. In that case, the court reviewed the
FDA’s post-Purepac decision against making an exception to
its longstanding rule that for every patent only one of two
approaches—a section viii statement or a paragraph IV certi-
fication—is appropriate. Although the FDA had indicated in
Purepac that on remand it would consider granting an excep-
tion to this rule by accepting TorPharm’s paragraph IV
certification, the agency ultimately decided against such an
exception, deeming the certification improper. The district
court found this decision neither arbitrary nor capricious,
explaining that the refusal ‘‘to make an equitable exception
from this rule was within the FDA’s discretion, and Tor-
Pharm has pointed to nothing in the statute or regulations to
cast doubt on the rule itself.’’ 

. Recognizing the narrow scope of review, the district
court concluded that the FDA’s rule deserved judicial defer-
                              15

ence ‘‘in the absence of some indication that it conflicts with
any of the constraints on the agency’s regulatory authority, is
inconsistent with the agency’s own prior pronouncements, or
is otherwise poorly reasoned or unpersuasive. There are no
such indications here.’’ 

TorPharm
gives us no basis for questioning the district court’s sound
reasoning.
   This, however, does not end our task, for the FDA did
more than simply deny TorPharm an exception to the agen-
cy’s rule. With Pfizer’s consent, it also removed the ’479
patent from the Orange Book. At that time, TorPharm was
defending against an infringement lawsuit that Warner-
Lambert had filed against it in response to its paragraph IV
challenge to the ’479 patent. Before delisting the patent,
therefore, the FDA had to determine, as required by its
regulation, whether to delay such action because TorPharm
had the potential to earn exclusivity by prevailing in the
already-initiated       lawsuit.      See      21      C.F.R.
§ 314.94(a)(12)(viii)(B) (‘‘A patent that is the subject of a
lawsuit [charging infringement] shall not be removed from
the list until FDA determines TTT that no delay in effective
dates of approval [of other ANDAs] is required TTT as a
result of the lawsuitTTTT’’).
   The FDA concluded that the regulation posed no bar to the
delisting. The district court’s opinion in Purepac combined
with Pfizer’s post-Purepac letter to the FDA stating that ‘‘at
all times, Warner-Lambert and Pfizer made clear to FDA
that the ’479 patent claimed the use of gabapentin (Neuron-
tin) for treating neurodegenerative diseases,’’ convinced the
agency that it had erred in believing that Warner-Lambert
claimed that the ’479 patent covered epilepsy. The FDA thus
decided that it should never have listed the patent in the
Orange Book and that no ANDA applicant had to submit, or
could maintain, either a section viii statement or a paragraph
IV certification regarding that patent. This meant that even
if TorPharm won its infringement lawsuit, it would not be
entitled to exclusivity. The FDA therefore decided that its
regulation did not prevent it from delisting the patent.
                              16

   Attacking this conclusion, TorPharm focuses on the fact
that the FDA, in explaining its view that it had to remove the
patent from the Orange Book, cited not only Purepac and
Pfizer’s letter, but also the Federal Circuit’s decision in
Warner-Lambert Co. v. Apotex Corp., 

(Fed.
Cir. 2003), the suit that Warner-Lambert filed against Tor-
Pharm. The Federal Circuit held that because the ’479
patent covered gabapentin’s use for treating neurodegenera-
tive diseases—a fact the two parties agreed about—Warner-
Lambert could not win an infringement suit against Tor-
Pharm, which sought to sell its drug as a treatment for
epilepsy. See 

(‘‘Because [TorPharm] is not sub-
mitting an application to sell a drug for treatment of neurode-
generative diseases, which is the only use covered by the
patent involved in this case, we conclude that [TorPharm] is
entitled to summary judgment of noninfringement.’’ (empha-
sis added)). According to TorPharm, the FDA’s use of
Warner-Lambert’s failed infringement lawsuit as a basis for
delisting the ’479 patent—and, more important, for denying
exclusivity—had the effect of eviscerating the incentive struc-
ture established by the FDCA. Pointing out that the Act
created the exclusivity period precisely to encourage compa-
nies to have infringement lawsuits brought against them,
TorPharm argues that it is absurd to tell a company which
won such a lawsuit that because its victory established the
challenged patent’s invalidity and hence that the patent never
needed to be challenged, the company was entitled to no
exclusivity.
   The district court agreed, deeming the FDA’s reliance on
the Federal Circuit’s decision ‘‘problematic.’’ 

n.15. According to the court, however, the
other two bases on which the FDA rested its conclusion—the
court’s ruling in Purepac and Pfizer’s letter to the FDA
confirming that ‘‘neither Pfizer nor Warner-Lambert ever
represented to FDA that the ’479 patent claimed an approved
use’’—were enough to support that conclusion. Because
these other bases eliminated or greatly ameliorated the dan-
ger of upsetting the FDCA’s incentive structure, the district
court concluded, the FDA properly delisted the patent. See

                              17

   Disagreeing with the district court, TorPharm argues that
reliance on either Purepac or Pfizer’s letter to the FDA
would still risk undermining the FDCA’s incentive structure
because both were tainted by Warner-Lambert’s lawsuit
against TorPharm. Although acknowledging, as it must, that
Purepac and the letter predated the Federal Circuit’s deci-
sion, TorPharm points out that they did not predate the
Illinois district court decision that the Federal Circuit af-
firmed. In that decision, the district court noted that the ’479
patent covered gabapentin’s use for treating neurodegenera-
tive diseases. See Warner-Lambert Co., 

, at
*1. TorPharm insists that both Purepac and Pfizer’s letter
were influenced by that ruling, after which the district court
here ‘‘had no choice but to conclude that the ’479 patent
claims an unapproved use.’’ Appellant’s Reply at 16.
   TorPharm misreads the TorPharm decision. The district
court in that case did not conclude that the ’479 patent
claimed an unapproved use. Rather, the court found that
given the evidence the FDA had before it at the time it made
its decision, it was arbitrary and capricious for the agency to
have found that Warner-Lambert claimed the ’479 patent
covered gabapentin’s use for treating epilepsy. Nothing in
the Illinois district court’s decision compelled that holding;
rather, the evidence before the FDA and the agency’s earlier
choice of use code compelled it. Had the evidence before the
FDA been different, or had the agency not given the patent a
use code corresponding to neurodegenerative diseases, it
would have been entirely possible for the district court here
to sustain the FDA’s initial decision despite the Illinois dis-
trict court’s ruling. Put another way, the two district courts
answered different questions: the Illinois district court con-
sidered whether Warner-Lambert could win an infringement
lawsuit against TorPharm given that the ’479 patent covered
neurodegenerative diseases and given that TorPharm wanted
to sell an epilepsy drug, while the district court here consid-
ered whether the FDA’s conclusion about what Warner-
Lambert claimed the ’479 patent covered ran counter to the
evidence before the agency.
                               18

   TorPharm also contends that the FDA could not rely on
Purepac because the case was wrongly decided. This makes
no sense. Not only did the FDA have no authority to ignore
Purepac, but as we have explained, the decision is unassaila-
ble. See supra pages 11–14. Because Purepac provided an
adequate basis for the FDA’s post-remand decision to delist
the ’479 patent, the agency’s reference to the Federal Cir-
cuit’s ruling, as the district court concluded, did not fatally
undermine its decision.
   TorPharm’s view of this issue—and, to a certain extent, of
this entire case—rests on its belief that ‘‘TorPharm played by
the rules, [while] Purepac didn’t.’’ Appellant’s Br. at 21. It
‘‘played by the rules,’’ it says, because it filed a paragraph IV
certification about the ’479 patent, thereby inviting an in-
fringement lawsuit, whereas Purepac filed a section viii state-
ment and then sued the FDA when the agency rejected that
statement. But no ‘‘rules’’ required Purepac (or anyone else)
to accept everything the FDA did, regardless of the lawful-
ness of the agency’s actions. Put simply, Purepac acted quite
properly: it (1) interpreted the ’479 patent as covering neuro-
degenerative diseases, as has every court to have looked at
the issue, the patent holder, and (eventually) the FDA; (2)
accordingly filed a section viii statement; and (3) refused to
accept the FDA’s unlawful rejection of that statement, a
refusal that subsequent events have vindicated. The fact that
TorPharm chose a different—and ultimately unsuccessful—
legal strategy means neither that Purepac flouted the rules
nor that TorPharm deserved exclusivity. Indeed, one might
well think that the equities, which TorPharm frequently
invokes, actually favor Purepac, the party that pursued the
correct legal strategy even though that strategy required it to
take on the heavy burden of charging the FDA with arbitrary
and capricious behavior.
   For its final argument, TorPharm claims that the FDA’s
actions following the Purepac remand reveal an inconsistency
with the agency’s actions regarding the drug mirtazapine.
Having nothing to add to the district court’s sound reasons
for rejecting this argument, see 

–86, we will affirm.
                               19

                             IV.
  This brings us finally to the ’482 patent. Recall that the
FDA awarded exclusivity to Purepac even though Purepac
delayed sending notice to Warner-Lambert until June 13,
2000, eighteen days after the FDA received Purepac’s para-
graph IV certification, whereas TorPharm mailed both its
certification to the FDA and its notice to Warner-Lambert on
the same day—June 13, 2000—the day Purepac finally sent
notice. TorPharm argues that the district court improperly
sustained this award of exclusivity. It also contends that the
court erroneously upheld the FDA’s decision to use a receipt
date rather than a mailing date in deciding when TorPharm
had submitted its amended ANDA.
   As to the first argument, TorPharm insists that the FDA
should have declared that Purepac’s delay in providing notice
rendered the company’s certification invalid. Had the FDA
done so, TorPharm would have been entitled to exclusivity, as
it would have been the first applicant to have submitted an
ANDA containing a valid certification. Instead, the FDA
ruled that if an applicant fails to provide notice at the same
time that it files its amended ANDA, the certification be-
comes effective only when the applicant ultimately provides
notice, rather than when the applicant files its amended
ANDA. Under this rule, applicants who do not immediately
give notice run the risk that another company will file a
certification and provide notice first, thereby winning the
right to exclusivity. By failing to give immediate notice,
moreover, such applicants delay FDA approval of their AN-
DAs, meaning that they must wait longer before they can
market their drugs. According to TorPharm, the FDA’s
decision to impose this penalty, instead of declaring Purepac’s
certification invalid for exclusivity purposes, effectively reads
the requirement of simultaneous notice and filing out of both
the statute and the regulation. We disagree.
   As the district court pointed out, the statute (as well as the
regulation, we note) ‘‘is in fact silent on the issue of what
follows from an applicant’s failure to follow the mandate of
simultaneity.’’ 

. Put anoth-
                              20

er way, the FDA recognized that Congress could have added
either of two equally plausible sentences to the statute: one
would have said ‘‘failure to provide simultaneous notice shall
render the paragraph IV certification invalid,’’ while the other
would have said ‘‘failure to provide simultaneous notice shall
delay the certification’s effective date until notice is provid-
ed.’’ That TorPharm would have preferred the FDA to fill
this gap by adopting the first approach hardly makes the
agency’s decision to adopt the second arbitrary or capricious.
The district court correctly and appropriately noted this
court’s longstanding recognition that ‘‘the breadth of agency
discretion is, if anything, at zenith when the action assailed
relates primarily not to the issue of ascertaining whether
conduct violates the statute, or regulations, but rather to the
fashioning of policies, remedies and sanctions.’’ Niagara
Mohawk Power Corp. v. Fed. Power Comm’n, 

,
159 (D.C. Cir. 1967), quoted in 

.
  TorPharm insists that the FDA’s choice of sanction violates
the FDCA because the statute links the simultaneity require-
ment to the award of exclusivity. ‘‘Certification, notice, and
exclusivity TTT are all bound up together in the statutory
scheme.’’ Appellant’s Reply at 22. Not so. Nothing in the
statute says that applicants earn exclusivity by simultaneous-
ly filing and providing notice. In fact, the simultaneity
requirement and the provisions regarding exclusivity appear
in different provisions of the statute. The simultaneity re-
quirement appears in FDCA section 505(j)(2), which lays out
the required elements of an ANDA. See 21 U.S.C.
§ 355(j)(2)(B)(iii). The exclusivity provisions are in FDCA
section 505(j)(5), which addresses FDA approval of ANDAs.
See 

The district court thus properly
concluded that the FDA’s choice of penalty permissibly filled
a statutory gap.
   As to the second issue, TorPharm argues that the FDA’s
decision to use mailing dates rather than receipt dates is
inconsistent with the agency’s use of mailing dates for materi-
als submitted to its Dockets Management Branch. The FDA
persuasively points out, however, that ‘‘materials relating to
                               21

NDAs and ANDAs are submitted to the agency’s Center for
Drug Evaluation and Research TTT , and [these] have consis-
tently been governed by the ‘date of receipt’ rule.’’ FDA Br.
at 30 n.7.
   Equally without merit is TorPharm’s assertion that the use
of receipt dates is arbitrary and capricious. TorPharm, the
district court noted, ‘‘points to nothing in the statute that
precludes the FDA’s date-of-receipt rule, or that mandates an
alternative mailbox rule.’’ 

.
Although it is true that ‘‘[n]either the statute nor the regula-
tion—neither of which even mentions ‘receipt’ dates—re-
quires’’ the FDA’s approach, Appellant’s Br. at 48 (citations
omitted), that (again) misses the point. The question is
whether either the statute or the regulation precludes the
FDA’s approach. Neither does. Each uses the verb ‘‘sub-
mitted,’’ which the FDA reasonably interprets to mean ‘‘re-
ceived’’ rather than ‘‘mailed.’’ Moreover, the FDA’s rationale
for using receipt dates—to avoid ambiguity that might arise
with mailing dates because of, for example, a conflict between
the date on the ANDA itself and the date of the postmark—
seems perfectly reasonable to us. We thus agree with the
district court that TorPharm has offered no basis for over-
turning the FDA’s reliance on receipt dates.
   TorPharm argues that the FDA’s use of receipt dates gave
Purepac ‘‘an unfair advantage’’ and that upholding the agen-
cy’s approach creates an incentive ‘‘to game the system by
certifying first.’’ 

TorPharm is right about the
incentive. An applicant that completes its amended ANDA
before completing its notice—as often happens because certi-
fications are simple one-page documents whereas notices are
typically quite long—can avoid a lag in the effective date of
its ANDA by sending the ANDA immediately rather than
waiting until it completes the notice. By not waiting, the
applicant increases the chance that the FDA will have re-
ceived the ANDA (including the paragraph IV certification)
when the applicant notifies, in which case the certification will
become effective as soon as the applicant mails the notice.
(The FDA relies on mailing dates for notices because appli-
cants send their notices to the patent holder rather than to
                               22

the FDA.) If the applicant instead waits to send its ANDA
until it completes its notice and can mail that as well, then the
certification will only become effective a day or more later,
when the FDA receives the ANDA in the mail. For this
reason, the FDA’s approach gives applicants an incentive to
disregard the statutory and regulatory mandates to provide
notice ‘‘when’’ and ‘‘at the same time’’ that they file their
amended ANDAs.
   Contrary to TorPharm’s argument, however, the existence
of this incentive does not mean that the FDA’s approach is
unlawful. As we explained, the FDA imposes a penalty on
those who notify after they filed their amended ANDAs, i.e.,
certifications become effective only upon notification. See
supra page 19. This penalty creates an incentive for compa-
nies to notify patent holders as soon as possible after filing
their ANDAs. In other words, the FDA’s penalty acts as a
counterweight to the incentive that TorPharm describes, en-
couraging companies to file ANDAs and provide notice ‘‘at
the same time.’’ Given the scope of the FDA’s discretion and
the benefits that the agency’s approach provides, any incen-
tive to delay that companies might have despite this counter-
weight provides no basis for invalidating the FDA’s approach.
  As to the contention that the FDA’s reliance on receipt
date creates an ‘‘unfair advantage,’’ TorPharm’s counsel ac-
knowledged at oral argument that the advantage existed only
as between TorPharm and Purepac, and that with the FDA’s
approach now clarified, every company will have the same
incentives and opportunities. Purepac’s ‘‘advantage,’’ which
derived simply from the fact that the FDA agreed with
Purepac’s interpretation of the statute rather than with Tor-
Pharm’s, hardly warrants setting aside the FDA’s decision.

                               V.
  The district court’s judgments are affirmed in all respects.
                                                     So ordered.