Research Policy 52 (2023) 104791

Contents lists available at ScienceDirect

Research Policy
journal homepage: www.elsevier.com/locate/respol

Short communication

The NBER Orange Book Dataset: A user’s guideI

Maya Durvasula d ,<,1 , C. Scott Hemphill a ,1 , Lisa Larrimore Ouellette b ,1 , Bhaven Sampat c ,1 ,
Heidi L. Williams d ,1
a NYU School of Law, United States of America
b Stanford Law School, United States of America
c
Department of Health Policy and Management, Columbia University, United States of America
d
Department of Economics, Stanford University, United States of America

ARTICLE INFO ABSTRACT

Dataset link: NBER Orange Book Dataset (Origi This paper introduces a newly digitized, open-access version of the Food and Drug Administration’s ‘‘Orange
nal data) Book’’—a linkage between approved small-molecule drugs and the patents that protect them. The Orange
Book also reports any applicable regulatory exclusivity that prevents competitive entry. We summarize the
Keywords:
FDA Orange Book’s coverage and discuss the opportunities and challenges associated with using these data for
Orange Book research. Empirical validations against various administrative datasets suggest that Orange Book records are,
Drug patents largely, complete and accurate. We conclude with a specific use case—calculating legal exclusivity periods for
Intellectual property drugs—to highlight the types of choices that researchers must make when using this resource.

1. Introduction
a public document called the ‘‘Orange Book’’.4 The Orange Book thus
provides the tight product-patent linkage that is normally missing—
Patents are widely used as measures of innovation. However, the
enabling researchers to investigate, for example, the spillovers of public
connection between a particular commercialized product and a par-
funding for research, changes in the effective exclusivity for brand-
ticular patent is typically unclear. Indeed, a given product might lack
name pharmaceuticals over time, the interplay between patent ex-
any patent protection, and a given patent might be associated with no
amination and litigation, the behavior of patent examiners, and the
product. Or more commonly, a product might fall within the scope of landscape of drug patents in other countries (Azoulay et al., 2019;
tens or even thousands of patents, and a patent might be associated Hemphill and Sampat, 2012; Lietzan and Lybecker, 2020; Lemley and
with many commercialized products. In general, there is no standard Tu, 2022; Frakes and Wasserman, 2022; Sampat and Shadlen, 2015).
or straightforward way to link products and patents.2 Each edition of the Orange Book provides a snapshot of unexpired
Pharmaceuticals are different. For any brand-name small-molecule patent protection at a moment in time. As patents on a drug expire and
drug marketed in the United States, the drug maker is generally re- new patents are issued, these changes are reflected in later editions.
quired to report applicable patent protection to the Food and Drug The Orange Book also provides a snapshot of unexpired regulatory ex-
Administration (FDA).3 The FDA makes this information available in clusivity granted by the FDA. For example, certain novel drugs receive

I We thank AnneMarie Bryson, Antara Joardar, Casey Lincoln, Tamri Matiashvili, and Maya Roy for excellent research assistance. We gratefully acknowledge
financial support from the National Institute on Aging, United States through grant P01AG005842 to the National Bureau of Economic Research, United States,
the National Science Foundation, United States through grant DGE-1656518, the Knight Hennessy Scholars Program, the Alfred P. Sloan Foundation, United
States, and the Smith Richardson Foundation, United States. Hemphill has at various times served as an expert witness in litigation involving Orange Book-listed
patents. Sampat has served as a consultant in litigation on issues relating to pharmaceutical patenting and innovation, and to the Medicines Patent Pool for work
on patents and access to medicines.
< Correspondence to: Department of Economics, Stanford University, 579 Jane Stanford Way, Stanford, CA 94305, United States of America.
E-mail address: mdurvas@stanford.edu (M. Durvasula).
1
All authors contributed to this work equally.
2
Some producers disclose a linkage by marking their product with the patent number or with the URL of a web page containing the relevant details. For
recent work using online disclosure as a source of linkages, see de Rassenfosse (2018).
3
Small-molecule drugs are relatively simple chemical compounds, which are derived by chemical synthesis. Biologic drugs, in contrast, are typically
manufactured using living organisms. These two classes of drugs are regulated separately in the United States.
4
The official name is Approved Drug Products with Therapeutic Equivalence Evaluations, and it is available at https://www.fda.gov/drugs/drug-approvals-and-
databases/approved-drug-products-therapeutic-equivalence-evaluations-orange-book.

https://doi.org/10.1016/j.respol.2023.104791
Received 1 November 2022; Received in revised form 13 April 2023; Accepted 14 April 2023
Available online 24 May 2023
0048-7333/© 2023 Elsevier B.V. All rights reserved.
M. Durvasula et al. Research Policy 52 (2023) 104791

five years of regulatory exclusivity that blocks the entry of generic to file a list of these patents with the FDA. Beginning in 1985, this
competition, even in the absence of any patents. Combining multiple patent information—including patent numbers and expiration dates—
editions reveals a comprehensive picture of patent and regulatory was included as an addendum to the Orange Book.
protection as it evolves over a drug’s lifecycle. To enable historical analyses, we obtained versions of each annual
In this paper, we introduce a newly digitized, open-access dataset Orange Book publication from 1985 to 2016 via a Freedom of Infor-
of Orange Book records from 1985 to 2016, and provide a ‘‘user’s mation Act request.7 We hand-entered or parsed the patent tables from
guide’’ to its opportunities and challenges for researchers.5 The dataset each of these annual publications to create digital files.
combines information from print editions in the early years of the Researchers who would like to work with up-to-date versions of
Orange Book (1985–1999) with digital data in later years (2000–2016). the Orange Book can do so by retrieving text files of annual editions
The dataset includes both patent protection and regulatory exclusivity. from the FDA’s website. The FDA’s web archive allows individuals to
This linked data between products, patents, and exclusivity enables the access historical versions of the website, including links to past annual
measurement and analysis of important features of the biomedical inno- Orange Book files, beginning in 2016.8 For each additional year of data,
vation ecosystem. For example, the Orange Book provides researchers researchers can access the associated text file containing records of
examining the pharmaceutical industry with opportunities to assess the unexpired patents and regulatory exclusivities held on approved drugs.
scope of patent and regulatory protection on new products, as well as to When appended, these records are sufficient to extend the coverage of
describe the geographic and financial origins of innovative technology. the digitized Orange Book data described here through the end of the
In Section 2, we provide background information about the contents most recent calendar year (see Appendix A for details).
of the Orange Book and the construction of this dataset. In Section 3, we
offer an overview of the data files. In Section 4, we assess the complete- 3. An overview of the data files
ness and accuracy of the dataset by comparing it to various external
datasets. Section 5 presents an Orange Book ‘‘use case’’—comparing
different measures of market exclusivity conferred by legal protection Key Takeaways
with a measure of generic approval—and Section 6 concludes.

2. Background on the Orange Book • Firms are required to list patents that cover drug sub-
stances, products, and methods of use, and the FDA lists
additional non-patent regulatory exclusivity periods.
• The NBER Orange Book Dataset includes full listings of
Key Takeaways
all patents (FDA-drug-patents.dta) and full listings of all
exclusivities (FDA-drug-exclusivity.dta) as they appear in
• Since 1985, sponsors of brand-name small-molecule each edition.
drugs have been required to provide a list of nearly all
associated patents to the FDA. These lists are published
annually as the ‘‘Orange Book’’.
A final digitized version of our Orange Book patent and exclusivity
• The NBER Orange Book Dataset includes Orange Book
files is available from the National Bureau of Economic Research. Four
editions published between 1985 and 2016. Researchers
data files—in Stata .dta format—are available.
interested in up-to-date versions can retrieve new annual
The patent data file, FDA-drug-patents.dta, contains full listings of
editions from the FDA’s website.
all patents associated with each product as they appear in each edition
of the Orange Book. Observations include a variable that captures
the edition from which the record was drawn. Each observation also
The Orange Book began as a collection of information about the sub- includes the New Drug Application (NDA) number issued by the FDA,
stitutability of approved generic drugs for brand-name small-molecule one or more product ‘‘active ingredients’’ (chemical components that
drugs, following the repeal of anti-substitution laws in many states. The create pharmacological activity), and the ‘‘proprietary name’’ (product
first edition was published on 31 October 1980—hence the pumpkin- trade name). The full dataset includes 5511 unique patents associated
colored cover. In 1984, Congress passed the Drug Price Competition with 2173 distinct NDAs.
and Patent Term Restoration Act—typically referred to as the Hatch– FDA regulations require firms to list patents that cover drug sub-
Waxman Act—which established the modern regime for competition stances, drug products, and drug methods of use.9 Beginning in mid-
between brand-name and generic drugs. Among several other changes, 2003, firms were required to flag submitted patents to indicate whether
the Act provides a pathway for a generic firm to market a competing, they protect drug substances (active ingredients) or drug products (‘‘a
therapeutically equivalent version of a brand-name drug without com- finished dosage form, e.g., tablet, capsule, or solution that contains a
pleting its own extensive and costly clinical trials. Instead, the generic drug substance’’).10 Accordingly, the data file includes flags for drug
firm can file an Abbreviated New Drug Application (ANDA) with the substances (‘‘DS’’) and drug products (‘‘DP’’) beginning in the 2004
FDA, which requires a demonstration of ‘‘bioequivalence’’ to ensure edition. As for method-of-use patents, beginning with the 1998 edition,
that the generic product will be absorbed by the body at the same rate the Orange Book contains a flag (‘‘use-code’’) and detailed use codes
and to the same extent as the brand-name drug. describing the approved indication or use covered by the patent. The
The ANDA provisions facilitate rapid generic entry.6 To provide
potential generic entrants with notice of patents that may stand in
the way of entry, the Hatch–Waxman Act requires brand-name firms 7
No Orange Book was published in 1986.
8
The web archive is available here: https://www.fda.gov/about-fda/about-
website/fdagov-archive.
5
The datasets—and associated documentation—are available here: https:/ 9
NDA holders are supposed to list any patents that ‘‘could reasonably be
/www.nber.org/research/data/orange-book-patent-and-exclusivity-data-1985- asserted’’ against manufacturers of the drug. 21 U.S.C. § 355(b)(1)(A)(viii).
2016. NDA holders are likely to list patents they own or exclusively license, but not
6
As the other half of the of Hatch–Waxman bargain, the Act also provides patents owned by unrelated third parties.
partial patent term restoration to firms marketing brand-name drugs for 10
Drug substance and drug product flags are reported for patents listed
portions of their patent term that were lost on the way to FDA approval. See after 18 August 2003. See https://www.fda.gov/drugs/drug-approvals-and-
below for additional detail. databases/orange-book-data-files for details.

2
M. Durvasula et al. Research Policy 52 (2023) 104791

Table 1
Patent data validation exercises.
Validation exercise Sample Findings
Patent coverage
Comparison to IQVIA/Ark NMEs approved between 1995 and 2017 76% of IQVIA/Ark ‘‘constraining’’ patents are listed in OB
Comparison to ParagraphFour.com Cases resolved (adjudicated or settled) since 1 91% of litigated patents are listed in OB
November 2003
Comparison to Hatch–Waxman patent term Drugs with PTE granted as of 2012 96% of drugs had extended patents listed in OB
extensions (PTE)
Expiration date accuracy
Inspection of records affected by TRIPS Records in OB editions 1994–1995 and 1995–1996 Only one change in OB expiration date inconsistent with TRIPS
Inspection of patents with PTE Patents granted PTE as of 2016 89% of patents have identical expiration in OB and PTO records
Inspection of patents with PTA Patents granted PTA as of 2014 OB records are accurate for majority of cases; see Section 4.3
for details
Comparison to maintenance fee payment dataset Patents with maintenance fee data as of 2022 46% of OB patents expire early; OB does not capture early
expiration

use codes are defined in FDA-patent-use-codes.dta. These codes may be have received any legal protections, including patents or regulatory
helpful in determining whether certain classes of patents affect generic exclusivities. Of the 2458 NDAs approved between 1985 and 2015,12
entry differently (see Section 5.2). We caution, however, that to our 1682 (68%) have at least one patent, and 1724 (70%) have at least
knowledge neither compliance with listing requirements nor accuracy one regulatory exclusivity. 1958 NDAs (80%) include at least one
of firm-supplied flags has been validated. patent or exclusivity reported in the Orange Book. The prevalence of
The exclusivity data file, FDA-drug-exclusivity.dta, contains full list- legal protection is unsurprising, given the documented importance of
ings of all regulatory exclusivities associated with each product. The intellectual property for R&D and product innovation in pharmaceutical
observations are organized in a manner that is analogous to the patent markets (Cohen et al., 2000; Levin et al., 1987).
data file. Each type of exclusivity is associated with a distinct code. The The set of NDAs includes both new molecular entities (NMEs) and
codes are defined in FDA-drug-exclusivity-codes.dta. line extensions. An NME is a drug with a novel active ingredient. Line
The FDA’s Orange Book product listing database contains additional extensions include new dosage forms and formulations associated with
information not captured in our dataset.11 A notable example is the previously approved products.13 NMEs are particularly likely to have
Products file, which identifies, for every FDA-approved drug product, legal protection. For the 796 NMEs approved over the same period,
the existence and identity of therapeutically equivalent alternatives. 678 (85%) have at least one patent and 751 (94%) have at least one
regulatory exclusivity in the Orange Book. 761 (96%) include at least
4. How complete and accurate are the patent and exclusivity one patent or exclusivity.
records? The focus of the Orange Book is small-molecule drugs, which are
drugs with relatively simple chemical structures that can be well char-
acterized by scientists. Biologic drugs—more complex products that
Key Takeaways often lack fixed chemical structures, such as therapeutic proteins and
vaccines—are largely excluded. These drugs present a distinct set of
• Nearly all new small-molecule drugs approved since scientific and regulatory challenges and have a different pathway to
1985 have some form of patent or exclusivity listed. regulatory approval. Biologic drugs are approved under so-called bio-
• Certain patents are ineligible for listing (such as man- logics license applications (BLAs), not NDAs.14 Prior to 2020, certain
ufacturing processes), and the Orange Book does not biologic drugs were approved under NDAs, rather than BLAs, and thus
record information for most biological drugs. have listed patents and exclusivities in archival versions of the Orange
Book.15
• Nearly all ‘‘important’’ patents for small-molecule
For researchers seeking to investigate patent protections for bio-
drugs—patents likely to constrain generic entry—are
logic drugs, we are unaware of any publicly available data source
listed.
that would contain information comparable to the Orange Book. Since
• Patent expiration dates are generally accurate and re-
2014, biologic drugs and ‘‘biosimilars’’ (analogous to small-molecule
flect changes to term length. Records are not, how-
generics) have been reported in an FDA document called, in a nod to
ever, consistently updated to reflect expiration due to
maintenance fee non-payment.
12
We obtained a count of NDAs using the FDA’s Drugs@FDA database:
In this section, we provide an overview of the coverage of patent https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm.
and exclusivity records in the Orange Book for drugs approved in the 13
Raw data on NDAs and NMEs can be downloaded via the database
United States. We also describe several validation exercises, which linked in Footnote 12. A frequently updated compilation of NMEs can also
are intended to assess the completeness and accuracy of Orange Book be downloaded from the FDA: https://www.fda.gov/drugs/drug-approvals-
patent and exclusivity records. Table 1 summarizes these validation and-databases/compilation-cder-new-molecular-entity-nme-drug-and-new-
biologic-approvals.
exercises. 14
Between 1985 and 2015, 85% of all NMEs were approved under NDAs.
Between 1985 and 1990, 6.3% of NMEs were approved under BLAs, a share
4.1. Drug coverage that has grown over time, to 11% between 1995 and 2000, 18% between 2005
and 2010, and 26% between 2015 and 2020. Data on NMEs were drawn from
The NBER Orange Book Dataset includes records for all small- the database linked in Footnote 12.
molecule drug products approved by the FDA under an NDA that 15
In March 2020, some biologics that were initially approved as NDAs were
reclassified as BLAs. Within this set of drugs that were reclassified, we identify
14 NMEs approved before 2017 with at least one patent and 22 NMEs with
11
Documentation and data files are available for download: https://www. at least one regulatory exclusivity listed in the Orange Book. See details here:
fda.gov/drugs/drug-approvals-and-databases/orange-book-data-files. https://www.fda.gov/media/119229/download.

3
M. Durvasula et al. Research Policy 52 (2023) 104791

the Orange Book, the Purple Book.16 However, the Purple Book does An important takeaway is that all patents listed in the Orange Book
not include comprehensive listings of patents associated with biologic may not be equivalent in terms of their exclusionary force or their ef-
drugs. Recent legislation requires some patents to be listed in the Purple fects on competition. In some research contexts, distinguishing between
Book, somewhat improving transparency for this class of drugs.17 primary and secondary patent protection may be important (Hemphill
and Sampat, 2012; Frakes and Wasserman, 2022; Gupta, 2021; Kyle
4.2. Patent coverage et al., 2020). In other contexts, there may be patents that are important
to track for research purposes that are not listed (because of non-
A key question is which patents are, in fact, listed in the Orange compliance with reporting requirements or because the patents fall into
Book.18 The answer varies across drugs and reflects an interaction of categories that cannot or need not be listed, as described above).24
legal requirements and firm strategy. A firm marketing a brand-name Below we report the results of several validation exercises in an effort
drug is required to list any patent containing at least one claim that to evaluate the completeness of the Orange Book dataset.
covers the drug’s active ingredient, its formulation, or any method of
use pertaining to an approved indication. Eligible patents issued after 4.2.1. Comparison to IQVIA/Ark
drug approval must be listed within thirty days of issuance.19 Some Numerous firms manually construct records of the ‘‘landscape’’
types of patent are ineligible for listing, however, such as patents on of drug patents, as part of competitive intelligence and freedom-to-
manufacturing processes, packaging, metabolites, and intermediates.20 operate analyses. These efforts typically draw on the Orange Book and
Firms face strong incentives to list all eligible patents because a integrate supplemental expert searches. In addition to Orange Book
generic firm that seeks to market a competing product prior to patent data, some researchers have used data from the IQVIA/Ark Patent
expiration is required to ‘‘challenge’’ every listed patent. The challenge Intelligence database (formerly, IMS LifeCycle Patent Focus) to identify
takes the form of a legal certification that each patent is invalid or not patents on particular drugs, which can be particularly useful for track-
infringed by the proposed generic product. These so-called ‘‘Paragraph ing patents on drugs not covered by the Orange Book (e.g., biologic
IV’’ challenges are common, often result in patent litigation, and are a drugs not subject to Hatch–Waxman listing requirements), non-listable
major source of generic entry.21 Listing patents confers an additional patents, non-U.S. patents on drugs, and non-granted patent applica-
advantage in the event of litigation. If the brand-name firm elects tions. IQVIA/Ark also provides a range of value-added information
to file an infringement lawsuit within 45 days of receiving notice of about patents beyond what the Orange Book provides, including spe-
a Paragraph IV challenge, FDA approval of the generic firm’s ANDA cific information about what the patents cover (consistent with the
is automatically blocked (‘‘stayed’’) for 30 months while the suit is point made above that not all drug patents are created equal). Ark
considered by a court.22 Patent Intelligence also performs ‘‘constraint analysis’’ to determine, in
Notably, the FDA does not conduct audits of any kind to deter- its own view, ‘‘whether or not the claims in the patent will prevent a
mine either the completeness or correctness of Orange Book patent generic or biosimilar version of [a drug] from entering a commercial
listings; instead, they defer to firm self-reporting. Previous research market whilst enforced’’ (Ark Patent Intelligence, 2019). The firm
uses Orange Book records to suggest a rise over time in ‘‘secondary’’ bases its determination on data collected from ‘‘product labels, . . . press
patenting listings for pharmaceutical products—those patents covering releases and clinical trial[s]’’. Each patent is classified into one of three
not the drug’s active ingredient but rather other peripheral aspects of categories. A patent is constraining if it completely blocks the ‘‘launch
the molecule (Kapczynski et al., 2012; Hemphill and Sampat, 2011). [of a] generic equivalen[t]’’. A patent is not constraining if it completely
However, these secondary patents are more likely to be challenged and fails to do so. And a patent is partially constraining if ‘‘it is possible to
litigated, especially for high-sales drugs (Hemphill and Sampat, 2012), develop a generic equivalent but it cannot be marketed for . . . particular
and when litigated to completion, the generic challenger typically product attributes’’—for example, a particular indication.
prevails (Hemphill and Sampat, 2013).23 For another paper (Kyle et al., 2020), one of the authors (Sampat)
obtained IQVIA/Ark data for 550 NMEs approved between 1995 and
2017 that had at least one Orange Book patent. There are 6522 unique
16
The official name is Lists of Licensed Biological Products with Reference patents in IQVIA/Ark for these drugs, but only 2692 listed in the
Product Exclusivity and Biosimilarity or Interchangeability Evaluations. Orange Book (in any edition through 2021). Orange Book patents
17
See https://www.bioprocessonline.com/doc/patent-transparency-for- are almost always listed in IQVIA/Ark as well, which is reasonable
biologics-biosimilars-the-revamped-purple-book-0001 for details. as the Orange Book is one of the inputs for IQVIA/Ark. However,
18
Note that we do not attempt to identify any drugs that, if all patents
approximately three-fifths of the time, an IQVIA/Ark patent associated
were properly listed, should appear in the Orange Book. As such, we focus on
with a given drug appears nowhere in the Orange Book. These unlisted
assessing the completeness of patent listings for drugs with at least one listed
Orange Book patent.
patents appear to be disproportionately from patent categories that
19
See 21 U.S.C. § 355(c)(2). are explicitly not allowed in the Orange Book; for example, 93% of
20
See 21 C.F.R. § 314.53(b)(1). what IQVIA/Ark categorizes as ‘‘process’’ patents are explicitly ex-
21
The name comes from the relevant ‘‘Paragraph’’ of the Hatch–Waxman cluded. IQVIA/Ark may also include patents on unapproved uses for
statute. See 21 U.S.C. § 355(j)(2)(A)(vii)(IV). The FDA maintains an up-to-date the molecule, or formulations or molecular forms not on the market.
list of Paragraph IV challenges. See https://www.fda.gov/drugs/abbreviated- Of the IQVIA/Ark patents listed as constraining entry for a given
new-drug-application-anda/paragraph-iv-drug-product-applications-generic- drug, 75% are listed as protecting the same drug in the Orange Book.
drug-patent-challenge-notifications. For method-of-use patents, instead of By contrast, only 26% of non-constraining patents are linked to the
filing a Paragraph IV certification, a generic applicant may assert that the
associated drug in the Orange Book. Taken together, comparison with
patent does not cover any use for which the applicant seeks approval. See 21
the IQVIA/Ark data suggests that there are many patents associated
U.S.C. § 355(j)(2)(A)(viii).
22
See 21 U.S.C. § 355(j)(5)(B)(iii). However, an ANDA can be approved
within the 30-month stay period if, before the stay’s expiration, a court enters a
decision in favor of the generic entrant on the question of patent infringement file a counterclaim in litigation to require a brand-name drug manufacturer
or validity. Additionally, the length of the stay may be altered by the court if to correct Orange Book patent ‘‘use codes’’ that inaccurately describe the
either party fails to cooperate in advancing litigation. coverage of a particular patent. A list of Orange Book patent listing disputes
23
In response to concern about improper Orange Book listings, the FDA is available here: https://www.fda.gov/drugs/drug-approvals-and-databases/
has created a process to challenge the accuracy of these patent listings. These orange-book-patent-listing-dispute-list.
changes were prompted in part by the Supreme Court’s holding (in Caraco 24
See this editorial on the consequences of unlisted process patents: https:
v. Novo Nordisk, 566 U.S. 399 (2012)) that a generic drug manufacturer may //patentlyo.com/patent/2006/05/editorial_the_o.html.

4
M. Durvasula et al. Research Policy 52 (2023) 104791

with FDA-approved drugs that are not listed in the Orange Book. 4.3. Patent expiration
However, the vast majority of patents that may be viewed as mattering
for competition—in the sense of constraining generic entry—appear in For each record, the Orange Book provides information about
the Orange Book. product-patent linkages and the expiration dates of all linked patents.
To assess accuracy, we investigate whether legally mandated changes
4.2.2. Litigation records in the expiration date of certain sets of patents are reflected in the data
Next, we investigate the extent to which pharmaceutical patents file. We examine the following four sources of change:
involved in litigation are reported in the Orange Book. In this con-
text, a similar picture emerges. As noted above, a common form of 1. Agreement on Trade-Related Aspects of Intellectual Property
litigation associated with pharmaceutical patents involves Paragraph Rights (TRIPS): To comply with the 1994 TRIPS Agreement, the
IV challenges, in which a generic firm challenges active Orange Book- standard U.S. patent term was shifted from a 17 year term that
listed patents held by brand-name firms in the process of preparing began upon patent issuance to a 20 year term than begins at
to enter with a competing product. A brand-name firm need not limit patent filing. This change took effect in June 1995. For patents
its litigation strategy to listed patents, however. For example, it might filed before 08 June 1995—including previously issued patents
assert infringement of a manufacturing process patent, which is not that were in force at the time of this change—patent term was set
eligible for listing. to be the later of the two potential expiration dates. For patents
To examine the prevalence of Orange Book-listed patents in brand- that had spent less than three years under review at the USPTO
generic litigation, we draw on records from ParagraphFour.com, which at the time of this change, TRIPS increased their patent terms. As
compiles information about both Paragraph IV and other patent dis- a result, the maximum increase in a patent’s term due to TRIPS
putes. Our data cover cases resolved (adjudicated or settled) since 1 is three years.
November 2003.25 There were more than 700 drugs associated with 2. Patent Term Extension (PTE): As noted above, PTE allows firms
these cases, and 2286 unique litigated patents. Of the litigated patents, to restore up to five years of patent protection for time that
214 (9.4%) were not listed in the Orange Book (in any edition through a product has spent during clinical trials and under regulatory
2021). Overall, these data suggest that for brand-generic disputes that review. In addition, patents that will expire before product
are litigated via the Paragraph IV channel, unlisted patents play a approval, or that are under review for PTE but may expire
relatively minor role. during the PTE determination process, are eligible to apply for
an interim extension.30 Interim PTE may be followed by full PTE,
4.2.3. Extended patents under Hatch–Waxman though this is not necessarily the case. If the patent does receive
The Hatch–Waxman Act allows a drug sponsor to delay patent full PTE, interim PTE is counted toward this extension.
expiration on a drug containing a new active ingredient. Such patent 3. Patent Term Adjustment (PTA): Since May 2000, patents have
term extension (PTE) permits the sponsor to recoup some of the time been eligible to receive a Patent Term Adjustment (PTA), which
lost during clinical trials and the FDA review process.26 A firm may seek extends patent term to account for delays caused by the USPTO
PTE for one patent per approved drug.27 A firm can receive a PTE of during patent examination.31
up to five years, and the remaining patent term after this extension 4. Maintenance Fee Non-Payment: Patent owners are required to
has been applied cannot exceed 14 years beyond the date of FDA pay fees to the USPTO to maintain their patent rights at 3.5-, 7.5-
approval. Extension time is calculated based on time spent in regulatory , and 11.5-year intervals after patent issuance.32 Non-payment
review plus half of the time spent in clinical trials, up to the five- results in termination of patent rights.
year maximum.28 Patents eligible for PTE must be in force when the
TRIPS. To determine whether changes associated with TRIPS are re-
drug is approved by the FDA and must cover the product or substance,
flected in Orange Book records, we examine the two Orange Book
a method of using the product, or a method of manufacturing the
editions in which this change is likely to be most pronounced: 1995 and
product.
1996, which reflect listings from the years 1994 and 1995 respectively.
In principle, a brand-name drug maker might choose to extend a
If the TRIPS changes impacted Orange Book listings, we would expect
patent not listed in the Orange Book, such as a manufacturing process
that patent expiration dates should shift forward in 1996, relative to
patent. Eisenberg (2012) notes that, in theory, this choice presents a
the 1995 expiration dates, by no more than three years. Of the 415
strategic dilemma for a firm. A patent on a method of manufacture
patents that were listed in the Orange Book in both 1995 and 1996, 159
might be filed later than (say) an active ingredient patent, and thus con-
(38%) changed their listed expiration dates between 1995 and 1996.
fer more years of patent protection. On the other hand, the claims may
154 patents reported a later expiration date in 1996 compared to 1995.
be weaker and less likely to impede generic entry. In practice, active
Among these 154 extensions, only 1 patent reported an expiration
ingredient patents appear to be a primary target of PTE. Hemphill and
extension that was greater than three years. The remaining expiration
Sampat (2012) find that for a sample of drugs that experienced generic
extensions are less than three years and, thus, are consistent with the
entry between 2001 and 2010, 79% of the active ingredient patents
TRIPS adjustment to patent terms.
were extended, compared to 13% of non-active ingredient patents. We
We conduct the same comparison exercise with two other sets of
compare the universe of Orange Book patents to a list of granted PTEs
years to confirm that these changes between 1995 and 1996 are due
maintained by the USPTO, with grant dates listed through 2012. We
to TRIPS, rather than standard year-to-year modifications. We consider
examine the set of 397 drugs that received PTE and that have at least
1990–1991 and 1999–2000, five years before and after the 1994–1995
one listed Orange Book patent. For these 397 drugs, 382 (96%) elected
editions. In 1989–1990, of the 338 patents appearing in both editions,
to extend a patent that is listed in the Orange Book.29
only 12 (4%) had any change in expiration. In years 2000–2001, 55
(6%) of 849 patents had different expiration dates in the two editions.
25
Last accessed: 3 March 2023.
26
PTE is codified at 35 U.S.C. § 156. In Section 4.3, we describe validation
of the quality of Hatch–Waxman/PTE data using USPTO expiration records. Orange Book. Based on our review of the remaining extended patents, nearly
27
This general rule has a narrow exception: where a drug has separate NDAs all correspond to biologic drugs (including vaccines) and medical devices.
approved on a single day, the firm may be eligible to extend multiple patents. 30
See 35 U.S.C. § 156(e)(2) and § 156(d)(5).
28
The statute and corresponding regulations at 37 C.F.R. § 1.775 contain 31
See 35 U.S.C. § 154(b).
additional details. For example, the extension may be reduced if the applicant 32
There is a six-month grace period after each deadline, during which the
did not act with due diligence during the regulatory review period. patent owner must pay a surcharge in addition to the maintenance fee. Data
29
Of the 803 unique trade names associated with extensions, 397 correspond on maintenance fee payment are drawn from USPTO records downloaded from
to Orange Book-listed drugs, and of the 809 patents extended, 510 are in the https://bulkdata.uspto.gov/data/patent/maintenancefee/MaintFeeEvents.zip.

5
M. Durvasula et al. Research Policy 52 (2023) 104791

Table 2
Discrepancies between USPTO patent term extension and Orange Book records.
Issue Affected patents Is OB correct?
OB does not report PTE 3998790; 4085225; 4154839; 4670444a ; 4689320; NO
4808605; 4898724; 4911920; 5008256; 5180668;
5234404; 5494903; 5510353; 5639443; 5418226; 5843901
OB miscalculates TRIPS extension 4284647; 4244946 NO
Miscellaneous OB errors 4105783a ; 4234571a ; 4567264a ; 4585597a ; 4687659 a ; NO
4845075a ; 4990517a ; 5387603a ; 5446194a
USPTO miscalculates TRIPS extension 3962432; 3927046; 3985758; 4001323; 4014986; 4056635 YES
USPTO reports only interim PTE, not full PTE 4977138; 5254556; 5407914 YES
Patent included on USPTO list, but no listed extended 4695578; 4762856; 5196444 YES
expiration date
Terminal disclaimer omitted from PTE term 4886812a ; 5656667a YES
calculation
USPTO does not reflect a Certificate of Correction that 4695590a YES
amended the term extension
Miscellaneous USPTO errors 3857952a ; 4661491a YES
Drug approved under a BLA and subsequently 5681818a N/A
reclassified

Notes: Table reports discrepancies between patent expiration dates reported in USPTO records (USPTO Patent Center and Hatch–Waxman
extension list) and dates listed in the Orange Book. Hatch–Waxman extensions are drawn from an 18 January 2016 copy of the patent term
extension list downloaded from the USPTO website (obtained using Internet Archive’s Wayback Machine). For each patent appearing in multiple
editions of the Orange Book, we use the latest applicable edition. For two patents (italicized), both OB and USPTO records are incorrect.
a
Denotes patents for which we report additional details on discrepancies in Appendix B.1.

Patent term extension. To determine whether the Orange Book reflects PTE, including those records that received both PTE and PTA. Any
PTE and interim PTE, we compare Orange Book records to a list of all discrepancies associated with this set of patents are investigated in
patents that received PTE between 1984 and 2016.33 Table 2.
Of the 4027 patents listed in the Orange Book from 1985 to 2016, We observe 943 Orange Book patents with reported PTA, which
402 (10%) appear in the PTE data and have non-missing patent ex- did not also receive PTE. Of these 943 patents, 366 (38%) had correct
piration dates. For these patents, we compare listed patent expiration expiration dates (expiration dates that appropriately reflected PTA)
dates across the Orange Book and the USPTO Patent Center database. in the Orange Book. In 144 cases (15%), the discrepancy between
357 (89%) have identical expiration dates in the Orange Book patents USPTO records and Orange Book records is exactly equal to PTA length,
dataset and the USPTO PTE data. For the remaining patents, we in- suggesting that the Orange Book was not updated to reflect PTA.
vestigate the source of the discrepancy. For 17 (4%), the Orange Book For the remaining 46% of observations (N = 434)—for which
expiration date is correct. The remaining 28 (7%) reflect a variety of USPTO and Orange Book expiration dates differed by more or less
Orange Book errors.34 These results are reported in Table 2, and further than the duration of PTA—we hand-checked a random 10% sample
details are available in Appendix B. Although researchers should take (N = 43). Within this set, Orange Book expiration dates were correct
note of cases in which Orange Book listings are inaccurate, our findings for 30 (70%). Discrepancies across the two datasets were explained by
suggest that Orange Book expiration dates affected by PTE are correct differences in the calculation of patent expiration dates and inconsis-
in 93% of cases. tent reporting of maintenance fee non-payment, terminal disclaimers,
and Certificates of Correction amending earlier decisions about PTA.36
Patent term adjustment. Next, we collect records of PTA from the Table 3 and Appendix B.2 provide details on these cases. As with PTE,
USPTO Historical Masterfile, which contains information on patents our findings suggest that the Orange Book reports accurate expiration
issued through 2014.35 dates for the majority of records. Appendix B summarizes our process
We focus on the set of patents that received only PTA. The pre- for determining correct expiration dates for patents and provides sug-
vious subsection assessed the accuracy of all patents that received gestions for researchers who may be interested in doing so for a larger
set of observations.
Finally, we examine PTA recalculation. Due to a USPTO error in
33
We accessed a version of this list current as of 18 January 2016 interpreting the relevant section of the federal code, a narrow set of
using the Internet Archive’s Wayback Machine, here: https://web. patents became eligible for a reconsideration of the duration of PTA.37
archive.org/web/20160118153156/https://www.uspto.gov/patent/laws-and- Eligibility was limited to patents filed no earlier than 29 May 2000 and
regulations/patent-term-extension/patent-terms-extended-under-35-usc-156. issued no later than 2 March 2010. Patent applicants were given 180
Prior to the Hatch–Waxman Act, Congress extended individual patents on two days from the issuance of their patent or, if the patent had not yet been
products, Nutrasweet (aspartame) and Forane (an anesthetic) under 35 U.S.C.
issued, two months from the initial determination of PTA to request
§ 155 and 35 U.S.C. § 155A. See Patent Terms Extended Under 35 U.S.C.
PTA recalculation. However, because this notice was issued in January
§ 155; 35 U.S.C. § 155A, https://web.archive.org/web/20170628065756/http:
//www.uspto.gov/patents/resources/terms/155.jsp (listing patents for these
2010, only patents issued between August 2009 and March 2010 were
two products). These products do not appear in our sample. Repeal of these realistically eligible for this recalculation. Petitions associated with
statutory provisions became effective in 2012. these recalculations are flagged in the USPTO’s Patent Center database
34
The errors include a failure to take account of PTE (16 patents), typo- with the event code ‘‘PTGR’’. We observe zero patents that received
graphical and similar clerical errors (7), miscalculation of TRIPS extensions
(2), and errors in expiration date calculation that are included in both the
Orange Book and USPTO data (2). Finally, one patent is associated with a 36
Terminal disclaimers are typically filed by inventors who hold patents
drug that was later reclassified under a BLA application; while early editions on several variants of the same technology. When later-filed patents may
of the Orange Book report accurate expiration dates, the drug is dropped from be obvious extensions of earlier-filed patents, inventors can ‘‘disclaim’’ (or
later editions and patent terms post-extension are never reported. dedicate to the public) any remaining patent term on the later-filed patent
35
The Historical Masterfile can be accessed here: https://developer.uspto. after the earlier-filed patent expires.
gov/product/historical-masterfile. Records of PTA are also available via the 37
See Wyeth v. Kappos, 591 F.3d 1364 (Fed. Cir. 2010); https://www.uspto.
USPTO’s Patent Center database: https://patentcenter.uspto.gov. gov/sites/default/files/patents/law/aipa/pta/wyeth_faqs_20100422.pdf.

6
M. Durvasula et al. Research Policy 52 (2023) 104791

Table 3
Discrepancies between USPTO patent term adjustments and Orange Book records.
Issue Affected patents Is OB correct?
OB does not reflect Certificate of Correction 8460704; 8470347; 8476010; 8545884; 8637512; NO
8664215; 8748573; 8802142
Maintenance fee non-payment 6923983; 7387793; 7405223; 7686786 NO
OB uses incorrect filing date to compute term 6669948 NO
USPTO calculates expiration using actual filing date; 6767901; 7700128; 7932268; 7964592; 8003673; YES
OB calculates expiration using effective filing date 8168616; 8193196; 8318802; 8324189; 8410131;
8415345; 8653061; 8746242; 8759350; 8790641;
8846650
Terminal disclaimer only recorded in OB 6960577; 7144861; 7320999; 7348362; 7384980; YES
7419973; 7452874; 7566705; 7683037; 7709444;
7816379; 7838027; 8672898; 8829017

Notes: Table reports sources of discrepancies between patent expiration dates reported in USPTO patent term adjustment data and dates listed
in the Orange Book. Patent term adjustment information is drawn from a 2014 copy of the USPTO’s Historical Masterfile, downloaded from the
USPTO website. Table reports sources of discrepancies for a randomly drawn 10% sample of all patents with different expiration dates across
the two datasets (N = 43). For three patents (italicized), USPTO records are also incorrect because expiration is calculated using actual instead
of effective filing date. Additional details are discussed in Appendix B.2.

PTA in our sample that also have the PTGR event code. We conclude In Section 5.1.1, we discuss these and other types of exclusivity in
that there are no Orange Book patents in our sample that received PTA greater detail.
that also received a PTA recalculation on account of this error. In contrast to Orange Book patent listings, which are self-reported
to the FDA by drug manufacturers, exclusivity periods are granted
Maintenance fee non-payment. We successfully link 4929 Orange Book
by the FDA itself and hence recorded directly by the agency. We
patents to USPTO records on maintenance fee payment. Of these
therefore expect Orange Book records of FDA-granted exclusivities to
patents, 2656 (54%) expire at full term (20 years), while 997 (20%)
be both complete and accurate. We nonetheless conduct two validation
expire at 12 years, 1212 (25%) expire at 8 years, and 64 (1%) expire
exercises.
at four years. Spot-checking of these records suggests that maintenance
First, we confirm that all drugs reported (separately) by the FDA to
fee non-payment is not reported in the Orange Book, as patents con-
have received ODE match Orange Book exclusivity records. Specifically,
tinue to appear despite having expired. This is unsurprising, given
that firms with expired patents face no incentive to continue reporting we make use of an FDA-compiled list of all drugs that have received an
changes to the FDA. Researchers interested in accurate measures of orphan drug designation.39 Given that the Orange Book focuses on FDA-
actual patent term should note, then, that patent expiration dates in approved drugs, we restrict this list to orphan designations for drugs
the Orange Book do not reflect this form of patent term truncation and that have actually been approved for marketing; we likewise restrict
that integrating additional USPTO records may be valuable. attention to drugs that received orphan drug designations before 2016,
as our Orange Book data is current through the end of 2015. For every
4.4. Exclusivity records such drug, we can identify a corresponding ODE award in the Orange
Book exclusivity file.
The second major component of the data file is a comprehensive Second, we similarly confirm that all drugs reported (separately)
set of FDA exclusivity records associated with each approved drug. by the FDA to have received pediatric exclusivity match Orange Book
Four are most important, in terms of their effectiveness in blocking exclusivity records.40 We observe pediatric exclusivity codes in the
competitive entry, particularly from generics: Orange Book for all drugs on the pediatric exclusivity list provided
separately by the FDA. Consistent with our expectations, product-level
1. New chemical entity (NCE) exclusivity: The FDA awards a five- Orange Book exclusivity records thus appear to be both complete and
year exclusivity period to brand-name drugs containing new ac- accurate.
tive ingredients. NCE protection takes the form of data exclusiv-
ity: a generic firm is prohibited from relying on the brand-name 5. Using the Orange Book: Calculating market exclusivity
firm’s clinical trial data.
2. Orphan drug exclusivity (ODE): The FDA awards a seven-year
exclusivity period upon the approval of a treatment for a rare
disease. ODE is an example of market exclusivity: a prospective Key Takeaways
entrant may not market the same drug, even if it conducts its
own clinical trials.38 • Patent and regulatory exclusivity addenda can be used
3. Pediatric extension: The FDA awards a six-month extension of to construct various measures of market exclusivity
existing exclusivities, including NCE and ODE, for conducting for small-molecule drugs. The appropriate strategy may
pediatric clinical trials. The extension also applies to unexpired depend on the research question at hand.
patents. • A case study of Eli Lilly’s Gemzar illustrates the subtleties
4. Generic exclusivity: Generics are eligible for their own form of of overlapping patents and regulatory exclusivities.
exclusivity. In particular, the first generic to file a Paragraph IV
patent challenge is potentially eligible for a 180 day period of
exclusivity that blocks the approval of other generics.
The Orange Book’s linkage between products and both patents and
exclusivities provides a unique opportunity to measure and analyze
38
See Thomas (2017) for a detailed discussion of the specific conditions
of each form of exclusivity. Thomas notes: ‘‘For many firms the distinction
between a data exclusivity and marketing exclusivity may be more apparent 39
Downloaded from https://www.accessdata.fda.gov/scripts/opdlisting/
than real. The expense of generating clinical data and other information oopd/listResult.cfm.
needed to obtain marketing approval from the FDA is prohibitive for many 40
Downloaded from https://www.fda.gov/drugs/development-resources/
firms’’. pediatric-exclusivity-granted.

7
M. Durvasula et al. Research Policy 52 (2023) 104791

Table 4
Common Orange Book linkages.
Dataset OB variable Linked variable Notes for use
Drugs@FDA NDA number NDA number Multiple distinct NDAs may be associated with
each OB patent, and multiple patents may be
associated with each NDA.
Medical Expenditure Panel Survey NDA number National Drug Code (NDC) A crosswalk from NDC numbers to NDA numbers
number and drug name is available from the FDA’s Structured Product
(brand-name, generic) Labeling system. Generic and brand-name merges
are often messy and may require manual
matching.
Paragraph IV challenges NDA number or drug NDA number or drug name
name
USPTO administrative records Patent number Patent number With considerable cleaning and manual matching,
it may be possible to link firm names in OB data
to USPTO assignee data.
Hatch–Waxman patent extensions Patent number Patent number
PatentsView Patent number Patent number
OECD Triadic Patent Families Patent number Patent number
Database

Notes: Table describes common linkages between the Orange Book and other datasets that capture aspects of biomedical innovation. Drugs@FDA data
are available at https://www.fda.gov/drugs/drug-approvals-and-databases/drugsfda-data-files. USPTO administrative records are available in many forms at
https://www.uspto.gov/ip-policy/economic-research/research-datasets. PatentsView data are available at https://patentsview.org/download/data-download-tables.
Waves of the Medical Expenditure Panel Survey are available at https://meps.ahrq.gov/data_stats/download_data_files.jsp. OECD Triadic Patent Families data
are available at https://data.oecd.org/rd/triadic-patent-families.htm. A list of Hatch–Waxman patent extensions maintained by the USPTO may be downloaded
from https://www.uspto.gov/patents/laws/patent-terms-extended. A list of Paragraph IV challenges maintained by the FDA may be downloaded from https:
//www.fda.gov/drugs/abbreviated-new-drug-application-anda/patent-certifications-and-suitability-petitions.

features of the biomedical innovation ecosystem. Table 4 describes Table 5

Orange Book exclusivity code categories.
common, useful linkages for researchers using Orange Book records,
Code Exclusivity type Duration Blocks generic entry
though this summary is neither exhaustive nor definitive.
For every small-molecule drug approved by the FDA after 1985, NCE New Chemical Entity Five years Yes
ODE Orphan Drug Exclusivity Seven years Yes
one can use regulatory exclusivity and patent addenda in the Orange PED Pediatric Exclusivity Six months Yes
Book to measure the effective exclusivity period.41 Here, we describe GAIN Generating Antibiotic Incentives Now Five years Yes
the patent and exclusivity data included in the Orange Book in more PC* Patent Challenge 180 days Yes
detail and note several features that need to be taken into account when CGT* Competitive Generic Therapy 180 days Yes

working with these records. We then provide one example illustrating D New Dosage Schedule Three years No
I New Indication Three years No
the ways in which researchers can make use of Orange Book resources
M Miscellaneous Three years No
to compute a drug’s period of market exclusivity. Finally, we investigate NC New Combination Three years No
more systematically how well various Orange Book measures capture a NDF New Dosage Form Three years No
drug’s experienced exclusivity period. NE New Ester/Salt Three years No
NP New Product Three years No
NPP New Patient Population Three years No
5.1. Additional details on Orange Book exclusivity and patent addenda NR New Route Three years No
NS New Strength Three years No

5.1.1. Exclusivity addendum

As noted above, small-molecule drugs approved by the FDA may
receive various forms of regulatory protection in addition to patent • Granted to drugs with new active ingredients.43
protection. We describe any FDA-assigned exclusivity that bars entry • Protects the active ingredient and runs concurrently with
of a generic competitor as a form of ‘‘generic-blocking exclusivity’’. other patents/exclusivities (with exceptions noted below).
Across editions of the Orange Book, a total of 1054 unique exclu- • Duration: five years.44
sivity codes appear in the exclusivity addendum, which are sorted into 2. Orphan drug exclusivity (ODE) under the 1983 Orphan Drug Act
sixteen categories by the FDA and summarized in Table 5. Of these and 21 C.F.R. § 316.31
sixteen categories, five confer generic-blocking exclusivity in the sense
defined above:42
43
The Ensuring Innovation Act of April 2021 amended the statutory
1. New chemical entity (NCE) exclusivity under the Hatch–Waxman
language from ‘‘active ingredient’’ to ‘‘active moiety’’, codifying the FDA’s
Act and 21 C.F.R. § 314.108(b)(2)
interpretation of ‘‘active ingredient’’. Historically, although statutory language
referred to ‘‘active ingredients’’, interpretations of this term differed. Com-
pounds in drugs may undergo chemical reactions in the body in order to
41
Note that the 1985 edition of the Orange Book contains only those patents induce a therapeutic effect. Alternatively, several drugs with the same central
and exclusivities that were in force at the time of publication. This implies that compound may have different compounds linked to them in ways that are not
Orange Book records will not indicate whether a particular drug was covered always clinically significant. ‘‘Active moiety’’ refers to the core molecule or
by a patent or exclusivity that expired before 1985. ion of a drug that is ‘‘responsible for the physiological or pharmacological
42
Table 5 describes a sixth category of generic-blocking exclusivity: com- action of a drug’s substance’’. Additional details are available here: https:
petitive generic therapy (CGT). The FDA Reauthorization Act of 2017 allows //www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfcfr/cfrsearch.cfm?fr=314.3.
the FDA to designate a drug with ‘‘inadequate generic competition’’ as a CGT. 44
A prospective generic entrant is allowed to file an ANDA with a Paragraph
Subsequent generic entrants may be eligible for expedited FDA review, as well IV certification after four years. If the challenge is resolved quickly, generic
as a 180 day exclusivity period upon ANDA approval. entry prior to the 5 year mark is possible.

8
M. Durvasula et al. Research Policy 52 (2023) 104791

Table 6
NDA #021546 (Rebetol/Ribavirin), PED extensions.
Application number Exclusivity type Exclusivity expiration Linked PED extension Corrected expiration
021546 NDF 29-Jul-2006 yes 29-Jan-2007
021546 PED 29-Jan-2007 – –
021546 ODE 29-Jul-2010 yes 29-Jan-2011
021546 PED 29-Jan-2011 – –

Notes: Table reports pediatric exclusivity extensions associated with NDA #021546 (Rebetol/Ribavirin). Columns 1–3 report
application numbers, exclusivity types, and exclusivity expiration dates for Rebetol, as they appear in the Orange Book
exclusivity addendum. Column 4 flags whether a particular non-PED exclusivity has an associated pediatric extension. Column
5 lists the corrected expiration date for the non-PED extension, which matches that listed for the linked PED extension in the
Orange Book.

• Granted to drugs that either (i) treat diseases for which the Researchers wishing to calculate a drug’s ‘‘legal protection period’’
relevant U.S. patient population is under 200,000 or (ii) empirically should be attentive to several details. First, not all regu-
involve research and development expenses that the firm latory exclusivities apply to every product within an FDA application.
does not reasonably expect to recover. For example, NCI exclusivities may apply to one product form of a drug
• Protects the drug when it is prescribed to treat a qualify- (i.e., a specific dosage or strength), but not others. Pediatric extensions,
ing disease or condition and runs concurrently with other by contrast, apply to every unexpired patent and exclusivity at the time
patents/exclusivities (with exceptions noted below). when the extension is approved, regardless of whether every product
• Duration: seven years. is involved in pediatric clinical trials.
Second, most (but not all) regulatory exclusivities run concurrently.
3. Pediatric exclusivity (PED) under Section 505A of the FDA Mod-
For example, if a firm receives a three-year NCI exclusivity and a seven-
ernization Act of 1997 and the Best Pharmaceuticals for Children year ODE exclusivity period on the same date, the drug receives a total
Act (BPCA) of seven years of protection. The exceptions are PED and GAIN, which
• Granted to drugs that have been used in pediatric studies, run consecutively to existing FDA-assigned exclusivities. If a drug has
following a written request by the FDA. no existing exclusivities, it receives no extensions.
• Extends existing exclusivity and unexpired patents. If a Table 6 provides an example of the data structure for Rebetol, an
drug has been granted NCE, PED is added as an extension. antiviral drug.46 Rebetol, in an oral solution dosage form, received FDA
If a drug has ODE, PED is added as an extension. approval on 29 July 2003. The Orange Book lists an NCI exclusivity
• Duration: six months. set to expire three years later, on 29 July 2006,47 and ODE exclusivity
set to expire on 29 July 2010. Both received PED extensions, which
4. Generating Antibiotic Incentives Now (GAIN) exclusivity under extended these expiration dates by six months, to 29 January 2007 and
Title VIII of the FDA Safety and Innovation Act (FDASIA) 29 January 2011, respectively.
• Granted to products that have received a Qualified Infec- One can construct an expiration date that takes into account PED
and GAIN extensions by determining which other patents and regula-
tious Disease Product designation (with some exceptions).
tory exclusivities are in force at the time of the PED/GAIN exclusivity
• Extends existing exclusivity and does not extend any
approval.48
patents. If a drug has been granted NCE, GAIN is added
as an extension. If a drug has ODE, GAIN is added as an
extension. 5.1.2. Patent addendum
• Duration: five years. The Orange Book patent files list every patent record ever appearing
in any Orange Book edition between 1985 and 2016. One important
5. 180-Day Exclusivity (also known as Generic Drug Exclusivity) issue to be aware of is that patent expiration dates listed in the Orange
under 21 U.S.C. § 355(j)(5)(B)(iv) Book data are unique within but not across years. This implies that
• Granted to a first generic applicant submitting a Paragraph researchers using the historical Orange Book records data need to take
IV patent challenge and thereby risking an infringement care to choose the conceptually appropriate expiration date for their
application.
lawsuit.
• Protects against entry of other generic applicants that have
filed Paragraph IV challenges. 5.2. Calculating exclusivity: An example
• Duration: 180 days from commercial marketing of the
generic drug (subject to forfeiture if, for example, the Calculating market exclusivity can be complex, given the over-
generic applicant fails to market the drug within 75 days lapping effects of patents and regulatory exclusivity. As an illustra-
of a favorable court decision). tion, consider Gemzar, a treatment for pancreatic cancer approved by
the FDA in 1996.49 Gemzar was protected by multiple patents and
Note that we do not designate various ‘‘new clinical investigation’’ regulatory exclusivities, which we consider in turn.
(NCI) exclusivities as blocking generic entry. Although these forms
of protection provide three-year periods of exclusivity to incentivize
new clinical research on existing drugs (e.g., a new dosage schedule, 46
NDA #021546 (trade name ribavirin).
clinical indication, or patient population),45 generic competitors may 47
In particular, a new dosage form.
be approved for any uses not covered by the exclusivity. In practice,
48
Note that expiration dates associated with PED extensions may appear to
be several days longer or shorter than six months. Statistical packages assign
these restrictions leave drugs vulnerable to generic competition.
the number of days in a month differently, and researchers should correct for
any such differences. Throughout our analysis, we confirmed that PED and
GAIN extensions were accurately linked to the regulatory exclusivities that
45
The PED and GAIN extensions discussed above also apply to NCI they extended.
exclusivity. 49
See NDA #020509.

9
M. Durvasula et al. Research Policy 52 (2023) 104791

Patents. The data file contains two listed patents: 4,808,614 and
5,464,826. The ’614 patent, covering Gemzar’s active ingredient (gem- Table 7
Orange Book patents associated with Gemzar (NDA #020509)
citabine hydrochloride), was granted on 28 February 1989. The ’826
Patent Edition Expiration Use Code
patent, covering a method of use, was granted on 07 November 1995.
Both patents were originally set to expire 17 years from issuance: 4808614 1997 28-Feb-2006
1998 28-Feb-2006
28 February 2006 and 07 November 2012, respectively. As shown in
1999 28-Feb-2006
Table 7, both patents first appear in the 1997 edition of the Orange 2000 15-May-2010
Book with their original expiration dates. 2001 15-May-2010
Later, the ’614 patent received a PTE of 1537 days,50 resulting in a 2002 15-May-2010
2003 15-May-2010
new expiration date of 15 May 2010. Table 7 shows that this extension 2004 15-May-2010
first appeared in the 2000 edition of the Orange Book. Both patents 2005 15-Nov-2010a
subsequently received a six-month pediatric extension, resulting in 2006 15-Nov-2010a
2007 15-Nov-2010a
effective expiration dates of 15 November 2010 (for the ’614 patent)
2008 15-Nov-2010a
and 07 May 2013 (for the ’826 patent).51 2009 15-Nov-2010a
First generic entry occurred on 15 November 2010,52 upon the 2010 15-Nov-2010a
expiration of the ’614 patent. Notably, generic entry was not blocked 2011 15-Nov-2010a
by the later expiring ’826 patent. This example illustrates the point 5464826 1997 07-Nov-2012 U-146
made earlier that not all patents are created equal. In particular, the 1998 07-Nov-2012 U-146
1999 07-Nov-2012 U-146
latest-expiring patent frequently does not determine when a brand- 2000 07-Nov-2012 U-146
name firm’s effective exclusivity period ends, given the prospect that 2001 07-Nov-2012 U-146
a court may rule that it is invalid or not infringed. In general, in the 2002 07-Nov-2012 U-146
language of Lemley and Shapiro (2015), patents are ‘‘probabilistic’’: 2003 07-Nov-2012 U-146
2004 07-Nov-2012 U-146
the asserted property rights are uncertain because they are subject to 2005 07-May-2013a U-146
challenge and litigation. 2006 07-May-2013a U-146
As discussed above, researchers should note that patent expiration 2007 07-May-2013a U-146
2008 07-May-2013a U-146
dates change across editions of the Orange Book within a single ap-
2009 07-May-2013a U-146
plication. In the Gemzar example, changes in patent expiration dates 2010 07-May-2013a U-146
arose due to a Hatch–Waxman restoration extension as well as six- 2011 07-May-2013a U-146
month pediatric exclusivity extensions, both of which extended patent 2012 07-May-2013a U-146
2013 07-May-2013a U-146
expiration dates. However, changes can also occur in the opposite
direction due, e.g., to the non-payment of patent renewal fees. Notes: Table shows Orange Book patents associated with Gemzar (NDA #020509). Use
code U-146 indicates that the patent is for ‘‘a method of treating susceptible neoplasms
As one descriptive exercise, we tabulate changes in Orange Book- in mammals’’.
listed patent expiration dates over time, summarized in Fig. 1 as the a
Indicates that the patent expiration date was affected by a pediatric exclusivity patent
number of months between the earliest and latest listed expiration date term extension, as denoted by ‘‘*PED’’ in the Orange Book.
for every patent that appears in at least one edition of the Orange Book.
This difference is 0 months for the median patent, six months at the
75th percentile of the distribution, and 165 months at the maximum
(resulting from an apparent typo in the Orange Book).
was delayed by two additional months. This result is due to a complex
Exclusivity. The data file contains a number of regulatory exclusivities interaction between patents and exclusivity—in particular, between the
for Gemzar, which are summarized in Table 8. Most notably, Gemzar later expiring ’826 patent and generic exclusivity.53
received protection as a new chemical entity thanks to its novel active As noted above, the first generic firm to file a Paragraph IV patent
ingredient. As discussed above, NCE exclusivity lasts five years from challenge is potentially eligible for a 180 day period of exclusivity that
approval, and thus expired in 2001, long before the expiration of blocks the approval of other generics. Eligibility for generic exclusivity
Gemzar’s first expiring patent. In this case, NCE exclusivity did not is considered on a strength-by-strength basis.54 For Gemzar, Hospira
impact generic competition. For other drugs with short-lived or weak was the first filer for the 2-gram strength, while Teva was the first
patents, by contrast, exclusivity might be highly relevant, in part given filer for the other two strengths. By the time the ’614 patent expired
its comparatively ironclad (as opposed to probabilistic) nature. in November 2010, the ’826 patent had been invalidated thanks to a
The data file also reports multiple instances of NCI exclusivity, successful lawsuit by Sun, another generic firm. One might therefore
which were lengthened by pediatric extensions. As noted above, NCI have expected immediate generic entry on all strengths. However, due
exclusivity is less effective in preventing generic competition. All NCI to Teva’s continued entitlement to the 180 day exclusivity period,
exclusivities expired before the first expiring patent. approval of other generic products was blocked as to the 1-gram and
200 mg strengths. Moreover, Teva was not obliged to immediately
Interaction of patents and exclusivity. The full story of Gemzar is more launch its product, and did not do so. Teva waited until January 2011
complicated than the foregoing summary suggests. First generic entry to launch a generic version of the 1 gram and 200 mg strengths, with
in November 2010 was confined to one strength of the drug: the 2-
gram version. For two other strengths (1 gram and 200 mg), entry
53
The details in this discussion are drawn from searching the Drugs@FDA
database for ‘‘gemcitabine hydrochloride’’, Eli Lilly’s 10-Q SEC filings during
50
See https://www.uspto.gov/patents/laws/patent-term-extension/patent- the litigation, and the judicial opinions reported at Sun Pharm. Indus. v. Eli
terms-extended-under-35-usc-156. Lilly & Co., 611 F.3d 1381 (Fed. Cir. 2010), Eli Lilly & Co. v. Sicor Pharms.,
51
Because pediatric exclusivity—unlike patent term extensions—does not 705 F. Supp. 2d 971 (S.D. Ind. 2010), and Eli Lilly & Co. v. Mayne Pharma,
change the patent’s actual expiration date, the new effective expiration dates 504 F. Supp. 2d 387 (S.D. Ind. 2007).
are denoted by ‘‘*PED’’ in the Orange Book. 54
Apotex v. Shalala, 53 F. Supp. 2d 454 (D.D.C. 1999), aff’d, No. 99-5231,
52
See ANDA #079183. 1999 WL 956686 (D.C. Cir. Oct. 8, 1999).

10
M. Durvasula et al. Research Policy 52 (2023) 104791

Fig. 1. Changes in Orange Book-listed patent expiration dates across editions.

Notes: Figure plots the number of months between the earliest and latest expiration dates listed for each patent in the Orange Book, based on records reported in the Orange Book
patent addendum. N = 5, 511.

Table 8
Orange Book exclusivities associated with Gemzar (NDA #020509)
Exclusivity code Definition of code Exclusivity expiration First edition Last edition
NCE New chemical entity 15-May-2001 1997 2001

I-234 For use in combination with cisplatin for the first-line treatment of 26-Aug-2001 1999 2001
patients with inoperable locally advanced (stage IIIA or IIIB) or
metastatic (stage IV) non-small cell lung cancer

I-428 For use in combination with paclitaxel for the first-line treatment 19-May-2007 2005 2007
of patients with metastatic breast cancer after failure of prior
anthracycline containing adjuvant chemotherapy unless
anthracyclines were clinically contraindicated

PED Pediatric exclusivity 19-Nov-2007 2005 2008

M-40 Additional information regarding clinical studies performed in 26-Apr-2008 2006 2008
pediatric patients with leukemia added to precautions

PED Pediatric exclusivity 26-Oct-2008 2006 2009

I-499 Use of Gemzar in combination with carboplatin for the treatment 14-Jul-2009 2007 2010
of patients with advanced ovarian cancer that has relapsed at least
six months after completion of platinum-based therapy

Notes: Table shows Orange Book exclusivities associated with Gemzar (NDA #020509). First and last editions listed indicate the first and last years of
Orange Book editions in which these exclusivity codes and expiration dates appeared in connection with NDA #020509. The exclusivity expiration dates
for all entries are constant across editions.

other generic competitors blocked by the 180 day exclusivity period 5.3. Combining records to measure exclusivity
until July 2011.55
As this example illustrates, the combined effect of patents and reg- Building on the data described in Sections 5.1.1 and 5.1.2, we
ulatory exclusivity is sometimes subtle. Here, the existence of a weak, investigate how well combinations of these measures actually capture a
late-expiring patent provided an opportunity for generic exclusivity to drug’s experienced exclusivity period as measured by observed generic
approval.
be put at play, ultimately delaying the onset of generic entry.
We identify the set of drugs with generic approvals by linking
Orange Book records with FDA administrative data recording drug
approval dates as well as approval dates for generic competitors.56 To
55
Under the applicable statute, Teva was required to launch its product
within 75 days of a final judgment of patent invalidity, or else forfeit its
entitlement to generic exclusivity. In the end, final judgment was entered 56
Specifically, we use Drugs@FDA records which are available for download
on 12 November 2010, and Teva launched just before forfeiture would have from: https://www.fda.gov/drugs/drug-approvals-and-databases/drugsfda-dat
occurred. a-files.

11
M. Durvasula et al. Research Policy 52 (2023) 104791

Fig. 2. NCE exclusivity vs. generic approval.

Notes: Figure plots the relationship between approval of a generic competitor and the following measure of exclusivity: NCE exclusivity, assigned by the FDA. Approval of a generic
competitor is measured using Drugs@FDA administrative data. NCE is reported in the Orange Book exclusivity addendum. Both the x- and y-axes are measured as time since the
drug’s initial approval. The 45˝ line is plotted in black, with two-year bands plotted as red dotted lines. An observation above the 45˝ line represents a drug for which a generic
competitor was approved before exclusivity expired. The sample includes drugs with both NCE exclusivity and observed generic approval (N = 277).

overlap with the coverage of our Orange Book data, we focus on drugs it should bind, in the sense of fully blocking generic approval. Fig. 2
approved by the FDA between 1985 and 2014.57 We focus on a set illustrates how we can test this idea on the 277 drugs in our sample that
of 1673 NDAs that correspond to the first approval of a unique set have both NCE exclusivity and any generic approval. Fig. 2 plots the
of active ingredients by the FDA.58 Any ANDA approved by the FDA number of years between a drug’s approval and NCE expiration (y-axis)
for a given set of active ingredients is designated, for the purposes of against the number of years until the drug experiences generic approval
this exercise, as a generic competitor.59 Especially for drugs with small (x-axis). Any observations below the 45˝ line would represent cases
markets and products that are challenging to reverse engineer, delays in in which generic approval occurred after NCE expiration. We include
generic entry are common (Hemphill and Sampat, 2012). A substantial
two-year bands around expiration.
number of drugs never experience generic entry even after expiration
Of these 277 drugs, 18 have an exclusivity period less than five
of all exclusivities, and these drugs are not captured by this exercise.60
Using this data combined with the Orange Book data, we construct years. These are the product of the FDA’s ‘‘umbrella’’ policy, under
four measures of nominal exclusivity to compare with actual time which NCE exclusivity applies not only to the first drug eligible for
to generic approval.61 Within our sample, the median drug receives exclusivity, but also to later-approved drugs containing the same active
13 years of market exclusivity—time before either all legal protections ingredient. The later-approved drugs are protected for the remainder of
expire or a generic competitor enters (whichever comes first).62 the five-year period, measured from the approval of the first drug.63
In addition, 29 drugs have an exclusivity period longer than five
Measure #1: NCE exclusivity. We first compare the timing of generic
years. 26 of these correspond to drugs approved between 1 January
approval to the expiration of NCE exclusivity. Because NCE protection
is granted and enforced by the FDA—in contrast to patents, which 1982 and 24 September 1984, the effective date of the Hatch–Waxman
are granted by the USPTO and challenged in courts—we expect that Act. Under a now-repealed provision of the Act, these drugs received
10 year ‘‘transitional’’ exclusivity terms.64 Two more drugs received
pediatric extensions, which lengthened the duration of NCE exclusivity
57
We aggregate all products approved under the same NDA into a single to five-and-a-half years.65 For the remaining drug, the expiration date
observation. appears to be an Orange Book error.66
58
This set includes NMEs — i.e., drugs with a novel active ingredient — as
well as new combinations of previously approved active ingredients.
59
Actual substitution of generic drugs is more complicated. State law and
private payers facilitate substitution by therapeutically equivalent generics, 63
For example, NDA #019537, covering the antibiotic Cipro (ciprofloxacin)
which as discussed in Section 2 requires a showing of bioequivalence. See in a tablet form, was approved on 22 October 1987. Its five-year NCE
the preface to the Orange Book for a more detailed discussion: https://www. exclusivity expired on 22 October 1992. NDA #019847, for Cipro in an
fda.gov/drugs/development-approval-process-drugs/orange-book-preface. The injectable form, was approved on 26 December 1990, and shares the
concept of equivalence employed here is less restrictive. same NCE expiration date (22 October 1992). The umbrella policy also
60
The FDA maintains a list of off-patent, off-exclusivity drugs without applies to fixed-dose combination drugs that contain the same active ingre-
an approved generic at https://www.fda.gov/drugs/abbreviated-new-drug- dient. See https://www.fda.gov/files/drugs/published/New-Chemical-Entity-
application-anda/list-patent-exclusivity-drugs-without-approved-generic. Exclusivity-Determinations-for-Certain-Fixed-Combination-Drug-Products.pdf.
61
Note that Hemphill and Sampat (2012) find that the correlation between 64
This provision, 21 U.S.C. § 355(j)(5)(F)(i), was repealed in 2021.
time to first generic approval and time to first generic launch, for a set of 112 65
See NDAs #018831 and #020192.
new molecular entities, is 0.97. 66
Zocor (NDA #019766) was approved on 23 December 1991 and appears,
62
This estimate is consistent with evidence reported in Kesselheim et al. first, in the 2002 Orange Book, with a listed NCE exclusivity expiration date
(2017), who find that most new drugs receive between 12 and 16 years of of 23 December 1997. Approval documents from the FDA confirm that NCE
market exclusivity. exclusivity expired in this case in 1996.

12
M. Durvasula et al. Research Policy 52 (2023) 104791

Fig. 3. Generic-blocking regulatory exclusivities vs. generic approval.

Notes: Figure plots the relationship between approval of a generic competitor and the following measure of exclusivity: all generic-blocking exclusivities assigned by the FDA.
Approval of a generic competitor is measured using Drugs@FDA administrative data. Generic-blocking exclusivities are reported in the Orange Book exclusivity addendum. Both
the x- and y-axes are measured as time since the drug’s initial approval. The 45˝ line is plotted in black, with two-year bands plotted as red dotted lines. An observation above
the 45˝ line represents a drug for which a generic competitor was approved before exclusivity expired. The sample includes drugs with at least one generic-blocking exclusivity
and observed generic approval (N = 280).

As expected, we observe no generic approval before NCE expiration. Measure #3: Earliest patent expiration. We next compare the timing of
As such, we interpret Measure #1 as a lower bound on a drug’s generic approval to the latest of (i) all generic-blocking exclusivities
exclusivity period. assigned by the FDA and (ii) the earliest expiring Orange Book patent.
Generic-blocking regulatory exclusivities, as in the case explored in
Measure #2: NCE, ODE, GAIN, and PED exclusivities. Next, we compare
Measure #1, should fully protect against generic entry. Existing litera-
the timing of generic approval to the latest of the following exclusiv-
ture suggests that early expiring patents on drugs may provide stronger
ities: NCE, ODE, and GAIN, with any associated PED extensions (see
property rights than those filed later in a drug’s lifecycle, as they protect
Section 5.1.1 for details). As was the case for NCE exclusivity, we expect
‘‘primary’’ features of the technology (Hemphill and Sampat, 2011;
to observe no generic approval before exclusivity expiration, as these
Branstetter et al., 2016; Gupta, 2021). We can construct this measure
exclusivity forms should fully bar generic approval.
for the 454 drugs in our sample period that both have at least one
Fig. 3 plots time to exclusivity expiration against time to generic
Orange Book patent and experience generic approval.
approval. Any observations below the 45˝ line would represent cases
Note, however, that the earliest expiring patent is not necessarily
in which generic approval occurred after exclusivity expiration. We
the patent that protects a drug’s active ingredient. Information on the
once again include two-year bands around expiration; 13% of ob-
active ingredient (‘‘primary’’) patent is not reported in the Orange
servations fall within the bands. Nearly all generic approval occurs
Book. Some teams, including Kapczynski et al. (2012) and Hemphill
after the expiration of generic-blocking exclusivities. In eight cases in
and Sampat (2012), have reviewed patent texts by hand to identify
which generic approval occurs before exclusivity expiration, at least
active ingredient patents. Similar efforts, perhaps drawing on novel
one product under the drug’s NDA was not protected by the relevant
machine learning and natural language processing tools, may allow
exclusivity. In many cases, these observations correspond to drugs that
researchers to integrate such information into their analyses.
were available for use before the passage of the Hatch Waxman Act
(e.g., atenolol, approved in 1981) but which received protections as Fig. 4 plots time to expiration against time to approval. 46% of
‘‘new’’ drugs following the introduction of regulatory exclusivities.67 observations fall within the two-year bands. Again, we observe a set of
Of the 280 drugs with at least one exclusivity in this set and ob- drugs with long nominal exclusivity periods (based on the combination
served generic approval, there are four cases for which the exclusivity of regulatory exclusivities and the earliest-expiring patent); specifically,
period appears to exceed 15 years. In each of these cases, the drug 35 of the 454 observations in this case have exclusivity periods that
received ODE. For example, Novantrone was approved on 23 December exceed 20 years. In 26 cases, the drug was approved before Hatch–
1987;68 however, it received ODE in 2000 that expired on 13 October Waxman’s effective date (i.e., approved before September 1984). These
2007. This appears as a 20 year exclusivity term in our data. drugs received patents, in some cases, decades after products were
Fig. 3 suggests that—in nearly all cases—this set of exclusivities initially approved, resulting in what appear to be extremely long ex-
does, in fact, block generic approval. We interpret this measure as a clusivity periods. For the remaining nine cases, we observe a similar
small extension of the ‘‘lower-bound’’ in Measure #1. pattern in which initial patents were not obtained until many years
after the drug had been approved. Consider NDA #018613, for Ovide.
The drug was approved on 02 August 1982, but the earliest record of
67
See NDA #018240. an Orange Book patent appears in 2010 for U.S. Patent No. 7,560,445,
68
See NDA #019297. which expires on 01 February 2027.

13
M. Durvasula et al. Research Policy 52 (2023) 104791

Fig. 4. Earliest expiring patent vs. generic approval.

Notes: Figure plots the relationship between approval of a generic competitor and the following measure of exclusivity: the latest of all generic-blocking exclusivity expiration
dates and the first expiring patent on a particular drug. Approval of a generic competitor is measured using Drugs@FDA administrative data. Exclusivity and patent records are
reported in Orange Book addenda. Both the x- and y-axes are measured as time since the drug’s initial approval. The 45˝ line is plotted in black, with two-year bands plotted as
red dotted lines. An observation above the 45˝ line represents a drug for which a generic competitor was approved before exclusivity expired. The sample includes drugs with at
least one Orange Book patent and observed generic approval (N = 454).

Fig. 4 suggests that generic approval is common both before and 6. Conclusion
after earliest patent expiration. When generic approval occurs after this
expiration date, this may be due to a combination of secondary patents The pharmaceutical industry is a unique context where, due to
(see Measure #4 below) and market forces that delay entry. FDA regulatory requirements, it is possible to link patents to products.
Since 1985, the FDA’s Orange Book has listed patent and exclusivity
Measure #4: Latest patent expiration. Finally, we compare the timing information for small-molecule drugs approved by the FDA. In this
of generic approval to the latest of all generic-blocking exclusivities paper, we introduce a digitized version of these records for the years
assigned by the FDA and the latest expiring patent in any Orange Book 1985 to 2016 (Appendix A describes how these data can be updated to
edition. To the extent that late-expiring patents are more likely to either include later years).
be invalidated through litigation or unenforced, we expect to observe Like many administrative datasets, the Orange Book was not created
generic approval before market exclusivity expiration in more cases for researchers. Nonetheless, it includes many variables—including
here than with Measure #3. cleaned patent numbers, FDA drug application numbers, and both
In Fig. 5, for a sample of 454 drugs, we observe that roughly half brand and generic drug names—that facilitate linkages with other
of drugs in our sample have generics approved before exclusivity expi- datasets that are commonly used in empirical analyses.
ration, as defined using the latest expiring patent. 34% of observations Whether the data are sufficiently accurate and complete for research
fall within the two-year bands around expiration. This is consistent with purposes will depend on the specific research question being asked. By
the idea that late-expiring patents are less effective at delaying generic design, the Orange Book focuses on small-molecule drugs and does not
approval and aligns with the findings in Hemphill and Sampat (2012), record information for most biological drugs. We find that, for 77% of
in which the authors show that, especially for high-sales drugs, late- NDAs and 96% of NMEs, the Orange Book reports at least one patent or
expiring secondary patents are disproportionately targeted in litigation. exclusivity. In other words, nearly all new drugs approved since 1985
When cases are litigated to completion, generics typically prevail. have some form of patent or exclusivity listed in the Orange Book.
In Fig. 6, we replicate Fig. 5 and highlight with red triangular Conditional on a drug appearing in the Orange Book, how complete
markers any drug for which, as of June 2020, the FDA has recorded are patent records? There is minimal review associated with the listing
at least one Paragraph IV challenge. Note that Paragraph IV challenges process; the FDA reports information provided by firms marketing
appear to explain many of the cases in which generic approval is brand-name drugs. As we note, rules associated with Hatch–Waxman
‘‘early’’ as compared to the date of last patent expiration. Weakly, this challenges and litigation create incentives for firms to report all eligible
is consistent with the idea that our Measure #4 is a reasonable ‘‘upper Orange Book patents (covering active ingredients, formulations, and
bound’’ on the duration of a drug’s exclusivity period. methods of use). Comparison to a commercial dataset, IQVIA/Ark,
Taken together, these four measures make clear that researchers indicates that there are many patents associated with specific drugs
using the Orange Book exclusivity and patent measures must be in- that are not listed; our review, however, suggests that these are mainly
tentional in their decisions about how to aggregate available data from categories not eligible for Orange Book listing. Of patents that
into effective exclusivity measures. Beyond the measures available in IQVIA/Ark suggests should ‘‘constrain’’ generic entry, roughly three-
the Orange Book itself, distinguishing between primary and secondary quarters are listed in the Orange Book. Litigation records indicate that
patents may be important, and the effects of these patents on generic more than 90% of patents asserted in litigation against potential generic
approval may vary depending on the incentives firms face to initiate entrants are listed in the Orange Book. Additionally, 96% of patents
patent challenges, including projected sales. that firms select for term extension—arguably, a revealed preference

14
M. Durvasula et al. Research Policy 52 (2023) 104791

Fig. 5. Latest expiring patent vs. generic approval.

Notes: Figure plots the relationship between approval of a generic competitor and the following measure of exclusivity: the latest of all generic-blocking exclusivity expiration
dates and the last expiring patent on a particular drug. Approval of a generic competitor is measured using Drugs@FDA administrative data. Exclusivity and patent records are
reported in Orange Book addenda. Both the x- and y-axes are measured as time since the drug’s initial approval. The 45˝ line is plotted in black, with two-year bands plotted as
red dotted lines. An observation above the 45˝ line represents a drug for which a generic competitor was approved before exclusivity expired. The sample includes drugs with at
least one Orange Book patent and observed generic approval (N = 454).

Fig. 6. Generic approval and Paragraph IV challenges.

Notes: Figure plots the relationship between approval of a generic competitor and the following measure of exclusivity: the latest of all generic-blocking exclusivity expiration
dates and the last expiring patent on a particular drug. Drugs with competitors that entered via Paragraph IV challenge are marked with red triangles. Both the x- and y-axes are
measured as time since the drug’s initial approval. The 45˝ line is plotted in black, with two-year bands plotted as red dotted lines. An observation above the 45˝ line represents
a drug for which a generic competitor was approved before exclusivity expired. The sample includes drugs with at least one Orange Book patent and observed generic approval
(N = 454).

measure of which patents matter to the firm—are listed in the Orange provide updates on maintenance fee non-payment. This is an important
Book. Taken together, we interpret these results as suggesting that the note for researchers, as nearly half of Orange Book listed patents are
Orange Book does seem to include nearly all of the ‘‘important’’ patents not maintained to full term.
on listed drugs. Patent numbers reported in the Orange Book are accurate and, as
We also find that patent expiration dates listed in the Orange Book such, can be merged with a variety of other patent datasets (including
are generally accurate and reflect changes to patent terms due to TRIPS; the USPTO’s PatentsView and the OECD Triadic Patent Dataset). NDA
Orange Book expiration dates coincide with expiration dates in patent numbers match cleanly to Drugs@FDA, which can be used to link drug-
term extension letters in roughly 90% of cases and also reflect patent specific information. Other fields, including sponsoring firm name, are
term adjustment in the majority of cases. While firms do seem to less standardized. Brand and generic drug names are available from
report these extensions to patent terms diligently, they do not appear to the Orange Book but, in our experience, direct string matches against

15
M. Durvasula et al. Research Policy 52 (2023) 104791

other datasets perform poorly and typically require manual matching Alternatively, researchers may prefer to download text files con-
procedures to improve accuracy.69 taining Orange Book patent and exclusivity records that are updated
In addition to patent information, we also examine the completeness monthly. The FDA makes the most recent version of the electronic
of Orange Book exclusivity information. As compared to other FDA Orange Book available for download.71 Researchers can use the Internet
sources, Orange Book records are both complete and accurate. This is Archive’s Wayback Machine and the FDA’s own web archive, which
unsurprising: unlike patent information, which is provided and updated covers October 2016 to present, to access historical versions of these
by firms, exclusivity information is drawn directly from FDA records. monthly files. By collecting and appending ‘‘snapshots’’ of the database
Our final exercise—combining patent and exclusivity data to cal- taken at regular intervals, it is possible to reconstruct the complete
culate legal exclusivity periods—is intended to capture the complex- listing of patents and exclusivities—including records that may appear
ities associated with using these data and to highlight the types of for only short periods of time before being de-listed.72
choices that researchers must make. Especially given concerns about
the proliferation of irrelevant or invalid drug patents, the gap between Appendix B. PTE and PTA discrepancies
‘‘nominal’’ and ‘‘actual’’ legal exclusivity may continue to grow. While
the Orange Book dataset provides a rich resource for understanding the Here, we outline the process that we use to determine the correct
pharmaceutical market, researchers will need to make choices based on expiration dates for Orange Book patents with PTE and PTA. Scaling up
specific questions of interest and, in some cases, to bring in outside such a procedure may be valuable for future efforts to audit the expi-
information when determining exactly which Orange Book patents ration dates of all Orange Book patents, beyond the samples examined
‘‘count’’. here.
There have been many calls to reform Orange Book listing practices To determine the correct expiration date in each case, we first
to achieve various policy goals. Hemphill and Sampat (2022) suggest collect the following for each patent: filing dates (actual and effective),
modifications that would prevent firms from listing irrelevant patents issuance date, PTE records, and PTA records. Determining a patent’s
(which may later be the target of costly patent challenges) and collect effective filing date can be complicated for child patent applications
more information from applicants about what, exactly, listed patents if, for example, the application in question is a continuation-in-part;
are intended to cover. Such changes would likely make the Orange researchers may need to determine effective filing dates on a case-
Book an even more valuable resource for researchers. by-case basis for such records.73 We also confirm that there are no
Moving forward, efforts to construct new product-patent linkages Certificates of Correction on file with the USPTO that alter PTE or PTA
will likely require new statutes—as were in place when listing require- determinations.74
ments for the FDA’s Purple Book were expanded in 2020. New authority Next, we determine if maintenance fees were paid by 4, 8, or
would also put to rest any argument that this information should be 12 years after the patent’s issuance; if not, the correct expiration
treated as a trade secret. Settings where listing requirements can be date is the date on which a maintenance payment was due. We also
tied to some corresponding benefit, as in the case of other products that review USPTO Patent Center records to determine whether a terminal
require premarket clearance, are likely to be the most straightforward disclaimer is on file; if so, the correct expiration date is the terminal
to implement. disclaimer date, plus any PTE and PTA extensions.
For the remaining cases, we determine whether the patent’s actual
Declaration of competing interest filing date was before 08 June 1995, the date on which TRIPS changes
to patent term calculation took effect. If a patent was filed before this
The authors declare the following financial interests/personal rela- date, its expiration date is the later of: {effective filing date + 20 years
tionships which may be considered as potential competing interests: C. + PTA + PTE, issuance date + 17 years + PTA + PTE}. If a patent was
Scott Hemphill reports that he ‘‘has at various times served as an expert filed after this date, its expiration date is its effective filing date plus
witness in litigation involving Orange Book-listed patents’’. This relation- 20 years and any associated PTA and PTE.
ship includes: paid expert testimony. Bhaven Sampat reports that he
‘‘has served as a consultant in litigation on issues relating to pharmaceutical B.1. PTE discrepancies
patenting and innovation and to the Medicines Patent Pool for work on
patents and access to medicines’’. This relationship includes: consulting We resolve 45 discrepancies in patent expiration dates between the
or advisory. USPTO Patent Term Extension (PTE) list and the Orange Book. Details
on specific, complicated cases flagged in Table 2 are below:
Data availability
• Patent 3857952: This patent was granted 31 December 1974.
We have shared the link to data. Data are available from the NBER. With a 17 year patent term and 2 year PTE, the expiration date
should be 31 December 1993, as reported in the Orange Book.
NBER Orange Book Dataset (Original data) (National Bureau of
We are unable to determine how the PTE date of 17 December
Economic Research)
1993 was obtained, as this patent is not available in the USPTO’s
Patent Center database. We assume that the PTE date is an error.
Appendix A. Updating data

The NBER Orange Book Dataset includes records from 1985 to 2016. 71
The current version can be accessed here: https://www.fda.gov/drugs/dr
Researchers may find it useful to append data from more recent years to ug-approvals-and-databases/approved-drug-products-therapeutic-equivalence-
these historical records. Annual editions, in PDF form, are available for evaluations-orange-book.
download from the FDA.70 Note that considerable processing may be 72
As of 05 September 2022, a Wayback Machine search for the following
necessary in order to turn these PDFs into ‘‘research-ready’’ data files. URL returned the FDA webpage ‘‘Approved Drug Products with Therapeutic
Equivalence Evaluations (Orange Book)’’, which includes a link to download
‘‘Orange Book Data Files (compressed) (ZIP)’’, the monthly updated text file:
69
Both firm names and drug names can be obtained from Drugs@FDA by https://www.fda.gov/Drugs/InformationOnDrugs/ucm129662.htm.
merging on the NDA number. 73
For our small number of discrepancies, we use USPTO Patent Center
70
The most recent annual edition is available here: https://www.fda.g records to make these determinations for each case. Scaling up this procedure
ov/drugs/drug-approvals-and-databases/orange-book-data-files. The FDA.gov may be possible using the USPTO’s bulk data on continuation patents.
archive allows researchers to retrieve previous annual editions: https://w 74
A list of all patents with Certificates of Correction is available here: https:/
ww.fda.gov/about-fda/about-website/fdagov-archive. Research and advocacy /www.uspto.gov/patents/search/authority-files/certificates-correction. To the
organizations also maintain archives of annual Orange Book editions. best of our knowledge, no machine-readable version of these corrections exists.

16
M. Durvasula et al. Research Policy 52 (2023) 104791

• Patent 4105783: This patent has a terminal disclaimer of term • Patent 5681818: This patent is associated with a drug that was
after 03 May 1994. With a 541 day extension, this yields the initially approved under a BLA. It was removed from the Orange
PTE expiration date of 26 October 1995 — the correct expiration Book before the Orange Book could reflect PTE.
date. The Orange Book lists an expiration date of 15 January
1997, which reflects what we believe is an earlier effective filing B.2. PTA discrepancies
date (which is somewhat unclear, due to the existence of a
continuation-in-part) of 23 July 1975 + 20 years + 541 days = Table 3 investigates discrepancies between Orange Book and USPTO
14 January 1997. There are often one day discrepancies between PTA records for a randomly drawn 10% sample (N = 43) of patents
the Orange Book and USPTO datasets. with different expiration dates across these two datasets. In 16 cases
• Patent 4234571: The correct expiration date is 11 June 2001. The (38%), patent expiration dates were calculated differently in PTA data
Orange Book lists 11 June 2011. and the Orange Book: in these cases, the Orange Book reports an
• Patent 4567264: This patent received five years of PTE and 1 expiration date based on the effective filing date of a patent, associated
year of interim PTE. The four year extension in the Orange Book with parent patents, and the USPTO reports an expiration date based
appears to be a typo. on its actual filing date. Orange Book records are correct in these cases.
• Patent 4585597: The PTE list reports a 1455 day term extension In 5 cases (12%), Orange Book records—as of the 2015 edition—do
from 16 June 2003, which should be 16 June 2007. The OB not include Certificates of Correction to PTA issued by the USPTO. For
reported date of 10 June 2007. Based on our review of available three patents, both errors are present: USPTO and Orange Book data use
documentation, we believe the OB date is an error. different filing dates for patents, and Orange Book records do not record
• Patent 4661491: The PTE list miscalculates the proper extension. a relevant Certificate of Correction. In 14 cases (33%), only the Orange
This patent had an initial expiration date of 27 May 2006 and Book correctly reports an expiration date updated to reflect a termi-
received a 1697 day extension (as confirmed by both the PTE list nal disclaimer. In the remaining five cases, Orange Book expiration
and a Certificate of Extension). However, the PTE list adds only dates are incorrect, as they either do not reflect maintenance fee non-
365 days to the expiration. The Orange Book date correctly adds payment (and associated patent expiration) or include a typographical
1697 days to the original expiration. error.
• Patent 4670444: The Orange Book listing does not include PTE,
but does include PED. The correct expiration date should be 09 References
June 2007 = 09 December 2006 + six months.
• Patent 4687659: The PTE list reports an expiration date of Ark Patent Intelligence, 2019. Constraint Analysis: Introduction of Concepts and
08 January 2007, which is the grant date (18 August 1987) Analysis.
Azoulay, Pierre, Graff Zivin, Joshua S., Li, Danielle, Sampat, Bhaven N., 2019. Public
+ 17 years + 873 days of PTE. The Orange Book reports an
R&D investments and private-sector patenting: Evidence from NIH funding rules.
expiration date of 04 May 2007, which we believe was obtained Rev. Econom. Stud. 86 (1), 112–152.
as follows: filing date (13 November 1984) + 17 years + 873 days Branstetter, Lee G., Chatterjee, Chirantan, Higgins, Matthew, 2016. Regulation and
= 05 April 2007; the listing then swaps month and day (05 April welfare: Evidence from paragraph IV generic entry in the pharmaceutical industry.
2007 becomes 04 May 2007). The correct expiration date, based Rand J. Econ. 47 (4), 857–890.
on our review of documentation associated with this record, is 05 Cohen, Wesley M., Nelson, Richard R., Walsh, John P., 2000. Protecting Their
Intellectual Assets: Appropriability Conditions and Why U.S. Manufacturing Firms
April 2007. Patent (or not). NBER, Working Paper # 7552.
• Patent 4695590: A Certificate of Correction clarifies that PTE Eisenberg, Rebecca, 2012. Patents and regulatory exclusivity. In: Oxford Handbook on
should be 712 days, instead of 122. The Orange Book is correct. the Economics of the Biopharmaceutical Industry. Oxford University Press, Oxford.
• Patent 4845075: This patent’s original expiration date was 04 Frakes, Michael D., Wasserman, Melissa F., 2022. Investing in ex ante regulation:
Evidence from pharmaceutical patent examination. Am. Econ. J. Econ. Policy.
July 2006, and it received a 1578 day extension. The PTE list
Gupta, Charu N., 2021. One Product, Many Patents: Imperfect Intellectual Property
reports 29 October 2010 as an expiration date, and the Orange Rights in the Pharmaceutical Industry. Working Paper.
Book lists 11 April 2009. 29 October 2010 is 17 years after Hemphill, C. Scott, Sampat, Bhaven N., 2011. When do generics challenge drug
patent grant plus 1578 days. 11 April 2009 is 18 years after patents? J. Empir. Leg. Stud. 8 (4), 613–649.
patent filing plus 1578 days. Based on our review of records, Hemphill, C. Scott, Sampat, Bhaven N., 2012. Evergreening, patent challenges, and
effective market life in pharmaceuticals. J. Health Econ. 31 (2), 327–339.
both 29 October 2010 and 11 April 2009 are incorrect. The
Hemphill, C. Scott, Sampat, Bhaven N., 2013. Drug patents at the supreme court.
Orange Book expiration date may reflect a typo. In this case, the Science 339 (6126), 1386–1387.
patent’s expiration date—calculated as 20 years after patent filing Kapczynski, Amy, Park, Chan, Sampat, Bhaven N., 2012. Polymorphs and prodrugs and
+ PTE—should be 11 April 2011, 20 years after patent filing plus salts (oh my!): An empirical analysis of ‘secondary’ pharmaceutical patents. PLoS
an extension. One 7 (12), URL https://papers.ssrn.com/abstract=2191728.
Kesselheim, Aaron S., Sinha, Michael S., Avorn, Jerry, 2017. Determinants of market
• Patent 4886812: A terminal disclaimer was filed. 08 October
exclusivity for prescription drugs in the United States. JAMA Intern. Med. 177 (11),
2010 is the correct expiration date after the terminal disclaimer. 1658.
• Patent 4990517: There are two different expiration dates for Kyle, Margaret, Sampat, Bhaven N., Shadlen, Kenneth C., 2020. TRIPS, Patents, and
this patent reported in the last Orange Book edition in which it Drug Prices in India. Working Paper.
appears. One expiration date—08 December 2011—appears to be Lemley, Mark A., Shapiro, Carl, 2015. Probabilistic patents. J. Econ. Perspect. 19,
75–98.
a 10 day typo of the correct PTE date, 18 December 2011. With a Lemley, Mark A., Tu, Sean, 2022. What litigators can teach the patent office about
6-month pediatric extension, the correct Orange Book date should pharamaceutical patents. Wash. Univ. Law Rev. 99, 1673–1731.
be 18 June 2012, not 08 June 2012. Levin, Richard, Klevorick, Alvin, Nelson, Richard, Winter, Sidney, Gilbert, Richard,
• Patent 5387603: The correct expiration date, listed in PTE data, Griliches, Zvi, 1987. Appropriating the returns from industrial research and
is 12 December 2018. The Orange Book lists 01 December 2018. development. Brook. Pap. Econ. Act. (3), 783–831.
Lietzan, Erika, Lybecker, Kristina, 2020. Distorted drug patents. Wash. Law Rev. 95
• Patent 5446194: 19 November 2013 is the correct expiration (3), 1317–1382.
date, given the length of PTE. Based on our review of available de Rassenfosse, Gaetan, 2018. Notice Failure Revisited: Evidence on the Use of Virtual
documentation, we believe 19 October 2013 is an error in the Patent Marking. NBER, Working Paper # 24288.
Orange Book. Sampat, Bhaven N., Shadlen, Kenneth C., 2015. TRIPS implementation and secondary
pharmaceutical patenting in Brazil and India. Stud. Comp. Int. Dev. 50 (2),
• Patent 5656667: In 2009, a terminal disclaimer was filed, dis-
228–257.
claiming the term after the expiration date of patent 5502077. Thomas, John R., 2017. Regulatory Exclusivity Reform in the 115th Congress. Report,
The Orange Book expiration date is correct, based on this terminal (R44951), Congressional Research Service.
disclaimer.

17