abstract. Pharmaceuticals is the sector most widely
thought to be in need of strong patent protection in order to sustain a robust level
of innovative activity. This Feature comprehensively seeks to revise that
assessment. It argues that a proper understanding of the actual informational
resources at play in drugs reveals that pharmaceutical innovation can,
significantly does, and entirely should proceed without any role played by
patents.
The
foundational plank of the argument is that innovation in pharmaceuticals
consists of not one but two distinct information goods: (1) knowledge of a
chemical or biological compound (the “compound information good”), and (2)
knowledge of a compound’s safety and efficacy for use in humans, as validated
by clinical-trial data (the “data information good”). It is the latter
information good, not the former, that is both the driver of the economics in this
sector and the apt focal point of innovation-policy rules. Indeed, a close
examination of how the doctrines of patent law map onto the pipeline of
pharmaceutical innovation reveals a set of radically sector-specific doctrines
that confer little protection during the preclinical research that generates
the compound information good, contrary to a common view. Meanwhile, for the
clinical testing that generates the data information good, revised
regulatory-exclusivity rules can and entirely should suffice. Indeed, the
protection presently afforded this good by patents is indirect, incomplete,
and—owing to a basic misalignment between the patent system’s focus and
sensible aims for innovation policy in this sector—haphazard and highly costly.
Consequently,
simply by phasing out patent protection for drugs and replacing it with a
revised form of regulatory exclusivity, we would reap large gains in social
welfare: better-tailored incentives, reduced access and duplication costs, and
significantly curbed wastes from gaming of the present system. Many of these
costs stem from “evergreening” practices and “me-too” drugs, which have both
been the subject of sharp criticism. The present analysis offers a deeper
diagnosis of the causes and extent of these problems, and it proposes more
effective, better-tailored solutions.
This
same analysis should also reorient broader debates in patent theory and
innovation policy more generally by revising our understanding of the special
case posed by drugs for innovation-policy support. The conventional view that
pharma presents an especially strong case for patent protection turns out to be
triply wrong. First, the innovation taking pride of place in judicial and
scholarly attention—the compound information good—presents no special case for
patents. Second, the innovation that does present a strong case for
innovation-policy support—the data information good—is both sidelined by the
patent system and in any case ill-suited for patent protection. Thus, the
special case presented by pharma is not for patents but for an alternative
innovation-policy intervention. Finally, the basis of that special case for
innovation-policy support lies in a regulatory regime rather than in any
generalizable economic or technological features of drugs.
author. Lecturer, University of California, Berkeley,
School of Law. Thanks to Yochai Benkler, Oren Bracha, and Terry Fisher for
invaluable comments on an earlier draft, and to Ken Ayotte, Colleen Chien, Seth
Davis, Dhammika Dharmapala, Aaron Edlin, Mark Gergen, Katerina Linos, Peter
Menell, Rob Merges, and Jonathan Simon for helpful discussion. For excellent
research assistance, I thank Garreth McCrudden, Caressa Tsai, and Will
Kirkland. I also wish to thank the editorial staff of the Yale Law Journal, in particular Deniz Arıtürk and Fred
Halbhuber, for terrific comments that significantly improved the Feature.
Research for this Feature was partly funded by a grant from the Institute for
New Economic Thinking (INET), for which I am grateful.
Introduction
Does pharma need patents? The consensus view among scholars
is a resounding “yes.” The pharmaceutical industry is widely agreed to be the
sector most in need of strong patent protection to sustain a robust level of
innovative activity. Study after study of the effects of
patents on innovation—be they empirical surveys asking firms in different
industries what they rely on to appropriate the benefits of innovation,
historical studies of long-term patterns of innovation and patent protection,
or synthetic theoretical-empirical treatments of the aggregate costs and
benefits of the patent system as a whole—agree that, whatever other conclusions
may be reached regarding the overall case for patent protection across the
economy, such protection is crucial for innovation in drugs. This conviction holds not only for
those most strongly endorsing the patent system as a whole, but also for those more uncertain
about the overall merits of patents.
Indeed, even the staunchest critics of the patent system in general accept that
pharma remains a crucial exception.
This Feature seeks to revisit that assessment
comprehensively. It argues that a proper
understanding of the actual informational resources at play in drugs reveals
that pharmaceutical innovation can, considerably does, and entirely should
proceed without any significant role played by patent protection. The foundational plank of the
argument is to underline how innovation in pharmaceuticals consists of not one
but two separate information goods: (1) knowledge of a new chemical or
biological compound, and (2) knowledge of the safety and efficacy of that
compound for use in humans, as validated by clinical trials. Moreover, not only is the latter
information good a separate innovation from the former; it is one very distinct in its risk-cost profile,
diverging sharply in those technical and economic features that are relevant to
innovation-policy analysis. What these features reveal is that the first
information good likely poses no stronger case for patent protection than
innovation in most other sectors, while fitting quite well a model of
decentralized, competitive development. Meanwhile, the second information good
does require strong innovation-policy support, while fitting better a model of
centrally coordinated development.
Two fundamental implications follow from this theoretical
distinction. First, the distinction reveals a new understanding of existing
patent practice in the pharmaceutical industry. Applying the insight of two
information goods discloses a dramatically new picture of how patent and
related laws map onto the pipeline of pharmaceutical innovation, including by
revealing a set of highly sector-specific patent doctrines applicable only to
pharma. The upshot of this picture is that patents provide only partial—and
largely unnecessary—protection over the first information good, and
indirect—and highly misaligned—protection over the second. Second, these
explanatory implications of the distinction justify a deep reform of
pharmaceutical innovation policy. A better innovation policy for this sector
would be to phase out patents altogether and replace them with an alternative
innovation-policy intervention, one better suited to the distinctive
technological and economic features of the second information good: a revised
system of “regulatory exclusivity.”
At the heart of pharmaceutical innovation lie two information
goods. The first is knowledge of a new drug product, which we may call the
“compound information good.” The
second is knowledge of that drug’s safety and efficacy for humans as evinced by
clinical-trial data, which we may call the “data information good.”
Generating the compound information good involves the exploration of a highly
uncertain possibility frontier: each step involves many risks—only about one in
a thousand candidate compounds make it through the drug-discovery phases of
“search, synthesis, and screening” to enter clinical trials—so as to warrant comparatively low
expenditures per step. By
contrast, generating the clinical information good is a comparatively low-risk,
high-cost endeavor: roughly one out of five to ten drugs that enter clinical
trials successfully navigate the process of testing and refinement to receive
Food and Drug Administration (FDA) approval,
while the costs of phase 1, 2, and 3 trials dwarf those of each step of
preclinical drug discovery. This sharp divergence in the
risk-cost profiles of these information goods carries two sets of crucial
implications for their apt innovation-policy treatment.
First, from a purely economic
point of view, it is the data information good—not the compound information
good—that is the driver of the industry’s innovation costs. While the cost of
drug development remains a topic of fierce controversy,
what is not controversial is that clinical-trial expenditures comprise the
lion’s share of the costs, running around 70% according to the industry’s own
preferred studies, and even higher for others. Indeed, a 2021 metareview of
twenty-two studies of drug-development costs conducted over the past four
decades found that over half (thirteen) of the studies reviewed did not even
consider preclinical drug-discovery expenditures significant enough to factor
in as a part of total costs.
In addition to their very different economic significance for
pharmaceutical innovation, these information goods also sharply differ in the technological features of the respective
innovation processes that generate them. Preclinical drug discovery, with its
high risks and lower costs, is well suited for a decentralized search, where
“many minds” are given free rein to explore various different avenues, even at
the risk of a fair bit of overlapping, duplicative activity. Clinical trials, on the other hand,
with their lower risks and high costs, are better suited for coordinated
development to curb duplicative efforts that would be highly wasteful at this
stage. In other words, preclinical research
should be a nonexclusionary zone, to enable many-minded exploration
unencumbered by proprietary barriers. Meanwhile, for clinical trials, some
mechanism is needed to call off the innovation race at their outset.
Integrating these distinct economic and technological features
of the two innovations leads to the following pair of conclusions. First, the
compound information good—generation of new knowledge of a chemical or
biological product or process—poses no special incentive case for patent
protection. Its share of overall industry innovation costs is relatively
modest. Further, what is the really relevant focus for innovation-policy
analysis is the differential between its average innovation costs and risks and
its average imitation costs and speed (i.e., the cost and time involved in
reverse engineering and being ready to manufacture a new or improved drug
product or process). And that differential is likely no greater than in many
other sectors where a combination of first-mover advantages and secrecy suffice
to ensure a relatively robust level of innovative activity.
In addition, patents also serve no useful “coordinating” function during the
research phase leading to the generation of the compound information good: its
comparatively high risks and low costs make that phase suitable for a
competitive, decentralized search.
Second, the data information good—generation of new clinical
results on a drug—does present a strong
case for an innovation-policy intervention, but it is one for which patents are
a highly unsuitable instrument. That strong case stems not only from the large
share of overall industry innovation costs taken up by this activity but also
from—what is again the relevant focus—the large difference between its average
costs and risks of generation and its average costs and speed of replication
(with the latter massively reduced by regulatory permission of imitator
piggybacking on innovator data). Yet the
patent system provides little to no direct protection over this information
good, as its doctrines center on the results of preclinical research, not
clinical testing. And it
is not only that patents currently sideline the protection of clinical data;
they also cannot effectively provide
such protection. Given the technological features of this innovation, it would
be untenable to try to reform the patent system to protect it; inquiries into
its desirability and validity are simply not ones that the patent system is
well suited to carry out.
Consequently, patents serve their two primary functions in
pharmaceutical innovation—coordinating innovation races and incentivizing innovative
activity—only indirectly, with respect to an information good, clinical data,
that they do not directly protect.
Meanwhile, for the information good that patents do directly cover—knowledge of
the compound—they play little to no coordinating role and only a secondary
incentive role.
A sounder innovation policy would be to replace the primary, yet indirect, role
played by patents over data information with a form of regulatory exclusivity
that specifically attends to the distinctive features of this innovation, while
at the same time phasing out the direct but secondary role patents play over
compound information.
The point of doing so is to bring our system of
innovation-policy rules into better alignment with the underlying innovations
that they seek to generate. This alignment would ensure that the rules directly
attend to the relevant features of the information goods they govern and that
they are better equipped to make the various tradeoffs facing any innovation
policy. In particular, such a reform would significantly improve the
performance of our innovation policy for drugs in tackling the two key
tradeoffs facing any incentive system that uses exclusionary rights (such as
patents or data exclusivity). First, it would reduce undue barriers to access
that exclusionary rights erect over those innovations that would have been
generated at lower levels of protection. Second, it would curb undue rent
dissipation—that is, wastefully duplicative innovative activity—that
exclusionary rights may incur for innovations that would have been incentivized
by a lower level of protection.
Specific versions of each of these concerns have been prominently voiced in the
critical literature on pharma, the first under the heading of “evergreening”
practices and the second under that
of “me-too” drugs. In
both cases, analysis of the distinct information goods—and of how existing
rules fail to align with their relevant features—immeasurably improves both our diagnosis of the precise causes
and extent of the problems and our
prospects for prescribing effective solutions.
In the case of evergreening and related practices such as
“reverse settlement agreements” (RSAs), this analysis identifies the generative
cause of such practices: the specific industry
structure of pharma that stems from the regulatory treatment of the data
information good. This
information-good analysis fills a gap in the literature by explaining why such
practices are, indeed, pharma-specific. The Feature then specifies better ways
of evaluating the extent of the social costs of such practices, anchored in the
distinction between the compound and data information goods.
Finally, this same information-goods analysis also points the way to reforms
that attack the problem at its root—the basic misalignment between patents and
data information—as opposed to proposals that seek only to remedy surface ills
with how patents currently operate.
And similarly for the duplication wastes incurred by me-too drugs, an analysis
focused on the distinction between generating new compounds and generating new
clinical data is better able to specify both the extent to which such drugs do
incur such wastes and how to tailor remedies for effectively curbing them.
In sum, an assessment of pharmaceutical innovation policy
that trains its attention on the data information good lying at its heart leads
to the following conclusions. The actual protection provided by patents over
the key information goods in pharmaceuticals is partial, indirect, and—owing to
a misalignment between what the patent system focuses on (the compound
information good) and what sensible innovation policy would center (the data
information good)—haphazard and highly costly. This protection would be
radically improved by replacing patents’ exclusionary rights with those of a
revised—streamlined and tailored—form of data exclusivity. Such exclusivity
should be streamlined to curb the gaming and administrative costs associated
with misaligned patents, and tailored to realign the system’s focus on the
incentives that matter—those pertaining to the costs, risks, and desirability
of generating different types of clinical data on drugs.
This analysis has major implications for lowering both the prices and the cost of drugs, and for improving both access to and
incentives for pharmaceutical innovation. In 2022, the United States spent $406
billion on retail prescription drugs.
One source of this high price tag, on which critics of the industry have
rightly focused their attention, is how RSAs and related evergreening abuses of
patents unduly drive up drug prices, with estimates of their effects ranging
between $3.5 billion to $6.2 billion in higher prices annually.
In response, the Federal Trade Commission (FTC) has called for reforms such as
the “delisting” of over 100 drug patents from the “Orange Book”—to remove one
evergreening obstacle to generic entry—as well as changing the antitrust burden
for establishing the legality of RSAs, to remove another.
The present analysis not only provides a firmer basis for such reforms than has
so far existed, but it also shows why they do not go far enough: not only
should some patents be delisted from
the Orange Book, but all such Orange
Book linkage should be
abolished, and similarly
for RSAs—they should be deemed per se, rather than merely presumptively,
anticompetitive.
Doing so would dramatically reduce barriers to access from trivial or modest
secondary drug patents and products.
At the same time, however, the foregoing
estimates of the costs of evergreening practices are incomplete because they do
not factor in possible incentive benefits from extended patent protection to be
weighed against its access costs. And these estimates do not factor in the
wastes associated with gaming the patent system to obtain such (indirect)
incentives. In addition, a focus on the role of evergreening—or
trivial or modest secondary drug patents or productions—in driving up industry
prices and costs needs to be supplemented with an analysis of the role of
me-too—or duplicative primary drug patents and products—in doing the same. For
both, the best metric of their costs is to step back
from specific cases, take a comprehensive view of the industry’s output, and
analyze the types and extent of innovation they represent. Such a review,
carried out here, reveals that, of the 2,872 new drugs approved in the years
1990 to 2023,
almost 70% were secondary products, and 86% of these were rated by FDA
not to hold out a significant advance. In other words, 60% of the industry’s output
consists of secondary products securing patent protection that is likely
incommensurate with the modest innovation they hold out. Moreover, of the
roughly 30% of output that consisted of primary products, over half (51%) were
similarly rated as standard—that is, held to be somewhat to highly duplicative
of already-available treatments.
Both the high access and duplication costs
incurred by evergreening practices and me-too drugs stem from the misaligned
incentives of the present system of innovation-policy rules in place for
pharmaceuticals. In each case, the cause lies in different aspects of how the
central innovation in pharmaceuticals, the data information good, is handled by
the present system of regulatory requirements, permissions, and data exclusivity.
And for both, the solution lies in the same domain: to replace patent
protection with a tailored system of regulatory exclusivity, one that retains
strong incentives for truly socially valuable forms of drug innovations while
curtailing them for others.
Turning from pharmaceutical innovation policy to broader
debates in patent theory, this analysis also provides a distinct explanation
for the consensus view that patents are especially important for
pharmaceuticals. The special case for protection presented by pharma, this
analysis reveals, is a regulatory
artifact rather than, as is commonly thought, the result of any
generalizable technological or economic features of the pharmaceutical
industry. That is, this case stems from the gap between innovation and
imitation costs with respect to the second, data information good, and not the
first, compound information good. More specifically, it is due to the
combined effect of two distinct
regulatory features with respect to data information: how regulatorily
mandated clinical trials massively drive up innovation costs, and how
regulatorily permitted piggybacking on clinical data massively drives down
imitation costs. Absent this combination, there is
little reason to believe that pharma would be very different—that is, with
respect to the compound information good—from other sectors in terms of the
ability of first-mover advantages and secrecy to sustain a robust level of
innovative activity. None of this is to query the regime of regulatory
requirements and permissions. Far from it. Rather, it is to just underline that
it is this regime that makes pharma
special, putting it in need of special innovation-policy support.
This point has crucial import for general
debates in patent theory. In those debates, pharma has long cast a shadow over
the standard conclusion that the overall case for patents—across the economy as
a whole—is uneasy,
and likely at its best for modest protection for small inventors at the
margins. Pharma has long operated as the key
exception to that general rule, one that, so long as it remained unexplained,
gnawed away at confidence in the rule. Showing that this exception can be not
only explained, but explained away,
reinforces the broader conclusion that for most sectors, strong patents are
likely not needed for robust innovation, a conclusion that may now be retained
in its original force, without qualification.
The rest of the Feature proceeds as follows. Part I lays the
theoretical foundations by setting out a framework for the analysis of
innovation policy, clarifying why all innovations need to be conceived as
information goods, identifying the two distinct (compound and data) information
goods at issue in pharmaceutical innovation, and specifying their divergent
technological and economic features as relevant to innovation policy in theory.
Part II then turns to analyzing how the two information goods are presently
treated by pharmaceutical innovation policy in practice. It begins with a
sketch of the technological and institutional pipeline of pharmaceutical
innovation, and the roles played by patents, FDA regulatory requirements, and
data exclusivity. It then details the coordination and incentive functions that
patents and data exclusivity do (or do not) play with respect to each of the
two information goods along the innovation pipeline. It shows that patents play
only a modest role in directly protecting the compound information good.
Meanwhile, patents serve more significant functions for the data information
good, but they do so only indirectly.
Part III then evaluates how well this system of indirect—and
thus misaligned—protection performs. It finds that for each of the two main
tradeoffs raised by exclusionary incentives—access costs and rent
dissipation—the system performs quite badly indeed. The undue access (and
gaming) costs incurred by “evergreening” practices and the duplication wastes
associated with “me-too” drugs are very high, and in each case they stem from
the basic underlying misalignment between patents and data information. The
most effective way to curb these costs, then, is not so much to improve how
drug patents work but rather to attack the problem at its root and eliminate
the basic misalignment by replacing pharma patents with a revised system of
tailored regulatory exclusivity. Finally, Part IV briefly canvasses three
issues broached by the present analysis that merit future investigation: how to
determine the precise duration and scope of regulatory-exclusivity protection;
whether and how to supplement such an improved system of regulatory-exclusivity
incentives on the “supply” side with better pricing (signals) on the “demand”
side; and whether the role of nonexclusionary innovation policies should be
expanded in this area.