Provided for non-commercial research and educational use only.

Not for reproduction, distribution or commercial use.

This article was originally published in Encyclopedia of Health Economics published by Elsevier, and the copy
attached is provided by Elsevier for the author’s benefit and for the benefit of the author’s institution, for non-
commercial research and educational use including without limitation use in instruction at your institution, sending it
to specific colleagues who you know, and providing a copy to your institution’s administrator.

All other uses, reproduction and distribution, including without limitation commercial reprints, selling or licensing
copies or access, or posting on open internet sites, your personal or institution’s website or repository, are prohibited.
For exceptions, permission may be sought for such use through Elsevier’s permissions site at:

http://www.elsevier.com/locate/permissionusematerial

Garrison L.P., and Towse A. Personalized Medicine: Pricing and Reimbursement Policies as a Potential Barrier to
Development and Adoption, Economics of. In: Anthony J. Culyer (ed.), Encyclopedia of Health Economics, Vol 2.
San Diego: Elsevier; 2014. pp. 484-490.

© 2014 Elsevier Inc. All rights reserved.

 Author's personal copy

Personalized Medicine: Pricing and Reimbursement Policies as a Potential

Barrier to Development and Adoption, Economics of
LP Garrison, University of Washington, Seattle, WA, USA
A Towse, Office of Health Economics, London, UK
r 2014 Elsevier Inc. All rights reserved.

Background: Barriers to Personalized Medicine expectancy) based on patient characteristics, whereas a

More than 10 years have passed since the completion of the first
sequencing of the human genome in 2001. Since then, there
has been continued and growing interest in the potential to use
this new genetic information to better predict patient response
to therapy. A variety of terms have been coined to describe this
potential: personalized medicine (PM), stratified medicine,
tailored medicine, and individualized therapy, among others.
This article will adopt PM as the descriptor, more because of its
currency and popularity than because of its more accuracy than
other terms. PM has been defined many a times as ‘‘providing
the right treatment to the right patient at the right time.’’
Arguably, that is the aim of all medical therapy: the important
difference in PM is the use of a new biomarker-based diagnostic
test to further define and identify a subgroup of patients (called
‘stratification’) for whom the treatment performs better – in
terms of either cost-effectiveness or benefit–risk balance.
Incentives for the development and use of PM raise a number
of interesting economic issues.
In the year 2000, Francis Collins, the current head of the
US National Institutes of Health, said: ‘‘In the next five to
seven years, we should identify the genetic susceptibility fac-
tors for virtually all common diseases – cancer, diabetes, heart
disease, the major mental illnesses – on down that list.’’
Clearly, this has not come to pass. In 2005, a more guarded
assessment in the report ‘Personalized Medicine: Hopes and
Realities’ from The Royal Society cautioned: ‘‘Pharmacoge-
netics is unlikely to revolutionize or personalize medical
practice in the immediate future.’’
Before considering potential reasons for this lack of ex-
pected progress, it is useful to define some relevant biological,
epidemiological, and clinical concepts and terms:
• Pharmacogenetics versus pharmacogenomics – the former
is the study of how people’s genetic makeup affects their
response to medicines, whereas the latter is the application
of genomic concepts to the development and clinical
application of pharmaceuticals.
• Genotype versus phenotype – the former represents a
person’s genetic makeup, as reflected by his or her deoxy-
ribonucleic acid (DNA) sequence, whereas the latter is an
observable trait or characteristic of an organism (that may
or may not be inherited).
• Germline versus somatic mutations – the former are
heritable variations, whereas the latter are acquired (e.g., in
cancer).
• Biomarkers – they are a broad array of biological indi-
cators, including genetic variants, proteins, and endogen-
ous metabolites, among others.
• Predictive versus prognostic biomarker – a prognostic
biomarker is used to project patient health (e.g., life
predictive biomarker predicts response to an intervention
(e.g., a drug).
The slower-than-expected progress over the past decade
could be for a number of reasons – scientific, regulatory, and
economic. It does seem clear that the science is more difficult
than first hoped. First, it is always important to remember that
the science behind drug development is complex and un-
certain. Almost 9 out of 10 new medicines under development
fail between Phase 1 and Phase 3. Despite increasing amounts
spent on drug development, industry productivity, as meas-
ured by approved new molecular entities, has been stagnant in
recent years. The most recent estimate from the UK Office of
Health Economics is that the average cost of drug develop-
ment has grown to approximately $1.5 billion US$2011 per
new compound. Clearly, the scientific unknowns and chal-
lenges are substantial.
Adding the parallel development of a predictive, bio-
marker-based test may reduce one scientific challenge but adds
another that involves its own uncertainties. Furthermore,
prediction based on genetic makeup is generally imprecise.
Although a small number of single mutations (monogenic
diseases) lead to specific health conditions (e.g., Huntington’s
disease), most complex diseases (such as diabetes) are affected
by a large number of genes. Although an entire genome can be
sequenced, it is only the beginning of understanding the
biological function of most genes. Although some traits, such
as height, are highly heritable, others are not: indeed, twin
studies indicate that genes account for only approximately a
quarter of the variation in lifespan – perhaps the ultimate
measure of health. Clearly, gene–environment interaction is a
very important influence. Regulation of drugs and devices may
also be a barrier: the approval pathway for new pharmacoge-
nomic tests has not been defined until recently and the
standards of evidence for clinical utility for combination
products (i.e., drug and test) are still being debated. Further-
more, there is not a level playing field between in vitro diag-
nostic tests (IVDs), which need regulatory approval for
marketing, and laboratory-developed tests (LDTs) also called
in-house tests (IHTs). The evidentiary requirements and
quality controls for IVDs are much greater. In Europe, public
health providers face a different regulatory regime that does
not require each test to be approved. Yet, these different tests
and providers compete in some PM applications.
But there are also some potential economic barriers that
could increasingly be a factor given growing efforts to control
medical spending in most developed countries. The incentive
issue for existing marketed drugs is fairly obvious. Once a new,
patented medicine is on the market with prices and re-
imbursement established around the world, the manufacturer
has a very limited economic incentive to discover the subset of

484 Encyclopedia of Health Economics, Volume 2 doi:10.1016/B978-0-12-375678-7.01210-4

 Author's personal copy
patients in whom the drug works best if reimbursement will
then be restricted to the subset. The problem is that re-
imbursement prices for drugs – especially outside the US –
tend to be rigid and inflexible during the period of the patent.
Thus, even if most of the aggregate health benefit – and hence
economic value – is concentrated in a subset of patients, the
manufacturer will receive or capture a much smaller share of
that aggregate value if reimbursement is restricted to that
subset but the reimbursed price does not rise to reflect their
higher average health gain. From a longer term dynamic per-
spective, if price inflexibility holds and there is a strong pos-
sibility that the targeting test would be forthcoming, then drug
manufacturers may have much less incentive to develop the
drug in the first place. This has long-term, dynamic impli-
cations for the incentives for R&D investments in PMs.
However, some targeting using tests might actually facilitate
R&D and the demonstration of efficacy and safety, although
this is difficult to predict a priori unless the mechanism of
action of the drug is closely linked to a biomarker, for ex-
ample, as was the case for imatinib (Gleevecs, Novartis) for
chronic myelogenous leukemia (CML).
The rigidity of diagnostic reimbursement could be an even
larger economic barrier – especially for companion diag-
nostics for drugs that are already on the market. In the US and
most EU health-care systems, reimbursement for diagnostics is
based on an administered pricing system linked imperfectly to
the expected marginal cost of production and distribution,
meaning that some diagnostics might garner profits, whereas
others might just break even or even lose money (but persist
for other business reasons). Thus, there is a limited incentive
to incur the substantial fixed costs of evidence generation that
would be necessary to demonstrate the clinical utility of the
biomarker-based test in combination with the drug.
This article focuses on economic issues related to pricing
and reimbursement policies as a potential barrier to the de-
velopment and adoption of new, innovative PM technology.
There are other important economic issues in PM, such as the
impact on drug development costs, the relation to physician
incentives to test and to follow test results, and issues related
to targeting as a strategy for late entrants into a drug class. The
next section summarizes the key theoretical issues. This is
followed by a discussion of some key examples of PMs that are
in use. The relevance of literature on the cost-effectiveness of
PMs is discussed next. The article ends by identifying six major
policy challenges. The general conclusion is that pricing and
reimbursement systems will need to implement more flexible
value-based rewards for PMs if the appropriate amount of
R&D and evidence generation is going to be supported.
Economic Incentives for Personalized Medicine
Development: A Framework
To this point, only 10–15 PM tests enjoy a significant volume
of use, and the amount of evidence about their health and
economic impact is limited. But many have been the subject
of a cost-effectiveness analysis (CEA) (Wong et al., 2010), and
most are considered cost-effective by usual standards – or they
would not be in use in health systems. For the purpose of
this article, the more interesting and relevant work is the
Personalized Medicine 485
theoretical analysis of economic incentives and their policy
implications for PM R&D.
This work has addressed two major questions. First, is it
likely that current market structures for innovative PMs and
their companion diagnostics will produce the optimal amount
of PM development and use? Second, if the current situation is
likely to be suboptimal, what could be done to improve it? It
is important to recognize that these two questions include not
only short-term questions about static efficiency but also long-
term questions about dynamic efficiency that affect the sus-
tainability of both PM and the patented medicines industry as
a whole, especially because PM is often cited as the ‘new
paradigm’ for drug development.
In an article published in 2002, Danzon and Towse de-
veloped a formal model of pharmacogenetic testing to exam-
ine the conditions under which the development of these
targeted interventions was likely to be socially optimal. The
model uses the common assumption of a societal willingness
to pay for health gains (i.e., a threshold amount), as is often
used in pharmacoeconomic CEA. However, they let the share
of the gain captured by the drug manufacturer vary. The es-
sential elements of their model are intuitively straightforward.
Imagine that a diagnostic test can stratify a patient population
that has previously received a drug into responders (R), those
who benefit from the drug, and nonresponders (N), those
who do not benefit. Before the test is available, both re-
sponders and nonresponders received the drug at price P. In
Danzon and Towse’s model, testing provides social value in
several ways, including (1) avoiding drug spending on the
nonresponders and (2) avoiding the costs and adverse health
effects of adverse events in nonresponders. Of course, the
value of the product among responders exists in either scen-
ario. The key drivers of the social value of testing are the
averted costs of adverse events times the share of N, and the
cost of testing both R and N groups. Testing will generally be
socially beneficial if the aggregate cost of testing is less than
costs savings from not using the drug plus avoiding adverse
events in N.
Their model identifies the key determinants of the in-
centives for drug manufacturers and payers to embrace phar-
macogenetic testing. They also note implications for test
developers and drug development more generally. Clearly, it
will be critical for drug manufacturers to be able to capture a
large share of the aggregate value created after the test is
introduced. This generally means that the price paid for re-
sponders in the testing scenario must rise roughly in pro-
portion to the rise in average patient benefit (due to
eliminating adverse events and/or nonresponders), compared
with the no-testing scenario, and also the price of test must be
modest, relative to the cost of treatment. Although payers
would generally prefer to pay less, in theory, they would be
willing to accept a situation where the total amount paid out is
the same if the net health benefits (i.e., among the responders)
are the same as long as the costs of testing and the savings
from avoiding adverse events in nonresponders are factored
in. Danzon and Towse conclude: ‘‘The willingness of payers to
award higher prices for targeted benefits ... will be essential to
retaining neutrality in investment incentives’’ (p. 10).
Danzon and Towse (2002) also note that it is possible that
the testing scenario could produce an eventual market size
 Author's personal copy
486 Personalized Medicine
that is so small (in terms of total revenues to the drug
manufacturer) that it will not be sufficient to justify the costs
of drug development if that cost is fixed (that is, independent
of the size of the patient population). In practice, however, the
US Food and Drug Administration (FDA) permits fewer and
much smaller trials for orphan drugs. And government sub-
sidies and other incentives may be needed in these situations
for socially optimal investment. They also suggest that
with free entry into the business of developing tests, manu-
facturers – with this new genetic knowledge – will have
an incentive to incorporate testing into the drug development
paradigm, which might also reduce the costs of that
development. This is because in most markets (with the
notable exception of the US), it is not possible to increase
prices once a drug is launched because of rigid government
price regulation. Thus, ex post targeting leads to lower volume
but does not guarantee an increased price to reflect the greater
health gain per patient in the smaller population. To date,
although it is clear that manufacturers are increasingly con-
sidering testing as part of drug development the aggregate
impact has not been large over the past 10 years.
Danzon and Towse assume the price and availability of the
test is a given. They focus on the social benefits of health gains
and of any reductions in cost. Considering many of these same
issues in a less formal analysis, Garrison and Austin (2007) build
on this analytical approach by analyzing the incentives for both
the diagnostic company and the drug company for the codeve-
lopment of companion diagnostic test (Dx) and a first-in-class
drug therapy (Tx). They also explore an additional potential
benefit of PM – the ‘value of knowing.’ They consider six scen-
arios that represent combinations of the following four factors:
1. whether Tx and Dx pricing reimbursement are value based
or cost based, and whether they are flexible over time;
2. timing – whether Tx is already on the market, i.e., ex post
versus ex ante;
3. whether intellectual property protection – to prevent
copycats – is a barrier to entry;
4. the competitiveness of insurance market over short versus
long term.
Although the overall result in terms of the importance of
flexible drug pricing and reimbursement is the same, Garrison
and Austin extend the model by adding some value creation for
the ‘value of knowing,’ i.e., that the test–drug combination will
be of higher value to the responders as they will know they will
benefit. Assuming that they are risk averse, this reduction in
uncertainty should give them greater peace of mind. This makes
the total pool of value created larger: how this aggregate value is
divided among patients, payer/insurers, drug manufacturers,
and test developers is an important question. They also em-
phasize that although pricing and reimbursement for new drugs
could be considered ‘value-based’ in the US and the key mar-
kets in the EU during the patent life, reimbursement for diag-
nostic tests tends to be a more rigid, administered pricing
system, which could be called ‘cost based.’
Their illustrative model considers a case where 20% of the
users are responders and 80% are nonresponders. In the ab-
sence of a test, the value created in the 20% is essentially
spread over the 100% (subject to adjustment for adverse
events in the nonresponders) by setting the price based on the
average benefit (including the nonresponders). In their base
case scenario, with no Dx available, the (patented) Tx captures
all of the value created (given the willingness-to-pay thresh-
old). The five other scenarios explore variations on the four
assumptions listed in this article while recognizing that the
total value created is now greater due to the value of knowing.
It does, then, become ex post a different zero-sum game, i.e.,
after the diagnostic becomes available – there is now a slightly
bigger pie to divide up. The implications of the six scenarios
are intuitive and straightforward.
If the price of the Tx is fixed, and the Dx enters the market
ex post, the revenues going to the Tx manufacturer would fall
dramatically. The Dx may capture this or not, depending on
whether its pricing is regulated and at what level. The Tx
manufacturer would suffer a severe revenue loss and has a very
limited incentive to encourage Dx entry – unless the value
could be recaptured by the Tx manufacturer by owning the Dx.
Indeed, this is the circumstance for most drugs already on the
market: neither Tx manufacturers nor Dx developers have a
strong incentive to develop a test that identifies responders.
Oncotype Dx for predicting breast cancer recurrence, which is
discussed below, is one exception: the manufacturer avoided
the cost-based reimbursement system in the US and charged
approximately $3500 for the test. It is true that some panels of
diagnostic tests might be lucrative under this system, but this
would seem to be less likely true for the novel complex
diagnostics needed for PM if their payment is based on the
summation of the expected costs of the analytic steps as op-
posed to a measure of the incremental health gain.
If the Tx manufacturer’s drug comes to market in combin-
ation with a companion test, then there are several different
possibilities. If the Dx is subject to marginal cost-based re-
imbursement, then the manufacturer will want to capture as
much value as possible through the drug price, which has much
more flexibility at launch in most countries. If the Tx manu-
facturer also owns the Dx, then, in theory, the value capture
could be split arbitrarily between the two. But, in practice, given
administered pricing for the Dx and strong intellectual property
protection for the drug, the incentive probably remains to
capture as much value through the drug price as possible.
These scenarios illustrate several key points about the
economic barriers that companion Dx–Tx products face. First
and foremost is that inflexible pricing and reimbursement,
which does not adjust to reflect value created, could under-
mine the rewards for developing PMs. Second, manufacturers
would ideally enter the market with a combination product
that has been clinically tested and validated ex ante as a
combination. Flexible, value-based pricing and reimburse-
ment would appear to be a necessary condition for en-
couraging optimal investments to produce more PM; of
course, it cannot guarantee a large volume of PM development
because scientific, regulatory, and drug development realities
represent constraints.
Personalized Medicine Products
Although some have been disappointed by the slow progress
in PM over the past 12 years, there has been a gradual accu-
mulation of PM products, so that now between 10 and 20
 Author's personal copy
Personalized Medicine 487

Table 1 Companion diagnostic testing in PM

Technology Economic and testing features

HER2 testing for breast cancer
BCR-ABL testing for chronic
myelogenous leukemia (CML)
Oncotype Dxs (Genomic Health)
for breast cancer recurrence
EGFR mutation testing in
nonsmall-cell lung cancer
(NSCLC)
HLA-B5701 allele testing for
abacavir in HIV
KRAS testing in colorectal cancer
PreDxs (Tethys Biosciences)
diabetes risk test
ALK mutation testing in NSCLC
A low-cost immunohistochemistry (IHC) (approx. $100–$200) test for human epidermal growth factor receptor
2 (HER2)-positivity was used in the initial clinical trial program and was provided by diagnostic companies.
Subsequently, a higher cost and more accurate test ($300–$500) was developed (called ‘FISH’) and is in use.
Approximately 80% of initial testing is done with IHC, with FISH retesting for patient with equivocal results.
The drug manufacturer receives nearly all of the economic value created by the combination from the drug
trastuzumab (Herceptins, Roche)
An example of an ex ante test (breakpoint cluster region-Abelson (BCR-ABL) gene) closely tied to the
development of the drug: large majority of value capture by the drug imatinib (Gleevecs, Novartis). A second,
BCR-ABL test is used to monitor for resistance and assignment to second-line therapies
An example of a relatively high-cost, value-capturing test aimed at avoiding unproductive chemotherapy
An example where the stratifying mutation (epidermal growth factor receptor (EGFR)) was identified in trials
that also included test-negative patients
Example of a test to identify patients who are more likely to suffer a severe adverse reaction to the HIV drug
abacavir (Ziagens, ViiV Healthcare)
The KRAS mutation predicts which patients will not respond to two different monoclonal antibody treatments
for colorectal cancer. The biomarker was identified after the products were on the market
This multimarker test identifies which prediabetic patients are at high risk of progressing to Type 2 diabetes: it
indicates whether to begin prophylactic treatment with metformin
Example of the drug crizotonib (Xalkoris, Pfizer) that targets a small subset (approximately 4%) of patients in
disease condition with significant unmet medical need. It offers substantial survival gains in the subset, but
with high testing cost per identified responder that must be factored in

notable PM products are available. Many of the PM products
have been in oncology where it has been possible to link
somatic mutations to chemotherapeutic response. Table 1
presents some examples of PM products that illustrate the
range of issues that arise in combining companion diagnostics
with drugs to achieve PM. More details on three of these ex-
amples are provided in the next three paragraphs.
Trastuzumab (Herceptins, Roche), a biological compound
for the treatment of human epidermal growth factor receptor 2
(HER2)-positive breast cancer, has been called the first ‘poster
child’ for PM. It was first approved in the US for the treatment
of metastatic breast cancer (MBC) in 1998. Longer term (3-year
follow-up) trials were initiated in early-stage breast cancer
(EBC) and marketing approval was received in 2004. It provides
a number of useful lessons as a case study in PM. First, these
combination products can have a long gestation and life cycle.
The potential of the HER mechanism was discovered in the
early 1980s, but it took approximately 15 years to yield a viable
compound combined with a predictive test. Second, the EBC
indication, which was approved 6 years later than the MBC
indication, produced much larger per-patient health gains
and benefited many more women in the aggregate. Because
the initial price of trastuzumab for treatment of MBC was set
closer to the implicit willingness-to-pay threshold for cost-
effectiveness (4US$100 000 per quality-adjusted life-year
(QALY)) in the US, treatment in the EBC indication was much
more cost-effective (oUS$30 000 per QALY) (Garrison et al.,
2007). So what appeared to be high-cost medicine for its
initial approved indication can be reasonably cost-effective
over the entire product life cycle. The company might well
have priced it higher if the favorable results in EBC would
have been anticipated. Another complexity of the trastuzu-
mab story is the persistence of basic issues about testing
strategy. Two general types of tests are in use – a cheaper
immunohistochemistry (IHC) test and a more expensive and
more accurate fluorescence in situ hybridization test (FISH).
IHC is the more common test and is used in 80% of all initial
tests in the US. Even after all these years of experience, the
optimal companion testing strategy is still under debate due
to uncertainty about real-world test performance: approxi-
mately 80% of women with breast cancer receive the IHC test
initially and some are retested with FISH.
Imatinib (Gleevecs, Novartis) is the unusual example of a
PM combination product developed through rational drug
design. It has been heralded as a virtual cure for many patients
with CML. It has the distinction of having one of the fastest
approval times by the FDA. The breakpoint cluster region-
Abelson (BCR-ABL) enzyme that promotes cancer cell devel-
opment appears only in cancer cells, can be identified by a
test, and can be blocked. Once again, the large majority of the
value created is captured by the drug, which has been esti-
mated to be cost-effective in the US with a cost per QALY of
approximately $50 000.
Oncotype Dxs for predicting breast cancer recurrence
could be considered the poster child for a value-capturing PM
test at approximately $3500 per test in the US. It is based on a
‘21-gene signature’ that was constructed through retrospective
analysis of historical tumor samples to generate a patented
index to predict the likelihood of recurrence. The major eco-
nomic benefit is that it avoids chemotherapy costs and side
effects (including the risk of death) in women with EBC. The
manufacturer was able to circumvent usual coding and pricing
practices in obtaining the US Medicare reimbursement. The
alternative would have been to use ‘code stacking’ of analytic
steps, such as RNA extraction, reverse transcription, gene
amplification, and interpretation and report. But this would
have resulted in a payment level of only approximately
US$540 (Gustavsen et al., 2010).
 Author's personal copy
488 Personalized Medicine
Personalized Medicine, Cost-Effectiveness Analysis,
and Pharmacoeconomics
In general, the methods of economic evaluations in PM are no
different than standard CEA. In the usual case in pharmacoe-
conomics, the cost-effectiveness of a new medicine is assessed
in a population with a particular disease (such as diabetes or
rheumatoid arthritis) for the subset of patients who use the
drug. A standard pharmaceutical CEA compares the impact of
the use of a new medicine on health outcomes (usually
measured in terms of QALYs gained) and on medical costs. The
impact of PM further segments this subpopulation through the
use of the biomarker-based test. The CEA, therefore, changes to
the question of the economic impact of the use of the com-
bination of the test and the treatment versus using the drug (or
another treatment) without testing in the full population. The
test is often called a companion diagnostic, and the pair has
been called a ‘codependent technology’ (e.g., by regulators in
Australia). The overall standard analysis in the field of phar-
macoeconomics remains ‘cost-utility analysis’ with the primary
metric being the incremental cost-effectiveness ratio (ICER) in
terms of cost per QALY gained, although it is not used by payers
in all jurisdictions, for example, Germany.
Given the cost of the test and the cost of the treatment, the
usual principles of CEA apply, except that the cost of the
testing and outcomes for all test recipients must be included in
the analysis. It has been argued that the QALY may not capture
the utility gain that patients may receive from the value of
knowing, i.e., the reduction in uncertainty in terms of whether
the medicine will work well for them. From a modeling ap-
proach, this reduction amounts to less uncertainty about the
value of the drug (assuming the test has high accuracy). If this
were to be included, some add-on or adjustment to the QALY
gain would be needed: there are various preference elicitation
techniques in economics that can be used to elicit willingness
to pay for such product attributes.
Besides creating value by reducing uncertainty, a targeted
Dx–Tx combination could create further value in a population
beyond what is captured in the QALY and ICER in, at least, two
other ways. First, suppose that some responders were non-
compliant because they were not sure whether or not they were
responding. A companion Dx could increase their confidence
about response and hence compliance with Tx, producing
better outcomes in those patients, and thus creating more value
at the aggregate level. Second, in the situation where the Dx is
not available, some patients who would be responders might
not actually choose to go to the doctor for the broader diag-
nosis. With an available Dx, however, they might seek this care.
Thus, overall uptake would increase. Measuring this aggregate
value gain would require a somewhat more sophisticated
model than is typically used in CEA for a typical patient: a
population-level uptake and use model would be needed, more
akin to a budget impact model. But it is feasible to model and
estimate these additional sources of value.
Regulatory and Policy Issues and Implications
There are several major scientific challenges in the develop-
ment of PM, including the low probability that a new
mechanism of action will work, the challenge of parallel de-
velopment of a new companion diagnostic, and the un-
certainty of genetic prediction of complex traits. Clearly, the
scientific unknowns and challenges are substantial. None-
theless, it is important to ask what can be done to optimize
the economic incentives. The following six potential chal-
lenges need to be researched, and, if appropriate, policy
changes made.
Flexible Value-Based Pricing for Drugs
Economic theory would suggest that payers and pricing and
reimbursement authorities should allow for flexible pricing
(both up and down) for drugs at launch and postlaunch if the
evidence suggests that they can be targeted in a narrower pa-
tient group or used in a number of different indications or
subgroups of different value. The UK 2009 Pharmaceutical
Price Regulation Scheme, for example, included (1) provisions
for ‘flexible pricing’ that allowed drug developers to seek ap-
proval from the National Institute for Health and Clinical
Excellence for higher prices when evidence of value increased
and (2) provisions for new indications to be launched at
different prices (higher or lower) than existing indications.
However, neither of these provisions has been used to date in
the UK. This may reflect uncertainty on the part of companies
as to how the policy would be applied or the linkage of the UK
market via parallel trade and reference pricing to other EU
markets where such provisions do not apply.
Regulatory Flexibility in Drug–Test Combination
Development
Given the evidence on drug development failure rates,
pharmaceutical companies can be expected to be resistant to
increases in development costs caused by adding a test into the
development program (notwithstanding the potential advan-
tages of ex ante vs. ex post stratification for the ability of the
pharmaceutical company to extract rents). In the case of the
drug–test companion development, one might, therefore, ex-
pect that it is important initially that the delivery of a proto-
type assay for use in Phase 3 development does not call for
significant investment in advance of being in position to rec-
ognize the efficacy or otherwise of the drug itself in Phase 2. To
achieve this would require flexibility in the regulatory hurdles
for tests as part of drug development. (And this links to the
importance for achieving socially optimal outcomes of having
an environment in which better quality follow-on tests are
incentivized to enter the market.) This flexibility in the drug
development process may require the payer’s pricing and re-
imbursement arrangements for drug–test combinations to
concentrate on the evidence from the Phase 3 drug develop-
ment trial and not require a patient randomization to use
the test (e.g., the ‘double randomized trial’ proposed in the
Australian government’s guidelines for codependent technol-
ogies.) As the model changes over time, drug developers may
not have to codevelop new tests and biomarkers but can draw
on existing ones. In this context, the key issue will not be for
the drug developer: instead, it will be necessary to ensure that
diagnostic developers have an incentive to bring improved
 Author's personal copy
tests to the market and will obtain value-based remuneration
for them.
Flexible Value-Based Reimbursement for Diagnostics
Testing that enables better choice of treatment or improved
disease management may generate downstream health effects
of extended life and/or improved quality of life. In addition, a
diagnostic test can enhance the level of information about a
specific clinical condition or health state, and so reduce or
eliminate uncertainty. This may have value independent of any
ability to improve treatment choice – although insurers’ will-
ingness to pay for this may be unclear. Historically, pricing
and reimbursement systems for diagnostics have been focused
on the expected cost of making and conducting the test (which
may depend on the technology platform used) and not
the value delivered. This has meant that the price of a new
diagnostic is often based on the price of existing tests (‘cross-
walking’) with similar clinical use or with similar character-
istics or on production cost based on analytic steps. For
example, in the US, a number of diagnostics are reimbursed
via code stacking by reporting a combination of reimburse-
ment Current Procedural Terminology codes describing
laboratory protocol stages. Theory would suggest that both
diagnostic-dedicated and drug health technology assessment
(HTA) processes should use a common, consistent, and
comprehensive approach to assessing value to enable the
development of pricing and reimbursement arrangements
capturing the full incremental value offered by those tech-
nologies. Policymakers also need to consider the incentive for
companies to invest in evidence collection to raise the stand-
ard of clinical utility data available to support the case for
using a companion test. Such incentives might come from
rewarding competing tests that supply evidence that they
bring improved health gain and information through greater
accuracy.
Dividing the Combination Value between the Drug and the
Diagnostic
Accepting for now the notion that flexible value-based pricing
is desirable for both an innovative PM and an innovative
companion diagnostic raises the question, how can the
problem of synergistic/codependent technologies noted above
be best resolved? As a starting point, it may be easier to start
with the case of the ex post, ‘standalone’ test. Recall that the
average development cost of a new drug is $1.5 billion
US$2011, and the productivity of drug R&D has been de-
clining over time. In contrast, the cost to develop a new bio-
marker-based assay is in the order of $10 million to $50
million – at the high end if additional clinical trials are needed
to validate the biomarker-based test. Consider a hypothetical
case where an existing biopharmaceutical is on the market and
earning $1 billion in annual revenues despite having only a
50% response rate. Suppose a reliable biomarker-based test is
invented that can predict the responders. In theory, the payer
would be indifferent to giving them $500 million and perhaps
a premium for the value of knowing (i.e., reduced uncertainty
for both the payer and the patient). Clearly, if the drug
Personalized Medicine 489
manufacturer thought in advance that this might happen, the
prospect of lower revenues would have reduced the likelihood
of them making the initial investment. One could argue,
however, that it was the drug manufacturer’s invention that
created the health gain in the responders. Then, the extra value
created by the Dx manufacturer consists of (1) avoiding any
adverse event treatment costs and any related health losses in
the nonresponders plus (2) the ‘value-of-knowing premium.’
(Note that if the premium were, say, 5% of total health gain, it
would be worth $50 million: an amount that could support
substantial evidence generation.) With flexible value-based
pricing, this logic would argue that the drug innovator’s price
would be roughly doubled (i.e., to $1 billion in revenues) and
the Dx innovator would receive a reward in relation to cost
savings and QALY gains in the nonresponders plus a premium
for the value of knowing (i.e., $50 million). This split is ar-
bitrary in a static sense, but it can be argued that it reinforces
dynamic efficiency by considering the relative size of the
investments needed to develop a drug versus a diagnostic.
Clearly, this issue deserves further research.
‘Follower’ Diagnostic Tests ‘Piggyback’ on Clinical Utility
Evidence
Evidence on drug effectiveness and value is generated by the
drug developer in order to obtain regulatory and pricing ap-
proval. The drug company has a patent (no one can copy the
product) and also regulatory exclusivity on the data they
generate (no one can reference the data to get regulatory ap-
proval by claiming that their product does the same). The
situation is more complicated in diagnostics: patents may be
less robust, regulatory hurdles are lower, and so it is more
difficult to prevent others appropriating the benefits of evi-
dence from studies the diagnostic manufacturer has paid for.
As a consequence, there may be underinvestment in evidence
generation. Data exclusivity could be given to evidence for
diagnostics. This would require ‘follow-on’ tests to replicate
the evidence generated by the ‘first-in-class’ test (as is the case
for drugs). However, under current diagnostic regulations in
the US and EU countries, LDTs (and in the EU any tests
provided by public health providers) have to meet lower
regulatory hurdles and so could not be prevented from pig-
gybacking on clinical utility evidence. In addition to the data
exclusivity requirement, there may, therefore, need to be an
expectation that payers will pay more for the test with the
stronger evidence base (i.e., that they will ignore tests without
an evidence base when making HTA). Of course, a balance has
to be struck, as with patenting. The objective is not to delay
competition to provide innovative tests but to ensure initial
providers have the potential to earn a return if they have
evidence to support claims of value. Research is clearly needed
on both effective incentives to prevent piggybacking for a least
a period of time and how long they should be put in place.
Flexibility in Diagnostic Test Evidence for Reimbursement
Given the US and EU regulatory environments, a pragmatic
approach could be taken to collecting evidence on the clinical
utility of diagnostics, for example, using small randomized
 Author's personal copy
490 Personalized Medicine
studies (if not built into Phase 3 of drug development), which
lead to conditional reimbursement approval and evidence-
based reimbursement rate to incentivize manufacturers plus
real-world, postlaunch data collection. To facilitate real-world
data collection, increased investment in national e-health re-
cords would, certainly, help, although the overall investment
case for such records is much greater than simply facilitating
assessment of the benefits (or otherwise) of developing and
validating diagnostic tests.
Conclusion
Although there is a general perception that the growth of PM
has been slower than hoped, it is not clear how much of this
perceived shortfall is due to scientific and regulatory con-
straints versus economic incentives. Many of the PM products
have been in oncology, where it has been possible to link
somatic mutations to drug response. The science is more dif-
ficult outside of oncology. There are, clearly, problems with
economic incentives in this area as well. However, the em-
pirical evidence on this is limited by the challenge of con-
structing the counterfactual. These may reflect the scientific
challenge: not enough drug-diagnostic combinations have
received regulatory approval to allow exploration of the issues
around commercial success. However, the relatively small
number of regulatory approvals may also reflect perceived
commercial limitations as well as scientific challenges.
Over the past dozen years, the thinking about PM pre-
dictive tests has evolved from thinking primarily about using
genetic mutations or combinations as predictors of clinical
response to a wider range of ‘biomarkers,’ and even to chan-
ging the name of the whole area to ‘molecular diagnostics.’
Owing to the rapidly falling cost of sequencing an individual’s
entire genome (e.g., projected to be less than $1000 in the near
future), a new set of economic issues are arising around the
question of how to use this plethora of data from whole
genome scans. These new economic issues have not been
addressed here but will need to be better understood in future
discussions of the economics of PM. In the extreme, the
emerging pharmacogenomics diagnostic industry will change
drastically, making it possible for a whole genome test to be
performed at birth on an individual. Still, testing for somatic
mutations would be needed and incentives would be required
to invest in providing evidence that a particular set of bio-
markers (including genes) is predictive of disease.
See also: Adoption of New Technologies, Using Economic
Evaluation. Biopharmaceutical and Medical Equipment Industries,
Economics of. Cost-Effectiveness Modeling Using Health State Utility
Values. Economic Evaluation, Uncertainty in. Information Analysis,
Value of. Patents and Other Incentives for Pharmaceutical Innovation.
Pharmaceutical Pricing and Reimbursement Regulation in Europe.
Policy Responses to Uncertainty in Healthcare Resource Allocation
Decision Processes. Pricing and Reimbursement of
Biopharmaceuticals and Medical Devices in the USA. Research and
Development Costs and Productivity in Biopharmaceuticals. Value of
Information Methods to Prioritize Research
References
Danzon, P. and Towse, A. (2002). The economics of gene therapy and of
pharmacogenetics. Value Health 5(1), 5–13.
Garrison, L. P. and Austin, M. J. F. (2007). The economics of personalized
medicine: A model of incentives for value creation and capture. Drug
Information Journal 41, 501–509.
Garrison, L. P., Veenstra, D. L., Carlson, R. J., Carlson, J. J. and Meckley, L. M.
(2007). Backgrounder on pharmacogenomics for the pharmaceutical and
biotechnology industries: Basic science, future scenarios, policy directions.
Report of the Pharmaceutical Outcomes Research Policy and Program.
Department of Pharmacy, University of Washington. Available at: http://
sop.washington.edu/porpp/general/reports.html (accessed 08.10.13).
Gustavsen, G., Phillips, K. and Pothier, K. (2010). The reimbursement landscape for
novel diagnostics. Weston, MA: Health Advances. Available at: http://
www.healthadvances.com/pdf/novel_diag_reimbursement.pdf (accessed
08.10.13).
Wong, W. B., Carlson, J. J., Thariani, R. and Veenstra, D. L. (2010). Cost
effectiveness of pharmacogenomics: A critical and systematic review.
Pharmacoeconomics 28(11), 1001–1013.
Further Reading
Blair, E. D., Stratton, E. K. and Kaufmann, M. (2012). The economic value of
companion diagnostics and stratified medicines. Expert Review of Molecular
Diagnostics 12(8), 791–794.
Davis, J. C., Furstenthal, L., Desai, A. A., et al. (2009). The microeconomics of
personalized medicine: Today’s challenge and tomorrow’s promise. Nature
Reviews Drug Discovery 8(4), 279–286.
Faulkner, E., Annemans, L., Garrison, L., et al. Personalized Medicine Development and
Reimbursement Working Group (2012). Challenges in the development and
reimbursement of personalized medicine – payer and manufacturer perspectives and
implications for health economics and outcomes research: A report of the ISPOR
Personalized Medicine Special Interest Group. Value Health 15(8), 1162–1171.
Garrison, L. P. and Austin, M. J. F. (2006). Linking pharmacogenetics-based
diagnostics and pharmaceuticals for personalized medicine: Scientific and
economic challenges. Health Affairs 25(5), 1281–1290.
Keeling, P., Roth, M. and Zietlow, T. (2012). The economics of personalized
medicine: Commercialization as a driver of return on investment.
New Biotechnology 29(6), 720–731.
Meckley, L. M. and Neumann, P. J. (2010). Personalized medicine: Factors
influencing reimbursement. Health Policy 94(2), 91–100.
Ramsey, S. D., Veenstra, D. L., Garrison, L. P., et al. (2006). Toward evidence-
based assessment of laboratory-based diagnostics and genetic test for health
plan reimbursement. American Journal of Managed Care 12, 197–202.
Trusheim, M. R., Berndt, E. R. and Douglas, F. L. (2007). Stratified medicine:
Strategic and economic implications of combining drugs and clinical
biomarkers. Nature Reviews Drug Discovery 6(4), 287–293.
Veenstra, D. L. (2009). The economic value of innovative treatments over the
product life cycle: The case of targeted trastuzumab chemotherapy for breast
cancer. Value in Health 12(8), 1118–1123.