Full text - Igitur

receipttrogleΛογισμικό & κατασκευή λογ/κού

29 Οκτ 2013 (πριν από 4 χρόνια και 12 μέρες)

243 εμφανίσεις





Sebastiaan Lommelaars

2011

The promise of
personalized medicine

Review of the possibilities and hurdles to achieve a cost
-
effective
personalized healthcare system.

Supervisor:
Prof. Dr. R. Bernards, PhD

Molecular Carcinogenesis

NKI
-
AVL




2

|
P a g e


Abstract

Until now we have approached cancer treatment from a histological point of view, fighting tumors
with indifferent cytotoxic drugs. Personalized medicine promises a much needed change to this out
-
dated treatment approach. Here I
review

both the promises and

obstacles
of

targeted therapeutic
s’
development
. Extra attention is put on the financial viability of personalized medicine. By modeling
the development costs and revenue possibilities I
describe how this approach can be cost
-
effective to
pharmaceutical c
ompanies and other parties involved. The individualization of cancer treatment will
greatly improve prognosis

saving patients the detrimental effects of systemic cytotoxic treatment
and relieving some pressure of the great socio
-
economic burden that cancer

puts on our society.





3

|
P a g e


Contents

Abstract
................................
................................
................................
................................
........

2

1. Introduction

................................
................................
................................
..............................

4

1.1 Conventional cancer therapy

................................
................................
................................

4

1.2 The promise of personalized medicine

................................
................................
..................

4

1.3 Cancer biomarkers

................................
................................
................................
...............

5

1.4 Biomarker discovery

................................
................................
................................
............

6

2. Obstacles to personalized medicine

................................
................................
............................

7

2.1 Technical difficulties with personalized medicine

................................
................................
...

7

2.1 Getting suitable tumor tissue
................................
................................
................................

8

CASE: Agendia, market entry of a genetic test
................................
................................
.....

9

2.3 Regulatory clarity
................................
................................
................................
................
12

2.4 Getting the right mindset

................................
................................
................................
....
14

3. Viability of personalized medicine

................................
................................
.............................
16

3.1 Adaptive clinical trials

................................
................................
................................
.........
16

3.2 Drug development costs
................................
................................
................................
......
19

3.3 Economic modeling of R&D

productivity

................................
................................
..............
20

CASE: Vemurafenib, small Phase III
clinical trial design
................................
.....................
20

3.4 Calculating the potential cost reductions of personalized medicine
................................
........
25

CASE: Iressa, Tarceva , Herceptin and the value of a biomarker.

................................
......
26

3.5 Revenues generated by personalized medicine
................................
................................
.....
36

4. Discussion

................................
................................
................................
................................
38

References
................................
................................
................................
................................
...
40

Glossary

................................
................................
................................
................................
......
44

Appendices

................................
................................
................................
................................
..
45




4

|
P a g e


1.

Introduction


1.1
Conventional c
ancer therapy

An estimated 10^16 cell divisions take place in a normal human body
over

the cou
rse of a lifetime,
with m
utations
occurring

about 10^
-
6 times per gene per cell division.
T
his implicates that
,

even in
the absence of external muta
gens such as tobacco smoke or X
-
rays,

every single gene is l
ikel
y to
undergo
mutation

on a staggering

10^10

separate occasions
. Fortunately
cells employ
complex
interconnected
regulatory
mechanisms

to help them maintain pr
ecise

control of the
integrity

of their
genome
. These

need to be disrupted before a cell can
give rise to a malignant tumor
.
Many
oncogenic driv
er

mutations alter components of signaling pathways that cause proliferative signals,
switching on cell growth,
regulate survival,
DNA replica
tion, and cell

division in the tumor. Common

example
s

are mutations in receptor tyrosine kinases, such as the Epidermal Growth Factor Receptor
(EGFR)
,

proteins of the downstream
RAS
family

or in p53, a well known
tumor suppressor protein
.
Other signaling p
athways act to inhibit proliferation, the best known example being the TGFβ
signaling pathway.

D
uring the progression of a tumor, cancer cells will
constantly
meet new barriers
to further expansion. For example when the tumor reaches a size of one or two m
illimeters
nutrients
and oxygen supply will become a limiting factor to
further growth. Each new

barrier, whether
physical or physiological,

must be overcome by the random acquisition of additional mutations.

The
ways in which a gene can be mutated are eno
rmously varied

(point mutations, translocations

and
copy number gains or losses). A
ny genetic accident that can increase, decrease,
or change the activity
of a signaling pathway

is likely to be found somewhere in the increasingly complex catalogue of
gene
t
ic

changes that occur in cancer

(Alberts et al., 2002)

This highly diverse roadway to malignancy
generates an immense
ly

heterogeneous pool of tumors, not only in terms of histology a
nd clinical
outcome, but also on

a molecular level.

As our understanding of these molecular processes
increases, we
are forced

to come to our senses and realize there will be no single cure for cancer,
ever.

We need to find ways
to

categoriz
e

patient populations

and treat their individual
tumor types

with
high

accuracy.


1.2
The promise of p
ersonalized medicine

Until now we have approa
ched cancer treatment fro
m a histological point of view, c
lassifying tumors
based on histo
patho
l
ogical indications

and fighting

them with indifferen
t c
ytotoxic d
rugs.
Chemoth
erapy

ignore
s

the
deep
er

molecular reasons that
lie

at the core o
f

tumor

progression and

destroy
s

any

rapidly
replicating cell in the body, i
ncluding healthy cells. The processes underlying
tumor progression, will also allow the tumor to rapidly adjust to
the
rapeutic agents, creating multi
-
resistant tumors that are no longer affected by conventional therapies.
T
he alternating

oncogenic
driver mut
ations make that seemingly similar tumors will
respond differently to therapeutic
interventions. These struggles
have made it painfully clear that we need to move towards more
specific
therapeutics
to
treat patients at an individual level
. We are now entering the first and most
important stage of personalized

medicine, where we will aim to
group patients into smaller

sub
-
populations, rationally predicting their response to newly designed targeted therapies.
The
individualization of cancer treatment will greatly improve prognosis for patients as treatments will
prove more effective and
spare

patients

from
the current ‘
trial and error’ approach of treatment
5

|
P a g e


selection
, saving vital time

and suffering
.

As cancer treatment puts

an ever growing fi
nancial burden
on our society
, the ability to rationally predict
treatment response will reduce the number of patients
receiving e
xpensive drugs without
any
clinical
benefit
. This socioeconomic

aspect

is expected to be a
strong driver
for patients, pharmaceutical companies and regulatory bodies to move towards
personalized medicine.


1.3
Cancer biomarkers

Despite tumor
heterogeneity, it is possible to classify tumors based
on the activity of key
components of the major carcinogenic pathways.
At this stage it would be far too laborious t
o
allocate each malignant

mutation individually.
A
s we are just beginning to understan
d the processes
underlying tumor growth it will take a long time before we
know the precise roles of each
abnormal
gene in these major pathways. With our current knowledge
we are able to predict therapy response
base
d on key components of cancer
-
associated

pathways. To
detect

this abnormal genomic activity
and predict treatment outcome
it is vital that we

look for new sets of cancer biomarkers, a
development that is already underway.
For example, Epidermal Growth Factor R
eceptor
(
EGFR
)

gene
mutations and increased EGFR copy number have been associated with favorable response to EGFR
tyrosine kinase inhibitors (EGFR
-
TKI)
like Tarceva and Iressa,
in patients with non
-
small
-
cell lung
cancer (NSCLC)

(
Liang et al., 2010
;
Massarelli et al., 2
007)
.

I
t is thought that the competitive binding
of Tarceva and Iressa with the EGF receptor inhibits proliferative downstream signaling.
With this
relatively global insight in the EGF pathway we are able to
improve
survival

for

most NSCLC patients
that ca
rry activating EGFR
mutations

(
Figure
1

AB
).

This
demonstrates that ‘pathway
activation biomarkers’ do not
require a complete understanding
of the pathway.
However
a
complicating factor in this
approach is that pathway
activation biom
arkers usually
cannot distinguish between
upstream or downstream
activation of the pathway.

When
the pathway is activated
downstream of the targeted
intervention point, the drug cannot
inhibit
the proliferative signals of
that

pathway.

For example
m
utation
s
in KRAS, PI3K,
and BRAF,
downstream effectors

of EG
F, have
been shown to predict poor response to TKIs and are
even
frequently mutated in the event of
a
cquired

resist
ance to
these
targeted therapies

(
Figure
1

C)

(Nguyen et al., 2009

; Ludovini et al.,
2011; Schmid et al., 2009; Sunaga et al., 2011)
.
This illustrates the need for strong biomarkers to
support proper use of targeted therapies. As our healthcare system move
s

towards more
personalized medicine the role

of biomarkers will increase accordingly.

Figure
1

Pathway activation

and inhibition
. A:
Activating mutations in the EGF

pathway result in proliferative
signals to the cell
.
These signals in term fuel the
tumors malignant growth. B: TKIs inhibit EGFR from transmitting its prolife
rative
signals, hereby reducing tumor progression C: When the EGF pathway is activated by
downstream components such as KRAS, PI3K or BRAF, TKIs cannot inhibit the
proliferative signals, so tumor growth continues unaffected by the treatment.

6

|
P a g e


Four different types of biomarkers can be distinguished
.

1.

Diagnostic biomarkers

help with tumor classification and most importantly can be used for
early detection of tumor growth. This type of biomarker is less re
levant to the development
of targeted therapies as it is used before treatment.

2.

Prognostic biomarkers

indicate

if patients need additional therapy after surgery of the
primary tumor. In case of very aggressive tumor tissue the physician might decide to giv
e
adjuvant treatment to reduce the chance of tumor recurrence.

3.

Predictive
biomarkers

will then aide the physician in the choice between different targeted
therapies
, predicting the patient’s response to the drugs.
This is the most important cl
ass of
biomarkers as targeted therapies usually come with a companion biomarker of this kind.


4.

Pharmacodynamic biomarkers

help in the last step of treatment choice, indicating to

the
physician the optimal dos
e for every individual patient.


1.4
Biomarker discove
ry

The key components of growth promoting pathways
are vital for the survival of the tumor.

Inhibiting
the up
regulated EGFR growth signal

with TKIs
in patients with NSCLC

frequently results in cancer cell
death. In fact this phenomenon of cancer cell death
after

sudden inhibition of growth promoting
signals is common amongst many types of cancer and is called “oncogene
-

or network a
ddiction


(
Tonon, 2008; Weinstein et al., 2008
; Sharma et al., 2010)
.

This makes the key
-
components

very
suitable targets for ta
rgeted therapies and thus valuable predictive biomarkers.
Thanks to rapid
developments of new genomic technologies,
cancer genomes can now be analyzed at
multiple levels

(
Figure
2
)
.

Large
-
scale
functional genetic screens
and genome
-
wide
RNA interference
are used for identification
of cancer associated genes

and D
NA
microarray analysis is used to follow gene
expression on a
high
-
throughput
genome
-
wide scale
.
Individual

m
utations in the
cancer genome can be found using large
-
scale DNA sequence analysis
whilst larger
copy number gains and losses can be
detected using comparative genomic
hybridization (CGH).
Finally, protein
biomarkers can be identified using tissue
micro ar
rays (TMAs) and reverse phase
protein lysate arrays (RPPAs)
. Relations
between
oncogenes

and the course of the
disease have already been found

using these
genome analysis techniques, ultimately
yielding promising biomarkers
(
Majewski et
al., 2011).

Figure
2

Methods of biomarker discovery.
There are a number of
promi sing ways for biomarker discovery, which are i ncreasingly proving
thei r i mportance for this aspect of personalized medicine.

7

|
P a g e


2.
Obstacles to personalized medicine


As the advantages of personalized
medicine become clearer and supporting
techniques more common

one might
wonder why so few targeted therapeutics
are used today. There are a number of
obstacles to their development that have
not been resolved and some of them will
prove difficult, even in the near future.
O
ne might say we are opening
Pandora’s
Box w
ith our search for targeted therapies.
By
fight
ing

cancer at a more specific level it
turns out that our knowledge of the driving
forces of malignancy

is not always
sufficient. So before we can truly start
performing personalized medicine a
number of obsta
cles need to be overcome

(
Figure
3
)
.



2
.
1

Technical difficulties with personalized medicine

By classifying and treating tumors based on pathway activation we overlook the specific genes and
proteins that might play a different role in individual tumors.
Upstream or downstream pathway
activation is an example we discussed before.

A.

Acquired drug

re
sistance.
Another problem with
the

pathway
based

approach is the fact that
tumors frequently acquire resistance to targeted therap
ies. With EGFR TKIs like G
efitinib and
Erlotinib we target the EGF pathway at the receptor, effectively knocking down the
mali
gnant EGF signal in patients with growth promoting mutations in EGFR. However the life
expectancy of these patients is still poor as many acquire resistance to the therapy at a later
stage

(Nguyen et al., 2009)
.

Inhibitory feedback systems that normally st
op DNA replication
in the event of
serious
mutations are lost well before the tumor is likely to be diagnosed. By
losing these and other
proofreading

mechanisms the tumor becomes
genetically instable
and
highly prone to mutations
. Mutations that bypass our point of intervention at the EGF
receptor or that occur downstream in the pathway will restore the malignant
downstream
EGF signal and
allow for further tumor growth
.



B.

Primary and secondary tumor genome.
Another
property

of mut
ation prone tumors
is the
fact that metastasis in a single patient are likely to continue mutating on their own.
Consequently not all metastasis will respond to the same targeted therapy and it would be
necessary

to diagnose
each individual growth promotin
g mutation. It would be impossible to
obtain biopsies from every met
astasis so for this approach to work we will need to rely on
non
-
invasive sampling techniques

(
Majewski et al., 2011)
.
Ultimately this
will help

physicians

determining

a cocktail of targeted therapeutics
that are combined to destroy all tumor tissue
as effective as possible.
These technical difficulties will be overcome as we gain more
Figure
3

Obstacles to
personalized

medicine.
These obstacles will be
di scussed in this chapter.

8

|
P a g e


experience with
t
argeted therapeutics over time. So for now we should focus our resea
rch
effort on targeted therapies and their
accompanying biomarkers.


2.1 Getting suitable tumor tissue

Because biomarkers are such a vital aspect of targeted therapies, their discovery is one of the main
goals for research. As we noted before, biomarkers are typically discovered by various ways of tumor
tissue analysis. It’s this tissue’s availability that
currently forms a bottleneck for swift discovery of
novel biomarkers.

A.

Tissue preservation.
After collection in the clinic, most tumor tissue is preserved through
Formalin Fixation and Paraffin Embedding (FFPE). This technique is well suited for
convention
al histopathological assessments as it preserves tissue architecture and allows for
easy storage of surplus samples in banks. FFPE, as a clinical standard, has generated a vast
repository of tissue material for long
-
term clinical studies, but unfortunately

it induces
chemical changes and degradation to the DNA, RNA and proteins, making the samples
suboptimal

for biomarker research.

The preservation
issue has raised voices to introduce
fresh tissue preservation as a new clinical standard in molecular patholo
gy which would be
better suited for biomarker discovery, but
this
comes at a price.
The US based National
Cancer Institute has launched a campaign to promote high quality biospecimen collection
and ba
nking for research purposes. In their

guides they
outlin
e the operational, technical,
ethical, legal and policy best practices
for
tissue repositories
.

N
ondiscrimination legislation
such as the US Genetic Information Nondiscrimination Act of 2008 (GINA) will also help by
encouraging people to donate specimens

b
y ensuring patients their genetic information
cannot be misused by any company such as their employer or insurance.
The global
introduction of

new processing and preservation method
s

require a tremendous effort and
moreover the logistics of fresh tissue preservation will be more complex, costly and time
consuming.
For now it’s worth
having a closer look at FF
PE samples and recently several
advances ha
ve

been made to overcome the degra
ding effects.
(Nirmalan et al., 2008; Berg et
al., 2010; Ralton et al., 2011)

This makes the use of FFPE tissue for biomarker research
superior to the use of fresh tissue for reasons such as cost, availability, standardized
technique, easy long
-
term storag
e and most importantly it enables retrospective studies,
thus vastly increasing development speed.


B.

Obtaining tumor tissue.
A second tissue related problem is the fact that it is not easy to
obtain tumor tissue from patients. Most patients enroll clinical
trials at a later stage of their
disease when the tumor has already spread to different sites in the body. Obtaining biopsies
from each metastasis would be impossible to achieve and scanning the bone marrow for
disseminated tumor cells has proven a difficu
lt and invasive procedure. This is why we need
to turn to less invasive methods of tumor cell collection. Recently there have been several
developments in this field that may prove capable of collecting enough circulating tumor
cells (CTC) to allow biomark
er research by methods we discussed before

(Gahan, 2010)
.
Because CTC levels are low, detection requires enrichment and density
-
gradient
centrifugation followed by separation. The two main approaches for CTC detection are
immunological assays using monoclo
nal antibodies or PCR
-
based assays both exploiting
tissue
-

and/or tumour
-
specific antigens. (Gerges et al., 2010 ; Müller et al., 2010) Aside from
the possibility to drastically simplify tumor cell collection, this technique holds another
promising prospec
tive. Early detection of CTCs might help identify patients in need of
adjuvant therapies after successful surgical resection of the primary tumor. Both applications
make it a very promising technique that should certainly be developed in the f
uture.
Anothe
r
option could be to analyze circulating cell
-
free DNA (cfDNA)

shed by tumor lysis. Combined
9

|
P a g e



Agendia is a Dutch molecular diagnostics company, founded in 2003 as a spin
-
off of the
Netherland’s Cancer Institute (NKI). They market the Symphony™ suite, consisting of
MammaPrint®, a breast cancer recurrence assay, BluePrint®, a molecular subtyping assa
y,
TargetPrint®, an ER/PR/HER2 expression assay, and TheraPrint®, a therapy selection
assay. The symphony™ suite helps physicians with various cancer related treatment
decisions.

CASE: Agendia, market entry of a genetic test


with massive parallel tumor sequencing efforts that are currently underway this method
could prove able of revealing many oncogenic driver mutations on a high
-
thro
ughput level
(Bernards, 2010). Similar to the analysis of CTCs, this technique can also be used for
diagnostic purposes in the clinic, as has already been show in colon cancer (Mead et al.,
2011).


C.

Sufficient patient samples.
The last
tissue related
diffic
ulty with biomarker discovery is
finding enough patient samples. To be able to significantly link gene
-
expression to disease
outcome, trials need at least 40 patients who respond positively to treatment and an equal
amount of non
-
responders. Currently most

phase II clinical trials do not hold such numbers
and increasing the number of patients would be very expensive and slow down development
time. It makes more sense to pre
-
screen for candidate genes that should be tested
in later
phase II clinical trials.
Recent developments have enabled us to screen for specific cellular
p
rocesses in mammalian cells using

functional genetic screens. This allows us to screen for
cancer
-
drug responses in the most unbiased way using “gain
-
of
-
function” and “los
s
-
of
-
function” c
onfigurations to identify promising target genes (Bernards, 2010). This approach
will narrow down the amount of genes that are to be tested in phase II clinical trials as
functional genetic screens can only identify causal relationships between genes and d
rug
response.

By pre
-
screening for cancer
-
related genes we can focus our resources on the most
promising entities, thereby reducing the amount of phase II clinical trials and patients

enrolling in them
.



The availability of high
quality tumor tissue will pose a significant obstacle to biomarker
development, but developments in this sector follow on
e

another in a rapid pace, showing promising
possibilities. As the sector matures, innovative clinical trials will also take some press
ure of the

tissue
issue


by reducing the number of trials and patients involved in development of biomarkers and their
subsequent therapies. Such adaptive clinical trials will allow scientists to test for multiple compounds
and
bio
markers in a single tria
l and focus towards entities that show the most promising results as
the trial prog
resses. I
nnovative trial designs

will pose a steep learnin
g curve for both pharmaceutical
co
mpanies and reg
ulatory bodies like the FDA. Adaptive trials will be discussed lat
er in this article

as
we continue here by illustrating the influence of regulatory bodies on personalized medicine.

10

|
P a g e



The MammaPrint® test is a prognostic biomarker

that helps predicting

the risk of breast
cancer recurrence in the first five years after
diagnosis, which is the period in which

chemotherapy produces most of its benefits to a patient. MammaPrint® was developed using
high
-
throughput microarray analysis of tumor tissue and determines

expression profiles of 70
genes to stratify patients at “hig
h risk” or “low risk” of breast cancer recurrence. This gives
physicians a more accurate tool to determine if adjuvant

chemotherapy would be advisable.

Because MammaPrint® is superior to conventional tool for selection of these “high risk”
patients the tes
t is able to reduce chemotherapy overtreatment by 27% (Van ’t Veer et

al.,
2002; Mook et al., 2007). Moreover MammaPrint® has been demonstrated to be
a cost
-
effective strategy to guide adjuvant

chemotherapy treatment

(Chen et al., 2010).

Hereby
MammaPrint
® saves patients the detrimental effects of systemic cytotoxic treatment and
relieves some pressure of the great financial burden cancer puts on our society.

MammaPrint® was the first IVDMIA to obtain 510(k) clearance from the FDA and was made
commercially available in 2008 in the US. Agendia has further established excellent coverage
for MammaPrint through Medicare (CMS) and private US health insurance compani
es.
Surprisingly Agendia is experiencing much greater difficulties in other countries such as the
Netherlands, where the company originated. MammaPrint® is CE
-
certified and already used
by multiple Dutch hospitals following its inclusion in the updated 200
8 guidelines of the Dutch
institute for healthcare improvement CBO. Several leading private insurance companies have
adopted a positive reimbursement policy towards MammaPrint® and the NKI
-
AVL, a
worldwide acknowledged center of excellence and leading inst
itution in advancing cancer
treatment and care, has declared MammaPrint® as their standard of care for breast cancer
treatment. Nonetheless the Dutch Insurance Governing Body (CVZ) has rejected
MammaPrint®’s inclusion in the basic health insurance package.

This decision has great
implications because if there is uncertainty about the ability to recoup the costs of an assay,
labs will not offer it. And if physicians must provide elaborate justification of medical
necessity, the tests will not be ordered. Kno
wing that coverage is a necessary condition to
commercial success, potential investors will hesitate to finance further developments, making
market entry very challenging, particularly to diagnostics companies
like Agendia,
that lack
the financial means of

big pharmaceutical companies.

CVZ’s rejection decision has been a major setback for Agendia, which is now focusing its
marketing efforts on the US, a clear illustration that more coherent worldwide regulation is
needed to assure that all patients worldwid
e receive state of the art cancer treatment. The
Dutch CVZ has declared they cannot grant full coverage to a product that is backed only by
retrospective validation and feasibility studies. Rather they await prospective validation in the
large MINDACT phas
e III trial (Microarray In Node
-
negative Disease may Avoid
ChemoTherapy), which will take till 2014 before generating its first results. Although the
validity and cost effectiveness of the MammaPrint® have been debated, the test has been
developed and exte
nsively validated using a retrospective approach. Also current clinical
results are very satisfactory, making it clear that more patients could benefit from the test.
Unfortunately, unlike the FDA and CMS, the Dutch CVZ is unable to cope with this innovati
ve
approach to cancer treatment and is reluctant to publicly stimulate its use. Agendia, which
has received multiple healthcare innovation awards, is challenged to great lengths in terms of
financing their attempt to further develop and market their produc
t portfolio.



11

|
P a g e



A recent attempt for an Initial Public Offering (IPO) was pulled by Agendia due to "an
extraordinarily volatile period in global capital markets resulting in high levels of uncertainty
and volatility", showing that the financing environ
ment for biotech firms is very challenging.

Agendia
claims

it has collaborated with Roche, Novartis, Sanofi and Pfizer in clinical trials
and is looking to add colon and lung cancer tests to its product portfolio. But to continue
research Agendia will have

to find ways to bridge their upcoming financing deficit. Even with
two of their products
already
generating revenue, they booked a 16.1M euro loss in 2010,
and after a first
-
quarter loss of 5.26 million euro in 2011, they had about 11M euro in cash on
Mar
ch 31. At the current burn rate Agendia would need new funding by the end of the year or
resort to other measures, like reducing staff and R&D activities. Further, Agendia could retry
the IPO, which was

estimated to generate 75M euro

in cash and increase t
heir lobby with the
Dutch government for additional grants and tax incentives.

In a recent elevator pitch Agendia points government officials to the fact that reducing
chemotherapy over
-
treatment will generate huge revenues because women return to work
mu
ch sooner instead of rehabilitating in the hospital or at home. In the Netherlands 13000
women develop breast cancer each year, of which 25% will metastasize at a later stage.
However as much as 75% of women receive adjuvant chemotherapy, meaning that 50%
of
all breast cancer patients receive unnecessary chemotherapy. According to Chief Scientific
Officer, Prof. Dr. R. Bernards, PhD, MammaPrint® could reduce chemotherapy
by 27%,
calculating to 45M euro

in labour productivity per year

(
Table
1
).



Table
1

Net savings to Dutch society after introduction of MammaPrint®.
The costs of MammaPrint® testing approximately
even out with the money saved by
a 27% reduction i n adjuvant chemotherapy. However MammaPri nt® wi l l reduce the
amount of l ost l abour years wi th 650 years annual l y, cal cul ati ng to 45M i n savi ngs to the Dutch soci ety per year.

Upon close examination we estimate the net

savings to be closer
to 65M euro

annually based on
recent research on the effects of chemotherapy of women’s working life. We estimate that
because of chemotherapy women have 18 additional weeks of absence from work (16
-
19 weeks
according to Drolet et al., 2005 and Lauzier

et al., 2008) instead of 13 weeks according to
Agendia. The actual total savings
could be

even higher as elder women (age > 51) are 1,9 times
more likely to go on long
-
term disability, stop working, or retire because of the effects of
chemotherapy (Hasset
t et al., 2009).




12

|
P a g e



Also, one year after diagnosis all breast cancer patients have lost 27% of their projected
annual wages (Lauzier et al., 2008), a number which could be reduced upon full introduction
of MammaPrint® in the Dutch healthcare system. Gi
ven the growing role of women on the
labour market and the aging of our workforce it seems logical for the Dutch government to
invest in healthcare products that can take some pressure of the social and economic burden
that cancer puts on our society


Lear
ning points:



T
he current regulatory framework relies on timelines and methods that are
mismatched to recent innovations in the diagnostics field, evidence requirements
are not in line with modern methods of biomarker development.



Diagnostics companies need

more government support to bridge their market
entry phase, buying time to earn back their development costs and reassure their
investors. Personalized medicine can introduce cost
-
effective changes to our
healthcare system, giving governments a good retur
n on their investments while
pushing cancer treatment to the next level.



2
.3 Regulatory clarity


Biomarkers

or
in vitro diagnostics (IVD
s
)

often come as complex multi
-
gene genomic tests that use
algorithms to calculate probable treatment outcome for patients. Such tests have been categorized
by the FDA as ‘in vitro diagnostic multivariate index assays’ (IVDMIA)
in the 2007 guidelines for
IVD
MIA
s
. However, the guidelines are still not codified by law and leave IVDs heavily debated. The
algorithms
,

do
ing

the math

for physicians
,

form a black box that cannot be independently validated
.
This

has raised voic
es for stronger regulation of

IVDs;

while others fear regulation would stifle
innova
tion. It may seem over
-
done to regulate the

blueprints of genomic tests, as long as they can
significantly improve treatment

(
Majewski et al., 2011)
. Ideally the regulatory trajectory for
biomarkers should f
ocus on three aspects of the test, namely analytical validity, clinical validity and
clinical utility

but also keep in mind the
associated ethical, legal and social implications
. The

US
Center for Disease Control and prevention (CDC) has developed the ACCE

model

(
Figure
4
)

that
is
composed of a standard set of 44 targeted questions

taking

all these factors into account

(Haddow et
al., 2003)
.
When
IVDs

prove viable in t
hese areas they should be granted market access, regardless of
their algorithms or other forms of internal
clockworks, protecting companies


intellectual properties
whils
t
safeguarding patients’ health and improving treatment outcome.




13

|
P a g e

















N
onetheless

the lack of coher
e
nt r
egulation should be resolved quickly
,

as it creates a danger
ous
situation with companies using loopholes to market their IVDs as so called ‘laboratory developed
tests’ (LDTs). LDTs are overseen by the 1988 Clinical Laboratory Improvements Amendments for
use
in a single laboratory
,
not as tightly regulated as

re
gular medical tools
.
Further, the effects of
variation in laboratory practices can greatly influence the reliability of these tests

(
Majewski et al.,
2011)
. Physicians basing vital
treatment
decisio
ns
on complex genetic tools that are so poorly
regulated is a recipe for disaster,

urgently requiring more stringent regulation policies.


Figure
4

ACCE Model Process for Evaluating Genetic Tests

(as published by CDC)



Analytic validity:
A test’s ability to accurately and reliably measure the genotype of interest. Analytic validity
focuses on the laboratory components
of testing, i ncluding analytic sensitivity, analytic specificity, laboratory
quality control and assay robustness.



Clinical validity
: A test’s ability to detect or predict the associated di sorder (phenotype), i ncl udi ng cl i ni cal
sensitivity (or the
clinical detection rate), cl inical specificity and positive and negative predictive values. Clinical
val i dity is affected by the preval ence of the di sorder, penetrance, anal yti c sensi ti vi ty and geneti c and
envi ronmental modi fi ers.



Clinical utility:
A test’
s ability to affect clinical decisions and patient outcomes in practice. Other el ements or
contextual factors to be considered include the natural history of the disorder, availability and effectiveness of
i nterventions, quality assurance, health ri sks of
testing or resulting i nterventions, financial impacts of testi ng,
adequacy of facilities to provide services, availability of patient and provider educati on and moni tori ng and
eval uati on of test performance i n practi ce.


14

|
P a g e


Another example
is

the lack of coherent
reimbursement
regulation
s
. The current procedures to
obtain reimbursement sta
tus for an IVD product tend to be complex and vary greatly from one
country to another.

In the US, t
he Center for Medicare and Medicaid Services (CMS) plays a lead role
in
setting

reimbursement for
IVD
s.
CMS

develops
coverage policy for over 46 million Med
icare
beneficiaries,
but
all private payers
also
benchmark CMS coverage decisions in establishing their own
policies.

However
,

the CMS fails to establish

clear
evidence criteria for coverage
,
basing their decision
on the fact if an IVD is

reasonable and
necessary for the diagnosis or treatment
” (CMS, 2006).

CMS
provides little specific guidance about the type and strength of evidence that would suffice, and
therefore the means by which CMS and its regional contractors make coverage determinations lacks
ev
idence standards that can be clearly understood by
IVD

innovators.

Thus
, to be reasonably certain
of success a
n

IVD

innovator must anticipate meeting a very high standard, which likely means
multiple prospective trials, each requiring numerous sites and a
large number of subjects

which is
particularly difficult when aiming for a smaller patient subgroup.

Investors require a clear view on
their future return on investment and the risks that are involved. Because current reimbursement
regulations obscure th
is

view
, R&D investments are now mainly driven by drug manufactures co
-
developing IVDs

for their targeted therapeutics.
The reimbursement framework relies on timelines
that progress far slower than the pace of
IVD development and

evidence requirements that a
re
mismatched with the clinical and economic realities of
IVD

development

(Parker, 2010)
.

Standards of
evidence must be made clear to companies in order to understand and
better
predict coverage
decisions. This will
mak
e

investments in IVDs more attractive
, thus stimulating the path towards
personalized medicine
.

Despite the remarks I

made on current
US

policies
,

it has to be noted that the FDA has taken a
leading role amongst government organizations worldwide in stimulating developments in the PM
sector.
The FDA’s Critical path
initiative in 2004

was the first major step
to make product
development more predicta
ble and less costly

(FDA, 2006)
.

Other governmental bodies such as the
EMEA were quick to adopt similar

policies
.
However it took till 2011 for the FDA to spen
d

$
25 million
USD to “
identify improved pathways to product development and approval for new tech
nologies that
offer promising new opportunities to diagnose,
treat, cure and prevent disease


(FDA, 2011). This
illustrates the FDA’s own realization that companies are
still

having great difficulties
finding their way
through the maze of regulatory obliga
tions before getting their products to the market.
Let alone the
regulatory differences between individual continents or countries.
So although governmental
incentives
to ease market access for targeted therapeutics and their companion diagnostics
are
impr
oving
,

there is still a long way to go.

Governmental bodies and pharmaceutical companies
should work together
to

clarify and
take away some of

the

risks associated with the large, upfront
investments in targeted therapeutics. Only then will
personalized me
dicine

be strategically favorable
and financially viable.


2
.
4

Getting the right mindset

The final obstacle is of
another kind, namely that some

pharmaceutical companies seem reluctant to
focus on
targeted therapies, clinging on to their ‘Blockbust
er

business model

in fear of losing profits
to reductions in their eligible patient subgroups
.

For the past 40 years they have been
relying on a
few blockbuster drugs for the bulk of their revenues, dictating their entire strategic direction.
Historically thi
s has allowed the industry to enjoy consecutive years of double
-
digit growth, but
15

|
P a g e


following recent developments in pharmacogenomics the blockbuster business model has come
under heavy pressure.
The model aims for the entire population, thus relying on very

large clinical
trials to demonstrate significant improvement among many non
-
responders and adverse

events
.
With capitalized costs of development currently
estimated
between
$161 million
USD
and $1,8
billion
USD
(Morgan et al., 2011) this model is just no
longer sustainable
. Even though the promises
of personalized medicine seem clear

and

the industry is

widely advocating their developments in this
field, they are

still reluctant to
make a
dedicated

switch

to targeted therapeutics. A good example of
this
pa
radoxical

behavior is the case of
the Sanofi
-
Aventis
Poly

(Adenosine diphosphate

Ribose)
P
olymerase 1 (PARP1),
inhibitor

Iniparib

(O’Shaughnessy et al., 2011)
.


Sanofi
-
Aventis recently
announced that a Phase III trial evaluating Iniparib in patients with m
etastatic triple
-
negative breast
cancer (mTNBC) failed to meet its primary endpoint. The 519
-
patient study failed to show significant
benefits on overall survival and
progress
ion
-
free survival from adding I
niparib
to standard
chemotherapy comprising
gemcitabine and carboplatin. These negative results came as bad news as
Iniparib had been projected to reach peak annual sales in excess of $1 billion

USD
. A closer look at
the
failing Phase III study however gives us an insight in Sanofi
-
Aventis reluctanc
e to dedicate the
drug solely to the
appropriate
patient

subgroup
.
Accounting for 15 to 20% of all cases of breast

cancer, triple
-
negative breast cancer

(TNBC)

shares

clinical and pathological features with hereditary

BRCA1
-
related breast cancers
, but requ
ires a biomarker for stratification.
Sanofi
-

Aventis failed to
determine a biomarker for
Iniparib
, which was not necessary to get through Phase II clinical trials,
probably because of the small patient cohorts.
But because they were unable to accurately se
lect
patients that would benefit from their drug in Phase III, this strategy did prove a mistake.
AstraZeneca with their Olaparib
, did go through the process of determining a biomarker

(
Lau et al.,
2009)

and has since had more success in BRCA1/BRCA2 positi
ve breast cancer.

It should be noted
that recent data indicate

that Iniparib is not
a good PARP inhibitor

at all
;

raising questions whether
this drug would even be successful in BRCA mutant breast tumors.

This example still illustrates the
point that some
pharmaceutical companies are reluctant to effectively narrow down their patient
population and target their drugs only to the patients that would benefit.




16

|
P a g e


3.
Viability of personalized medicine


There is an ongoing debate on the financial viability of personalized medicine. Given the obstacles
that we have already described it is clear that there still is a long way to go. So before committing
ourselves to this new way of treatment it is wise to g
o over the financial viability
of
personalized
medicine in order not to waste our recourses. There are a number of changes to the conventional
way of drug development and its revenue models, which
I
think can turn out very favorable for all
parties involve
d. First
,

I
will focus on changes in the development of drugs and later we will see how
personalized medicine can generate a good return on investment.

3.1
Adaptive clinical trials

Most clinical trials with cancer drugs follow the same development path. Fi
rst their safety and
pharmacodynamic properties are evaluated in phase I clinical trials with late
-
stage cancer patients.
Then follow single
-
arm or sometimes randomized phase II trials that focus on specific tumor types to
test the workings of the compound
. When all is well, companies will seek regulatory approval
through phase III trials and bring the drug to the market. This is a long and painstaking process that
costs pharmaceutical companies many resources, especially in late
-
stage clinical trials. Foll
owing the
‘blockbuster business model’, investments in failed compounds are to be earned back with
blockbuster drugs that do succeed, indirectly raising the costs of clinical trials tenfold. Surprisingly
the relative amount of compounds that make it to the

market is still as low as it was 25 years ago. In
fact there is a 90% failure rate in reaching phase III from phase I (Blair, 2010)
.

In an attempt to stimulate more innovative
clinical trial designs the Food and Drug
Administration (FDA) has released the critical
path initiative in 2004 and critical path
opportunity list in 2006. This was done in
response to stagnation in the developmen
t
of
New Molecular Entities (NMEs
)

and the
ever growing amount of compounds that
enter trial but never make it to the market

(Woodcock, 2005
). The FDA makes strong
recommendations promoting adaptive
clinical trials and the potential use of
Bayesian
statist
ical methods
for the development of new compounds and their accompanying
biomarkers

(FDA, 2006)
.

This issue is further stressed by the European Medicines Agency (EMEA),
who issued a similar paper in 2006 concerning confirmatory clinical trials with flexibl
e design and
analysis plan

(EMEA, 2006)
.
The FDA defines adaptive trials as “a study that includes a prospectively
planned opportunity for modification of one or more specified aspects of the study design and
hypotheses based on analysis of data (usually i
nterim data) from subjects in the study.” This means
that the accumulating data is analyzed at pre
-
defined timepoints, following evaluation of the
hypothesis and possible adaptations to optimize the future course of the trial. This ought to be done
under s
trict statistical testing to ensure proper conduct (FDA, 2010). Here
I

see regulatory concerns,
fearing for the integrity and validity of clinical trials when allowing high levels of flexibility. This in
Figure
5

Validity and integrity
of adaptive clinical trials.

17

|
P a g e


turn will pose a threat to pharmaceutical companies
who will be hesitant to go through the effort of
designing expensive adaptive trials with the risk of being rejected due to regulatory ambiguity. To
overcome this pitfall the FDA encourages pharmaceutical companies to seek earlier and more
comprehensive in
teraction with regulatory bodies like themselves.

The following
types of adaptive trials can be distinguished, based on the parameters that can be
adj
usted
during

the trial;

the
adaptive
randomization design,
the
group sequential design,
the
sample
size re
-
estimation design,
the
drop
-
the
-
loser design,
the
adaptive dose finding (e.g., dose e
levation
)
design,
the
biomarker
-
adaptive design,
the
adaptive trea
tment
-
switching design, the hypothesis
-
adaptive design, the

adaptive seamless
phase II/III

trial design, and the

multiple adaptive design.
More details on these trial designs have been described by
Chow et. al, 2008
, making it clear that
adaptive trials can play a vital role in clinical research and development. The biomarker
-
adaptive
design is

obviously of specific interest to the progress of personalized medicine given that biomarkers
play such an important role for targeted therapeutics.
But
I expect that
other specific traits of
adaptive clinical trial design

are vital tools in keeping perso
nalized financially viable.

Advantages of adaptive designs
.
The most clear advantage of adaptive trials is the fact that they
allow for ‘on the go’ modifications such as changing endpoints or dose adjustments in response to
results that could not be fores
een at the start of the trial, as long as they are within the
pre
determined boundaries
of variability
. Moreover the adaptations will be pre
-
approved by
regulatory bodies and ethics committees so there is no need to file protocol amendments

when
changes are

executed
. Logistics for changing treatments or doses can also be planned upfront.
And
finally there is

a broad regulatory acceptance, particularly in case of exploratory a
daptive design
clinical trials
for prove of concept studies
(Mahajan et al., 2010).
Adaptive trials would allow testing
of more doses in the same
Phase

I and II

trials, helping physicians to better understand the effect of
the novel compound and the clinically relevant doses. This allows pharmaceutical companies to set
up more effective P
hase III trials and reduce compound attrition at this stage or maybe worse, at
later stages of the development process.
Thus
, t
rial subjects are used more efficiently and fewer
subjects are given ineffective compounds, ineffective doses or doses that are u
nnecessarily high. So
pharmaceutical companies waste less on unsuccessful compounds and can quickly re
-
assign
resources to alternative drugs within their pipeline

(Mahajan et al., 2010).
Unlike today, this would
save patients from exposure to harmful or in
effective doses and experimental treatments, thus
presenting a strong advantage to patients as well. T
he overall effect
of adaptive trial designs
will be
that drug development
becomes less time consuming, cheaper and more favorable to all parties
involved.


Disadvantages and risks associated with adaptive designs
.
Given their complexity, the
implementation of adaptive trials can pose a risk for pharmaceutical companies new to this
approach. The highest risk would be making mistakes due to inexperience or fa
iling the trial when
not being properly aligned with governmental agencies. Further, the use of Bayesian statistical
analysis is compulsory after adaptations rather than free choice and Bayesian statistical methods are
still considered non
-
standard.
Anothe
r risk is that ‘on the go’
changes based on unblinded data may
jeopardize the credibility of the study. Even EMEA’s
paper

has highlighted the
risk

of damaging the
integrity of a trial due to frequent interim analyses.

There is also the risk of

adapting trials too early,
thereby jeopardizing the overall

study findings. Above all, overall

regulatory acceptance
and
experience
is still far from sight

(Mahajan et al., 2010).

18

|
P a g e


Before getting started
, the EMEA underlines two important issues to minimiz
e the risks associated
with ‘on the go’ adaptations to clinical trials. First of all there should be a strong need to reevaluate
the study objectives.
S
econdly the number of interim analyses should be well founded by forehand.
Researches should find a bala
nce between the need to assess the accumulating data at interim
points, while maintaining the integrity of the trial. “
Routinely breaking the blind should be avoided,
particularly when it can be foreseen that insufficient information will be available for
stopping the
study because of proven efficacy or futility or meaningful safety concerns of the experimental
treatment


(EMEA, 2006).

Recently, the Biomarkers Consortium

launched a pioneering multi
-
agent adaptive clinical trial to
treat breast cancer, inte
nded to give several investigational drugs to treat breast cancer together at
the same time, under a project named “Investigation of Serial Studies to Predict Your Therapeutic
Response with Imaging And Molecular Analysis (I
-
SPY 2 TRIAL).” The Biomarkers Co
nsortium is a
unique public

private partnership led by the Foundation for the National Institutes of Health (NIH).
In this trial, adaptive design will enable researchers to use early data from one set of patients to
make decisions about which treatments mi
ght be more useful
for patients later in the trial

and
eliminate ineffective treatments more
quickly (The Biomarker Consortium, 2010; Jones, 2010).
Further the trial has been designed to match patients to drugs that will likely respond, based on the
molecu
lar characteristics of their tumors, found earlier in the study. This revolutionary approach is
supposed to greatly reduce the number of patients necessary to significantly demonstrate the clinical
end
-
points of the study. This principle will be elaborated

on in the Vemurafenib case, which applies a
similar approach.

So as this type of clinical trials is in its infancy
, caution is still advised.
C
ompanies should not turn to
adaptive designs as a solution for poor planning

in an attempt to save their trials
.
Also the potential
for improved efficiency comes at a price; compared with more traditional trial designs, adaptive
approaches require more work and additional effort during planning, implementation, execution,
and repo
rting
(Quinlana, 2010).
Although, presently, there might be some problems in the execution
of adaptive designs, with the release of
the
draft guidance for industry on adaptive design clinical
trials, more and more companies are bound to use adaptive design
ed clinical trials
,

thus making the
drug development process shorter and cheaper.

Figure
6

Comparison between conventional trial and adaptiv
e design trial.
Adaptive clinical trials differ from conventional
tri als i n a number of ways, such as their

flexibility, sizes and clinical end points.
(Modified from Mahajan et al., 2010)

19

|
P a g e


3.
2

Drug development costs

The pharmaceutical industry is under great pressure, demonstrated by more or less flat share prices
for the past 7 years and disappointing innovat
ive behavior.
Between 1990 and 2005, 920 cancer
compounds underwent

clinical trials, yet only 32 were approved

(Reichert et al., 2008),
without a
dramatic increase in R&D productivity, today's pharmaceutical industry cannot achieve sufficient
innovation to

replace the loss of revenues due to upcoming patent expirations. Key patent
expirations between 2010 and 2014 have already been estimated to put more than
$
209 billion
USD
in annual drug sales at risk, resulting in
$
113 billion
USD
of sales being lost to
generic substitution
(EvaluatePharma, 2009). Among the challenges faced by pharmaceutical companies, the most
important aspect will be to raise R&D productivity and efficiency. In order to evaluate the changes
personalized medicine can have on this process

we need to have a model that is able to generate a
well founded calculation on the precise costs of drug development and that is able to cope with the
different variables that PM will change.
For
example the Boston Consulting group estimates that
before genomics technology, developing a new drug has cost companies on average $880 million, 15
years from start to finish, with about 75% attributed to failures along the way. By applying genomics
techno
logy, companies could on average realize savings of nearly $300 million
USD
and two years per
drug, largely as a result of efficiency gains. Representing a 35% cost and 15% time saving (BCG, 2001).
Strong debate and variation surrounds all calculations on
the actual costs of drug development.
Estimates
of the costs to bring a drug to market
vary between
USD$92 million

cash ($161 million

USD
capitalized) to
$883.6 million
USD
cash ($1.8 billion
USD
capitalized)

(Morgan et al., 2011) (
Figure
7
).





Figure
7

Estimates of the components of drug development costs from 5 leading studies.
Esti mates vary between $92
mi l lion USD

cash ($161 mi llion USD ca
pitalized) to $883.6 mi llion USD cash ($1.8 billion USD capital i zed). A more compl ete
overvi ew of recent studi es can be found i n appendi x

2
. (Modi fi ed from Morgan et al., 2011).


20

|
P a g e



Metastatic melanoma has a poor prognosis, with the median survival for patients with stage
IV melanoma ranging from 8 to 18 months after diagnosis, depending on the substage.
Vemurafenib (PLX4032) is a pot
ent inhibitor of mutated BRAF.

It has marked antitu
mor
effects against melanoma cell lines with the BRAF V600E mutation but not against cells with
wild
-
type BRAF. In a recently published study
with only 675 patients
that compared
Dacarzine with Vermurafenib it demonstrated median progression
-
free survival
of 5.3 months
in the Vemurafenib group and 1.6 months in the dacarbazine group. At 6 months, overall
survival was 84% (95% CI, 78 to 89) in the vemurafenib group and 64% (95% CI, 56 to 73) in
the dacarbazine group. Further, the difference in confirmed resp
onse rates between the two
study groups (48% vs. 5%) was highly significant. (Chapman et al., 2011)

Significantly demonstrating a strong survival difference with only 675 patients in a Phase III
clinical trial is unseen in the industry. Normal Phase III cl
inical trials range between 1000
-
5000 patients, but with the power of patient stratification this number can be greatly reduced.
Smaller Phase III clinical trials mean a strong reduction in development costs. But maybe
even more important is that fact that

companies can charge premium prices for drugs with
such strong difference in survival and high response rates.



CASE: Vemurafenib, small Phase III clinical trial design

















3.
3
Economic model
ing

of R&D productivity

Even studies that estimate a more modest total capitalized cost of development like Adams and
Brantner, who calculate an average of
$1074.3

million USD
for the period of
1989 and 2002
, found
that some New Molecular Entities (NMEs
)

take well over

$2.5

bill
ion

USD
to develop (Adams and
Brantner, 2006).
For the following calculations
I

stick to the model proposed by Paul et al, which
differentiates itself mainly on high attrition rates and high

capitalized development costs of
$1799.6
million

USD to bring a d
rug to market
.
This number is much higher than the generally accepted $1
billion USD, but it has to be noted that $1 billion USD is the result of studies on the period between
1990 and 2000 and that development costs tend almost to double every 10 years fo
r the past 50
years (Appendix
1
, Morgan, 2011).
This model has been constructed using recently
published

R&D
performance data from a group of 13 large pharmaceutical companies, provided by the
Pharmaceutical Benchmarking Forum

(Paul et al., 2010)
. Further

this model has been designed to
calculate the effects of strategic changes on the total capitalized costs

of drug developments and will
conveniently serve
my own calculations. Thus I feel the $1.8 billion USD has been well
substantiated
,
especially as can
cer drug development has particularly high attrition rates and relative high
associated costs (Booth et al., 2003).
In this model
,

clinical development (Phases I

III) accounts for
approximately 63% of the costs for each NME launched, and preclinical drug d
iscovery accounts for
32%. The process of discovering and developing an NME on average required approximately 13.5
years (yearly averages ranged from 11.4 to 13.5 across 2000

2007).

Notice how the value of money
makes up more than half of the investment in

a NME by comparing out of pocket costs to capitalized
21

|
P a g e


costs. This effect buffers the relative high costs of late
-
phase clinical trials compared to early stage
development because this money is invested much earlier in the project and is thus ‘more expensi
ve
money’
.



R&D productivity can be simply defined as the relationship between the va
lue (medical and

commercial) created by any NME

and the investments that are required for its development. With
this definition in mind, Paul et al., have created the 'pharmaceutical value equation', which includes
the key elements that determine both the
efficiency and effectiveness of the drug discovery and
development process for any given pipeline). For example, having sufficient pipeline
WIP

is crucial
given the substantial
attrition rate
s
. However
, increasing
WIP

(especially late
-
phase) alone will
und
oubtedly increase
C

and may also increase
CT
, which could further reduce
P
, thus diminishing
productivity.



Figure
8

P
h
armaceutical value equation.
R&D Producti vity (
P
) can be viewed as a function of the elements compri si ng the
numerator


the amount of scientific and clinical research being conducted simultaneously, desi
gnated here as the work in
process (
WIP
), the Probability of Technical Success (
p
(TS)) and the Value (
V
)


di vided by the elements in the denominator,
the Cycl e Ti me (
CT
) and Cost (
C
). Thus, if one could increase the p(TS) (that is, reduce attrition) for a
ny given drug candidate,
P would increase accordingly. If one could cut development time (CT) or costs (C), the productivity (P) would increase even
more.

(Modified from Paul et al., 2010)

Figure
9

R&
D model for the costs of successfully developing and marketing a NME.
The model defines the distinct phases
of drug di scovery and development from the i nitial stage of target
-
to
-
hi t to the fi nal stage, l aunch. R&D parameters
i ncl ude: the probability of suc
cessful transition from one stage to the next (
p
(TS)), the phase cost for each project, the cycle
ti me required to progress through each stage of development and the cost of capital, refl ecti ng the returns requi red by
shareholders to use their money during

the lengthy R&D process, al so named the val ue of money. Wi th these i nputs
(darker shaded boxes), the model cal cul ates the number of assets (work i n process,
WIP
) needed i n each stage of
development to achieve one NME l aunch. Based on the assumpti ons for s
uccess rate, cycl e ti me and cost, the model
further calculates the 'out of pocket' cost per phase as well as the total cost to achieve one NME l aunch per year (US
$
873
mi l lion). Lighter shaded boxes show calculated values based on assumed inputs. Capital i zi
ng the cost, to account for the
cost of capital during this period of over 13 years, yi elds a 'capitalized' cost of
$
1,778 mi l l i on per NME l aunch.

(Modi fi ed
from Paul et al., 2010)

22

|
P a g e


With this model we can investigate which parameters hav
e the strongest contribution to the
capitalized cost of one NME of
$
1.78 billion

USD.
As
Figure
10

demonstrates, attrition rates p(TS) of
late phase
clinical trials have the biggest impact on R&D efficiency. In the

baseline model, Phase II
p
(TS) is 34% (66% of compounds fail in Phase II). If Phase II attrition increases to 75% (a
p
(TS) of only
25%), then the cost per NME increases to
$
2.3 billion

USD
,
or an increase of 29%. Conversely, if
Phase II attrition decreases from 66% to 50% (that is, a
p
(TS) of 50%), then the cost per NME
decreases by 25% to
$
1.33 billion

USD
.

Similarly, our baseline value of
p
(TS) for Phase III molecules is 70%; that is, an a
ttrition rate of 30%. If
Phase III attrition can be reduced to 20% (80%
p
(TS)), then the cost per NME will be reduced by 12%
to
$
1.56 billion

USD
.

Work in progress
(
WIP
).
Companies

need to have enough WIP or products in the
ir pipeline to secure
their R&D p
roductivity. Given the high attrition rates a company would need 8.6 WIPs to get a single
NME to market. However with conventional strategies an increase in WIP would also result in an
increased CT, diminishing the positive effects on P,
especially if deve
lopment resources become rate
-
limiting.

To overcome this, companies should try to allocate more resources to early stage
development, ideally by redirecting them from entities that are doomed to fail in Phase III (or even
Figure
10

Parametric sensitivity analysis.
Cal culates the capitalized cost per launch based on assumptions for the
model's
parameters (the probability of technical success (
p
(TS)), cost and cycl e ti me, all by phase). When baseline values for each of
the parameters are applied, the model calculates a capitalized cost per launch of US
$
1,778 million. parameters are vari ed

from 50% l ower and 50% higher relative to the baseline value for cost and cycl e time and approximatel y pl us or mi nus 10
percentage points for p(TS). Once cost per l aunch i s cal cul ated for the hi gh and l ow val ues of each parameter, the
parameters are order
ed from highest to lowest based on the relative magnitude of impact on the overal l cost per l aunch,
and the swings in cost per l aunch are plotted on the graph. At the top of the graph are the parameters that have the
greatest effect on the cost per launch,

with positive effect i n blue (for example, reducing cost) and negati ve effect i n red.
Parameters shown l ower on the graph have a smal l er effect on cost per l aunch.

(Modi fi ed from Paul et al., 2010)


23

|
P a g e


Phase IV). Adaptive clinical trial
s designed for targeted therapeutics are ideally suited for this as they
allow testing of more doses in the same
Phase

I and II

trials, allowing companies to
prepare for
more
efficient Phase III trials and reduce attrition rates at later stages. After the
first results surface,
resources can be directed towards the most promising WIPs, while abandoning trial arms that seem
unsuccessful.
Given the C and CT of a single Phase III unit of WIP ($150 million

USD
), almost 10 Phase
I molecules ($15 million

USD
) can

be developed for the same cost, ideally through to proof
-
of
-
concept

studies (Paul et al., 2010). Biomarkers and surrogate end points are
inextricably linked to this
approach.

Another way to substantially increase early WIP is through outsourcing.
Traditi
onally, large
pharmaceutical companies have
managed

discovery, development, manufacture and
commercialization of their medicines
mainly in
-
house
.

Today,
virtually all elements of R&D
can be
partnered or outsourced
to
substantially improve R&D productivity
by affordably enhancing the
pipeline from early discovery through to launch.
This

will theoretically allow greater access to
intellectual property, molecules, capabilities, capital,
knowledge

and

of course, talen
t (PWC, 2008).
Complex ICT infrastructures u
sed
in
personalized medicine will
facilitate this process of outsourcing
and even collaboration.

Value (V)
.
Patients, physicians
,

payers
and healthcare o
fficials
apply different criteria to determine
the value of
a
new drug.
P
ayers
will be

increasingly interested
in
clinical trial data that prove the
efficiency, but also cost effectiveness of a certain NME. Patients on the other hand will be primarily
inte
rested to be cured from their disease and retain a high quality of life. This high value creation is
exactly the promise of PM. Targeted therapeutics will prove more effective, sparing patients the ‘trial
and error’ approach of treatment selection, saving
vital time and suffering. Clearly PM can increase a
company’s R&D productivity here.

Cycle time (CT).

Reducing CT has long been the management goal of any production system.
However in the unpredictable pharmaceutical R&D setting, proven concepts cannot be

broadly
adapted. Again we make claim for the use of adaptive

and seamless Phase II and III study designs
that
can

prove extremely useful in reducing clinical
CT
, generally by reducing

non
-
value
-
added wait times
between phases of development

(Paul et al., 2010)
.
Further,
71% of oncology drug approvals were
given a priority review rating by the FDA, in contrast to

40% for other new drugs. Also s
ponsors of
oncology drugs were much more often able to take advantage of at least one of the FDA’s p
rograms
to speed development (subpart E, accelerated approval, fast track) (DiMasi et al., 2007).

Finally there
is the critical path initiative and following measure by the FDA aimed at facilitating market entry for
innovative compounds. Especially in earl
y development, time saving will have great impact on the
capitalized
cost
s

and overall time
-
savin
g will leave a longer period to get a return on
investments

(ROI)

before patents run out.

Costs (
C)
Unit cost reductions, like
CT

reductions, can be leveraged

to improve productivity but
without the implementation of PM it will be difficult to greatly reduce

C

of R&D activities
. However
as the cases of Vemurafenib and the
I
-
SPY 2 TRIAL

clearly demonstrate
,

targeted therapeutics
does

yield
the
possibility to gre
atly reduce development costs in Phase III clinical trials. By selecting the
right patient subgroups for a drug in clinical trials, based on their genetic makeup, compounds will be
sure to show great response rates and improved survival differences. Follow
ing this principle, clinical
trials require far less patients to enroll to demonstrate significant results, even if the difference in
24

|
P a g e


survival is initially not so strong. Vemurafenib was able to demonstrate significant results using only
675 patients, inst
ead of 1000
-
5000 commonly used in Phase III trials. Another way of cost reduction
is by finding alternative financing
methods (Douglas, 2008), like attracting additional cash flows.
As
we demonstrated in the
Agendia/
MammaPrint case, PM will relieve some of

the ever growing
pressure that cancer puts on our society. It would be advisable for governments to install financial
incentives for the development of personalized medicine, thus reducing the costs of R&D
development with sponsorships.

Probability of tec
hnical success (p(TS)).

By now it
will

be crystal clear

that reducing the attrition rate
of drug candidates in clinical development represents the greatest opportunity for pharmaceutical
R&
D, and arguably for sustaining the viability of the entire industry. As
the

sensitivity analyses
in
Figure
10

show
s,

reducing Phase II and III attritio
n are the strongest levers
to reduce

the costs per
NME

(Paul et al., 2010)
.

Non
-
technical attrition as a result of strategic or commercial decisions is not
relevant in this case. Instead
, lack of efficacy
and low margins of safety
are

the major causes of P
hase
II and III
technical
attrition

(Kola & Landis, 2004). There are two solutions to this problem. First is
better target selection as demonstrated by the I
-
SPY 2 TRIAL.
When

our understanding of genetic
onset of disease increases we will be able to much
better predict which segments of patient
populations will respond to certain targeted therapeutics

and aide us in the decision to
commit
substantial time and resources

to a drug
.
The second solution to attrition is in

the routine pursuit of
early proof of
concept
studies, especially in Phase I, for which biomarkers and surrogate endpoints
can often be employed.

These measures will be necessary to make early, but well informed
'go/no
-
go' decisions
.




25

|
P a g e


3.
4

Calculating the potential cost reductions of personalized medicine

The economic R&D model proposed by Paul et al. allows us to make accurate calculations on the
effects that PM can have on the capitalized development costs for targeted therapeutics. To do

this
we will make a number of assumptions and adapt the chances to the baseline model

Figure
10
.

Cost reducing:






$1135M



p(TS)

Phase II to 50%




$500M



p(TS)

Phase III to 80%




$200M



CT Phase II
-
III reduced by one year


$100M



p(TS)

Submission to launch to 100%


$125M



C Phase III reduced to 75M



$175M



CT Submission to launch reduced by 0.5 year

$35M


Cost
increasing:





$400M



p(TS) Phase I to 50%




$350M



C Phase I increased to 20M



$50M



Total potential capitalized savings:



$735 million

USD

Baseline
capitalized cost per NME:



$1799.6 million

USD

New c
apitalized costs per NME:



$1065 million

USD



The m
odeling
reveals

that the development costs of a targeted therapeutic can be reduced by ~40%
from $1799 million
USD
to $1065 million

USD
if all the potential benefits for PM development can be
achieved. Such large reductions in development costs built a strong case for pharmaceutical
companies to explore these promising possibilities and start adopting measures to keep their
businesses sust
ainable in the future.

Figure
11

Estimating new cost per NME for personalized medicine.
When al l the potenti al benefi ts of personal i zed
medi cine are achieved the model suggests that the development costs of a targeted therapeutic can be reduced by ~40%
to $1065 mi l lion USD.
Some costs of

development decrease, with a total of $1135 mi llion USD, while others i ncrease with a
total of $400 mi l l i on USD.

26

|
P a g e



To get a good impression of the effects of a biomarker on the development, market approval
and sales of a drug we compare the stories of Iressa, Tarceva and He
rceptin. Iressa and
Tarceva are both EGFR Tyrosine Kinase Inhibitors (TKIs), indicated for treatment of Non
-
Small Cell Lung Cancer (NSCLC) and Herceptin is a monoclonal antibody directed against
H
uman
E
pidermal growth factor
R
eceptor
2
-
positive (HER2+) breast cancer cells.




Iressa
(Gefitinib) is an AstraZeneca TKI that received FDA´s accelerated market
approval in 2003 but failed to demonstrate significant survival benefits after which the
FDA decided to change its indication. Ever sin
ce, AstraZeneca has been struggling to
gain market approval by searching for a biomarker that could successfully stratify
patients that could benefit from the drug. Recently AstraZeneca has been able to
demonstrate EGFR to be a good biomarker and gained ma
rket approval for Iressa in the
EU and USA.



Tarceva

(Erlotinib hydrochloride
) is marketed in the US by Astellas Pharma´s OSI
Pharmaceuticals and Genentech, a wholly owned
subsidiary

group of Roche, who does
the marketing elsewhere. Although Tarceva is clin
ically comparable to Iressa (CVZ,
2010) it did demonstrate a significant survival benefit and was granted market approval
by the FDA in 2004. Even without an appropriate biomarker Tarceva has been the only
TKI on the market for the following 5 years, furth
er illustrating pharmaceutical companies
struggling with biomarkers for their drugs.



Herceptin

(
Trastuzumab) is marketed by Roche and was FDA approved in 1998 as one
of the first drugs to leverage the power of genetics. Its success is largely attributable
to
its accompanying biomarker as it is only prescribed to patients whose genetic tests
reveal an over
-
expression of the HER2 protein, an indication aggressive cancer that is
responsive to treatment by the drug. Herceptin can be considered an ultimate succe
ss
story of targeted therapy that has generated huge revenues for its developer and is very
successful in treat its indicated patients.


Lung cancer.
According to the World Health Organization, there are more than 1.6 million
cases worldwide of lung and br
onchial cancer each year, causing approximately 1.4 million
deaths annually. According to the National Cancer Institute, lung cancer is the leading cause
of cancer
-
related death worldwide. Approximately 75 to 80% of all cases of lung cancer are
non

small
-
c
ell lung cancer (NSCLC). Advanced
-
stage NSCLC is currently considered an
incurable disease for which standard chemotherapy provides marginal improvement in
overall survival at the expense of substantial morbidity and mortality (Cataldo et al.,2011;
Hansen
et al., 2002).


CASE: Iressa, Tarceva , Herceptin and the value of a biomarker.




27

|
P a g e



The recently developed TKIs, Iressa and Tarceva, are targeted therapies that are indicated
for treatment of NSCLC in patients with activating EGFR mutations. EGFR is a receptor
protein that extends across the cell membrane. EGF binds
the extracellular part of EGFR
leading to activation, which triggers a complex signaling cascade that leads to accelerated
cell growth, division and metastasis. It is estimated that as many as one in ten (10%) lung
cancer patients in the Western population

and one in three (30%) Asian patients with lung
cancer have NSCLC with EGFR activating mutations (Rosel et al., 2009; Mitsudomi et al.,
2006) making them suitable patients for Iressa and Tarceva.

Iressa.
Iressa, the first EGFR TKI for NSCLC was granted ac
celerated approval by the US
FDA in May 2003 for use only after two rounds of standard treatment with platinum
-
based
drugs and docetaxel (Cohen et al., 2004), based on promising results of phase I/II studies
(Fukuoka et al., 2003; Kris et al., 2003). Howev
er, in the post
-
marketing Phase III study (Trial
709 or ISEL) Iressa failed to produce a survival benefit among NSCLC patients compared to
placebo (Thatcher et al., 2005). Retrospective tumor analyses by the Massachusetts General
Hospital and the Dana
-
Farb
er Cancer Institute showed that eight of nine patients who had
responded to Iressa had EGFR mutations. No mutations were detected among the seven
patients who did not respond to the drug. Using cell culture experiments they hypothesized
that EGFR might hav
e been a distinctive biomarker amongst these patients. “Iressa was a
targeted therapy before the target was really known,” remarked Matthew Meyerson, M.D.,
Ph.D., a member of the Dana Farber team. So, in 2005 the FDA issued a new label for Iressa
limiting
its use to “patients with cancer who in the opinion of their treating physician are
currently benefiting (AstraZeneca estimated 15000 patients) or have previously benefited
from Iressa treatment.” On December 17, 2004, AstraZeneca announced the disappointi
ng
ISEL results and subsequent label change in a press release and a “Dear Doctor” letter
(Appendix
2
). Further, AstraZeneca withdrew its European Marketing Authorization
Application (MAA) for gefitinib to treat patients with NSCLC from the EMEA. This resu
lted in
a reduction of 58% in new prescriptions written for Iressa, and 86% of physicians treating
NSCLC modified their treatment practice.

The case of Iressa clearly demonstrates the detrimental effects of marketing a targeted
therapeutic whilst failing t
o appoint an appropriate biomarker. Sales figures for Iressa
plummeted and have not been recovering so far. In fact in 2009 the FDA has partially
withdrawn Iressa from the market forcing AstraZeneca to focus on the EU market in trying to
revive Iressa, thi
s time with an accompanying biomarker. Iressa sales figures will be
compared to Tarceva later in this case. Aside from clear losses in sales figures,
AstraZenecas company image has been severely damaged by measures such as their press
release,
the

“Dear Do
ctor” letter sent to ~141,000 physicians and other healthcare providers,
advertisements in major medical and oncology journals and direct communications to all
known Iressa patients.

It is difficult to attach an exact financial figure to the fact that pati
ents,
physicians and other healthcare officials lose confidence in a company and its products. In
2004 the pharmaceutical industry spent 24.4% of its incomes on promotion, versus 13.4% on
R&D, as a percentage of US domestic sales of $235.4 billion USD (Gag
non et al., 2008). The
fact that pharmaceutical companies spend almost twice as much on marketing compared to
R&D does give us good insights on the size of the damage to AstraZenecas branding efforts
because of the Iressa label change.




28

|
P a g e



Tarceva.
Alongside ISEL ran a Phase III study for Tarceva (BR.21) that compared it with
placebo for patients failing at least one chemotherapy regimen. Tarceva turned out to be
superior to placebo for progression
-
free survival and objective response rate. Interesti
ngly,
the HRs for Tarceva in the two patient subsets in which Iressa appears to confer benefit,
Asians and never
-
smokers, were 0.61 and 0.42, respectively (Shepherd et al., 2005). In
contract to ISEL, Tarceva did prolong survival for most of the other pati
ent subsets, although
the sample size is too small in some of the subsets to draw a definitive conclusion. Tarceva
was approved by the FDA in November 2004(Johnson et al., 2005) and in the European
Union in September 2005 as monotherapy for the treatment o
f patients with locally advanced
or metastatic NSCLC after failure of at least one chemotherapy regimen. The evidence
suggests that Tarceva is more efficient than Iressa even though they were at the time
compared like Cola and Pepsi, almost completely simi
lar. Today the difference is considered
to be mostly attributable to the fact that Tarceva was dosed at its maximum
-
tolerated dose
(MTD) (Shepherd et al., 2005), while Iressa was dosed at about one third of its MTD (Cohen
et al., 2004). Given the fact that

EGF is an important growth promoting signaling pathway
that is expressed even in healthy cells, finding a growth inhibiting effect on a tumor upon
strong inhibition of EGFR seems logical. Although this dosage explanation is still debated we
consider this
to be the primary reason for the varying results of both drugs. Due to this minor
difference in clinical strategy, Genentech was able to access the market without AstraZeneca
competing for a share of the profits for the following 5 years.


For both compan
ies it has been fairly difficult to determine the right biomarker for their EGFR
TKI as the precise roles of EGFR and other components of this pathway have long been
debated (Dziadziuszko et al, 2006; Liang et al., 2010). Approximately 70% of NSCLCs with
E
GFR mutations (exon 19 deletions or the exon 21 L858R) attain responses to EGFR Iressa
and Tarceva, with improved response rate (RR), progression
-
free survival (PFS) and in some
reports overall survival (OS) (Gaughan et al., 2011). The European Randomized
Trial of
Tarceva vs. Chemotherapy (EURTAC) Phase III study was stopped early because it met its
primary endpoint, demonstrating that activating EGFR mutations were a suitable biomarker.
Today Tarceva has been approved in over 90 countries and used to treat

more than a
quarter of a million patients (Roche year reports). Similarly, after a painstaking approval
process AstraZeneca was able to revive Iressa based on data from the Phase III INTEREST
study and the Phase III IPASS study, which compared Iressa with

doublet chemotherapy
(carboplatin/paclitaxel) in 1st line NSCLC patients, a marketing authorization for Iressa for
the treatment of EGFR mutation positive advanced NSCLC patients (all lines of therapy) was
granted by the EMEA in June 2009, followed by the

European launch in July (AstraZeneca
year reports). Interestingly, aside from targeting EGFR mutation positive patients the UK
National Institute of health and Clinical Excellence (NICE) has also demanded that Iressa
can only be prescribed at a fixed pric
e in agreement with UK patient access schemes (NICE,
2010).




29

|
P a g e







So, what if AstraZeneca had co
-
developed a biomarker for EGFR activating mutations and
would
have been granted continued market approval in 2005 by the FDA? What would have
been the influence on its revenues compared to the current situation? Iressa would have
taken the market by storm, crushing Genentech’s attempt to market Tarceva without a
comp
anion diagnostic. When we compare both companies year reports it becomes clear that
Iressa could have had a much better market position and higher revenues without the 2005
label change (
Figure
12
1). Of course it is difficult to compare the marketing strategies of two
different companies and their drugs, but for this case study we see two possible scenarios
that can give us an indication on the added va
lue of a biomarker for Iressa.




Situation A
would be our primary assumption that Iressa takes the market by storm.
Positioning themselves as first on the market in 2003 would
-
in marketing terms
-

be
enough to built their brand and achieve at least a 2:1 m
arket share compared to Tarceva
(Ries and Trout, 1981). Given that in this case Iressa has a biomarker and Tarceva
initially not, we further assume that Genentech would have a hard time getting Tarceva
approved at all, as the standard of care for Iressa wo
uld be much higher. So presumably
Tarceva would enter the market even later than it did now, leaving it safe to say that
Iressa could achieve revenues similar to Tarceva in the current situation, basically
switching revenues for both drugs.




Figure
12

Yearly revenues of Iressa and Tarceva in Million US Dollars.
In 2010 Iressa generates $311 mi l l i on USD wi th
accumulated revenue of $2305 mi llion USD. Tarce
va generates $1325 mi llion USD i n 2010 wi th accumulated revenue of
$6156 mi l l i on USD. Tarceva has thus generated $3851 mi l l i on USD more than Iressa.


30

|
P a g e





Situation B
would be a more conservative scenario in which Genentech quickly learns
from AstraZenecas

success and develops a biomarker of its own for Tarceva. Their
market entry would be postponed even further, but given the fact that Tarceva has
performed better than Iressa in the past and assuming that part of Tarcevas success can
be attributed to bette
r R&D and marketing strategies of Genentech, we estimate that in
the seven years since Iressa was launched both drugs reach a 50% market share. So as
Iressa enters the market ~2 years earlier the accumulated revenues for Iressa will still be
higher than th
ose of Tarceva.










Thus, if AstraZeneca would have developed Iressa with an effective biomarker, this
biomarker should have been valued between $2631 and $3851 million USD in 2010 and
increase annual revenues by a staggering 263%
-

426%. A good
biomarker would probably
exclude a substantial number of patients that in the past did receive Iressa or Tarceva
without a positive effect on their disease. But as both drugs have a biomarker today and
physicians have been testing for EGFR status even befo
re this was legally required we
assume this case to be an accurate estimation in demonstrating the added value of a
biomarker.


Learning points:



The initially failed attempt to market Iressa without a biomarker was very
damaging to AstraZenecas image and b
randing activities.



A biomarker for Iressa would have been valued $2.5 and $4 billion USD in 2010
and increase annual revenues by 250%
-

425%.




Table
2

Biomarker value for Iressa.
Based on the revenues of Iressa and Tarceva i n the current situation we esti mate the
val ue of a biomarker for Iressa under the conditions described for Situati on A and Si tuati on B. In Si tuati on A there
i s a
swi tch i n revenues between Iressa and Tarceva, generating an accumulated bi omarker val ue of $3851 mi l l i on USD and
i ncreasing 2010 revenues for Iressa to $1325 mi llion USD. In situation B both drug reach a 50% market share pl ateau i n
2010, generating a
n accumulated biomarker value of $2631 mi llion USD and i ncreasing 2010 revenues for Iressa to $818
mi l l i on USD.

31

|
P a g e



Comparing Iressa and Tarceva to Herceptin.
A weighted pooled analysis of 60 available
studies has been done to evaluate the clinical outcome in patients with EGFR
-
mutated
NSCLC who
were treated with chemotherapy or EGFR TKIs. In this analysis, the overall
median PFS was 13.2 months with Tarceva, 9.8 months with Iressa and 5.9 months with
chemotherapy (Paz
-
Ares et al., 2009). This illustrates that TKIs are still not performing as
wel
l as one might expect for such novel targeted therapeutics. Especially when compared to
Herceptin, which is able to actually cure patients of their disease rather than prolonging their
PFS. So if we want to compare Herceptin to the EGFR TKIs we should to t
ake into account
variables like differences in the disease or the total number of patients etc. For instance, in
contrast to HER2 in most breast cancers, EGFR expression in NSCLC is often
heterogeneous, varying between different regions within a single tum
or and between primary
and metastatic tumors in the same patient (Eberhard et al., 2008)

Another major problem with
EGFR TKIs is that most tumors
initially respond but over time
(median of 6
-
12 months) most
tumors develop acquired
resistance. Two major
me
chanisms of resistance have
been identified; T790M mutation
in 50% of EGFR
-
mutated
patients with TKI resistance and
the amplification of the MET
oncogene, present in 20% of TKI
-
resistant tumors. In half of the
cases of the MET oncogene the
T790M is coexist
ent. It is
possible that other kinases (such
as insulin
-
like growth factor
-
1
receptor [IGF
-
1R]) might also be
selected to bypass EGFR
pathways in resistant tumors.
The growing preclinical data in
EGFR
-
mutated NSCLCs with
acquired resistance to Iressa or

Tarceva has spawned the initiation of clinical trials testing novel EGFR inhibitors that in vitro
inhibit T790M (Neratinib, XL647, BIBW 2992, and PF
-
00299804), MET, or IGF
-
1R inhibitors in
combination with EGFR TKIs, and heat shock protein 90 inhibitors (N
guyen et al., 2009). So if,
in the future, EGFR TKIs are administered earlier in the disease process, tumors will be less
genetically instable and as a whole contain less cell that might undergo mutation. Together
with a better understanding of the genetic

background of acquired resistance it seems
plausible that Iressa and Tarceva become a lot more efficient and have successes that are
similar to Herceptin today.




Figure
13

Mechanisms of acquired r
esistance to EGFR TKIs in EGFR

m
utated
NSCLC.
Secondary resistance
due to
mutations i n T790M and MET amplification.

32

|
P a g e



In 1998, the FDA approved Herceptin as one of the first targeted therapeutics. It is pr
escribed
only to patients whose genetic tests reveal an over
-
expression and amplification of the HER2
gene, which is highly amplified (two
-

to five
-
fold compared to normal) in 25% to 30% of
breast cancers. HER2 amplification along with receptor protein ove
r
-
expression is a central
driving force in breast tumor growth and indicates more aggressive tumor behavior and poor
prognosis (Hicks et al., 2005). As a result of HER2 induced growth and progression, primary
tumor masses generally consist of a homogeneous

pool of HER2
-
amplified cells and this
HER2+ status is maintained in tumor metastases (Carlsson et al., 2004; Paik et al., 1990)
Herceptin is a monoclonal antibody directed against the defective HER2 protein, effectively
taking out only tumor cells, whilst

not inhibiting the growth of cells expressing normal levels of
HER2 (Mass et al., 2005). Breast cancer is the second leading cause of cancer deaths in
women today (after lung cancer) and is the most common cancer among women. According
to the World Health

Organization, worldwide about 1.4 million women will be diagnosed with
breast cancer annually and about 458,000 will die from the disease.

In this case study we compare Tarceva to Herceptin according to a publication of Sophie
Kornowski
-
Bonnet, General Ma
nager Roche Paris (Konowski
-
Bonnet, 2009) on a Roche
forum meeting. In France the primary variable in deciding which therapies are reimbursed is
the quality of the drug. France uses a rating system called the “Amelioration du Service
Medical Rendu” (ASMR)
or evaluation of therapeutic benefit. This is expressed as a number
between 1 (top/best rating) and 5 (worst rating). So drugs with a better rating can demand a
higher price and will gain faster market access. This “quality of treatment” rating system can
be clearly seen in the reimbursement price of a product and its sales. Herceptin for example
has a 79 percent market share penetration in France, while Tarceva has a market share of
20
-
25 percent. The primary reason is that Herceptin uses diagnostics to st
ratify patient
populations, while Tarceva at the time did not have a proper biomarker. Tarceva sales show
a plateau and even decrease in year 5 after market entry because of shorter treatment
duration (in the majority, non
-
responder patients) and no deeper

market penetration. In their
model Roche assumes a 60% smaller patient population, targeting only those patients most
likely to respond. Despite a smaller population, the modeled sales turnover is much higher
when Tarceva uses diagnostics. Longer treatmen
t duration, coupled with high response in
the targeted population, leads to a higher ASMR rating and higher initial price for the drug.
This price doesn’t erode over time, because of the perceived high value of the drug by pricing
authorities

(
Figure
14
)
.




33

|
P a g e














The reverse is also interesting, without a

stratifying test for Herceptin, Roche could offer
treatment to every breast cancer patient as an option. Although the patient population would
significantly increase, sales would decrease because treatment duration in the majority of
patients would be com
paratively short. The ASMR rating would be lower, resulting in a lower
initial price. Market penetration would also shrink because physicians would not value
Herceptin as high as they do today. So without a biomarker, Herceptin would actually have
les
s rea
l market value, as shown in
Figure
15
.


Figure
14

Tarceva modeled with a predictive test
.
Despite a smaller populati
on, the modeled sales
turnover for

Tarceva
i s much higher when using a biomarker.


Figure
15

Herceptin modeled without its predictive test.

Wi thout a biomarker, Herceptin would have
much l ess real market value.

34

|
P a g e



According to Roche’s hypothesis, adding a biomarker to Tarceva would generate a very large
value, as demonstrated in
(
Error! Reference source not found.
)
. If we take these figures for
rance and extrapolate them to worldwide cancer incidence rates we see that the comparison
between Herceptin and Tarceva is not as farfetched as might seem. Sal
es of Tarceva indeed
reach a plateau in about five years after launch and the yearly revenues of Herceptin show
remarkable similarities to both tests with patient stratification in
(
Figure
14
)
and

(
Figure
15
)
.













To be comprehensive in this comparison we need to verify the total number of pa
tients
eligible for Herceptin treatment and compare this to the total number of lung cancer patients
that would be suited for EGFR TKI treatment. The number of patients that would likely
respond to treatment with Herceptin is approximately 350.000 per year
. A lot more than the
200.000 patients
I estimated to be
suited for EGFR TKI treatment yearly, but still the
numbers are of a comparable proportion.


Worldwide yearly breast cancer incidence*



1.400.000


25% of patients suited for Herceptin per year



350.000

Worldwide yearly lung cancer incidence*



1.600.000

75% NSCLC







1.200.000

Asia: 360.000 with 30% EGFR+





108.000

Rest: 840.000 with 10% EGFR+





84.000

Total patients suited for EGFR TKI treatment per year.


200.000


*
As indicated by the W
HO for 2008.




Figure
16

Yearly revenues Iressa, Tarceva and Herceptin

35

|
P a g e



So as we noted before EGFR TKIs have some very specific issues with tumor heterogeneity
and acquired resistance that need to be resolved in the future for the
drugs to reach their
desired effect. But assuming these obstacles will be overcome and a very conclusive
biomarker (or set of biomarkers) is found for EGFR TKIs, we can state that Tarceva could
have reached peak annual sales similar to those of Herceptin,
adjusted for patient subgroup
sizes of both drugs.
















After estimating the potential yearly revenues for EGFR TKIs (Situation C), modeled to the
success of Herceptin, we can return to our initial calculation of the added value of a
biomarker f
or Iressa. This ‘back of an envelope’ calculation might not be very precise, but it
does give us an indication of the order of magnitude that potential revenues for targeted
therapeutics with an appropriate biomarker are in. For example, since its launch i
n 1998
Herceptin has yielded accumulated revenues of over $25 billion USD and achieved peak
annual sales of over $4 billion USD since 2007 (Roche annual reports).




Figure
17

Yearly revenues for Iressa, Tarceva, Potential EGFR TKI and Herceptin.
Potenti al yearl y
revenues for EGFR TKIs as a functi on of Hercepti n yearl y revenues. Fi gures from before 2005 were
extrapolated as
there was no real potential market at this poi nt because Tarceva was not yet on the
market.

36

|
P a g e









Thus, if AstraZeneca developed Iressa with a very conclusive biomarker (or set of
biomarkers), this biomarker would have been valued
between $2631 and $8888 million USD
in 2010 and increase annual revenues by a staggering 263%
-

806%.



Learning points:



The initially failed attempt to market Iressa without a biomarker was very
damaging to AstraZenecas image and branding activities.



A co
nclusive biomarker for Iressa would have been valued between $2.5 and $9
billion USD in 2010 and increase annual revenues by 250%
-

800%.




3.
5

Revenues generated by personalized medicine

Now that we have seen how personalized medicine can significantly
reduce development costs it
makes sense to have a look at the revenue side of the equation. Pharmaceutical companies are
anxiously clinging to the ‘blockbuster business model’ claiming they need blockbuster drug revenues
to earn back their investments. In
the past this model has been very successful, but following recent
developments in genomics, the idea that ‘one size fits all’ is no longer sustainable. Rather companies
should try to stratify patients and tailor drugs to meet their specific needs. This ‘n
ichebuster business
model’ has already proven to be very successful.

In the ‘Iressa, Tarceva, Herceptin’ case I calculated the potential value of a biomarker for the
targeted drug Iressa. Although a biomarker shrinks patient subgroups, the
corresponding dr
ug will be
able to obtain a larger market share and prove a higher value to
all
parties involved. A conclusive
biomarker for Iressa would have been valued between $2.5 and $9 billion USD in 2010 and increase
annual revenues by
~
250%
-

800%. However there are other aspects of tailoring medicine to the right
patients that have proven to be profitable for pharmaceutical companies and will continue to pose
financial incentives for the implementation of personalized medicine.

Table
3

P
otential biomarker value for Iressa.
Si tuation C
demonstrates a potential biomarker value of almost USD$9 billion
and an i ncrease in annual revenues of 800%.

37

|
P a g e


Gleevec
, Imatinib Mesylate.
In patients with Chronic Myeloid Leukemia (CML) an abnormal protein
continuously induces white blood cell production. The chimerical Bcr
-
Abl protein expressed by CML
cells has constitutive tyrosine kinase activity, which is essential f
or the pathogenesis of the disease.
Gleevec, an ATP
-
competitive selective inhibitor of Bcr
-
Abl, has unprecedented efficacy for the
treatment of CML. Most patients with early stage disease achieve durable complete hematological
and complete cytogenetic remi
ssions, with minimal toxicity
(
Deininger

et al., 2003)
. The first reason
for its success is the fact that Bcr
-
Abl is a truly tumor selective target that is absent in wild type cells.
Secondly, in CML, Bcr
-
Abl is a single oncogenic alteration that sustains
the malignant phenotype, so
inhibition of this protein is sufficient to treat the disease (Blagosklonny et al., 2003). What makes the
case of Gleevec so interesting is the fact that it aims at a very small proportion of the patient
population, but is still

able to generate very high revenues. First, because of its high response rate
among stratified patients and excellent curing capabilities for a serious disease like CML, Novartis can
justify premium pricing for this drug. Secondly, Gleevec is able to turn

CML into a chronic disease,
almost completely restoring patients’ quality of life, but at the same time addicting them to the drug.
This life time ‘addiction’ has unprecedented effects on the annual sales figures for Gleevec which
have risen to $
4.265
mil
lion

USD

in 2010 (Novartis year reports).

Another interesting aspect of Gleevec is the fact that Novartis has been able to market it for other
diseases as well. In February 2002, the FDA granted Fast
-
Track approval for the treatment of specific
patients wi
th
Kit
-
positive

inoperable and/or metastatic GastroIntestinal Stromal Tumors (GIST). So
while initially targeted for CML, a thorough understanding of the underlying causes of the disease
has allow Novartis to market their product for other diseases. This p
rinciple to enlarge the patient
population for targeted therapeutics has been demonstrated throughout the industry.





38

|
P a g e


4.
Discussion


Because our population is rapidly aging we can expect absolute cancer number
s

to rise dramatically
in the coming years. N
ew healthcare approaches are urgently needed to temper the spiraling costs of
cancer treatment and built sustainable healthcare business models. At current rates we might have
to face decisions on how much our society is actually willing to pay to improve
cancer patients’
quality of life. Disappointing developments in the pharmaceutical sector show that improving
treatment standards is a painstaking and expensive undertaking. In fact, the biggest reductions in
cancer incidence rates are not the result of th
erapeutics, but of prevention and to some extent of
diagnostics. Proper use of food preservatives has tempered stomach cancer; colon
o
scopy and better
diets reduced colon and rectum cancer. Lung cancer, the highest cancer related cause of death is
mainly ca
used by smoking. So it is gen
erally accepted that prevention is

the
most effective way

to
fight cancer
. But this does not mean that we can linger, first of all because prevention cannot
generate results quick enough to face our current challenges and
secondly because there will always
be a significant proportion of cancer incidence that cannot be prevented. Personalized medicine is
particularly suited for the field of oncology as tumors show high rates of heterogeneity and
aggressiveness that require a

great deal of stratification and target
ed

therapeutics for optimal
individual treatment. Oncologists, patients and payers are eager for new treatments that insure
better quality of life, even at premium
prices. The promises of personalized medicine are ev
ident, but
indeed one of the more uncertain aspects of personalized medicine is whether the anticipated
benefits will be realized at an acceptable cost. Proper quantification of the costs and benefits of
targeted therapeutics and their cost effectiveness a
re obviously of interest to all stakeholders.
In this
article I have built a case for the financial viability of personalized medicine, but it has to be noted
that this sentiment is not universally shared (Leeder and Spielberg, 2009).
Recently released ana
lyses
by the Deloitte Center for Health Solutions suggest that the
return on investmen (
ROI
)

depends on
particular scenarios and is different among different stakeholders. In their most relevant scenario
they estimate the ROI for patients, payers (insuranc
e companies),

biotech
/pharma and diagnostics
firms, after investing in the co
-
development of a drug and a diagnostic.

Especially the payer doesn’t
fare particularly well, never reaching a positive ROI in this scenario and in other scenarios facing ROI
real
ization over a six
-
year period. But as Deloitte note themselves, this is
due to
a model limitation.
The framework fails to reflect other benefits important to the payer, such as providing coverage for
treatment that is safer and more efficacious, and for i
ncreasing sales to personalized medicine
candidates to increase throughput of patients and resulting benefits attributed to personalized
medicine (Deloitte, 2009). Also diagnostics companies seem to face a grim future by never reaching
their
breakeven

poin
t. Every consumer who has the condition will receive a diagnostic test. This
represents significant
sales
revenue
s

but is not sufficient to offset the large R&D expense to develop
the test. Diagnostic companies need higher pricing, lower costs or other str
ategies like government
funding or co
-
funding by pharmaceutical companies. However the model proposed by Deloitte is not
designed to adept modest changes in the
ir

assum
ptions and is in fact not at all

compliant with some
of the far reaching implications of

personalized medicine that we have proposed in this paper. So
although frequent reality checks are very favorable, I am of the opinion that cost effective business
models can be formed for personalized medicine and others share this opinion (Blair, 2009).


By lowering size, time and failure rates, especially of late phase clinical trials, development costs for
targeted therapeutics can be greatly reduced. Adaptive clinical trials will play a pivotal role in this
process. Justified by the communal benefits
that PM has to offer, governments should put incentives
in place to further stimulate PM development. Incentives include tax benefits, alternative patent
protections like prolonged exclusivity,
subsidizing and funding research and investing in grid service
s
and data exchanges to connect scientists and clinicians and increase data
-
mining capacities. Such
changes should ease development further and allow pharmaceutical and diagnostic companies to
39

|
P a g e


develop new targeted therapeutics at af
fordable cost
s
.
My estim
ations
reveal that the development
costs of a targeted therapeutic can be reduced by ~40% from $1799 million USD to $1065 million
USD if all the potential benefits for PM development can be achieved. At the same time revenues can
be sustained despite narro
wing patient groups. In fact I believe that the development of a biomarker
could greatly increase annual revenues. Even the number of patients can be increased by opening
new markets for a drug as has been demonstrated by Gleevec. By reducing development t
imes, drugs
will have a longer window of protection against generics, together with prolonging patients’ lives and
thus prolonged treatment periods so drug revenues can be greatly increased. Finally by proving
to
supply
an adequate, safe and cost
-
effective

treatment option, companies can get premium prices for
their targeted therapeutics. In the case of Iressa a biomarker would have been valued between $2.5
and $9 billion USD in 2010 and increase annual revenues by a staggering 250% to 800%. These
findings
built a strong case for pharmaceutical companies to explore the possibilities of personalized
medicine and start adopting measures to keep their businesses sustainable in the future.



In fact, some pharmaceutical companies are starting to explore the co
-
d
evelopment of a biomarker
for their drugs. Pharmaceutical companies generally have two options for biomarker development;
combining diagnostics and therapeutics under one roof, or collaborating with external diagnostic
companies. The integrated approach mi
ght be the most promising as in
-
house diagnostics expertise
is likely to make it easier to involve diagnostics in the development process and it should benefit
most from any cost
-
savings or revenues increase throughout the entire process. Also, r
ather than

developing targeted therapeutics in their labs, most of the major pharmaceutical companies are
turning to small biotech firms to fill their pipeline.
Roche is generally considered to be the leader in
terms of personalized medicine and diagnostics co
-
development. Not only does the company have
several drugs on the market that come with a companion diagnostic (Herceptin, Tarceva covered in
this article) i
t also has a PM policy in place since the beginning of 2009 that stipulates that every
single product in their pipeline should have an associated program to identify biomarkers.
Furthermore they have the advantage of having an in
-
house diagnostics unit. Bu
t the number of
mergers and acquisitions has greatly increased throughout the entire industry. Merck KGaA and
Novartis are now closely behind Roche. Merck has gained experience through Erditux (cetuximab)
and a high percentage of pipeline drug
s

with biomar
ker programs and Novartis has founded a
molecular diagnostics unit to facilitate biomarker research in its oncology franchise.


Despite the enormous potential for the industry and society, many challenges remain for
personalized medicine to become a mainst
ream medical practice. In addition to technical and
financial obstacles there are many practical challenges ahead as well. There are regulatory issues
surrounding targeted therapeutics and tests so companies still face uncertainty over the evidence
require
ments. Other key challenges include needs related to information gathering, sharing and
interpretation, like standards in electronic medical records and dealing with the privacy and ethical
issues of DNA collection. Further, we need to develop new strategi
es to educate practitioners and
patients/consumers
so they can make informed choices about a therapy. Understanding both the
benefits and limitations of personalized medicine will be of vital importance. Only if we can
overcome the obstacles discussed thro
ughout this paper will we be able to fully realize the potential
of patient stratification and targeted therapeutics.
The individualization of cancer treatment will
greatly improve prognosis for patients as treatments will prove more effective, saving vita
l time and
suffering.
P
ersonalized

medicine has the power to revolutionize the practice of medicine and greatly
enhance treatment outcome for many individual patients.
I truly believe

the effects of such an
industry
-
wide change to be cost
-
effective and hig
hly favorable to society.





40

|
P a g e


References


Adams C., and Brantner

V (2006) Estimating The Cost Of
New Drug Development: Is It Really $802 Mi l lion?
Health
Affairs, 25
, no. 2 (2006): 420
-
428


Al berts B, Johnson A, Lewis J, Raff M, Roberts K, Walter P
(2002), Mol ecular biology of the cell, fourth edition.


BCG (2001)
A Re
volution in R&D: How Genomics and
Genetics are Transforming the Biopharmaceutical
Industry.
(Boston Consulting Group, Boston,
Massachusetts, 2001).

Berg D, Hi pp S, Malinowsky K, Böllner C, Becker KF.
(2010) Mol ecular profiling of signalling pathways in
for
malin
-
fixed and paraffin
-
embedded cancer ti ssues.
Eur J Cancer
. 2010 Jan;46(1):47
-
55.


Bernards R.(2010) It's diagnostics, stupid.
Cell
. 2010 Apr
2;141(1):13
-
7.


Bhatt AN, Mathur R, Farooque A, Verma A, Dwarakanath
BS. (2010) Cancer bi omarkers
-

current
perspectives.
Indian J Med Res
. 2010 Aug;132:129
-
49.


Bl agosklonny MV, Darzynkiewicz Z. (2003) Why Iressa
fai led: toward novel use of kinase inhibitors (outlook).
Cancer Biol Ther
. 2003 Mar
-
Apr;2(2):137
-
40.


Bl air (2010) Molecular diagnostics and personali
zed
medi cine: value
-
assessed opportunities for multiple
stakeholders.
Personalized Medicine

(2010) 7(2), 143

161.


The Bi omarkers Consortium (2010) The Biomarkers
Consortium launches I
-
SPY 2 breast cancer cl inical tri al.
(Press release) 2010 Mar 17.

Downloaded from:
http://i spy2.org/docs/ISPY2_Press%20Release_3
-
15
-
10.pdf

Bl air (2008) Assessing the Value
-
Adding Impact of
Di agnostic
-
Type Tests on Drug Development and
Marketing,
Mol Diag Ther

2008; 12 (5): 331
-
337

Booth B, Gl assman R, Ma P.(2003)
Oncology's

trials.

Nat
Rev Drug Di scov. 2003 Aug;2(8):609
-
10.

Bowalekar (2011) Adaptive designs in clinical trials.
Perspect Clin Res.

2011 Jan;2(1):23
-
7.

Carl sson J, Nordgren H, Sjostrom J (2004) HER2
expression i n breast cancer primary tumors and
corresp
onding metastases: Ori ginal data and literature
revi ew,
Br J Cancer

90:2344
-
2348, 2004.


Catal do, Gibbons, Pérez
-
Soler, Quintás
-
Cardama, (2011)
Treatment of Non

Small
-
Cell Lung Cancer with Erl otinib
or Gefi tinib.
N Engl J Med
2011; 364:947
-
955 March 10.


Chapman PB, Hauschild A, Robert C, Haanen JB, Ascierto
P, Larki n J, Dummer R, Garbe C, Testori A, Maio M, Hogg
D, Lori gan P, Lebbe C, Jouary T, Schadendorf D, Ribas A,
O'Day SJ, Sosman JA, Ki rkwood JM, Eggermont AM,
Dreno B, Nolop K, Li J, Nelson B, Hou J,

Lee RJ, Flaherty
KT, McArthur GA; BRIM
-
3 Study Group. (2011) Improved
survi val with vemurafenib in melanoma with BRAF
V600E mutation. N Engl J Med. 2011 Jun
30;364(26):2507
-
16. Epub 2011 Jun 5.


Chen E, Tong KB, Malin JL.
(2010) Cost
-
effectiveness of
70
-
gene MammaPri nt signature i n node
-
negative breast
cancer.
Am J Manag Care.

2010 Dec 1;16(12):e333
-
42.


Chow and Chang (2008), Adaptive design methods i n
cl i nical tri als


a review.
Orphanet Journal of Rare
Diseases
2008,
3
:11


CMS (2006) Gui dance for the
Public, Industry, and CMS
Staff: National Coverage Determinations with Data
Col l ection as a Condition of Coverage: Coverage with
Evi dence Development. Document Issued on July 12,
2006.

Cohen MH, Wi lliams GA, Sri dhara R (2004) United States
Food and Drug Ad
ministration Drug Approval summary:
Gefi tinib (ZD1839; Iressa) tablets.
Clin Cancer Res

10:1212
-
1218, 2004.

CVZ (2010) Comparing Erl otinib and Gefitinib,
downloaded from:
http://www.cvz.nl/binaries/live/cvzinternet/hst_content
/nl/documenten/cfh
-
rapporten/2
010/cfh1001+gefitinib+iressa.pdf


Dei ninger MW, Druker BJ. (2003) Specific targeted
therapy of chronic myelogenous leukemia with i matinib.
Pharmacol Rev
. 2003 Sep;55(3):401
-
23. Epub 2003 Jul
17.


Di Masi

JA, Grabowski HG. (2007) Economics of new
oncol ogy drug development. J Cl in Oncol. 2007 Jan
10;25(2):209
-
16.


Douglas, Mi tchell (2008) ASSESSING RISK AND RETURN:
PERSONALIZED MEDICINE DEVELOPMENT & NEW
INNOVATION PARADIGM, Ewing Marion Kauffman
Foundation
, 2008, Downloaded from:
http://www.kauffman.org/research
-
and
-
pol icy/assessing
-
risk
-
and
-
return.aspx


Drol et M, Maunsell E, Mondor M, Bri sson C, Bri sson J,
Mâsse B, Deschênes L. (2005) Work absence after breast
cancer diagnosis: a population
-
based study.,
C
MAJ.

2005
Sep 27;173(7):765
-
71.


Dzi adziuszko R, Hirsch FR, Varella
-
Garcia M, Bunn PA Jr.
(2010) Sel ecting l ung cancer patients for treatment with
epi dermal growth factor receptor tyrosine kinase
41

|
P a g e


i nhibitors by i mmunohistochemistry

and fluorescence in
si tu hybridization
--
why, when, and how? Clin Cancer Res.
2006 Jul 15;12(14 Pt 2):4409s
-
4415s.


Eberhard, Giaccone, Johnson (2008)
Bi omarkers of
Response to Epidermal Growth Factor Receptor
Inhibitors in Non

Small
-
Cell Lung Cancer Working Group:
Standardization for Use in the Clinical Trial Setting,
Journal of Clinical Oncology
, Vol 26, No 6 (February 20),
2008: pp. 983
-
994.


EMEA (20
06)
Reflection paper on methodological issues
in confirmatory clinical trials with flexible design and
analysis plan.

London, UK:. European Medicines Agency.
Report no: Doc. Ref. CHMP/EWP/2459/02.

Eval uatePharma (2009) Alpha World Preview 2014.
Downloaded
from:
http://www.evaluatepharma.com/worldpreview2014.as
px


FDA (2006) Innovation or Stagnation: Cri tical Path
Opportunities List. Washington DC, USA: Food and Drug
Admi nistration; 2006. Downloaded from:
http://www.fda.gov/downloads/ScienceResearch/Specia
l T
opics/Cri ticalPathInitiative/CriticalPathOpportunitiesRe
ports/UCM077258.pdf


FDA (2010) Gui dance for industry: Adaptive design
cl i nical tri als for drugs and biologics. Washington DC,
USA: Food and Drug Administration; 2010. Downloaded
from:
http://www.fda.gov/downloads/Drugs/GuidanceCompli
anceRegulatoryInformation/Guidances/UCM201790.pdf
[ci ted on 2010 Apr 30]


FDA (2011) 2011 FDA Budget Summary, downloaded
from:
http://www.fda.gov/AboutFDA/ReportsManualsForms/R
eports/BudgetReports/2011/defau
lt.htm

Fukuoka M, Yano S, Giaccone G (2003) Mul ti
-
institutional
randomized phase II trial of gefitinib for previously
treated patients with advanced non
-
small cell lung
cancer.
J Clin Oncol

21:2237
-
2246, 2003.

Gagnon MA, Lexchin J (2008) The cost of pushin
g pills: a
new estimate of pharmaceutical promotion
expenditures in the United States,
PLoS Med.

2008 Jan
3;5(1):e1.

Gahan (2010)

Ci rculating nucleic acids in plasma and
serum: diagnosis and prognosis in cancer,

EPMA Journal
,
2010, 1:503

512


Gaughan

EM, Costa DB.(2011) Genotype
-
driven
therapies for non
-
small cell lung cancer: focus on EGFR,
KRAS and ALK gene abnormalities,
Ther Adv Med Oncol.

2011 May;3(3):113
-
25.


Gerges N, Rak J, Jabado N. (2010) New technologies for
the detection of ci rculating
tumour cells.
Br Med Bull
.
2010;94:49
-
64. Epub 2010 Apr 23.


Hassett, MD, MPH ; O’Mal ley, PhD Keating, MD, MPH
(2009) Factors Influencing Changes in Employment
Among Women Wi th Newly Diagnosed Breast Cancer,
Cancer

2009;115:2775

82.


Haddow JE, Palomaki

GE. (2003) ACCE: A Model Process
for Eval uating Data on Emerging Genetic Tests. In:
Human Genome Epidemiology: A Sci entific Foundation
for Usi ng Genetic Information to Improve Health and
Prevent Di sease.
Oxford University Press,

pp. 217
-
233,
2003.


Hansen HH. (2002) Treatment of advanced non
-
small cell
l ung cancer,
BMJ
2002, 325:452
-
453.

Hi cks DG, Tubbs RR (2005) Assessment of the HER2
status i n breast cancer by fl uorescence i n situ
hybri dization: A technical revi ew with interpretive
gui delines.
Hum
Pathol

36:250
-
261, 2005.

Johnson JR, Cohen M, Sri dhara R (2005) Approval
summary for Erl otinib for treatment of patients with
l ocally advanced or metastatic non
-
small cell l ung cancer
after failure of at least one prior chemotherapy regimen,
Clin

Cancer Res

11:6414
-
6421, 2005


Jones D. (2010) Adaptive trials receive boost.
Nat Rev
Drug Discov.

2010;9:345

8.


Keckl ey PH, Dhar A, Underwood HR. (2009) The ROI for
targeted therapies: a strategic perspective. Washington,
DC: Del oitte Center for Health
Solutions, 2009.
Downloaded from:
www.deloitte.com/dtt/cda/doc/content/us_chs_ROIforT
argeted Therapies_January2009%281%29.pdf


Kol a I, Landis J (2004) Can the pharmaceutical industry
reduce attrition rates?
Nat Rev Drug Discov.

2004
Aug;3(8):711
-
5.


Kornowski
-
Bonnet, Discussing the (as yet) untested
business case behind PHC (2009) Downloaded from:
http://www.roche.com.pk/media030809a.html


Kri s MG, Natale RB, Herbst RS (2003) Efficacy of gefitinib,
an i nhibitor of the epidermal growth factor receptor
tyrosi ne kinase, i n symptomatic patients with non
-
small
cel l lung cancer.
JAMA
290:2149
-
2158, 2003.

Lau, Dry, Harbron, Knights, Riches, Mangena, Avis,
Brown, Runswick, Dearden, O'Shaughnessy and
O'Connor (2009)
Identification of gene expression
bi omarkers
that predict sensitivity to the PARP i nhibitor
ol aparib

.
Mol Cancer Ther

2009;8(12 Suppl ):C119
.


Lauzi er S, Maunsell E, Drolet M, Coyl e D, Hébert
-
Croteau
N, Bri sson J, Mâsse B, Abdous B, Robidoux

A, Robert J.
(2008) Wage l osses i n the year after breast cancer:
42

|
P a g e


extent and determinants among Canadian women.
J Natl
Cancer Inst.

2008 Mar 5;100(5):321
-
32.


Leeder JS, Spielberg SP (2009)
P
ersonalized

medicine
:

reality

and

reality

checks
.
Ann
Pharmacother.
2009 May;43(5):963
-
6.

Li ang Z, Zhang J, Zeng X, Gao J, Wu S, Li u T (2010)
Rel ationship between EGFR expression, copy number
and mutation i n lung adenocarcinomas.
BMC Cancer.

2010 Jul 19;10:376.


De Luca A, Normanno N. (2010) Predictive biomar
kers to
tyrosi ne kinase i nhibitors for the epidermal growth
factor receptor in non
-
small
-
cell lung cancer.
Curr Drug
Targets.

2010 Jul;11(7):851
-
64.


Ludovi ni V, Bi anconi F, Pi stola L, Chiari R, Mi notti V,
Col ella R, Giuffrida D, Tofanetti FR, Siggillino

A, Flacco A,
Bal delli E, Iacono D, Mameli MG, Cavaliere A, Cri nò L
(2012)
Phosphoinositide
-
3
-
Kinase Catalytic Al pha and
KRAS Mutati ons are Important Predictors of Resistance
to Therapy wi th Epidermal Growth Factor Receptor
Tyrosi ne Kinase Inhibitors i n Pa
tients with Advanced
Non
-
small Cell Lung Cancer.
J Thorac Oncol
. 2011 Feb 12.


Lynda D Ralton, Graeme I Murray (2011) The use of
formalin fixed wax embedded tissue for proteomic
analysis,
J Clin Pathol
2011;64:297
-
302
doi:10.1136/jcp.2010.086835


Mahajan

and Gupta

(2010) Adaptive design clinical trials:
Methodology, challenges and prospect.

Indian J
Pharmacol
.
2010 August;
42
(4)
: 201

207.


Majewski IJ, Bernards R. (2011) Taming the dragon:
genomic biomarkers to individualize the treatment of
cancer.
Nat Med.
2011 Mar;17(3):304
-
12.


Massarelli E, Varella
-
Garcia M, Tang X, Xavi er AC, Ozburn
NC, Li u DD, B
ekele BN, Herbst RS, Wistuba II (2007)

KRAS
mutation is an important predictor of resistance to
therapy with epidermal growth factor receptor tyrosine
k
i nase inhibitors i n non
-
small
-
cell lung cancer.
Clin
Cancer Res.

2007 May 15;13(10):2890
-
6.


Mass RD, Press MF, Anderson S (2005) Evaluation of
cl i nical outcomes according to HER2 detection by
fl uorescence i n situ hybridization i n women with
metastatic breast cancer treated with Trastuzumab,
Clin
Breast Cancer
6:240
-
246, 2005.


Mead R, Duku M, Bha
ndari P, Cree IA. (2011) Ci rculating
tumour markers can define patients with normal colons,
benign polyps, and cancers.
Br J Cancer
. 2011 Jun 28.
doi: 10.1038/bjc.2011.230.


Mi tsudomi T, Kosaka T, Yatabe Y. (2006) Bi ological and
cl i nical i mplications of EG
FR mutations i n l ung cancer,
Int
J Clin Oncol

2006;11:190

8


Mook S, Van't Veer LJ, Rutgers EJ, Pi ccart
-
Gebhart MJ,
Cardoso F. (2007) Individualization of therapy using
Mammaprint: from development to the MINDACT Tri al.
Cancer Genomics Proteomics
.

2007 May
-
Jun;4(3):147
-
55.


Morgan

S.
Grootendorst P., Lexchin
e J.,

Cunningham
a C.

(2011) The cost of drug development: A systematic
revi ew.
Health Policy

100 (2011) 4

17.


Mül l er V, Al ix
-
Panabières C, Pantel K. (2010) Insights i nto
mi nimal residual diseas
e in cancer patients: i mplications
for anti
-
cancer therapies.
Eur J Cancer
. 2010
May;46(7):1189
-
97. Epub 2010 Mar 27


Nguyen KS, Kobayashi S, Costa DB. (2009)
Acquired
resistance to epidermal growth factor receptor tyrosine
ki nase inhibitors i n
non
-
small
-
cell lung cancers
dependent on the epidermal growth factor receptor
pathway.

Clin Lung Cancer.

2009 Jul;10(4):281
-
9.


NICE (2010)
Gefi tinib for the first
-
line treatment of
l ocally advanced or metastatic non
-
small
-
cell l ung
cancer
, Downloaded from
:
http://www.nice.org.uk/nicemedia/live/13058/49880/4
9880.pdf


Ni rmalan NJ, Harnden P, Selby PJ, Banks RE.
(2008)
Mi ni ng the archival formalin
-
fixed paraffin
-
embedded
ti ssue proteome: opportunities and challenges.
Mol
Biosyst.

2008 Jul;4(7):712
-
20.


O'Shaughnessy J, Osborne C, Pi ppen JE, Yoffe M, Patt D,
Rocha C, Koo IC, Sherman BM, Bradley C. (2011) Ini parib
pl us chemotherapy i n metastatic tri ple
-
negative breast
cancer. N Engl J Med. 2011 Jan 20;364(3):205
-
14. Epub
2011 Jan 5.


Pai k S, Hazan R, Fishe
r ER (1990) Pathologic findings
from the National Surgical Adjuvant Breast and Bowel
Project: Prognostic significance of erbB
-
2 protein over
expression i n primary breast cancer,
J Clin Oncol

8:103
-
112, 1990.


Parker (2010) The Adverse Impact of the US
Rei m
bursement System on the Development and
Adopti on of Personalized Medicine Diagnostics.


Paul SM, Mytel ka DS, Dunwiddie CT, Persinger CC,
Munos BH, Lindborg SR, Schacht AL.(2010) How to
i mprove R&D productivi ty: the pharmaceutical industry's
grand challenge
.
Nat Rev Drug Discov.

2010
Mar;9(3):203
-
14. Epub 2010 Feb 19.


Paz
-
Ares L, Soulières D, Melezínek I, Moecks J, Keil L,
Mok T, Rosell R, Kl ughammer B. (2010) Cl i nical outcomes
i n non
-
small
-
cell l ung cancer patients with EGFR
mutations: pooled analysis.
J
Cell Mol Med.

2010
Jan;14(1
-
2):51
-
69. Epub 2009 Dec 8.


PWC (2008) The changing dynamics of pharma
outsourcing in Asia: are you readjusting your sights?,
Pri cewaterhouseCoopers. 2008, Downloaded from:
http://www.pwc.com/gx/en/pharma
-
life
-
sci ences/outsourc
ing/index.jhtml


43

|
P a g e


Qui nlana, Gaydosb, Macac and Kramsd (2010)
Barriers
and opportunities for i mplementation of adaptive
designs i n pharmaceutical product Development,

Clinical
Trials

2010;
7:
167

173


Rei chert JM, Wenger JB. (2008) Development trends for
new cancer therapeutics and vaccines.
Drug Discov
Today

2008;13:30

37.


Ri es A. and Trout J. (1981)
Positioning: The Battle for
Your Mi nd,
McGraw
-
Hill, 1981

Rosell R, Moran T, Queralt C (2009) Screeni
ng for
epi dermal growth factor receptor mutations i n l ung
cancer,
N Engl J Med

2009;361:958

67.


Schmi d K, Oehl N, Wrba F, Pi rker R, Pi rker C, Fi lipits M
(2009) EGFR/KRAS/BRAF mutations i n primary l ung
adenocarci nomas and corresponding l ocoregional l ymph
n
ode metastases.
Clin Cancer Res.

2009 Jul
15;15(14):4554
-
60.


Sharma SV, Settleman J (2010)
Exploiting the balance
between life and death: targeted cancer therapy and
"oncogenic shock".
Biochem Pharmacol.

2010 Sep
1;80(5):666
-
73.


Shepherd FA, Pereira JR, Ci uleanu T (2005) Erl otinib in
previ ously treated non
-
small
-
cell l ung cancer,
N Engl J
Med

353:123
-
132, 2005.


Sunaga N, Shames DS, Girard L, Peyton M, Larsen JE,
Imai H, Soh J, Sato M, Yanagitani N, Kaira K, Xi e Y, Gazdar

AF, Mori M, Mi nna JD (2011)
Knockdown of oncogenic
KRAS i n non
-
small cell l ung cancers suppresses tumor
growth and sensitizes tumor cells to targeted therapy
.
Mol Cancer Ther
. 2011 Feb;10(2):336
-
46.


Thatcher N, Chang A, Parikh P (2005) Results of a phase
III placebo
-
controlled study (ISEL) of gefitinib (IRESSA)
pl us best supportive care (BSC) in patients with
advanced non
-
small
-
cell lung cancer (NSCLC) who had
recei ved 1 or 2 pri or chemotherapy regime
ns.
Proc Am
Assoc Cancer Res

2005; Late Breaking Session:LB
-
6.


Tonon G.

(2008)
From oncogene to network addiction:
the new frontier of cancer genomics and therapeutics.
Future Oncol.

2008 Aug;4(4):569
-
77.


V
an 't Veer LJ, Dai H, van de Vi jver

MJ, He YD, Hart AA,
Mao M, Peterse HL, van der Kooy K, Marton MJ,
Wi tteveen AT, Schreiber GJ, Kerkhoven RM, Roberts C,
Li nsley PS, Bernards R, Friend SH., (2002), Gene
expression profiling predicts clinical outcome of breast
cancer.
Nature.

2002 Jan 31;41
5(6871):530
-
6.


Wei nstein IB, Joe A.(2008) Oncogene addiction.
Cancer
Res.

2008 May 1;68(9):3077
-
80; di scussion 3080.


WHO (2008) Worl d Health Organization cancer statistics
by Gl obocan, Downloaded from:
http://gl obocan.iarc.fr/factsheets/cancers/lung.as
p
http://gl obocan.iarc.fr/factsheets/cancers/breast.asp


Woodcock J. (2005) FDA i ntroductory comments: cl inical
studies design and evaluation issues.
Clinical Trials

2005;
2: 273
-
275
.





44

|
P a g e


Glossary


New molecular entity
(NME).
A medication containing an active ingredient that has not been
previously approved for marketing in any form in the United States. NME is conventionally used to
refer only to small
-
molecule drugs, but in this article we use the term as a shorthand to refer

to both
new chemical entities and new biologic entities.

Capitalized cost

This is the out
-
of
-
pocket cost corrected for cost of capital, and is the standard
accounting treatment for long
-
term investments. It recognizes the fact that investors require a ret
urn
on research investments that reflects alternative potential uses of their investment. So, the
capitalized cost per drug launch increases out
-
of
-
pocket costs by the cost of capital for every year
from expenditure to launch.

Out
-
of
-
pocket cost

This is th
e total cost required to expect one drug launch, taking into account
attrition, but not the cost of capital.

Cost of capital

This is the annual rate of return expected by investors based on the level of risk of the
investment.




45

|
P a g e


Appendices

Appendix
1

Estimated costs of drug development.
Modi fied from Morgan, 2011

46

|
P a g e



Appendix
2

“Dear Doctor letter” by AstraZeneca.
Sent to practitioners upon market withdrawal of Iressa.