Biomarkers in oncology drug development: rescuers or troublemakers?

Review

10.1517/17425250802432263 © 2008 Informa UK Ltd ISSN 1742-5255 1391All rights reserved: reproduction in whole or in part not permitted

Biomarkers in oncology drug development: rescuers or troublemakers? Estelle Marrer † & Frank Dieterle Novartis Institutes for BioMedical Research, Translational Sciences, Novartis, Klybeckstr. 141, CH-4002 Basel, Switzerland

Oncology is considered as the pioneer indication for the clinical application of molecular biomarkers. Newly developed targeted anticancer therapies call for the implementation of molecular biomarker strategies but even novel cytotoxic treatments use biomarkers for the assessment of efficacy and toxicity. Biomarkers may play several roles in the progression of a drug from research to personalised medicine. In particular biomarkers are used to understand the mechanism of action of a drug, monitor the modulation of the intended target, assess efficacy and safety, adapt dosing and schedule, select patients and prognosticate the clinical outcome. Nowadays, the use of biomarkers in oncology is still challenged as only a limited number of oncology drugs on the market have a companion biomarker test to be mandatorily performed before treatment. This is in contradiction with the current major investment the pharmaceutical sector is devoting to bio-marker identification and development. What are the measurable milestones and outcomes of these investments? How does biomarker development contribute to reaching the ultimate goal of finding the right molecules for the right targets at the right doses and schedules for the right patients? This review provides a critical overview of recent salient achievements in the identification and development of biomarkers.

Keywords: biomarker use , challenges , co-development , diagnostics , successes , targeted therapies

Expert Opin. Drug Metab. Toxicol. (2008) 4(11):1391-1402

1. Introduction

Pharmaceutical companies are nowadays confronted with a severe face wind: the total number of approved drugs has halved since 1999 whereas the total R&D budget has increased threefold. Oncology particularly is a challenging indication with only 5% of drugs proceeding from first in human to marketing approval [1] . Many compounds fail late and expensively in Phase III clinical trials because of a lack of efficacy and toxicity. Here, again oncology sticks out with the smallest overall efficacy rate compared with the other therapeutic disease areas: efficacy of the treatment is shown only in 25% of patients ( Figure 1 ).

One of the major challenges in oncology drug development is that despite a valuable therapeutic activity the survival benefits are still modest and the safety challenging. Herceptin ® (Genentech, San Francisco, CA, USA) introduced a shift in paradigm: from cytotoxic agents to targeted therapies supported by personalised medicine. Whereas targeted therapies aim at stopping the growth of cancer cells by interfering with specific targets involved in carcinogenesis and tumour growth, cytotoxic therapies destruct cells. Many genes/proteins have been identified as playing key roles in tumour growth increasing dramatically the number of potential drug targets and, therefore, the diversity of available treatments. As a consequence,

1. Introduction

2. Prognostic biomarkers

3. Predictive biomarkers

4. Pharmacodynamic biomarkers

5. Safety biomarkers

6. Bottlenecks in

biomarker development

7. Expert opinion

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.

Biomarkers in oncology drug development: rescuers or troublemakers?

1392 Expert Opin. Drug Metab. Toxicol. (2008) 4(11)

among the cancer drugs now entering clinics, most will gain regulatory approval as first in class agents.

It is widely accepted that cancer is not one disease but a multitude of complex diseases involving different molecular backgrounds within different organs. Consequently, targeted therapies should open new perspectives but new challenges. New needs have emerged such as the molecular pathological characterisation of the tumour and new entities to be able to bring the right drug to the right patient. These new entities are commonly referred to as biomarkers.

The different phases of drug development in oncology involve different categories of biomarkers:

Prognostic: predict the course of disease and ultimately the • clinical outcome independent of any treatment modality Predictive: identify patients most likely to respond or not • and/or least likely to show adverse events in the context of a specifi c treatment PD/MoA (pharmacodynamic/mode of action): are most • likely to be used in research or early development to verify if a drug reaches and inhibits its target, examine the mechanism of action, gain insights into PK/PD (pharmacokinetic/pharmacodynamic) correlations and optimise dose and schedule Safety biomarkers: help in monitoring and managing • potential on/off target adverse effects.

Ideally the different biomarkers should provide the proof of concept for target modulation, confirm the preclinical hypotheses underpinning the drug, identify the most

responsive patients, optimise dose and schedule, and ultimately help in decision-making including go/no-go and pharmaceutical risk management. Thus, > 50% of the approved oncology therapies have predictive biomarkers associated [2] . Despite this list of important attributes and functions, the disclosed clinical use of biomarkers in oncology has been sporadic and in the literature critically discussed with only the same few examples of achievements constantly cited. These key achievements, for example, HER2, are nowadays getting old and start be considered as the low-hanging fruits. Did these success stories generate too high expectations? Most biomarkers find applications during the entire course of drug development, most successfully in the groove of early development. Because the key performance indicators for biomarkers take into account solely their use as surrogates, the overall return of investment is considered poor. This is due to the fact that only few of them make their way to the market for diagnosis or prognosis as mandatory biomarkers in the context of approved drug labels.

The well-known old-timers will be rediscussed in the context of their use and then critically re-examined focusing not on the theoretical ideal case but on the practical issues and challenges of daily drug development and how those might be overcome in future.

2. Prognostic biomarkers

Prognostic biomarkers are associated with clinical outcome independent from treatment. Clinical staging, tumour

0

10

20

30

40

50

60

70

80

90

100

Oncolo

gyAlzh

eimer

’sIn

cont

inenc

e

HCVOste

opor

osis

Rheum

atoid

arth

ritis

Diabet

esAsth

ma

Cardia

c arry

thm

iasSch

izoph

renia

Depre

ssion

Analge

sics (

Cox2)

Figure 1 . Percentage of response to treatment for different therapeutical areas. The need for tailored therapies in oncology is highlighted (black) according to Cossman [39] . HCV: Hepatitis C virus.

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.

Marrer & Dieterle

Expert Opin. Drug Metab. Toxicol. (2008) 4(11) 1393

grading and classification are well-known examples of prognostic markers. Prognostic tests are often based on molecular readouts involving new technologies, that is, genetics, genomics and proteomics. The molecular changes screened can be at the:

DNA level 1. Amplifi cations • Mutations (insertion, deletions, SNPs) • Translocations • Epigenetic events (methylation, acetylation) •

RNA level 2. Gene expression profi les (mRNA) • miRNA •

Protein level 3. Protein expression • Phosphorylation status • Protein fusions •

The two most cited examples of prognostic tests, which entered the clinics, are based on mRNA signatures and used primarily in breast cancer. The decision to use adjuvant chemotherapy in patients with early stage breast cancer is based on tumour size, lymph nodes involvement and some molecular determinants, that is, HER2 expression and estrogen receptor expression. Using the current guidelines, overtreatment with cytotoxic chemotherapies can occur at high frequency especially in low-risk breast cancers. A patient stratification and prediction of outcome can nowadays be achieved with gene expression profiling. The first test aims at determining probability of early stage breast cancer recurrence in 5 – 10 years after treatment. The test is called MammaPrint ™ (Agendia, Amsterdam, The Netherlands) and was approved by the FDA in 2007. This DNA microarray-based diagnostic kit measures the gene expression levels of 70 genes in breast cancer tumour biopsies. The score derived from these 70 gene expression profiles screened in one go (constituting one signature), will help determining the risk of recurrence and with it the need of adjuvant therapy [3] . The application of the test on malignant breast needle biopsies contributes to enhance the success of a treatment by defining the patient’s responsiveness in a regimen. Initially, the test was developed for the identification and selection of younger breast cancer patients at low risk for distant metastasis, for whom systemic treatment could be avoided. At present a substantial number of studies highlight the further utility of the test with a potentially wider patient eligibility, for example, distant recurrence after adjuvant treatment in older breast cancer patients [4] . The clinical utility is now being evaluated in several clinical trials versus current clinical standards for treatment assignment such as Saint Gallen guidelines, or Dutch CBO (Dutch Institute for Healthcare Improvement) guidelines.

The second test, the Oncotype DX assay (Genomic Health, Inc., Redwood City, CA, USA), has been validated in women with lymph node-negative, estrogen receptor-expressing

breast cancer. It is a 21-gene panel assay that has been associated with chemotherapy response on the basis of RT-PCR (PCR after reverse transcription of RNA). Its development was based on retrospective studies collected from trials using various chemotherapy agents [5] . Few breast cancer studies have been used to genomically profile breast samples of patients treated with Tamoxifen ® (AstraZeneca, London, UK) and/or Tamoxifen plus chemotherapy before surgery. The generated profiles have been associated with the likelihood of pathologic complete response.

These assays aim at helping optimising risk classification and the prediction of recurrence, consequently, improving the selection of patients for adjuvant chemotherapy. There are a number of continuing initiatives, which are less known, and need to find their opportunities for validation in clinical trial [6] (e.g., Aviara H/I for assessment of endocrine resistance and risk of recurrence in certain breast cancers or Aviara MGI for objective measurement of tumour grade and prediction of chemosensitivity in ER+ (estrogen receptor-positive) breast tumours. The future of this molecular prognostic tests and their wide-spread routine use depends on the improvement of the overall survival rates in on-going clinical breast cancer trials using these tests [7] .

For the development of prognostic tests in oncology, the test readouts need to be correlated with the clinical outcome pointing to different challenges:

Time: the identifi cation of biomarkers relying on • retrospective studies needs prospective trials for validation. Additionally, the value of the test being assessed by the overall survival of the patients can only be measured 5 years after the trials comparing success of treatment with or without test guidance. Trial size: multiple testing and multivariate readouts • imply a large number of patients to be tested to achieve signifi cance. Ethics: how to deal with prognoses of bad outcome? •

The potential that these tests can be developed with retrospective data offers the opportunity to identify new molecular biomarkers and associated tests independent from drug development and, therefore, can even be developed after launch. These initiatives are often collaborations (e.g., in investigator driven studies or Phase IV studies), which allows to put together substantial amounts of data from ‘unlimited’ number of patients without the hurdle of depending on the enrolment in registration trials.

3. Predictive biomarkers

Predictive biomarkers are nowadays the most exploited and successfully applied biomarkers. Predictive biomarkers are known under a number of terms:

Stratifi cation biomarkers • Treatment guidance biomarkers •

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.



Theranostics • Companion diagnostics • Drug test co-development •

Predictive biomarkers forecast response or lack of sensitivity to a specific therapy. The application of these markers is also known under the buzz-word ‘personalised medicine’. Predictive biomarkers help in selecting the right drug for the right patients.

To this end the development of the biomarker and associated test needs to be completely integrated into the drug discovery and development processes, especially if the test is required by the drug label (e.g., Her2 for Herceptin treatment, see below). Consequently timing is critical: data from preclinical studies and from the literature should help the early biomarker identification. Limited numbers of patients treated with the experimental drug are available for biomarker validation and test qualification (Phase I: 40 – 50 of patients, Phase II: low 100s of patients, Phase III 1000+ of patients). In the ideal case, a validated test should be ready by entry into Phase III to demonstrate clinical utility. This means that the test needs to be developed and validated during Phase I and II.

So far, all approved companion diagnostics directly query the status of the drug target or downstream pathway markers, which will become increasingly important.

Examples involving the following biomarkers are discussed in this order:

Her2/neu 1. Pml-Rara 2. Bcr-Abl 3. C-kit 4. B-Raf 5. EGFR 6.

3.1 Her2/neu The first example of a diagnostic test associated to the use of a drug as a requirement from a regulatory perspective, is the Her2/neu diagnostic test for choice of treatment with Herceptin. Not only has it become part of the drug label but the FDA and EMEA approval of trastuzumab (Herceptin) as a front-line therapy in combination with the chemotherapy paclitaxel for the treatment of Her2 positive metastatic breast cancer was based on the request to develop a predictive test to select the right patients. As background information, Herceptin is a humanised monoclonal antibody blocking Her2/neu receptors. Overexpression of Her2 occurs in ∼ 25% of breast cancer, and this is the potential target population for the therapeutic antibody. Two technologies are used to assess the overexpression of Her2: i) immunohistochemistry (IHC), which measures the overexpression of Her2 at the protein level; and ii) fluorescence in situ hybridisation (FISH), which assesses the Her2 gene copy number. The first test developed by Dako in 1998 is called Herceptest ™ (DakoCytomation, Carpinteria, CA USA). The FISH-based

test PathVysion ™ , has been developed by Vysis (now Abott Molecular, Des Plaines, IL, USA). The latest test is a comprehensive gene copy study through silver in situ hybridisation designed by Ventana.

3.2 Pml-Rara Leukaemia should not be seen as one single disease but as a heterogeneous collection of complex disorders associated to particular molecular events leading to the leukemogenesis. The in-depth characterisation of the cancer type with histo-pathology readouts can nowadays be complemented with the use of new molecular methods (genetics, genomics, proteomics markers) to guide the choice of therapy. For example, patients with pml-rara translocation-related leukaemia, respond to all- trans retinoic acids, whereas Philadelphia chromosome-related leukaemia patients respond to imatinib [8,9] .

3.3 Bcr-Abl The best example of genetic diagnostic testing, which specifically directs targeted therapy in oncology, is the detection of bcr-abl translocations in patients with chronic myelogenous leukaemia (CML). Bcr-abl translocation between chromosome 22 and 9 (Philadelphia chromosome) guides the selection of imatinib mesylate (Gleevec ™ or Glivec ™ , Novartis AG, Basel, Switzerland) as therapy, which has been proven to inhibit the proliferation of Bcr-abl expressing hematopoietic cells [10] .

Biomarkers for leukaemia have traditionally been assessed by cytogenetic methods but nowadays genotype-based analyses deliver more predictive information: resistance to imatinib treatments are linked to mutations within the kinase domain of bcr-abl or gene amplification. The two main approaches to overcome this resistance are imatinib dose escalation and the use of alternative more potent tyrosine kinase inhibitors, such as nilotinib (Novartis) [11] . A screening of the mutations in leukaemia patients has guided the identification and development toward a second generation of bcr-abl inhibitors (nilotinib) [12-14] .

3.4 C-kit The detection of C-kit protein in tissue biopsies supports the diagnosis of gastrointestinal stromal tumours and is used to decide on patients’ eligibility for imatinib mesylate therapy [15] . Imatinib mesylate therapy is indicated for the treatment of C-kit-positive unresectable and/or metastatic malignant gastrointestinal stromal tumours. The corresponding test ‘C-kit pharmDx’, a qualitative IHC kit system has been developed and commercialised by DakoCytomation to detect C-kit protein (CD117 antigen) expression in tissue biopsies. It was approved by the FDA in 2005, whereby the label for Gleevec indicates its use for ‘information only’.

3.5 B-Raf Activating mutations in the B-Raf oncogene are linked to 65% of malignant melanoma, and numerous other

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.

Marrer & Dieterle


malignancies. In particular the B-Raf V600E missense mutation has been reported in > 80% of B-Raf alleles associated with constitutive activation of B-Raf kinase. Plexxicon together with Roche Pharmaceutical developed a highly selective inhibitor of B-Raf V600E in combination with a companion in vitro diagnostic test. The mutation positive patients should highly benefit from the treatment. High doses of the compound can be used because of the presence of the target only in the tumour cells; in addition B-Raf V600E screening will select the right patients.

3.6 EGFR The third example of an FDA required test is EGFR pharmDX ™ (DakoCytomation), an assay used to identify colorectal cancer patients eligible for treatment with cetuximab (Erbitux ™ , ImClone, New York, NY, USA). EGFR is also seen as a predictive biomarker for treatment of head and neck cancer with cetuximab (Erbitux ® ), for pancreatic cancer and NSCLC treated with erlotinib (Tarceva ® , Genentech), and for NSCLC treated with gefitinib (Iressa ® , AstraZeneca, Södertälje, Sweden). The use of EGFR screening in the previously mentioned contexts is recommended by the FDA for information only.

The power for effective predictive biomarkers is illustrated with EGFR in NSCLC therapies, that is, gefitinib and erlotinib. Early NSCLC clinical trials with gefitinib and erlotinib revealed partial responses in ∼ 10% of treated patients with higher rate observed in East Asians, females or non-smoking patients. The systematic sequencing of the EGFR gene from the responders revealed a series of somatic gain-of-function mutations highlighting the dependence of 10% of responders on a mutated EGFR kinase. The overall incidence of EGFR mutations in NSCLC among clinical responders to gefitinib or erlotinib is 77%, compared with 7% in NSCLC patients’ refractory to EGFR kinase modulators. Interestingly, 10 – 20% of patients without EGFR mutations show a partial response to gefitinib pointing to other molecular determinants of drug sensitivity.

Most retrospective studies have reconfirmed the sensitivity of somatic EGFR-mutant NSCLCs to targeted therapies (gefitinib or erlotinib) with 50 – 80% of responders. These responses may occasionally be durable (> 3 years) but in most cases they only last for 6 – 12 months before relapse [16] . Drug resistances are associated to T790M mutation [17] in the EGFR kinase domain, to KRAS mutation (key player in the downstream pathway) [18] , compensatory mechanisms such as c-MET amplification and activation of ERBB3 signalling [19] . KRAS is an interesting example, in which mutation is an important predictor of resistance to therapy with EGFR tyrosine kinase inhibitor in NSCLC and in colon cancer. Patients with both EGFR mutation and increased EGFR copy number have a chance of objective response which is greater than 99.7%, whereas patients with KRAS mutation independently of increased EGFR copy number have a risk of disease

progression of > 96.5% [18] . Recently, it has been shown that also in the context of metastatic colorectal cancer KRAS mutation influences treatment outcomes: in the combination of standard chemotherapy with cetuximab, the addition of cetuximab is significantly beneficial exclusively in patients with KRAS wild type [20] . In a clinical context the methods of choice for EGFR and KRAS mutation analysis is PCR (polymerase chain reaction) amplification using intron-bases primers as described [21] . As far as for EGFR copy number analysis, a FISH is performed using commercial assays, for example from Vysis [18] . There are indications breaking the correlation between EGFR positive tumours and positive response to treatment with EGFR tyrosine kinase inhibitors. In cetuximab/panitumumab trials, Timar et al. showed that the response rate obtained with anti-EGFR antibodies in chemoresistant colorectal cancer patients was independent of the level of EGFR expression [22] . Alternatively, hsp90 inhibitors can be used in this challenging situation. Nevertheless, the experience with EGFR kinase inhibitors has yielded a strong correlation between somatic receptor mutations and increased response in patients, a relationship that could lead to first line or adjuvant use of these drugs in patients harbouring EGFR mutations [10] . The clinical role of EGFR mutation analysis as treatment guidance for NSCLC patients remains an open question, awaiting solid validation in prospective studies with in-depth histological and molecular tumour analysis.

4. Pharmacodynamic biomarkers

PD biomarkers should be indicative of modulations of drug-targets and/or downstream signalling pathways. These biomarkers help in determining on-target activity and, therefore, can guide the selection of dose and schedule. These markers are tightly related to the therapeutic target(s) to be inhibited. Most of their applications are already exploited in preclinical settings and early clinical trials and play a crucial role in internal decision-making processes. Here it is absolutely mandatory to validate the assays and have them ready for first in man studies.

The applications of pharmacodynamic markers can also give mechanistic insights into the MoA, interrogating the status of the molecular target (mutation, overexpression, occupancy). They allow to answer the first and most basic question: does the drug reach the targeted tissue and block the intended target? If so, PD markers can be used to collect quantitative information on exposure, kinetics of inhibition of the target, percentage of inhibition of the target etc. One can summarise the qualities of PD markers in three points:

activity achieved on the intended molecular target • (e.g., inhibition of kinase substrate, phosphorylation of the downstream substrate) or modulation of the corresponding biochemical pathway (downstream of pathway activity)

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.



production of the desired biological effect (changes of • apoptosis, invasion, angiogenesis) clinical response (tumour regression, time to • progression, survival).

The following example will highlight the use of pharmaco-dynamic biomarkers not only as the proof that the drug hits its target but also as a guide for dose selection and along with dose selection, going one step further, a personalised selection of drug dose: imatinib mesylate, as described previously, blocks the protein kinase activity of BCR-ABL. In the dose escalation phase, the inhibition of the target (BCR-ABL as pharmacodynamic marker) in the tumour cells of imatinib-treated CML patients was monitored. This allowed demonstrating that the same doses block the target and induce clinical remission without inducing side effects. Similarly it has been shown that the percentage of inhibition of BCR-ABL kinase activity correlates with clinical outcome. This means that for a certain dose, the percentage of inhibition of the target will vary from patient to patient. Coming back to the dose escalation, a pharmacodynamic marker could help to select the starting dose, which would be different from one patient to the other, and the escalation slope respecting safety windows [23,24] .

Most of the current targeted anticancer agents undergoing clinical trials are kinase inhibitors. Therefore, the modulation of the kinase activity by the drug can directly be assessed by measuring the phosphorylated status of substrates downstream of the intended target. As an example, the inhibition of mTOR by RAD001 can be monitored determining the level of the phosphorylated form of the ribosomal protein S6. Similarly, pEGFR (phosphorylated EGFR) can be used as a pharmacodynamic biomarker for the monitoring of efficacy of gefitinib treatment, with a tumour shrinking correlated to reduction in pEGFR (p = 0.001).

A pharmacodynamic marker can overpass the scope of its predefined use and become also a safety biomarker especially in oncology, in which the observed adverse events are directly related to the percentage of inhibition of the target. An illustrative example here is the 20s proteasome inhibitor bortezomib (Velcade ™ , Millenium, Cambridge, MA, USA). In a Phase I trial, it has been demonstrated that the percentage of inhibition of the 20s proteasome in whole blood can be directly correlated with toxicity and dosing. Adverse events (neuropathy, fatigue, and diarrhoea) were observed with 70% of inhibition of the 20s proteasome activity, which corresponds to a dose of 3 – 3.5 mg given on day 1 and on day 4 every other week [25] .

5. Safety biomarkers

Another field of application of biomarkers is the prediction of metabolism and exposure. Most of the validated markers recommended for testing, such as thiopurine methyltransferase

or UDP-glucuronosyltransferase 1A1 (UGT1A1), CYP450s are used for susceptibility of toxicity assessments. It has been translated into a diagnostic test (AmpliChip ™ CYP450, Roche, Basel, Switzerland) using two cytochromes P450 (Cyp2c9 and Cyp2d6) to genotype patients [26] . For example, tamoxifen is extensively metabolised after oral administration. It is a substrate of cytochrome P450 3A, 2C9 and 2D6, and an inhibitor of P-glycoprotein. Regarding CYP2D6, the FDA recommends to screen for patients with reduced activity in this pathway (poor metabolisers) who will have higher plasma concentrations of tamoxifen compared with people with normal activity (extensive metabolisers). Similarly, an FDA tag for UGT1A1 in the context of nilotinib treatment states “A pharmacogenetic analysis of 97 patients evaluated the polymorphisms of UGT1A1 and its potential association with hyperbilirubinemia during Nilotinib treatment. In this study, the (TA)7/(TA)7 genotype was associated with a statistically significant increase in the risk of hyperbilirubinemia relative to the (TA)6/(TA)6 and (TA)6/(TA)7 genotypes. However, the largest increases in bilirubin were observed in the (TA)7/(TA)7 genotype (UGT1A1 * 28) patients”. The linkage of a significant portion of the hyperbilirubinaemia risk to UGT1A1 polymorphism supports the larger finding that the bilirubin elevation is biochemically driven but is not a consequence of some more severe organ complication. The hyperbilirubinaemia observed on treatment with nilotinib is clinically manageable (even not leading to treatment interruption), as no events of grade 4 (> 10 times upper limit of normal) were observed and the bilirubin elevations were reversible [27] .

It is known that cancer treatment, especially cytotoxic drugs, for example, cisplatin or methotrexate, induce organ toxicity. The kidney is for both drugs the main toxicity target organ. Here, new urinary renal safety biomarkers (e.g., kidney injury marker 1, clusterin, cystatin C, β 2 -microglobulin) have been recently identified, characterised and qualified for monitoring renal drug-induced toxicity earlier and more sensitive than the peripheral clinical tests used now (serum creatinine and blood urea nitrogen) [28,29] .

6. Bottlenecks in biomarker development

The fact that the same examples of successful clinical biomarkers are repeatedly cited does highlight that there is only a limited number of qualified and approved biomarker references in the field, consequently indicating that there must be major hurdles all along biomarker development path. It is reasonable to assume that challenges have already taken place at the early phase. The identification of a potential biomarker should start concomitantly with drug discovery activities. The initial steps usually involve generation of data coming from different sources, among others genetics, genomics, proteomics, metabonomics and

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.

Marrer & Dieterle


protein modifications. These data are often generated with preclinical animal models and, therefore, the availability of the right disease model and also its proper use regarding human relevance is crucial. The need for de novo generation of an animal model can significantly delay or even stop the biomarker development. The different disciplines used for the identification of the biomarker of interest do not necessarily provide an entity translatable to human. For example, a change in mRNA levels within an animal organ needs to be ‘transformed’ into a secreted protein measurable in the human periphery. From fundamental science generation to first in man proof of concept, major steps need to be accomplished which requires a multi-disciplinary, multi-institution strategy. Large research consortia, retrospective clinical trial data and population-based sample biobanks allow to link biomarker-generated data with patient observational information of the relevant pharmacological phenotypes. This involves significant investments and, therefore, the biomarker-related intellectual property needs to be protected latest at that time point.

A second important step after the first clinical proof of concept is the concretisation of a fully validated assay to qualify the biomarkers in Phase II studies. At every single step, the technical aspects of analytics and assay development will come into the picture adding complexity and costs. In addition, the analytical methods used in discovery may often not have the same level of analytical precision and reproducibility compared to the requirements in clinical settings [30] . The tissue handling, for example, sampling, preparation and conservation, adds significant variance [31] . Even validated assays with regulatory approval can perform poorly in non-experienced users’ hands. Either central

laboratories have to be highly specialised in running these validated assays, or the assays will have to be ‘simplified’ to allow a broadly use even directly at bedside. Approximately 20% of HER2 assays had a different outcome when they were re-evaluated in a high-volume central laboratory [32] . Looking at all these implications of costs, resources and manpower, the biomarker has to be worth the effort. But even then, its development can be jeopardised by the failure of its associated drug, bringing the big question of early co-development versus late engagement into the light especially when considering that only 5% of the oncology drugs proceed from first in human to approval [1] . A remark about late-stage engagement: if biomarkers are discovered during Phase II, their biological qualification and assay development/validation cannot be ready for implementation without potentially delaying the start of Phase III. For complex qualification and validation of genomic predictive markers ( in vitro diagnostic multivariate index assays, IVDMIA), often the Phase II trials offer the first opportunity for correlation with adequate statistical power because of sufficient patients being exposed to efficacious doses of the investigated drug. Going further along drug and biomarker development, the final validated bed-side device or validated assay performed by a central lab need to be available for Phase III trials, which aim at demonstrating the clinical utility of the biomarker and its nature (prognostic, predictive, pharmacodynamic, safety). Randomised studies involving patients with different backgrounds (diseases, metabolism, genetics, record of treatment) will be required to provide the ultimate distinction between predictive and prognostic markers by comparing novel therapies with standard of care ( Figures 2 and 3 ).

Non-biomarkerbased strategy

Biomarker basedstrategy

New drugtreatment

Randomization

Biomarker test

Standard ofcare

New drugtreatment

Standard ofcare

Randomization

+ -

Figure 2 . Trial designs for qualifi cation of stratifi cation biomarkers: this design provides measurable outcomes to compare the new drug to the standard of care and to evaluate the utility of biomarker-guided treatment choices.

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.



7. Expert opinion

Many contexts from drug development to clinical proof of concept point to the use of biomarkers as being extremely helpful and powerful. This is particularly true in oncology, in which the new trend is focusing on targeted therapies. In this new paradigm, the monitoring of the effect on the target or downstream pathway players should be a must. In addition, these new treatments come with high costs rendering a biomarker-based stratification strategy economically viable because of the possibility to minimise treatment failure. A cancer patient, to whom the wrong drug is given, will face disease progression during the treatment time impacting overall survival rate and more costs (late-stage disease, side effects of wrong treatment). Cancer can be a highly stratified group of pathologies; therefore, overall efficacy is usually extremely low whereas some patients with the cancer subtype matching the treatment show great benefits. Another point to consider is the safety; in some patient groups treatments can be of high risk. If the developed biomarker(s) and the associated drug is a perfect match, this duo will guide the choice of treatment, dosing and schedule, therefore, minimising adverse events and increasing survival. A representative example here is Herceptin, which may never have made it to the market without its Her2 companion diagnostic test. Another example, Erbitux, an EGFR inhibitor (targets the extracellular domain of the protein) has drawn a lot of attention because of its high price associated to a lack of efficacy. A negative first set of clinical data were counterbalanced by an obvious patient stratification based on the presence or absence of EGFR in the context of colorectal cancer. Today, Erbitux is

having in its label the following specification “Patients enrolled in the clinical studies were required to have immuno-histochemical evidence of positive EGFR expression using the DakoCytomation EGFR pharmDx ™ test kit”.

Unfortunately, nowadays we have only few available examples illustrating the co-development of drugs with biomarkers. There is a high medical need to generate more of them because the overall survival rates for many common malignancies are still low. Comparatively, the costs for the biomarker development is low; typically < 15% of the costs of clinical trials, whereas the overall development costs still exceed $1 billion per new drug application. The return of investment of a drug is usually not driven by the biomarker costs except for situations in which the biomarker development is responsible for the delays in the launch of the drug impacting initiation and completion of clinical trials. Also, some of the biomarker development costs can be amortised over backups and follow-up drugs sharing the same target (e.g., BCR-ABL kinase inhibitors) or even sometimes having targets belonging to the same pathway. It needs also to be considered that the development of a companion diagnostics for a drug already on the market can prolong the lifetime of the drug by a new patent covering the combined use of drug and test.

The specific costs of biomarker development need to be integrated into the overall picture of patients’ benefit and healthcare system’s economical value. Unfortunately, for every drug not achieving regulatory approval, the co-development costs of the diagnostic test and/or device will be lost. Most of the biomarkers are tightly linked to a drug-specific context taking into account the indication (disease), the target and the molecular characteristic of the

New drugtreatment

Standard ofcare

New drugtreatment

Standard ofcare

20% 10%20% 10%Prognostic

10% 10%30% 10%Predictive

20% 10%40% 10%Predictive +Prognostic

Biomarker test

+ -

Figure 3 . Trial designs for qualifi cation of stratifi cation biomarkers to investigate their prognostic and predictive power. Three examples associated to their 5-year overall survival rate are compared.

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.

Marrer & Dieterle


cancer type, and have to be revalidated or even redeveloped for every new set of parameters. Also, an inappropriate performance of the test can jeopardise the drug. On the other hand, increased efficacy and reduced side effects in subjects selected through the diagnostic test will help to lower the attrition rate in late clinical development. Efficacy is increased in subjects selected through the diagnostic test, thus regulatory approval can be reached faster. With few exceptions, approval of biomarkers by health authorities requires an evaluation of the clinical utility with prospective studies and not with samples banked from previous clinical trials. This means that either the biomarker must be ready for testing before starting the corresponding clinical studies or more clinical trials with the focus on the evaluation of the biomarker needs to be considered. Segmenting the market may impair the economical viability of the drug. Segmenting also could mean that some patients may be wrongly screened out, especially if the scientific proof of utility of the marker does not fully cover every applicability/indication/potential interference. Yet, segmented indications can be economically viable because a subpopulation of responders can only be recognised with statistical relevance, if the subpopulation represents a reasonable fraction in a clinical trial [33] . Although at first sight it may be less attractive for industries to have restrains in terms of patients to be treated, the superiority of treatments in terms of personalised efficacy:safety ratio is an undeniable competitive advantage. Biomarkers will provide better and more sensitive measures for the competition of drugs. As a consequence medicine will be guided by molecular-based diagnostics, which will help to build the marketing strategies.

Going back to drug development, the general perception is that there is little current evidence that biomarkers have impact on early development. Unfortunately, the research part of drug development is very rarely mentioned with regard to biomarkers, although a lot of efforts are done around PD/MoA biomarkers. Moreover, for the targeted therapies, the adverse events observed can be related to on-target effect(s); therefore, the same biomarkers can be used for the monitoring of efficacy and safety depending on their associated thresholds (expression window). The use of these biomarkers for internal decision making at an early stage is most of the time not referred to later on. Even as late as ‘first in man’ a significant amount of monitoring is done using biomarkers for the proof of concept and internal decision making. But often these biomarkers do not reach the companion diagnostics step. Especially, as it often represents a competitive advantage for the company not to advertise the use of these biomarkers for early attrition or candidate selection and optimisation of drugs. There is no key performance indicator or statistics to measure how often a biomarker helped to decrease the risk of the subsequent development costs or helped in accelerating the attrition of a ‘bad’ drug candidate. There is also no

retrospective estimation of premature termination based on misinterpreted biomarker data. The controversial discussion between Ratain and Goulart shows how much of a challenge it is to identify and select the right biomarkers for the right clinical purposes [34-36] . In 2007, the article by Goulart on the use and role of biomarkers in Phase I oncology trials reviewed the literature from 1991 to 2002. The use of biomarkers in the clinical drug development so far contributed to Goulart’s optimistic conclusions that despite challenges (minimal impact on dose and schedule selection) biomarkers have a strong potential to positively affect new drug development. This was heavily challenged by Ratain and Glassman in their editorial especially with respect to the general value of biomarkers in Phase I trials, the costs and the difficulties of implementation (e.g., invasive tissue biopsies). Before discussing different points, it has to be considered that the period taken into account (1991 – 2002) by Goulart in his review may be just before the real ‘biomarker takeoff ’. In Phase I, biomarkers have been referred to as being the key for proof of concept, dose selection and schedule selection. Biomarkers as standard clinical observations have been used for a long time, for example, blood count for toxicity but are nowadays not referred to as biomarkers because they are not being monitored by new sophisticated molecular technologies. Biomarkers usually come with challenges for a clear go/no-go decision:

no 100% specifi city • no 100% sensitivity • technical challenges (sensitivity and reproducibility of the • method used, inter-method comparability, interferences) ‘black boxes’ in terms of scientifi c progresses • and understanding biological interferences (e.g., diseases, genetic background, • environmental infl uences).

All these factors can turn a yes/no decision into a grey area. Practically, the complexity of biomarkers can result in their misuse or correct use but wrong connections leading to overinterpretation. For example inappropriate correlations between the modulation of a MoA/PD biomarker as a sign of target inhibition and efficacy (end point) or the extrapolation of an effect observed in the periphery to a localised impact on the tumour. It has also been mentioned that the implementation of biomarkers in inappropriately designed studies can have a number of negative impacts. For example the identification of biomarkers based on a single study without further confirmation results in an unavoidable chance of false positive selections of biomarkers. Also, if only biomarker-stratified patients are treated with a drug without having appropriate control arms, it cannot be concluded if the drug is efficient in other non-eligible patient populations or if the biomarkers are rather prognostic instead of being purely predictive (by using the trial design shown in Figure 3 ). Both trial designs shown in Figures 2 and 3

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.



are designed to test the clinical utility of biomarker-guided treatment regimens. Depending on the scientific evidences supporting the use of the new biomarker, its validation and the level of confidence clinicians will have in its use for therapy guidance, one of these two schemes will be considered. For example, a biomarker with strong clinical evidences (retrospective investigations), Figure 3 , in which two out of four treatment arms are treated the wrong way with respect to the biomarker readout would be the wrong choice. Conspicuous examples include the initially restricted development of cetuximab only in EGFR positive tumours and similarly Erlotinib trials only in patients with EGFR mutations. To prevent the underuse of a drug, which could have a beneficial effect in biomarker-negative screened populations, the biomarker should be tested also in the non-biomarker-based treatment arm shown in Figure 2 .

For a defined situation, how much evidence and trust do we need in biomarkers for clinical decision making? Biomarker-based stratification means possibly keeping patients away from a beneficial drug. In the oncology field, it can mean putting patients at risk by testing the negative predictive value of the biomarker (do biomarker-negative patients really not respond to treatment?). Testing all options in trial arms may inflate the biomarker costs and the number of patients to be enrolled finally impacting the overall recruitment time and costs.

Another ethical concern to consider is the overuse of prognostic biomarkers, especially the genetic prognostic biomarkers. Molecular diagnostics of disease prevalence means that drugs will intend to prevent in first line and cure only if necessary, which could be extremely beneficial. But this opens the debate on ethics. Should the prevalence for known, characterised and testable cancers be tested? When? Does highest benefit imply testing right after birth? What kind of psychological impact will this new information have, especially if the cancer is diagnosable but not yet curable? Ethical questions will have to be identified and discussed not only among experts but in society to define the most appropriate way for a better living knowing that progress cannot be stopped.

Talking about progress, what are the new trends for biomarkers in oncology? As developed above, a strong limitation of biomarker identification and proper use is linked to missing pieces of science. The new molecular-based therapies involve an in-depth understanding and characterisation of the disease, its origin, molecular basis, pathological manifestation and so on. Genetics and molecular pathology are the disciplines of choice to help better biomarker strategies by allowing strict tumour typing

assessment by comprehensive assessment of tumour molecular genotype/phenotype focused on molecular alterations/pathways at diagnosis. Transcriptomics analyses and proteomics will provide biomarkers correlating to histologic patterns and architecture. All these approaches focus on the molecular analysis of the tumour itself, which is an invasive procedure for the patient. Hence, peripherally accessible biomarkers will gain importance in future. Recently, the existence of rare circulating tumour cells in the blood of cancer patients has been demonstrated in an increasing number of publications [37] . Instead of studying primary tumours themselves, it may then be possible to characterise the cancer indirectly by monitoring genetic alterations in the circulating tumour cells or changes of the serum proteins. Tumour cells can be recovered by immunoaffinity purification using antibodies but the low number of cells (5 – 10 cells per 7.5 ml blood) can be a technical challenge [38] .

There are not only major progresses visible in terms of new disciplines and technologies around biomarkers but also the constant increase of publications citing new promising candidates (e.g., bcl2 amplifications, FGFR3 mutations, b-Raf mutations, PTEN mutations/deletion, Ras mutations, BRCA mutations, p53 mutations, PI3 kinase mutations, Topoisomerase II amplification/deletion) for molecular medicines in oncology gives striking evidence that this field just starts to emerge.

If biomarker development could be assimilated to an exact science such as mathematics, the following equation could summarise today’s situation:

(Great Concept Some Impactful Applications)Biomarker

Many Challenges to Still Overcome

+=

and, like Moore’s law, the addition of n (should n have a positive or negative sign?) as the number of years will tell us where we go.

(Great Concept Some Impactful Applications)Biomarker

Many Challenges to Still Overcome

n+=

Acknowledgements

The authors are indebted to S-D Chibout and J Meyer for reviewing the manuscript. This publication engages the view of the authors solely.

Declaration of interest

The authors are employees of Novartis AG, Basel, Switzerland.

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.

Marrer & Dieterle


Bibliography Papers of special note have been highlighted as either of interest (•) or of considerable interest (••) to readers.

1. Kola I, Landis J. Can the pharmaceutical industry reduce attrition rates? Nat Rev Drug Dev 2004 ; 3 : 711 -5

2. FDA website of validated genomic biomarkers. Available from: http://www.fda.gov/cder/genomics/genomic_biomarkers_table.htm

•• Current status of biomarkers appearing in drug labels and their recommended use.

3. Gutman S. Offi ce of In Vitro Diagnostic Device Evaluation and Safety, US Food and Drug Administration, Lichtenfeld L, American Cancer Society, FDA teleconference, 6 February 2007

4. Wittner BS, Sgroi DC, Ryan PD, et al. Analysis of the MammaPrint breast cancer assay in a predominantly postmenopausal cohort. Clin Cancer Res 2008 ; 14 : 2988 -93

5. Paik S, Tang G, Shak S, et al. Gene expression and benefi t of chemotherapy in women with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol 2006 ; 24 : 1 -10

6. Ross JS, Hatzis C, Symmans WF, et al. Commercialized multigene predictors of clinical outcome for breast cancer. Oncologist 2008 ; 13 : 477 -93

7. Bueno-De-Mesquita JM, Van Harten WH, Retel VP, et al. Use of 70-gene signature to predict prognosis of patients with node-negative breast cancer: a prospective community-based feasibility study (RASTER). Lancet Oncol 2007 ; 8 (12): 1079 -87

• The fi rst FDA-approved test based on gene expression profi les taken into consideration for decision making in a clinical context.

8. Haferlach T, Bacher U, Kern W, et al. Diagnostic pathways in acute leukemias: a proposal for a multimodal approach. Ann Hematol 2007 ; 86 (5): 311 -27

9. Bolufer P, Colomer D, Gomez MT, et al. Quantitative assessment of PML-RARa and BCR-ABL by two real-time PCR instruments: multiinstitutional laboratory trial. Clin Chem 2004 ; 50 (6): 1088 -92

10. Druker BJ. Imatinib as a paradigm of targeted therapies. Adv Cancer Res 2004 ; 91 : 1 -30

•• Imatinib is a real breakthrough in oncology.

11. Kantarjian H, Cortes J. BCR-ABL tyrosine kinase inhibitors in chronic myeloid leukemia: using guidelines to make rational treatment choices. J Natl Compr Canc Netw 2008 ; 2 : S37 -42.

12. Kantarjian H, Giles F, Wunderle L, et al. Nilotinib in imatinib-resistant CML and Philadelphia chromosome-positive ALL. N Engl J Med 2006 ; 15 : 2542 -51

13. Druker BJ. Circumventing resistance to kinase-inhibitor therapy. N Engl J Med 2006 ; 15 : 2594 -6

14. Shah NP. Advanced CML: therapeutic options for patients in accelerated and blast phases. J Natl Compr Canc Netw 2008 ; 6 (Suppl 2): S31 -6

15. Heinrich MC, Corless CL, Demetri GD, et al. Kinase mutations and imatinib response in patients with metastatic gastrointestinal stromal tumor. J Clin Oncol 2003 ; 21 : 4342 -9

16. Sharma SV, Bell DW, Settleman J, Haber DA. Epidermal growth factor receptor mutations in lung cancer. Nat Rev Cancer 2007 ; 7 : 169 -81

17. Yun CH, Mengwasser KE, Toms AV, et al. The T790M mutation in EGFR kinase causes drug resistance by increasing the affi nity for ATP. Proc Natl Acad Sci USA 2008 ; 105 (6): 2070 -5

18. Massarelli E, Varella-Garcia M, Tang X, et al. KRAS mutation is an important predictor of resistance to therapy with epidermal growth factor receptor tyrosine kinase inhibitors in non-small-cell lung cancer. Clin Cancer Res 2007 ; 13 (10): 2890 -6

19. Bean J, Brennan C, Shih JY, et al. MET amplifi cation occurs with or without T790M mutations in EGFR mutant lung tumors with acquired resistance to gefi tinib or erlotinib. Proc Natl Acad Sci USA 2007 ; 104 (52): 20932 -7

20. Van Cutsem E, Lang I, D’haens G, et al. KRAS status and effi cacy in the fi rst line treatment of patients with metastatic colorectal cancer (mCRC) treated with FOLFIRI with or without cetuximab: the CRYSTAL experience. Presentation at the 2008 ASCO Annual Meeting, 2008, Boston

21. Shigematsu H, Lin L, Takahashi T, et al. Clinical and biological features associated with epidermal growth factor receptor gene mutations in lung cancer. J Natl Cancer Inst 2005 ; 97 : 339 -46

22. Timar J, Lang I. The RAS paradoxon of the EGFR-targeted therapy of colorectal cancer. Magy Onkol 2008 ; 52 (2): 185 -91

23. Luo FR, Yang Z, Camuso A, et al. Dasatinib (BMS-354825) pharmacokinetics and pharmacodynamic biomarkers in animal models predict optimal clinical exposure. Clin Cancer Res 2006 ; 12 (23): 7180 -6

24. Quintás-Cardama A, Cortes J. Management of patients with resistant or refractory chronic myelogenous leukemia. Oncology (Williston Park) 2008 ; 22 (4): 430 -7

25. Hamilton AL, Eder JP, Pavlick AC, et al. Proteasome inhibition with bortezomib (PS-341): a phase I study with pharmacodynamic end points using a day 1 and day 4 schedule in a 14-day cycle. J Clin Oncol 2005 ; 23 (25): 6107 -16

26. De Leon J, Susce MT, Murray-Carmichael E. The AmpliChip CYP450 genotyping test: integrating a new clinical tool. Mol Diagn Ther 2006 ; 10 : 135 -51

27. Singer JB, Shou Y, Giles F, et al. UGT1A1 promoter polymorphism increases risk of nilotinib-induced hyperbilirubinemia. Leukemia 2007 ; 21 (11): 2311 -5

28. Dieterle F, Marrer E, Suzuki E, et al. Monitoring kidney safety in drug development: emerging technologies and their implications. Curr Opin Drug Discov Devel 2008 ; 11 (1): 60 -71

29. European Medicines Agency, EMEA. Available from: www.emea.europa.eu/htms/human/mes/biomarkers.htm

30. Dieterle F, Marrer E. New technologies around biomarkers and their interplay with drug development. Anal Bioanal Chem 2008 ; 390 (1): 141 -54

31. Woude GF, Kelloff GJ, Ruddon RW, et al. Reanalysis of cancer drugs: old drugs, new tricks. Clin Cancer Res 2004 ; 10 : 3897 -907

32. Wolff AC, Hammond ME, Schwartz JN, et al. American Society of Clinical Oncology/College of American Pathologists guideline recommendations for human epidermal growth factor receptor 2 testing in breast cancer. J Clin Oncol 2007 ; 25 (1): 118 -45

33. Jonathan KC. Knowles discusses the impact of pharmacogenomics on market segmentation. Nat Rev Drug Discov 2004 ; 3 : 822

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.



34. Goulart BHL, Clark JW, Pien HH, et al. Trends in the use and role of biomarkers in phase I oncology trials. Clin Cancer Res 2007 ; 13 (22): 6719 -26

• A very controversial but extremely valuable discussion between two groups of experts (Goulart BHL et al. versus Ratain MJ et al. ) in biomarkers and their clinical use. See also [35] .

35. Ratain MJ, Glassman RH. Biomarkers in phase I oncology trials: signal, noise or expensive distraction? Clin Cancer Res 2007 ; 13 (22): 6545 -8

• A very controversial but extremely valuable discussion between two groups of experts (Goulart BHL et al. versus Ratain MJ et al. ) in biomarkers and their clinical use. See also [34] .

36. Workman P. Biomarkers in early clinical trials: the committed and the skeptics. Clin Cancer Res 2008 ; 14 (8); 2512

37. Maheswaran S, Sequist LV, Nagrath S, et al. Detection of Mutations in EGFR in Circulating Lung-Cancer Cells. N Engl J Med 2008 ; 359 (4): 366 -77

38. Sawyers CL. The cancer biomarker problem. Nature 2008 ; 452 : 548 -52

•• The best overview of opportunities, challenges and limitations in the cancer biomarker fi eld.

39. Cossman J. Companionship on the Critical Path, Presentation at the Ventana Symposium, 2008 ; Paris

Affi liation Estelle Marrer † & Frank Dieterle † Author for correspondence Novartis Institutes for BioMedical Research, Translational Sciences, Novartis, Klybeckstr. 141, CH-4002 Basel, Switzerland Tel: +41 616962095 ; Fax: +41 616966212 ; E-mail: [email protected]

Exp

ert O

pin.

Dru

g M

etab

. Tox

icol

. Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

UB

Gie

ssen

on

12/1

7/14

For

pers

onal

use

onl

y.

Documents

Biomarkers in oncology drug development: rescuers or troublemakers?