PERSPECTIVES Table Biomarkers of established or potential clinicalutility to guide therapy Tumour type Biomarker Potential clinical use Breast Steroid hormone receptors elect hormone therapy biomark ct patient Breast HER2 elect trastuzumab use Breast Oncotype Dx gene profile KRAS mutation status use lteinstabit and KR w are fa Non-small cell lung EGFR mutation totreatment.Why toidentify Non-small celllung ERCCI the heter Clioblastoma MGMT methylation Guide temozolomide use ne [BOX 1 Melanoma BRAF V600E mutatio Select therapy ECfRepige2'gWCiectpIeeptocfERCCrctonrepaircrosrcompltemenmationgroupiHER2.abo e downstream path ntly,increasing attention ble.Initi the trea way to cally active metabo and endoxife CYP2D6 is involved in the ective ical studies to thos the clinical benefit of gefitinib when added phic in the populati ling groups anardl apy for adv .Slow metabo drug was effectivel United the risksa lizers may bea 0 n inva tance of activ ng EGFR muta tions to Erbtux Bristol-Myers Squibb,Merck might b likely to b shoud routiny be tested for CYP2D6 nd morere cent studie haveoed the mistry rather than deba ilen is a ce of tu ER2) s the rt nce of the nee of such evidence was suffic ent for Euro onclu sive or unreliable results again lice an c LC who have same exten rker assays are require narkers and clinica trial design Contrast the nent of and drug 0g RegaherYeeaalrilaoiatn EGFR pa uce regulatory (ressa:Astrazeneca)for treating on and both the challenges becomes clear. the FDA 364 MAY 2010 VOLUME 9 www.nature.com/reviews/drugdisc 2010 Matmillan Publishers Limited.All rizhts reserve Food and Drug Administration (FDA) label for irinotecan to recommend initiating treat￾ment at a reduced dose for patients with the *28 genotype. More recently, increasing attention has been paid to the pharmacogenetics of tamoxifen, a highly effective drug for the treatment and prevention of breast cancer. Tamoxifen is metabolized by cyto￾chrome P450 2D6 (CYP2D6) to biologi￾cally active metabolites: 4-OH tamoxifen and endoxifen. CYP2D6 is involved in the metabolism of many commonly used drugs and, like UGT1A1, the gene is polymor￾phic in the population, yielding groups of patients who are limited or extensive metabolizers of tamoxifen16. Slow metabo￾lizers may be at increased risk of cancer progression when receiving tamoxifen therapy17 and alternative treatments, such as aromatase inhibitors, might be preferred in such individuals. Whether or not women should routinely be tested for CYP2D6 genotype before receiving tamoxifen is a matter of ongoing debate, but the contro￾versy illustrates the importance of the need for better understanding of the genetics of drug metabolism before assuming that all patients will benefit from treatment to the same extent. Biomarkers and clinical trial design A better understanding of cancer biology and drug metabolism has enormous potential to improve the efficiency of drug development. The clinical development of gefitinib (Iressa; AstraZeneca) for treating NSCLC illustrates both the challenges and opportunities in the development of targeted therapies for cancer. In the case of gefitinib, the target (EGFR) was known and ample preclinical and early clinical data existed to indicate that the target was inhibited at pharmacologically achievable drug concentrations that were clinically tolerable. Initial clinical trials showed promising, even dramatic, results in some patients18 and the drug received marketing approval in the United States of America according to the accelerated approval pathway. Subsequently, large-scale, prospective, randomized Phase III trials failed to confirm the clinical benefit of gefitinib when added to standard chemotherapy for advanced NSCLC19,20 and marketing approval for the drug was effectively withdrawn in the United States and Europe. All of these actions occurred before the recognition of the importance of activating EGFR mutations to identify individuals who are likely to benefit from treatment with this class of agents6 , and more recent studies have confirmed the importance of tumour genotyping to iden￾tify likely responders21. Indeed, the weight of such evidence was sufficient for European regulatory authorities to once again license gefitinib for marketing for the treatment of patients with advanced NSCLC who have an activating mutation of the EGFR tyrosine kinase. Contrast the clinical development of gefitinib to that of trastuzumab (Herceptin; Genentech) — a molecule that also inhibits EGFR pathway signalling but the develop￾ment of which was restricted to study of patients with HER2 overexpression — and the value of having a validated biomarker to improve the efficiency of drug development becomes clear. Many ways to fail Despite these important advances in understanding tumour biology and using biomarkers to identify and select patients who are likely to benefit from or be resistant to treatment, there remain few examples of clinically useful biomarkers that can identify drug sensitivity and predict clinical benefit. Indeed, clinically useful biomarkers, such as HER2 and KRAS, are far more useful to identify patients who are unlikely to respond to treatment. Why is it so difficult to identify positive predictive biomarkers? Once again, the challenge lies primarily in understanding the heterogeneity of cancer and the plasticity of the cancer genome (BOX 1). Tumours with drug-sensitizing mutations can simultaneously harbour or develop drug resistance mutations as in the case of the EGFR T790M mutation22, or there may be downstream pathway-activating muta￾tions as in the case of KRAS23. Activation of a parallel pathway that circumvents a pharmacological block is known to occur as in the case of MET amplification, thereby causing resistance to small-molecule EGFR inhibitors24. In addition, pathway blockade can result in feedback upregulation of the pathway to overcome the block25. The opportunities for biomarker-directed drug development are exciting, offering the potential to limit enrolment of candidates for clinical studies to those most likely to benefit from the treatment under study. This enables the design of studies that demonstrate larger effects, but with smaller numbers of patients required26. However, the risks associated with this approach are considerable. An invalid biomarker (for example, EGFR expression for cetuximab (Erbitux; Bristol-Myers Squibb, Merck Serono, ImClone Systems)); a suboptimal technology to assess the biomarker (for example, immunohistochemistry rather than fluorescence in situ hybridization to assess HER2); or a technically difficult assay that provides inconclusive or unreliable results can all confound the clinical trial design. Therefore, confidence in the biological rele￾vance of the marker and extensive analytical validation of biomarker assays are required to move forward. Regulatory challenges Biomarker-driven clinical trials also intro￾duce regulatory challenges when the aim is to co-develop a biomarker assay with the drug. This is because both the test and the drug must meet regulatory standards for marketing approval and clinical use. Within the FDA, review of in vitro diagnostic tests Table 1 | Biomarkers of established or potential clinical utility to guide therapy Tumour type Biomarker Potential clinical use Breast steroid hormone receptors select hormone therapy Breast Her2 select trastuzumab use Breast Oncotype Dx gene profile Assess prognosis; select chemotherapy Colon KrAs mutation status Guide eGFr-specific antibody use Colon Microsatellite instability Assess prognosis or utility of 5-fluoruracil adjuvant treatment Non-small cell lung eGFr mutation Guide selection or use of eGFr tyrosine kinase inhibitors Non-small cell lung erCC1 select platinum-based chemotherapy Glioblastoma MGMT methylation Guide temozolomide use Melanoma BrAF v600e mutation select therapy eGFr, epidermal growth factor receptor; erCC1, excision repair cross-complementation group1; Her2, also known as erBB2; MGMT, methyl guanine methyltransferase. Pers P ectives 364 | MAY 2010 | vOLUME 9 www.nature.com/reviews/drugdisc © 2010 Macmillan Publishers Limited. All rights reserved