PERSPECTIVES OPINION

Implementing personalized cancer care Richard L. Schilsky Abstract | Implementing personalized cancer care requires a sound understanding of cancer genomics, familiarity with the analytical methods used to study cancer, knowledge of the mechanisms of action of targeted drugs, and ways to assimilate and understand complex data sets. Perhaps the greatest challenge is obtaining the drugs predicted to be beneficial based on the genomic profile of a patient’s tumour. A potential solution is creation of a national facilitated access programme and registry for off-label use of targeted anti-cancer drugs. Within such a programme, patients could receive the targeted agent matched to the genomic profile of their tumour. Physicians would receive guidance in interpretation of complex genomic tests and access to drugs. Pharmaceutical companies, payers and regulators would receive data on off-label drug and test use and clinical outcomes to inform their research and development plans and coverage decisions and to track real-world safety. Although recently launched prospective clinical trials will determine the true benefit of matching drugs to genomic alterations, the approach proposed here will facilitate delivery of personalized medicine services to participating patients while at the same time making observations that allow us to learn from each patient to inform clinical care and future research initiatives. Schilsky, R. L. Nat. Rev. Clin. Oncol. advance online publication 1 April 2014; doi:10.1038/nrclinonc.2014.54

Introduction Clinical oncologists have long recognized that each patient with cancer is unique. Clinical presentation of the disease varies from person to person; its natural history follows a unique course in each individual; the needs, preferences and tolerances of patients differ; and the outcomes of care vary widely even when the same course of treatment is delivered. Personalized cancer care is not a new concept, but its application has been enabled and enhanced by a far greater understanding of the biological under­pinnings of cancer than ever existed before. Our deeper understanding of cancer at the genetic level, coupled with the increasingly widespread availability of affordable technologies to interrogate the genome, the transcriptome, the proteome and other aspects of the tumour and host is leading to new paradigms of cancer care that rely increasingly on tumour molecular profiling throughout the course of care to refine prognosis and inform treatment Competing interests The author declares no competing interests.

decisions.1,2 A new generation of drugs, both anti­bodies and small molecules that target specific molecules and pathways crucial for cancer survival has revolutionized the treatment of some cancers and given hope to many patients with cancers once considered hopeless. Evidence is building through reports of small research studies and clinical anecdotes to suggest that patient outcomes are improved when a specific targeted agent can be matched to a molecular profile.3–7 Because the molecular profile of each person’s tumour is likely unique, personalized cancer care now implies far more than an understanding of each person’s preferences, it requires understanding the unique v­ulnerabilities of each individual tumour. Implementing personalized cancer care thus requires a sound understanding of cancer genomics; an appreciation of the strengths, weaknesses and limitations of the analytical methods available to study cancer; a recognition of the sources of variability in molecular testing, including the impact of variability in tissue handling and processing; a deep knowledge of the mechanisms of action of targeted drugs

NATURE REVIEWS | CLINICAL ONCOLOGY

and a way to collect, assimilate and understand complex data sets in the context of a busy clinical oncology practice. With the increasing availability of single gene and multiplex mutation analysis, large-scale next-­generation sequencing, transcriptomic analysis and other tests now offered by hospital, cancer centre and commercial laboratories, oncologists increasingly struggle with decisions about what tests to order for what purpose, how to understand and interpret the test results and how to convert the information provided by the tests to meaningful clinical decisions that help patients. I will discuss several key e­lements of implementing personalized cancer care services in an oncology practice setting and offer solutions to some of the obstacles that currently exist to making the promise of personalized m­edicine available to many patients.

Fit for purpose Biospecimen collection An essential tenet of molecular testing is that the test performed be appropriate to provide the required information about the biological state of interest to inform a clinically relevant decision in a specific patient population, that is, that the test be ‘fit for purpose’ in the intended use population. So too must the biospecimen obtained from the patient be fit for the analytical requirements of the test to be performed. It is necessary that clinicians are aware of the requirements of molecular tests with respect to the amount of tissue (or blood, serum, plasma, etc.) needed to provide a sufficient amount of the analyte of interest, type of specimen necessary to provide the requisite amount of tumour tissue; and the tissue handling, storage and shipping requirements to preserve the analyte(s) of interest. Often the clinician who obtains the tissue, such as an interventional radio­logist, is not the physician who orders the test; therefore, close communication between these clinicians is needed to insure that the right specimen is obtained in the optimal way for the intended analysis. The logistics are often complex, and clear guidance from laboratory testing facilities and professional societies to enable development of the standard operating pro­cedures that are essential to insure that high quality specimens ADVANCE ONLINE PUBLICATION  |  1

© 2014 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES are obtained and handled appropriately is often lacking. While some institutions have developed tissue handling guidelines for their own use, this situation could be remedied with the development and dissemination of specimen acquisition guidelines for commonly used molecular tests by national organisations, such as the College of American Pathologists (CAP). Such guidelines should be readily available in interventional radiology suites, operating rooms, pathology departments and other clinical locations where tissue specimens are commonly obtained or processed for molecular testing. An excellent existing resource is the National Cancer Institute Best Practices for Biospecimen Resources guidance document.8

Molecular workup A large and growing array of molecular diagnostic tests is offered to oncologists that claim to be useful to refine prognosis, select therapy, monitor treatment effectiveness or detect disease recurrence. Before ordering a test, oncologists should determine how the information provided by the test will inform their clinical decision-making. Important test characteristics, defined by several agencies and expert panels include: analyti­ cal validity (how well the test detects the analyte of interest), clinical validity (how well the test identifies clinically meaningful phenotypes), and clinical utility (usefulness of the test to inform a clinical decision that improves patient outcomes).9,10 Although of greatest importance to patients, demon­ stration of clinical utility is not required to achieve regu­l atory approval to market a test in the USA or to obtain reimbursement for its use. As a result, it may be difficult for physi­cians to determine the value of a test that is offered for the molecular assessment of a tumour. It is incumbent then on professional societies to develop and disseminate evidence-based guidelines for cancer molecu­lar diagnosis that include recommended markers, testing platform, definitions of positive and negative tests, optimal report content and test limitations or confounders. Indeed, the ASCO and CAP have offered such guidelines for hormone receptor and HER2 testing in breast cancer,11,12 and CAP, the Association of Molecular Pathologists (AMP) and the International Association for the Study of Lung Cancer (IASLC) have collaborated on the publication of guidelines for EGFR mutation and ALK translocation testing in adeno­ carcinoma of the lung.13 Such guidelines,

if widely implemented and coupled with labora­tory proficiency testing and certification will provide confidence in the reliability of test results delivered to oncologists. As more complex genomic analysis—such as whole genome and exome sequencing— is introduced into clinical practice it will be necessary to set standards for such testing that assure its clinical utility. The Institute of Medicine has recently published a report on ‘omics-based’ tests that recommends best practices for translation of an omics‑based discovery into a validated omics-based test for clinical use.14 In f­ollow-­up to this rep­ ort, the National Cancer Institute (NCI) published a checklist of criteria that can be used to determine the readiness of an omics-based test to guide clinical care in the context of clinical trials,15 an important step in e­stablishing the clinical utility of a new test.

What is actionable? Clinical interpretation The clinical interpretation of genomic tests is challenging even for experts in the field, and in many cancer centres this is now performed in the context of molecular tumour boards that integrate expertise from molecular pathologists, genomics experts and even bioinformaticians into the traditional multi-disciplinary clinical tumour board. The goal of such tumour boards is to identify actionable genomic events in a patient’s tumour, that is, gene variants that inform or recommend a therapeutic action, typically matching a drug to a putative drug target.16 However, there is not yet universal agreement on the definition of ‘clinically actionable’ with some authors suggesting that the type of clinical effect, the strength of evidence of the effect and the size of the effect must all be considered in the clinical context.17 Given the complexity of genomic information, how is the community clinical oncologist, who might not have access to the expertise of a molecular tumour board, to interpret the data and make an informed treatment recommendation? Clinical reports of genomic analysis vary greatly in content, format, breadth and depth of analy­ sis of the genomic data, that is, in the level of evidence to support the biological importance and clinical relevance of the finding, and in the use of decision support tools to assist the physician in arriving at a clinical recommendation. Various cancer gene databases have been developed that provide curated information about clinically significant gene variants and their relationship to

2  |  ADVANCE ONLINE PUBLICATION

disease outcomes or drug therapies, sometimes including information about available clinical trials,17and some commercial firms offer clinical interpretation of genomic tests. It remains to be seen whether and how physi­cians incorporate these resources into the flow of their clinical practice and whether they prefer that such tools be presented passively or actively in the context of the electronic medical record. It is clear, however, that many physicians will require assistance to determine if a specific genomic variant is clinically actionable. For example, the potential of the variant as a target might not be easily recognized or might be modified by the presence of other variants in the tumour. In addition, the specific course of treatment to pursue requires a deep understanding of the molecular pharmacology of available targeted agents, including their potential toxicities, in order to make the drug–target match with the greatest potential to benefit the patient without causing severe or unexpected side effects. Ideally, a trusted and curated database of actionable genomic variants could be established by an agency such as the US NCI or a professional society such as ASCO to insure that all oncologists have ready access to compre­ hensive information that is presented in an easily understood way to practitioners who must assimilate and use the information in the course of a busy clinical practice. For particularly challenging cases, ASCO is considering the development of a web-based, interactive, molecular tumour board that could provide access to expert consultation from clinical oncologists and molecular pathologists with expertise in genomics who provide in-depth discussion of the clinical and genomic features of the case, including potential treatment options, for educational purposes.

Making the match Perhaps the greatest challenge currently facing oncologists is obtaining the drug or drugs predicted to be beneficial based on the molecular profile of the patient’s tumour. In the great majority of cases such drugs will either be marketed agents that could be prescribed off label or investigational agents that are available only through participation in a clinical trial or, rarely, through a company-sponsored expanded access programme. In each case, the physician and patient face many complex problems. Will off-label use of a marketed drug be reimbursed by the patients’ insurance coverage? Is an appropriate clinical trial www.nature.com/nrclinonc

© 2014 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES accessible to the patient and are they eligi­ ble to participate? Is an expanded access programme offered by the drug sponsor and how does the patient learn about it and participate? Potentially beneficial treatment for a patient may be delayed by weeks or months as doctors engage in protracted discussion with drug companies and payers about access to treatments for their patients. Seeking access to drugs also places burdens on insurers and drug companies that receive frequent calls from patients and physicians seeking coverage determinations, information about clinical trials or compassionateuse programmes. Even when the drug can be obtained and the patient is treated, the outcomes of that treatment are not captured in any database that can inform other patients, treating physicians, payers, drug sponsors or regulators about the utility of the approach.

Access programme and registry A solution that potentially benefits all stakeholders is creation of a national facilitated access programme and registry. The programme would focus initially on enabling off-label prescribing of approved targeted therapies that are administered to patients with advanced cancer who have exhausted standard treatment options and for whom treatment is selected based on results of a genomic profile of their tumour performed in a Clinical Laboratory Improvement Amendments (CLIA)-certified laboratory using an a­nalytically validated test.

How might it work? The proposed programme, depicted in Figure 1, requires close collaboration among physicians, pharmaceutical companies and payers, as well as an honest broker to validate the treatment plan and maintain a robust IT infrastructure to capture patient characteristics and outcomes. If genomic profiling of the patient’s tumour has been performed using an analyti­cally validated test performed in a CLIA-certified laboratory, the patient would be offered enrolment in a national registry protocol that captures their demographic and clinical information. Based on the results of genomic profiling and the physi­cian’s clinical assessment of the patient, the treating physician would submit a proposed treatment plan and request for a targeted agent (commercially available but proposed for off-label use) to an honest broker third party, a ‘national formu­lary’. The drug dose and schedule would be consistent with the approved drug label to

minimize the risk of unexpected or severe toxicities. The physician’s request would be reviewed based on pre-determined and transparent criteria to determine if the treatment plan is scientifically rational, clinically reasonable, safe in the proposed patient and potentially beneficial. If approved, the drug would be prescribed for administration to the patient for a defined period, for example, 3 months, before disease re-assessment using conventional clinical criteria. At the time of re-assessment, the patient outcomes assessed by the treating physician (both efficacy and toxicity) would be recorded in the registry database. Treatment would continue until disease progression or treatment intoler­ance and the patient would be followed for survival.

Requirements for implementation This programme can succeed only if there is interest, commitment and partnership among all involved parties. Patients must con­sent to have their data capture­d in the registry, physicians must be willing to pre­ pare the request for the drug and the ration­ ale for its use and to submit the required outcome data for each patient (physi­cian payment could be made contingent on this by participating payers), pharmaceutical companies must be willing to provide their marketed drugs for off-label use either at no cost or at reduced cost under the terms of an expanded access programme, and payers must be willing to cover the medical costs associated with treatment. Although regulatory approval is not required for off-label prescribing of marketed agents, traditional expanded access programmes do require FDA approval and it is likely that both sponsors and regulators will prefer FDA approval of an ‘umbrella expanded access programme’ that includes multiple drugs in a single programme. Approval of the protocol by a national institutional review board would be obtained to facilitate enrolment at the point of care. ASCO could identify an independent third party to serve as the honest broker that organizes review of the proposed treatment plans. ASCO could host the outcomes registry. There are now more than 25 commercially available drugs that target molecular pathways that are frequently aberrant in human tumours. Table 1 lists some of the targeted agents approved for use in the US, the labelled indication and the putative drug target(s). A more-comprehensive list that includes drugs in clinical development can be found at My Cancer Genome.18 Clearly,

NATURE REVIEWS | CLINICAL ONCOLOGY

Genomic profiling performed using analytically validated assay in CLIA certified laboratory Patient is registered in national registry such as CancerLinQ Physician submits genomic profile and drug request with rationale

ED

APPROV

Honest broker reviews request against pre-determined guidelines and approves use of drug

Patient receives drug treatment Outcomes entered in registry Cost of care reimbursed

Figure 1 | Facilitated access programme and registry. Creation of a national facilitated access programme and registry requires close collaboration among physicians, pharmaceutical companies and payers. Genomic profiling of the patient’s tumour using an analytically validated test performed in a CLIA-certified lab, would allow the patient to be offered enrollment in a national registry protocol that captures their demographic and clinical information. Based on the results of genomic profiling, the treating physician would submit a proposed treatment plan and request for a targeted agent (commercially available but proposed for off-label use) to an honest broker third party, a ‘national formulary’. The drug dose and schedule would be consistent with the approved drug label. The physician’s request would be reviewed based on predetermined and transparent criteria to determine if the treatment plan is scientifically rational, clinically reasonable, safe in the proposed patient and potentially beneficial. If approved, the drug would be prescribed for administration to the patient for a defined period, for example, 3 months, before disease re-assessment using conventional clinical criteria.

there are drugs available that target many of the most commonly occurring pathway aberrations in human solid tumours, such as mutations in or overexpression of EGFR, ERBB2, VEGFR, BRAF, MEK, MET, ALK, KIT, FGFR and mTOR, and hundreds of new targeted drugs are in clinical develop­ ment, some of which will eventually make it to market.

Who benefits? All stakeholders potentially benefit from establishing this programme. Patients receive the targeted agent matched to the molecular profile of their tumour and stand to personally benefit if the treatment ADVANCE ONLINE PUBLICATION  |  3

© 2014 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES Table 1 | FDA-approved targeted agents for cancer treatment Drug

FDA approved indication

Target(s)*

Afatinib

NSCLC

EGFR

Axitinib

RCC

VEGFR

Bosutinib

CML

Bcr–abl

Cabozantanib

MTC

RET, VEGFR, MET, TRKB, TIE2

Cetuximab

Colon, NSCLC, HNC

EGFR

Crizotinib

NSCLC

EML4–ALK, ROS1, MET

Dabrafenib

Melanoma

BRAF V600E

Dasatinib

CML

Bcr–abl, SRC, cKIT, PDGFR

Erlotinib

NSCLC

EGFR

Everolimus

RCC, breast, pNET

mTOR, TSC1, TSC2

Ibrutinib

MCL

BTK

Imatinib

CML, GIST

Bcr–abl, cKIT

Lapatinib

Breast

EGFR

Nilotinib

CML

Bcr–abl

Panitumumab

Colon

EGFR

Pazopanib

RCC, STS

VEGFR, PDGFR, FGFR, KIT

Pertuzumab

Breast

HER2

Regorafenib

Colon

VEGFR, TIE2, PDGFR, RET, cKIT

Ruxolitinib

Myelofibrosis

JAK1, JAK2

Sorafenib

RCC, HCC, DTC

BRAF, KIT, FLT‑3, RET, VEGFR, PDGFR

Sunitinib

RCC, GIST, pNET

PDGFR, VEGFR, KIT, FLT‑3, RET

Temsirolimus

RCC

mTOR

Trametinib

Melanoma

MEK1, MEK2

Trastuzumab

Breast, gastric

HER2

Trastuzumab emtansine

Breast cancer

HER2

Vandetanib

MTC

RET, EGFR, VEGFR, TIE2

Vemurafenib

Melanoma

BRAF V600E

Vismodegib

BCC

SMO

*Drug targets obtained from review of FDA approved package insert for each drug. Abbreviations: BCC, basal-cell carcinoma; BTK, Bruton tyrosine kinase; CML, chronic myeloid leukaemia; DTC, differentiated thyroid cancer; GIST, gastrointestinal stromal tumour; HCC, hepatocellular carcinoma; HNC, head and neck cancer; MCL, mantle cell lymphoma; MTC, medullary thyroid carcinoma; NSCLC, non-small-cell lung cancer; pNET, pancreatic neuroendocrine tumour; RCC, renal cell carcinoma; STS, soft-tissue sarcoma.

is effective. Physicians receive guidance in interpretation of complex genomic tests and access to drugs. Pharmaceutical companies receive data on off-label drug use and clinical outcomes to inform their research and development plans and life-cycle manage­ ment of their products. Payers receive data on off-label use of drugs, as well as molecular diagnostic tests along with patient outcomes to inform future coverage decisions, and regulators receive data on extent of offlabel drug and test use, patient outcomes and real world safety data. With this programme, it becomes possible to deliver personal­ized medicine services to every patient while at the same time making observations that allow us to learn from each patient to inform future research initiatives.

What are the risks? The approach proposed here, that is, matching drugs to genomic alterations in patients with advanced cancer, has yet to be proven efficacious in prospective, randomized controlled trials. Thus, there is certainly a risk that patients will experience toxicity without any treatment benefit. The same is true for any off-label treatment programme in a patient who has exhausted all standard treatment options. Anecdotal evidence and small trials indicate that matching drugs to genomic profiles of the tumour can offer benefit to patients. 3–7 For example, the MOSCATO‑01 (Molecular profiling in Cancer for Treatment Optimization) trial is being conducted at the Institut Gustave Roussy in patients with treatment-resistant

4  |  ADVANCE ONLINE PUBLICATION

progressive metastatic cancers with lesions accessible to biopsy to perform molecular profiling by comparative genomic hybridization and whole-genome sequencing. Preliminary results of the first 129 patients were presented at the 2013 ASCO annual meeting.6 Progression-free survival (PFS) using therapy based on genomic assessment was compared to the PFS for the most recent therapy on which the patient had just experienced disease progression (PFS ratio). Of 129 consented patients, 111 (86%) had a tumour biopsy. An actionable target was identified in 52 patients (40%). Among the 25 patients treated according to their genomic assessment, five had partial responses (20%), 14 cases had stable disease (56%) and three cases of progressive dis­ ease (12%) were observed. The authors concluded that the result merits further testing and the trial is continuing. Concern also exists that launch of the programme proposed here will divert patients away from participation in clinical trials. However, many patients do not have ready access to clinical trials, are not eligible or choose to not participate. Clinical trial enrolment of adult patients with cancer remains at 3–5% of potentially eligible patients, despite the best efforts of many in the o­ncology community. There is also a risk that an inexperienced clinical oncologist will fail to understand or will misinterpret the results of complex genomic profiling tests and propose treatment plans that are inappropriate or excessively toxic, particularly if untested combinations of targeted agents are prescribed. Context dependency of the importance of specific mutations is also acknowledged as a potential limitation to this approach, that is, a drug that is effective when targeted against a certain mutation in one tumour type might not be similarly effective against the same mutation in a different tumour type. Of course, there is also concern that the clinical outcome data captured in the registry will be incomplete, inaccurate or biased and, therefore, its utility to make reliable clinical inferences about the usefulness of matching drugs and genomic profiles will be limited. Finally, there is some risk that pharmaceutical companies, insurance companies or both will not wish to participate in the programme proposed. Ongoing discussion with representative companies from both groups suggests this will not be the case as both seem to appreciate the potential value of this approach. www.nature.com/nrclinonc

© 2014 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES Risk mitigation strategies The programme proposed here will be conducted using an institutional review boardapproved registry protocol that is overseen by a steering committee to be appointed by ASCO. Using this approach will allow a number of strat­egies to be put in place to mitigate the risks described earlier. Specific eli­gibility cri­teria will limit enrolment to patients with advanced cancer for whom standard treatment options are no longer available; patients must have acceptable performance status and relatively normal organ function; must have recovered from the toxic effects of previous treatment and must provide informed consent that will include a description of the unproven nature of the approach being taken to select their t­reatment as well as the potential risks of treatment. Patients who are eligible to participate in clinical trials will be strongly encouraged to do so. To help guide the treating physician’s treatment recom­mendation, the protocol will contain a definition of ‘actionable’ genomic variants for purposes of the registry as well as criteria for drug selection (Box 1). Only single-agent treatment will be permitted in the initial phases of this programme and drug prescribing must follow the FDA approved label for dose, route and frequency of administration as well as toxi­city management. Only drugs specifically committed to the programme by the drug sponsor(s) will be eligible to be prescribed. Furthermore, the ‘honest broker’ to be selected or convened by ASCO will review the proposed treatment plan to insure that it conforms to the drug selection criteria specified in the protocol before treatment commences. The registry protocol will also specify the clinical efficacy and toxi­city end points to be collected but these will (at a minimum) include tumour response, PFS and overall survival defined according to conventional criteria as well as all serious adverse events and grade 3–5 toxic effects. Data release An important goal of this programme is to regularly release information about the performance of drugs targeted to specific genomic alterations in particular tumour types. As clinical inferences are made, it will be important to inform the participants in the programme, and the oncology commu­ nity more broadly, about which drugs seem to be working in which context and which drugs are not effective. This can be accomplished by regular review of the data for drug-variant matches once a sufficient

Box 1 | Proposed definition of clinically actionable variants and criteria for drug selection Definition of clinically actionable variants ■■ Gene variant is the target of an approved drug for any cancer indication ■■ Activating mutations in genes upstream of the molecular target of an approved drug ■■ Inactivating mutations in genes that result in unique susceptibility to a specific molecular intervention (for example, BRCA1 mutation and PARP inhibitors) ■■ Other genes of interest that have appropriate justification for selection based on published scientific evidence regarding susceptibility to a specific molecular targeted therapy Criteria for drug selection ■■ Level 1: Agent met a clinical end point (objective response, progression-free survival or overall survival) in a clinical trial testing the agent in the patient’s tumour type harbouring the mutation of interest ■■ Level 2: Agent is commercially available for use in any tumour type with the specific genomic variant identified in the patient’s tumour ■■ Level 2: Agent demonstrated evidence of clinical activity against the patient’s tumour type based on published literature ■■ Level 3: Agent demonstrated preclinical evidence of antitumour activity and evidence of target inhibition in model systems of patient’s tumour type

number of patients have been enrolled in each ‘match’ for a given tumour diagnosis. The protocol that governs this programme will include a statistical analysis plan that will specify the number of events that must be observed in each ‘match’ to declare the drug worthy of further study in that context or to recommend against continued use of the drug in that setting. Once the requisite number of events has been observed, the data will be reviewed by the programme’s steering committee, which will determine whether the data are sufficiently robust for public release. In this way, the programme can regularly report on both promising and disappointing outcomes that can help inform clinical practice, coverage decisions and research hypotheses.

Conclusions Clinical trial data have begun to emerge that suggest that patients have better outcomes when their treatment is assigned based on matching drugs to potential genomic targets in their tumour.3–6,19 Such reports have fueled interest among patients and physicians to use genomic profiling as a guide to treatment planning in patients with advanced cancer when standard treatment options have been exhausted, and commer­cial labora­tories have begun to market such profiling tests to onco­ logists. Clinical reports to date suggest that 30–70% of advanced solid tumours harbour potentially ‘actionable’ mutations or other gene variants depending on the definition of ‘actionable’ applied,3,20,21 but the outcomes of patients treated based on such tests remain largely anecdotal or unknown. The programme proposed here is not an intervention trial, that is, no attempt will be made to direct patient care or to study a particular drug

NATURE REVIEWS | CLINICAL ONCOLOGY

treatment. Nor is this programme strictly an observational cohort study in that there is the potential for interaction with physicians through clinical decision support s­ervices. The primary goal of this programme is to facilitate patient access to m­arketed agents used off label that are predicted to benefit them based on analy­sis of their tumour genomic profile and, importantly, to then observe which treatments benefit which patients by capturing their outcomes in a registry. By providing clinical decision support to oncologists in the analysis of tumour molecu­ lar profiles, and through use of an honest broker to validate the physician’s treatment plan, this programme seeks to enable all oncologists to have access to the same information and thoughtful analysis that is currently available only in some of the country’s leading cancer centres.16,22 By providing access to targeted agents, at least those that target common mutations, patients with advanced cancer will have an opportunity to benefit from a treatment that is potentially effective, but may not be easily accessible due to cost, cover­age or regulatory obstacles. Central to the success of this programme will be participation from the pharmaceutical industry so a sufficiently broad portfolio of targeted agents directed against the most commonly occurring pathway aberrations can be established. Support from the payer community will also be essential to insure that the medical care costs of patients receiving these off-label therapies are covered. By securing commitments for drugs and reimburse­ment as part of this programme, doctors and patients will be able to avoid lengthy and tedious negoti­ations with pharma­ceutical and insurance c­ompanies that consume v­aluable staff resources and delay treatment. ADVANCE ONLINE PUBLICATION  |  5

© 2014 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES Precedent for a conceptually similar programme was established with the National Oncologic PET Registry (NOPR) that was begun at a time when the clinical utility of PET scanning for the assessment of cancer was not well established, but the techno­logy was widely available. This situation led to substantial clinical usage and generation of multiple claims for reimbursement, but without any mechanism to learn whether or not the PET scan informed clinical decision-making. The NOPR was established by several professional s­ocieties to capture information on whether and how the scan impacted physician decision-­ making. 23 The Center for Medicare and Medicaid Ser vices (CMS) agreed to re­i mburse for the scans if the ordering physi­cian submitted information pre-scan and post-scan regarding their treatment plan and whether it changed as a result of performing the scan. Results captured in the NOPR supported the clinical utility of PET scanning for most tumour types, which led CMS to issue a national coverage determination that a limited number of scans would be reimbursed for most cancer patients.24 By capturing information on the clinical character­istics of a cancer patient, the molecular profiling test performed, the clinical decision based on that test, the treatment administered and the patient outcomes, payers will have a wealth of information to inform coverage decisions. Similarly, the pharmaceutical companies that participate in this programme will receive far more information about offlabel use of their drugs than would customarily be available and could potentially use this information to make informed decisions about whether to pursue secondary i­ndications for marketed agents. The ASCO Cancer Research Committee has embraced this proposal and appointed a steering committee of oncology experts to develop the implementation plan that is briefly described herein. We will likely begin with a pilot programme that involves a small number of oncology practices, a single payer and a few drugs. It will not be easy. Some fundamental issues will have to be considered, such as whether any molecu­ lar profiling test performed in a CLIAcertified laboratory will be acceptable; how clinical decision support will be provided to those physicians who need it; and what rule set will drive treatment recom­ mendations if multiple drugs are available or if more than one actionable gene variant is detected in a tumour. The NCI is

organizing a prospective clinical trial examining a similar concept, Molecular Analysis for Therapy Choice (MATCH).25 This protocol will be multiple single arm phase II studies conducted under a single umbrella protocol and investigational new drug application. Patients with advanced-stage cancer and accessible tumour will a undergo biopsy, and molecular profiling will be performed in a network of CLIA-certified laboratories using common techno­logy and procedures to examine a panel of 200 to 300 commonly aberrant genes. A ‘level of evidence’ approach to identify targets and assign drugs will be employed to match the treatment to the gene variant. For example, gene variants that are the targets of FDAapproved drugs will be considered to have the highest level of evidence supporting the match. The MATCH protocol will have dual primary end points of objective response rate and 6‑month PFS and will use a Simon 2-stage design26 to declare a drug active or not against the gene variant being targeted. The drug portfolio will include primarily investigational agents. This clinical trial is similar in structure but vastly different in objectives from the programme we propose. The MATCH trial is a phase II trial that seeks to efficiently determine the activity of investigational drugs against genomic targets as part of an overall development strategy for a new agent. Our programme is not about studying investigational drugs, it is about informing care; that is, insuring that oncologists practicing in any setting have access to information and support tools to enable them to make the best possible treatment recommendations to patients with advanced cancer whose tumours have been analysed to reveal their molecular portrait. Our programme does not seek to determine what can work in the clinical trial setting, but what does work in real world patients with advanced-stage cancer by recording the outcomes of patients now being treated in the era of genomic medicine. This programme also seeks to facilitate that care by providing access to targeted agents for off-label use. Most importantly, this programme provides a mechanism to capture information that is already being generated every day in the clinical practice of onco­ logy, but is hidden in medical records. It will also allow us to analyse and then learn from that information for the benefit of all stakeholders involved in cancer drug develop­ ment, drug regulation, cancer care and healthcare reimbursement, but particularly for the benefit of patients.

6  |  ADVANCE ONLINE PUBLICATION

A new generation of prospective clinical trials exemplified by the recently launched NCI M‑PACT study and the soon to be launched Lung Cancer Master Protocol,27,28 will provide the definitive test of the value of matching drugs to genomic alterations. Obtaining these results will likely take years. Until then, the programme proposed here will allow us to capture real world data that, while imperfect, might still be informative to guide the development of this novel t­reatment strategy. American Society of Clinical Oncology, 2318 Mill Road, Suite 800, Alexandria, VA 22314, USA. [email protected] 1.

Garraway, L. A. Genomics-driven oncology: framework for an emerging paradigm. J. Clin. Oncol. 31, 1806–1814 (2013). 2. MacConaill, L. E. & Garraway, L. A. Clinical implications of the cancer genome. J. Clin. Oncol. 28, 5219–5228 (2010). 3. Tsimberidou, A. M. et al. Personalized medicine in a phase I clinical trials program: The MD Anderson Cancer Center Initiative. Clin. Cancer Res. 18, 6373–6383 (2012). 4. Arnedos, M. et al. Array CGH and PIK3CA/AKT1 mutations to drive patients to specific targeted agents: A clinical experience in 108 patients with metastatic breast cancer. Eur. J. Cancer 48, 2293–2299 (2012). 5. Andre, F. et al. Comparative genomic hybridization array and DNA sequencing to direct treatment of metastatic breast cancer: a multicenter, prospective study (SAFIR01/ UNICANCER). Lancet Oncol. 15, 267–274 (2014) 6. Hollebeque, A. et al. Molecular screening for cancer treatment optimization (MOSCATO 01): A prospective molecular triage trial—Interim results [abstract]. J. Clin. Oncol. 31 (Suppl.), a2512 (2013). 7. Iyer, G. et al. Genome sequencing identifies a basis for everolimus sensitivity. Science 338, 221 (2012). 8. National Cancer Institute. Best Practices for biospecimen resources [online], http:// biospecimens.cancer.gov/global/pdfs/ NCI_Best_Practices_060507.pdf (2007). 9. Deverka, P., Messner, D. & Dutta, T. Effectiveness guidance document: Evaluation of clinical validity and clinical utility of actionable molecular diagnostic tests in adult oncology [online], http://www.cmtpnet.org/wp-content/ uploads/downloads/2013/07/ CMTP_MDx_EGD07‑17‑2013.pdf (2013). 10. Teutsch, S. M. et al. The evaluation of genomicapplications in practice and prevention (EGAPP) initiative: methods of the EGAPP Working Group. Genet. Med. 11, 3–14 (2009). 11. Hammond, M. E. H. et al. American Society of Clinical Oncology/College of American Pathologists guideline recommendations for immunohistochemical testing of estrogen and progesterone receptors in breast cancer. J. Clin. Oncol. 28, 2784–2795 (2010). 12. Wolff, A. C. et al. Recommendations for human epidermal growth factor receptor 2 testing in breast cancer: American Society of Clinical Oncology/College of American Pathologists Clinical Practice Guideline Update. J. Clin. Oncol. 31, 3997–4013 (2013).

www.nature.com/nrclinonc © 2014 Macmillan Publishers Limited. All rights reserved

PERSPECTIVES 13. Lindeman, N. I. et al. Molecular testing guideline for selection of lung cancer patients for EGFR and ALK tyrosine kinase inhibitors: Guideline from the College of American Pathologists, International Association for the Study of Lung Cancer, Association of Molecular Pathology. J. Thorac. Oncol. 8, 823–859 (2013). 14. Micheel, C. M., Nass, S. J. & Omenn, G. S. (Eds) Evolution of Translational Omics: Lessons Learned and the Path Forward (The National Academies Press, 2012). 15. McShane, L. M. et al. Criteria for the use of omics-based predictors in clinical trials. Nature 502, 317–320 (2013). 16. Meric-Bernstam, F., Farhangfar, C., Mendelsohn, J. & Mills, G. B. Building a personalized medicine infrastructure at a major cancer center. J. Clin. Oncol. 31, 1849–1857 (2013). 17. Van Allen, E. M., Wagle, N. & Levy, M. A. Clinical analysis and interpretation of cancer genome data. J. Clin. Oncol. 31, 1825–1833 (2013).

18. My Cancer Genome. Anticancer agents [online], http://www.mycancergenome.org/content/ other/molecular-medicine/anticancer-agents/ (2014). 19. Von Hoff, D. D. et al. Pilot study using molecular profiling of patients’ tumors to find potential targets and select treatments for their refractory cancers. J. Clin. Oncol. 28, 4877–4883 (2010). 20. MacConaill, L. E. et al. Profiling critical cancer gene mutations in clinical tumor samples. PLoS ONE 4, e7887 (2009). 21. Iyer, G. et al. Prevalence and co-occurrence of actionable genomic alterations in high-grade bladder cancer. J. Clin. Oncol. 31, 3133–3140 (2013). 22. Tran, B. et al. Cancer genomics: technology, discovery and translation. J. Clin. Oncol. 30, 647–660 (2012). 23. Hillner, B. E. et al. Impact of Positron Emission Tomography/Computed Tomography and Positron Emission Tomography (PET) alone on expected management of patients with cancer: initial results from the National Oncologic PET Registry. J. Clin. Oncol. 26, 2155–2161 (2008).

NATURE REVIEWS | CLINICAL ONCOLOGY

24. Center for Medicare and Medicaid Services. Proposed decision memo for Positron Emission Tomography (FDG) for solid tumors (CAG‑00181R4) [online], http://www.cms.gov/ medicare‑coverage‑database/details/ nca‑proposed‑decision-memo.aspx?NCAId=263 (2013). 25. Conley, B. A. Precision cancer medicine exceptional responders NCI-MATCH [online], http://deainfo.nci.nih.gov/advisory/ctac/ 1113/Precision%20Cancer%20Medicine.pdf (2014). 26. Simon, R. Optimal two stage designs for phase II clinical trials. Control. Clin. Trials 10, 1–10 (1989). 27. US National Library of Medicine. ClinicalTrials.gov [online], http://clinicaltrials.gov/ct2/show/ NCT01827384?term=MPACT&rank=3 (2014). 28. Friends of Cancer Research. Lung cancer master protocol activation announcement [online], http://www.focr.org/sites/default/files/ Lung%20Cancer%20Master%20Protocol%20 Activation%20Announcement%20Slides.pdf (2013).

ADVANCE ONLINE PUBLICATION  |  7 © 2014 Macmillan Publishers Limited. All rights reserved