Precision medicine in oncology drug development: a pharma perspective

REVIEWS

Precision medicine in oncology drug development: a pharma perspective Simon J. Hollingsworth AstraZeneca, Innovative Medicines & Early Development – Oncology, Unit 310 – Darwin Building, Cambridge Science Park, Milton Road, Cambridge CB4 0FZ, UK

A rapid expansion in precision medicine founded on the potential for durable clinical benefit through matching a drug to a predictive marker used to select patients has driven the development of targeted drugs with accompanied companion diagnostics for patient selection. Oncology has been at the forefront, with the improvements in patient survival notable. Increasing numbers of molecular subgroups require an equally increasing number (and new generation) of highly selective agents targeting inevitably lower incidence molecular segments, posing significant challenges for drug development. Innovative trial designs (umbrella or basket studies) are emerging as patient-centric approaches and public–private partnerships, cross-industry, government and non-profit sector collaborations are enabling implementation. Success will require continued innovation, new paradigms in oncology drug development and market approval and continued collaboration.

The landscape The landscape in oncology drug development, particularly early clinical development, is evolving rapidly with recent years seeing the notable expansion of precision medicine (variously termed stratified medicine, personalised healthcare, etc.), the central premise of which is to offer greater potential for durable clinical benefit by matching a drug (and its mechanism-of-action) to a predictive marker used to select patients. The understanding of tumours in unprecedented molecular detail, the ‘hallmarks of cancer’ [1–6], coupled with modern drug development enabling specific targeting of the implicated pathway or mechanism (hallmark) and development of diagnostic technologies to identify patients (by markers) have collectively enabled notable improvements in survival rates for some cancers [7–14]. The more we have learnt about the molecular mechanisms that drive these tumours, and have thus been able to develop drugs targeted to these mechanisms, the greater the benefit seen in patient survival. Lung cancer, particularly exemplified by non-small-cell lung cancer (NSCLC) [7,8], has seen great development (and success), enabled by the increasing understanding of its molecular subclassification [2,3]. For example, the elucidation of the relationship between epidermal growth factor receptor (EGFR) E-mail address: [email protected]. 1359-6446/ß 2015 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.drudis.2015.10.005

mutation and response to EGFR tyrosine kinase (EGFR-TK) inhibition [15,16] underpinned the development and subsequent approval of IRESSA1 (gefitinib; AstraZeneca) for treatment of adults with locally advanced or metastatic NSCLC with activating mutations of EGFR-TK – following the completion of the Phase III studies IPASS (NCT00322452) [17,18] and INTEREST (NCT00076388) [19,20]. Similarly, and more recently, the identification of anaplastic lymphoma kinase (ALK) fusion oncogene as a molecular driver in some adenocarcinomas [21–23] underpinned the development of the ALK tyrosine kinase inhibitor XALKORI1 (crizotinib; Pfizer) [24] and its approval for treatment of patients with late-stage, locally advanced or metastatic NSCLC expressing the abnormal ALK gene – the approval including a companion diagnostic test for the ALK gene (i.e. the Vysis ALK Break Apart FISH Probe Kit) [25]. These are only selected examples in lung cancer but they have particularly marked the developing oncology landscape and its continued evolution towards an increasing number of patient–tumour groups identified by (increasingly complex) diagnostics to enable coupling to molecularly targeted drugs. For up-to-date approvals see the FDA [26–28] and the European Medicines Agency (EMA) [29]. The drivers for precision medicine are clear, and widely and robustly discussed, for examples see [30,31]. For patients (and physicians) advantages include durable clinical benefit, reduced

www.drugdiscoverytoday.com 1 Please cite this article in press as: Hollingsworth, S.J. Precision medicine in oncology drug development: a pharma perspective, Drug Discov Today (2015), http://dx.doi.org/10.1016/ j.drudis.2015.10.005

Reviews  POST SCREEN

Drug Discovery Today  Volume 00, Number 00  October 2015

DRUDIS-1690; No of Pages 9 REVIEWS

Drug Discovery Today  Volume 00, Number 00  October 2015

exposure to noneffective drugs and potential to exploit current scientific and technological advances. For the pharmaceutical industry, the potential to tackle core challenges in discovering and developing better and more efficacious medicines, reducing attrition in drug development and reducing development costs are particularly beneficial.

Precision medicine in oncology drug development Reviews  POST SCREEN

For oncology drug development the increasing number of molecular subgroups requires a portfolio approach with an equally increasing number of agents targeting inevitably lower incidence segments. Not only is there a need for a new generation of highly selective molecularly targeted agents, the drivers for precision medicine, but there is also a need to understand the mechanisms that enable their development once discovered. The intent here is not to review modern oncology drug discovery, rather the development of a candidate drug once ready for clinical trials. The increasing number of highly selective, molecularly targeted agents generates significant challenges for the more established drug development process. For example, current patient screening approaches are suboptimal and are not effective for patients, physicians or industry. Low frequency events are difficult to find, diagnostics sample quality and/or quantity compromise multiple analyses, there is poor patient and/or physician experience with cycles of repeat diagnostics, and current regulatory requirements for randomised trials and to validate companion diagnostics are challenging. Fig. 1 illustrates some of the reality of conventional screening approaches if used in an early clinical development study. The figures are a stark and surprising reality. To run a study testing a new candidate drug in a patient subpopulation selected by a molecular marker with a 2% incidence one would need to screen 78 patients for every one patient recruited to the study, with a standard 20 patient Phase I expansion in this patient

subpopulation requiring screening of at least 1560 patients in total. Equally staggering is the cost – if using a relatively simple diagnostic (costing US$1000 per assay) the screening costs alone to find these 20 patients would be in the order of US$1.8 million. Significantly, the patient experience is extremely poor with cycles of disappointment being first considered for a trial only then to fail screening (not having the marker for recruitment), requiring repeat biopsies (if possible) to enable analysis of the next marker, and with limited drug options. From a clinical trial operational viewpoint, this is complex and unwieldy. This type of approach is clearly not sustainable. We need innovative trial designs better suited to development of targeted molecules and that, in turn, require innovation and change in those aligned areas integral to the implementation and success of such trials – new generation multiplex diagnostics coupled with diagnostic standards and methods standardisation, regulators willing to engage with new types of (different) datasets and improved patient access to diagnostics and candidate drugs. Addressing many of these points has seen the emergence of umbrella (within tumour types, selected by different markers for single or multiple candidate drugs) and basket (across tumour types, often selected by single marker or for a single candidate drug) studies in early clinical development. Whereby, rather than using serial, single diagnostics to align patients to different trials, a single multiplex diagnostic is used to assign patients to different candidate drugs (or trial arms) within the same trial. Designs vary but the principle remains: ‘select the trial for the patient, not the patient for the trial’. Such studies offer greater options for patients, with candidate drugs aligned closely to their tumour characteristics and significant efficiencies made possible in screening and patient flow. In theory, this should reduce timelines and costs of clinical development and provide a more patient-centric and sustainable way forwards for drug development.

Recruitment to study Phase I expansion study in a molecular subgroup Patient selection marker at 2% incidence

Selection by standard IHC or DNA diagnostic, costings ~US$1125 (indicative costs)

15% test fail – technical reasons

Example for an pharma sponsored study Patients screened / recruited Screening $ / patient recruited (assay, processing, logistics, reporting)

For 20 patient study

Operational delivery

Physician experience

15% patient drop-out – clinical reasons Patient experience = Patient recruited

78 screened / 1 recruited

US$88 235

1560 patients screened US$1.8 million Complex Limited options – keep screening if tumour material allows Cycle of disappointment Up – potential trial Down – not eligible Repeat biopsies Limited drug options

Drug Discovery Today

FIGURE 1

The reality of conventional screening approaches in an early clinical development study. For a Phase I expansion in a group of patients selected based on a molecular marker with 2% incidence, using a standard diagnostic approach (e.g. IHC or DNA test, with indicative cost), and indicative rates of screening failure and patient drop-out. 2

www.drugdiscoverytoday.com Please cite this article in press as: Hollingsworth, S.J. Precision medicine in oncology drug development: a pharma perspective, Drug Discov Today (2015), http://dx.doi.org/10.1016/ j.drudis.2015.10.005

DRUDIS-1690; No of Pages 9

Clinical trials The evolution of clinical trial designs for personalising medicine has centred understandably around the growth in development of targeted agents, enabled significantly by the in-parallel advances and development of high-throughput technologies that have permitted an unprecedented level of detail to be gained from tumours, and the emergence of a global level of activity in tumour screening programmes. The BATTLE (Biomarker-integrated Approaches of Targeted Therapy for Lung Cancer Elimination; NCT00409968) trial [32] was the first prospective molecularly stratified adaptive randomised clinical trial. With the objective of assessing utility in certain marker – therapy associations, chemorefractory NSCLC patients (n = 255) were adaptively randomised to five drugs based on tumour molecular markers analysed from fresh (mandated) tumour specimens. Using a Bayesian model for real-time analysis of efficacy and an 8-week disease-control rate as the primary endpoint, the study demonstrated a notable benefit from Nexavar1 (sorafenib; Bayer and Onyx Pharmaceuticals) in mutant-KRAS patients. The subsequent and ongoing BATTLE trials take this concept further: the BATTLE-2 trial (NCT01248247) will recruit n = 450 advanced lung cancer patients progressing on first-line chemotherapy to a number of drugs or drug combinations based on analysis of 11 biomarkers; the BATTLE-FL (front-line) trial (NCT01263782) will

REVIEWS

use molecular profiling to allocate treatment-naive metastatic and/or recurrent EGFR wild-type lung cancer patients to one of three drugs in combination with doublet chemotherapy. Similarly, but based in breast cancer, the I-SPY trial (NCT00043017) identified combined clinical, genomic and imaging predictive markers for early cancers (n = 221) treated with neoadjuvant chemotherapy [33–36], and the subsequent I-SPY 2 trial (NCT01042379) will examine the same stage 3 breast cancer patient population randomised between standard neoadjuvant chemotherapy or the same combined with a targeted agent based on tumour molecular profile. We are now seeing a global level of innovation in this area (Table 1 identifies selected examples) and a broad range of initiatives and studies ongoing that span from early-phase clinical signal seeking studies (typically Phase IIa studies) once a drug candidate has an available recommended Phase II dose and schedule (e.g.: MATRIX National Lung Cancer Trial, Cancer Research UK – in NSCLC multidrug, multi-pharma, Phase IIa study with patients selected via a nationwide screening programme) [37] through to later-phase studies with intent to regulatory interactions and drug registration (e.g.: Lung Master Protocol, Friends of Cancer Research US – in squamous cell lung cancer, multidrug, multipharma; Lung-MAP, NCT02154490) [38]. Given the need (often) to identify a reasonable number of patients in lower incidence

TABLE 1

Selected precision-medicine-based studies being conducted globally Area

Study

Tumour(s)

Phase and design

Identifier

USA

Lung-MAP BATTLE

Squamous lung NSCLC

Phase II/III randomised Umbrella – route to 4x Phase II randomised

BATTLE-2 BATTLE-FL I-SPY I-SPY 2 NCI-MPACT NCI-MATCH

NSCLC NSCLC Breast cancer Breast cancer All All

Phase II randomised Phase II randomised Phase II diagnostic study Phase II randomised Phase II stratified, non-randomised Screening, route to Phase II

NCT02154490 NCT00409968 (umbrella) NCT00411671 NCT00411632 NCT00410059 NCT00410189 NCT01248247 NCT01263782 NCT00043017 NCT01042379 NCT01827384 –

Global

V-BASKET

All

Phase II stratified, non-randomised

NCT01524978

Asia

VIKTORY LC-SCRUM Japan-SCRUM

Gastric cancer NSCLC Lung

Screening, route to Phase II Screening, route to Phase II/III Screening, route to Phase II/III

NCT02299648 – –

EU

CREATE AURORA SPECTAColor SPECTALung WINTHER

Selected Breast cancer Colorectal cancer Lung All

Phase II stratified, non-randomised Screening, route to Phase I/II/III Screening, route to Phase I/II/III Screening, route to Phase I/II/III Stratified, non-randomised

NCT01524926 NCT02102165 NCT01723969 NCT02214134 NCT01856296

France

MOSCATO SHIVA MOST SAFIR 01 SAFIR 02 Lung SAFIR 02 Breast

All All All Breast cancer NSCLC Breast cancer

Screening, route to Phase I/II Phase II stratified, controlled Phase II stratified, randomised Screening, route to Phase I/II Phase II stratified, randomised Phase II stratified, randomised

NCT01566019 NCT01771458 NCT02029001 NCT01414933 NCT02117167 NCT02299999

UK

Lung-MATRIX CRUK SMP1 FOCUS 4

NSCLC Selected Colorectal cancer

Phase II stratified, non-randomised Screening, feasibility Phase II/III randomised

EudraCT: 2014-000814-73 – EudraCT: 2012-005111-12

Abbreviation: NSCLC, non-small-cell lung cancer. Further details on each study are available via ClinicalTrials.gov (NCT) or European Clinical Trials Register (EudraCT) identifier/number.

www.drugdiscoverytoday.com 3 Please cite this article in press as: Hollingsworth, S.J. Precision medicine in oncology drug development: a pharma perspective, Drug Discov Today (2015), http://dx.doi.org/10.1016/ j.drudis.2015.10.005

Reviews  POST SCREEN

Drug Discovery Today  Volume 00, Number 00  October 2015

DRUDIS-1690; No of Pages 9 REVIEWS

Drug Discovery Today  Volume 00, Number 00  October 2015

Reviews  POST SCREEN

segments, most approaches bring together members of tumourbased consortia, industry, charity and other partners involving national-level screening initiatives for example: AURORA; Breast International Group in metastatic breast cancer, EU screening initiative (NCT02102165) [39]; the French UNICANCER-sponsored SAFIR studies in NSCLC (SAFIR02_Lung; NCT02117167) and metastatic breast cancer (SAFIR02_Breast; NCT02299999); and LC-SCRUM (and SCRUM-Japan), the Japanese nationwide screening initiatives in lung, colorectal and gastric cancers [40].

Making it happen Implementation of such studies is not trivial and requires integration of multiple complex components at the simplest levels – the processes central to how we: (i) find and recruit patients in selected segments (subpopulations); (ii) ensure accuracy and consistency in the diagnostics used to identify patient segments, developed and maintained to common standards, using the latest technological developments to enable efficiency and cost-effectiveness; (iii) enable broad access for patients and physicians to the latest developmental agents offering multiple rather than limited options; and (iv) develop and embed a regulatory framework that enables study conduct and the use of different types of data package generated to seek market approval. The need to screen and find sufficient numbers of patients with tumours with a specific marker has seen the generation of many cooperative groups within and across regions (e.g. AURORA; NCT02102165) [39], the SPECTA programmes (SPECTAColor; NCT01723969 and SPECTALung; NCT02214134) [41,42] and the collective involvement of multiple pharma, for example NCI-MATCH [43], MATRIX [37], to allow availability of a broader range of candidate drugs and wider collaboration from tumour-specific consortia, diagnostics and regulatory groups and major charities and other

(1) Diagnostic system • • • • • • • • • • •

Platform Screening/selection algorithm Broad patient profiling Sample efficient Robust data generation Cost-effective Transferable, widely deployable Works to agreed standards Viable development route Support regulatory interactions Support for markets

Broad and robust tumour profiling for patient selection

interested parties. This is central to success but it does further increase the complexity of such approaches. For early-phase clinical signal seeking studies (Phase Ib, Phase IIa), as a general guide, there are a number of aspects that can significantly enable implementation.  A diagnostic system (platform, screening or selection algorithm) that enables broad but robust tumour–patient profiling and provides viable development routes for larger global studies, regulatory interactions and support for markets.  An early development protocol that is flexible allowing change to emerging science and understanding of patient and tumour markers, and/or a confirmatory development protocol permitting regulatory interactions using different types of datasets.  An operational machinery that allows studies to be conducted over diverse groups and geographical areas with aligned and efficient regulatory and ethics processes, patient screening and recruitment and ability to distribute multiple candidate drugs to multiple sites in a cost-effective and efficient manner. Even with the simplest of studies this is complex and, although studies typically require individual case solutions, we suggest some general considerations that can help with these challenges (more detail is provided in Fig. 2 with particular relevance to early clinical development).  Development and implementation of marker panels and nextgeneration sequencing.  Modular and rolling protocol designs with separate but integrated screening and clinical trial components, permitting study arms (of tumour-marker–drug-candidate pairs) to be flexibly added or removed.  Using single protocols, with single regulatory and ethics submissions, hub-and-spoke models for clinical sites and centralised pharmacies.

(2) Protocol • Single or aligned protocol • Aligned and efficient review – centralised regulatory/ethics • Flexible • Modular • Rolling – open ended • Adaptable to emerging science • Allows different datasets • Allows regulatory interactions

Flexible protocol with central review

(3) Operational delivery • Hub-and-spoke models • Must accommodate – diverse groups and geographical areas • Centralised pharmacy – to enable cost-effective delivery of multiple drugs to multiple sites • Highly collaborative working – across many different groups • Good partners

Centralised pharmacy and collaborative working

Drug Discovery Today

FIGURE 2

Considerations for successful delivery of precision medicine studies. Successful delivery of precision medicine (umbrella, basket, multidrug, portfolio) studies, particularly in early clinical development, requires innovation across clinical trial design and implementation – key aspects to consider for diagnostics, clinical protocol and operational delivery. 4

www.drugdiscoverytoday.com Please cite this article in press as: Hollingsworth, S.J. Precision medicine in oncology drug development: a pharma perspective, Drug Discov Today (2015), http://dx.doi.org/10.1016/ j.drudis.2015.10.005

DRUDIS-1690; No of Pages 9 REVIEWS

Early clinical stage signal seeking studies are now readily implementable, although they do require region-specific approaches and consideration. Often there is a trade-off to be made between aspects such as complexity, needed resources, strategy, cost, regulatory environment, partner capabilities, drug distribution needs, among others. A fundamental consideration is protocol design, particularly when considering strategy for regulatory interactions and how best to enable delivery of the clinical design. For example, the MATRIX [37] study in NSCLC contains 21 trial arms at the outset, each the combination of candidate drug and marker to select patients, contained within a single protocol, and was designed to be able to add-in or take-out arms and drug–marker pairs once the specific hypothesis has been tested. This places a notable burden on the review process (i.e. the need for potentially many amendments to accommodate the changing in arms is complex and could understandably raise many questions during the review and approval process). It does however contain everything within a single protocol, and so is a practical and self-contained approach. By contrast, screening programmes such as AURORA (NCT02102165) [39] or VIKTORY (NCT02299648) are standalone and used to direct selected patients to other (aligned) trials under separate protocols. Here, if multiple arms (drugs and markers) are contained within a single ‘receiving protocol’ there can be much the same complexity as seen in MATRIX. However, each drug and marker pair combination can also be written as a single protocol in themselves, which offers a simple and single protocol-by-protocol basis that could be more favourable for the review and regulatory process, although it is time and resource intensive given the potential for multiple protocols. Protocol designs vary greatly, but key design aspects drive (and offer) some core advantages and disadvantages (see Fig. 3 for selected examples).

Single protocol Screening protocol Core inc-/exclusion criteria Core trial design Supplement 1 Supplement 2 Supplement 3 Supplement 4 Supplement 5

Early-stage clinical studies can be shaped to look for clinical signals across a broad range of selected patient populations within the same study, thereby testing multiple hypotheses in different study arms. By contrast, later-stage confirmatory clinical studies that are typically used for regulatory submission and to seek market approval (usually Phase III studies) differ by being focused on generating data in a large enough cohort of a specific patient subpopulation to confirm a clearly defined and controlled hypothesis, along with the requirement of using a market-ready companion diagnostic (i.e. one with a finalised and locked-down design that will not change and represents the assay to be used on the market in routine clinical practice). However, many of the same considerations remain, and we are also seeing the increased availability (and use) of broad-based screening programmes as a feeder mechanism to channel patients to such studies. Perhaps the most advanced later-stage clinical trial is the Lung-MAP master protocol in NSCLC (NCT02154490), which is an adaptive Phase II/III study designed with registration intent, and tests candidate drugs from multiple pharmaceutical companies in a public–private collaborative partnership with cross-industry, government and non-profit sector involvement [38]. Here, although the final output differs (i.e. seeking market approval rather than signal seeking), the core components for success remain the same (Fig. 2). There will be an increasing diversity and complexity in how clinical studies are designed, developed and implemented to support precision-medicine-based drug development, and there will be a need to continue to develop and refine the processes involved. In some respects, the natural direction of travel could be a move towards a more community-based drug development paradigm with the successful development of a candidate drug to market approval resulting from an integrated process across a more diverse

Individual protocols Screening protocol

Screening protocol

Protocol 1

Core inc-/exclusion criteria Core trial design

Protocol 2

Sub-protocol 1

Protocol 3

Sub-protocol 2

Protocol 4

Sub-protocol 3

Protocol 5

- Self contained - Complex if many arms - Requires amendment to add/remove arms - Complex legal and contractual aspects if multiple partners - Not easy to adapt - Regulatory aspects need consideration

Modular protocol

- Flexible - Easily adaptable - Regulatory friendly - Easy for multiple partners - Nonintegrated - Resource intensive

Sub-protocol 4 Sub-protocol 5

- Flexible (vs single protocol) - Adaptable - Enables multiple partners (easier vs single protocol) - Regulatory aspects need consideration

Drug Discovery Today

FIGURE 3

Example options for protocol designs including some key design features to consider. Designs will vary considerably and choice to use a particular design will depend on a balance of numerous aspects such as complexity, resources needed, strategy, cost, regulatory environment, partner capabilities, drug distribution needs, among others. www.drugdiscoverytoday.com 5 Please cite this article in press as: Hollingsworth, S.J. Precision medicine in oncology drug development: a pharma perspective, Drug Discov Today (2015), http://dx.doi.org/10.1016/ j.drudis.2015.10.005

Reviews  POST SCREEN

Drug Discovery Today  Volume 00, Number 00  October 2015

DRUDIS-1690; No of Pages 9 REVIEWS

set of partners than might have traditionally been the case. Significant progress has been made, more notably, in the earlier stages of clinical development where perhaps the adoption and implementation of stratified medicine approaches is easiest and more acceptable. With this greater experience, the future challenges are also becoming clearer.

Companion diagnostics Reviews  POST SCREEN

A central component in the clinical success of a precision medicine drug candidate lies in the ability to identify the right patient robustly and consistently (i.e. have an accurate and reproducible means to identify the marker used for patient selection). Whatever the assay, process or diagnostic algorithm used, the results need to be reliable and reproducible despite different operators, laboratories, geographical locations, and so on. Consequently, the drug–diagnostic interaction has taken on increasing importance, with the companion diagnostic (including the end-to-end process of sample acquisition, processing, assay, reporting, etc.) taking on a gatekeeper role to drug use. The intent here is not to offer a comprehensive review of the companion diagnostic codevelopment pathway, technical considerations and challenges – these are comprehensively covered elsewhere [44] – but to comment briefly in the context of, and relevance to, oncology drug development. A number of core challenges are central: (i) whereas technological advances such as next-generation sequencing (NGS) have enabled us to examine tumours in unprecedented detail, the ability to understand which pieces of information generated are meaningful, relevant and could drive a certain clinical intervention is significantly challenging, and the ability to reproduce findings from such exquisitely sensitive assays is notably difficult; (ii) the co-development pathway often requires or prefers employment of a diagnostic that is further advanced in its development process than the drug candidate, requiring significant up-front investment, and before any solid understanding about whether the drug candidate might have clinical activity or that this is the appropriate marker with which to select patients for that drug – so often the development phases of the drug and diagnostic are not parallel and aligned and are usually out-ofsync. This is a common position in the co-development process. The technical requirements for the diagnostic to be accurate, robust, reproducible, etc. sit alongside the need to demonstrate unequivocally that it selects a patient subgroup for a differential clinical benefit – so there is a circular interplay between accurate and reproducible identification and understanding of predictivity and clinical utility. The level of evidence supporting the use of a diagnostic (or marker) to select patients is intrinsically linked to the ability to identify and select those patients, in turn linked to the need to ensure sufficiently robust development of the diagnostic, which is often driven out of the understanding that it is predictive. It is complex. However, potential solutions might arise in and from the use of technologies such as NGS, and through separate modular development of the technical and clinical components required for regulatory approval and market use. For example, the platform technology (technical module piece) could be developed up-front and approved for market use, and subsequently markers added to the menu of the platform when partnered with the clinical trial data demonstrating clinical utility (clinical module piece) – so the process becomes slightly less 6

Drug Discovery Today  Volume 00, Number 00  October 2015

misaligned and relies mainly on development of the clinical trial dataset. This is an approach being taken by key NGS platform providers, is employed in many umbrella and basket study initiatives and is aided by a changing regulatory environment programme [for example the NCI-Molecular Analysis for Therapy Choice (NCI-MATCH) Trial] [43] in solid tumours and lymphomas utilising a nationwide network of CLIA laboratories alongside close consultation with authorities and regulators. Similarly, the CRUK Stratified Medicine Initiative and MATRIX National Lung Trial [37] uses an integrated nationwide health service pathology network using selected central pathology hubs to provide appropriate diagnostic outputs as well as the mechanism for diagnostic development and a route to support transfer to routine clinical use, again in close consultation with government bodies, authorities and regulators. From a clinical trial perspective, because multiplex or high density approaches such as NGS offer a viable means to identify patient subgroups for many [rare(er)] variants via a single diagnostic assay they have seen widespread employment in screening programmes and associated clinical (umbrella, basket, etc.) trials. The technical requirements for assays are widely understood and the precision medicine community as a whole is aware of the need for technically suitable assays (e.g. validated, etc.) to support clinical trials. As a consequence, most programmes, screening initiatives and trials have appropriate measures built in as a core component and companion diagnostic development and refinement as key deliverables. However, although there is the acknowledgement of the need for robust screening processes (sample quality, diagnostics, reporting, etc.), this does not automatically support a companion diagnostic co-development pathway as might be required for regulatory approval – the rarity of some variants does not necessarily support a more traditional codevelopment approach. Potential solutions for many of these challenges could again come from the employment of moremodern precision medicine clinical trial designs, as well as further development of collaborative and co-operative frameworks across business areas. Such solutions, however, are going to need innovation and adjustment to current practices and requirements. For example, data on (rare) variants could be derived from a variety of different sources, initiatives or trials that would help solve the issues around generating a sufficient body of clinical evidence to support a drug–marker paired use. But, this would require cross-verification and/or -validation between screening modalities to ensure consistency in the identification and selection of patients, effectively a cross-comparison of diagnostic technologies and practices, as well as an ability to combine datasets from different studies in a way that was acceptable for decision making and (potentially) regulatory interactions. It is perhaps not in the interests of diagnostics companies to compare their technologies directly, which is understandable, but we must move to a position whereby patients are selected by marker rather than technology system or platform (e.g. differing NGS approaches). This is not simple to resolve, and will require new ways of working between and among the relevant parties (pharma, diagnostic companies and others). The ultimate end users are healthcare systems, physicians and patients – so there needs to be a viable means of developing companion diagnostics that can support an ever increasingly complex formulary but that

www.drugdiscoverytoday.com Please cite this article in press as: Hollingsworth, S.J. Precision medicine in oncology drug development: a pharma perspective, Drug Discov Today (2015), http://dx.doi.org/10.1016/ j.drudis.2015.10.005

DRUDIS-1690; No of Pages 9

can be transferred to routine clinical use in a cost-effective and sustainable way. There is the potential to see a staged continuum whereby drug and diagnostic development starts concurrently via an umbrella or basket trial used as the vehicle for gaining the clinical evidence required, and the underlying screening framework provides the ‘seeder’ to add to the diagnostic menu for later routine clinical deployment. This however would require a more integrated process from start to finish with all parties aligned, collaborative and co-operative. How we might move towards this position is unclear, and not the main focus of this commentary, but there are some considerations that would move the agenda forwards: agreement on diagnostic standards that could be widely employed for benchmarking; broader datasets cross-comparing technologies; greater sharing of samples and datasets; pathology standards and training; and greater public–private sector collaboration. Directly related are the costs for sample acquisition, processing, analysis and management and the logistical requirements for delivering these. The needs are clear: mechanisms to ensure diagnostic robustness, accuracy, standards and quality control (QC); larger sample datasets with cross-platform comparison and verification or validation; and frameworks to enable these pieces to happen, alongside viable and widely deployable process(es) and a viable cost model. Potential solutions could lie in the continued development and employment of cross-sector initiatives, perhaps via pharma and diagnostic companies in conjunction with regional governments, healthcare systems, hospitals and hospital networks and/or geographical networks, or the emergence of newer models for commercial vendors to offer integrated and systemised approaches; although commercial approaches can have ethical challenges to overcome which again would require new models of collaboration or co-operation. Many current initiatives (Table 1) see national and international bodies and charities front-ending diagnostic development and screening costs to help attract pharma and drug development sectors, or the use of framework or participation fees to help offset these costs; leading to cost-sharing approaches to balance investment across the parties. This is a viable route forwards but would also benefit further from cross-initiative sharing of data (and perhaps resources) which would help develop viable routes to use composite datasets to support diagnostic development and companion diagnostic regulatory filings. Different engagement from regulatory authorities on such datasets would also be significantly enabling. The landscape is evolving but ultimately a viable healthcare system model will require evolution of models around cost-sharing and process if they are to be cost-effective and deliverable. Should sample and diagnostic analysis and development be borne solely by the investigating pharma or drug development company? A better solution would be partnership with the early drug development process providing a seeder for ultimate deployment in the healthcare system – and perhaps even greater cross-pharma and drug development company partnership to maximise opportunities for operational, logistical and cost efficiencies.

Concluding remarks The challenges ahead The successful leverage of precision medicine (i.e. improvements in patient survival) will require integration across all involved

REVIEWS

parties and processes: drug development companies, diagnostic companies, cancer-specific groups and consortia, advocacy groups, regulators, policy and policy makers, governments, reimbursement mechanisms, clinical professionals (i.e. physicians, pathologists, nurses, etc.) and the plethora of other entities, companies and bodies central to the development and implementation of clinical care. We also need healthcare systems that are geared to cope with the complexity that precision medicine will bring, and the systems built ready to enable delivery. This is likely to require change in almost every aspect of current practice. For drug development companies, the requirements at their simplest level are clear: (i) an easily deployable mechanism that allows you to unearth clinical signals quickly or fail early and easily and in a cost-contained manner; (ii) a viable route forwards through later clinical development; (iii) a mechanism for market approval of such agents; and (iv) markets and healthcare systems able to use the medicine effectively once ready. Regarding points (i) and (ii), the emergence of innovative clinical trial designs such as umbrella and basket studies, accompanied by the technological advances in diagnostics and notable increase in public–private partnerships, cross-industry, government and non-profit sector collaborations, are providing viable solutions and an emerging landscape tailored to the needs in development of precision medicine products. However, although the mechanisms for running trials are developing, there is a parallel need for greater understanding in the use of such datasets in highly selected patient subgroups. For example, key questions arise about the effects of the drug candidate on biomarker (diagnostic)-negative patients and the need to show that differential clinical benefit is caused by patient selection and the ability to generate a reasonable understanding of drug safety given the low number of patients exposed. Up-scaling trials using the same umbrella or basket framework for screening and trial conduct could be used to generate larger datasets, perhaps in a staged or gated manner following discovery of an early clinical signal. However to address concerns over biomarker- or diagnostic-negative patients a suitable negative comparator group would need to be included despite the counter argument that you might be exposing patients to a therapy without a clear and justifiable mechanistic rationale – this is a difficult challenge. Alongside this, we would need a better understanding on what would constitute a reasonable number of patients exposed, in order to understand safety concerns. A greater safety understanding will inevitably require more patients. Further developments on how best to understand and interrogate the reproducibility, informitivity and predictivity of single-arm study data will be significantly enabling. Regarding point (iii), although the processes for seeking market approval for a candidate drug are established and clear, the current requirements and guidelines underpinning these are more focused around a standard development pathway and less adapted to the needs of precision medicine. Consider the scenario: if we find a clinical signal generated in an early phase signal seeking study (e.g. an umbrella or basket study) how do we move this forwards to a trial supportive of market authorisation? Few examples exist [outside XALKORI1 (crizotinib; Pfizer)] [45,46], and there are significant challenges around a number of core components including:

www.drugdiscoverytoday.com 7 Please cite this article in press as: Hollingsworth, S.J. Precision medicine in oncology drug development: a pharma perspective, Drug Discov Today (2015), http://dx.doi.org/10.1016/ j.drudis.2015.10.005

Reviews  POST SCREEN

Drug Discovery Today  Volume 00, Number 00  October 2015

DRUDIS-1690; No of Pages 9 REVIEWS

 lower incidence segments cannot be evaluated using tradi-

Reviews  POST SCREEN

tional trial designs and data packages, so how can we bring together a number of data packages from several similar studies for consideration;  prospective patient selection in multiple studies can come from differing diagnostics (and platforms) and there is an inability to bridge or cross-validate the companion diagnostic;  different studies (or tumour types or indications) use different or composite endpoints, so how do we use these in understanding if trials are successful? How can we gain regulatory approval, etc.? Furthermore, whereas the main experience to date in this area has been in the development of monotherapy treatments, it is widely suggested that the future of oncology therapy will be based in treating patient segments with combinations of agents, and/or agents used in close sequence to one another – collectively adding yet further and significant complexity. Current guidelines do not offer sufficient insight into how these challenges should be tackled and how best to work with regulators in defining studies and the data packages needed for market approval. The current reliance on ‘in-flight’ scientific advice and/or guidance from the regulatory bodies carries a perceivable risk in the development process, and parties will need a more preferable solution of open and regular dialogue to ensure that together we do what is needed to make precision medicine routinely available. This is, however, recognised by the regulatory authorities and more-recent developments, such as the FDA Breakthrough Therapy designation and Guidance for Industry on Expedited Programmes for Serious Conditions – Drugs and Biologics [47], are an important step forwards and help to provide a vital framework to support precision medicine development. Recent cases have seen ZykadiaTM (ceritinib; Novartis) receive marketing approval based on an extended Phase I study of only 163 patients [48], and in Japan Alecensa1 (alectinib; Roche/Genentech) received marketing approval based on data from fewer than 100 patients [49]. Therefore, the landscape is evolving. Regarding point (iv), by far the biggest challenge, although outside the scope of this review, this will require engagement and change from regulators, in policy and from policy makers, from governments, within reimbursement mechanisms, in education and adoption in clinical professionals and many more within development and implementation of clinical care. We will need healthcare systems that can deliver an increasingly diverse formulary based on an increasingly complex diagnostic pathway. This will touch on all aspects from the basics of the patient record upwards. Although not the subject of this commentary, it is perhaps necessary to say that central to the forward development of precision medicine is a new generation of highly selective, molecularly targeted agents. This area, although benefiting most obviously from a greater and deeper understanding of biology and disease pathophysiology, and advances in technology and clinical capabilities, will probably follow the clinical phase development process and emerge in a new way of doing things. A more ‘hybridised’ world between pharmaceutical companies, academia, small – medium enterprises (SMEs), etc. is developing, as is the willingness or desire to work collectively and collaboratively to find

8

Drug Discovery Today  Volume 00, Number 00  October 2015

solutions to treating cancer, particularly in the ever increasing complexity of precision medicine.

Changing the landscape Development of policy, process, education, new systems and understanding of pharmaco-economics, value and savings in lifetime patient costs will of course enable the further development and use of precision medicine as an emerging mainstay of modern oncology clinical practice. However, the world today is increasingly driven from a consumer standpoint and the historical deferment of the patient to the consulting clinical opinion is also changing. Direct to consumer molecular profiling of tumours (for example via companies such as 23andMe, deCODEme, Navigenics), for a review see [50], the breadth of information available from electronic resources and the emergence of patient-led consortia [51] will undoubtedly help shape the interaction and environment. From deep profiling of their tumours, patients can already take the discussion to their treating physician regarding which clinical studies are likely to give them the best possible chances of survival [52]. The ability of drug development companies to make the information on current trials more easily available is also a significant advantage for all parties. Perhaps an inevitable step forwards from this will be the emergence of patient consortia making their tumour profile data directly available to drug development companies and so, in effect, brining about a new paradigm of how patients might get directed to suitable clinical trials. Our scientific understanding increases inexorably and we are perhaps entering an era where each individual tumour is assessed in great depth for its biology at each clinical intervention and throughout treatment, with therapy offered according to the specific biology rather than by line of treatment or established standard of care. The nascent understanding that tumour course might be directly affected by treatment modality, and that different treatments might elicit different pathways or developments, furthers the notion that future success in treating tumours will be better derived from a different approach – truly individual- and tumour-specific precision medicine.

The journey has begun Precision medicine has evolved rapidly in recent years to reality, as has our understanding of how to implement this in the clinic – more than ever we can look to a time when this is a clinical reality for many more patients with cancer. The development of the frameworks and processes involved in precision medicine is also notably enabling for and aligned to modern oncology drug development. There is an obvious and reciprocal relationship that can help drive the changes needed in all areas and for mutual gain. There are of course still significant hurdles to overcome, particularly for example in how we can use the datasets generated by umbrella and basket studies to lead to market approval of new precision medicine products, and tackling these challenges will require working in collaboration and across many boundaries. However, in an evolving era of precision medicine, we have now made significant steps forwards in generating truly patientcentric clinical trial designs within the drug development process, and can really say we are now able to select the trial for the patient rather than the patient for the trial.

www.drugdiscoverytoday.com Please cite this article in press as: Hollingsworth, S.J. Precision medicine in oncology drug development: a pharma perspective, Drug Discov Today (2015), http://dx.doi.org/10.1016/ j.drudis.2015.10.005

DRUDIS-1690; No of Pages 9 Drug Discovery Today  Volume 00, Number 00  October 2015

REVIEWS

1 Hanahan, D. and Weinberg, R.A. (2011) Hallmarks of cancer: the next generation. Cell 144, 646–674 2 The Cancer Genome Atlas Research Network (2014) Comprehensive molecular profiling of lung adenocarninoma. Nature 511, 543–550 3 Cancer Genome Atlas Research Network (2012) Comprehensive genomic characterization of squamous cell lung cancers. Nature 489, 519–925 4 Cancer Genome Atlas Research Network (2014) Comprehensive molecular characterization of gastric adenocarcinoma. Nature 513, 202–209 5 Cancer Genome Atlas Network (2012) Comprehensive molecular characterization of human colon and rectal cancer. Nature 487, 330–337 6 Cancer Genome Atlas Network (2012) Comprehensive molecular portraits of human breast tumours. Nature 490, 61–70 7 Maemondo, M. et al. (2010) Gefitinib or chemotherapy for non-small-cell lung cancer with mutated EGFR. N. Engl. J. Med. 362, 2380–2388 8 Ja¨nne, P.A. et al. (2015) AZD9291 in EGFR inhibitor-resistant non-small-cell lung cancer. N. Engl. J. Med. 372, 1689–1699 9 Ledermann, J. et al. (2014) Olaparib maintenance therapy in patients with platinum-sensitive relapsed serous ovarian cancer: a preplanned retrospective analysis of outcomes by BRCA status in a randomised phase 2 trial. Lancet 15, 852–861 10 Shaw, A.T. et al. (2013) Crizotinib versus chemotherapy in advanced ALK-positive lung cancer. N. Engl. J. Med. 368, 2385–2394 11 Sosman, J.A. et al. (2012) Survival in BRAF V600-mutant advanced melanoma treated with vemurafenib. N. Engl. J. Med. 366, 707–714 12 Slamon, D. et al. (2011) Adjuvant trastuzumab in HER2-positive breast cancer. N. Engl. J. Med. 365, 1273–1283 13 Demetri, G.D. et al. (2002) Efficacy and safety of imatinib mesylate in advanced gastrointestinal stromal tumors. N. Engl. J. Med. 347, 472–480 14 Kris, M.G. et al. (2014) Using multiplexed assays of oncogenic drivers in lung cancers to select targeted drugs. JAMA 311, 1998–2006 15 Lynch, T.J. et al. (2004) Activating mutations in the epidermal growth factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib. N. Engl. J. Med. 350, 2129–2139 16 Paez, J.G. et al. (2004) EGFR mutations in lung cancer: correlation with clinical response to gefitinib therapy. Science 304, 1497–1500 17 Mok, T.S. et al. (2009) Gefitinib or carboplatin-paclitaxel in pulmonary adenocarcinoma. N. Engl. J. Med. 361, 947–957 18 Fukuoka, M. et al. (2011) Biomarker analyses and final overall survival results from a Phase III, randomized, open-label, first-line study of gefitinib versus carboplatin/ paclitaxel in clinically selected patients with advanced non-small-cell lung cancer in Asia (IPASS). J. Clin. Oncol. 29, 2866–2874 19 Kim, E.S. et al. (2008) Gefitinib versus docetaxel in previously treated non-small-cell lung cancer (INTEREST): a randomised Phase III trial. Lancet 372, 1809–1818 20 Douillard, J-Y. et al. (2010) Molecular predictors of outcome with gefitinib and docetaxel in previously treated non-small-cell lung cancer: data from the randomized Phase III INTEREST trial. J. Clin. Oncol. 28, 744–752 21 Soda, M. et al. (2007) Identification of the transforming EML4–ALK fusion gene in non-small-cell lung cancer. Nature 448, 561–566 22 Bang, Y. et al. (2010) Clinical activity of the oral ALK inhibitor PF-02341066 in ALKpositive patients with non-small cell lung cancer (NSCLC). J. Clin. Oncol. 28 (Suppl. 18), 3 23 Kwak, E.L. et al. (2010) Anaplastic lymphoma kinase inhibition in non-small-cell lung cancer. N. Engl. J. Med. 363, 1693–1703 24 Shaw, A.T. et al. (2012) Phase III study of crizotinib versus pemetrexed or docetaxel chemotherapy in patients with advanced ALK-positive non-small cell lung cancer (NSCLC). Eur. Soc. Med. Oncol. Meet Abstr. LBA1 PR 25 Vysis ALK BAP FISH Probe Kit instructions. Available at: http://www.accessdata.fda. gov/cdrh_docs/pdf11/P110012c.pdf 26 FDA Drug Approvals And Databases. Available at: http://www.fda.gov/Drugs/ InformationOnDrugs/ 27 FDA List of Cleared or Approved Companion Diagnostic Devices (In Vitro and Imaging Tools). Available at: http://www.fda.gov/MedicalDevices/ ProductsandMedicalProcedures/InVitroDiagnostics/ucm301431.htm

28 FDA Nucleic Acid Based Tests. Available at: http://www.fda.gov/MedicalDevices/ ProductsandMedicalProcedures/InVitroDiagnostics/ucm330711.htm 29 European Medicines Agency (EMA) approvals. Available at: http://www.ema. europa.eu/ema/index.jsp?curl=pages/regulation/landing/ human_medicines_regulatory.jsp&mid 30 Report The Academy of Medical Sciences. Available at: https://www.acmedsci.ac. uk/viewFile/51e915f9f09fb.pdf 31 Report Association of the British Pharmaceutical Industry (ABPI). Available at: http://www.abpi.org.uk/our-work/library/medical-disease/Documents/strat_med. pdf 32 Kim, E.S. et al. (2011) The BATTLE trial: personalizing therapy for lung cancer. Cancer Discov. 1, 44–53 33 Esserman, L.J. et al. (2012) Chemotherapy response and recurrence-free survival in neoadjuvant breast cancer depends on biomarker profiles: results from the I-SPY 1 TRIAL (CALGB 150007/150012; ACRIN 6657). Breast Cancer Res. Treat. 132, 1049– 1062 34 Esserman, L.J. et al. (2012) Pathologic complete response predicts recurrence-free survival more effectively by cancer subset: results from the I-SPY 1 TRIAL – CALGB 150007/150012, ACRIN 6657. J. Clin. Oncol. 30, 3242–3249 35 Hylton, N.M. et al. (2012) Locally advanced breast cancer: MR imaging for prediction of response to neoadjuvant chemotherapy – results from ACRIN 6657/ISPY TRIAL. Radiology 263, 663–672 36 Lin, C. et al. (2012) Locally advanced breast cancers are more likely to present as interval cancers: results from the I-SPY 1 TRIAL (CALGB 150007/150012, ACRIN 6657, InterSPORE Trial). Breast Cancer Res. Treat. 132, 871–879 37 MATRIX National Lung Cancer Trial, Cancer Research, UK. Available at: http:// scienceblog.cancerresearchuk.org/2014/04/17/stratified-medicine-and-the-lungcancer-matrix-trial-part-of-a-cancer-care-revolution/ 38 Lung-MAP, Lung Master Protocol, Friends of Cancer Research, US. Available at: http://www.focr.org/lung-map 39 Zardavas, D. et al. (2014) The AURORA initiative for metastatic breast cancer. Br. J. Cancer 111, 1881–1887 40 Japan National Cancer Centre. Available at: http://www.ncc.go.jp/en/index.html 41 EORTC, Colorectal Cancer Screening Platform, SPECTAColor. Available at: http:// spectacolor.eortc.org/ 42 EORTC, SPECTALung Available at: http://www.eortc.org/news/eortc-throughspectalung-participates-in-european-consortium-validating-blood-basedcancer-biomarkers/ 43 National Cancer Institute (NCI), US, NCI-MATCH (Molecular Analysis for Therapy Choice). Available at: http://www.cancer.gov/researchandfunding/areas/ clinical-trials/nctn/match 44 Fridlyand, J. et al. (2013) Considerations for the successful co-development of targeted cancer therapies and companion diagnostics. Nat. Rev. Drug. Discov. 12, 743–755 45 Shaw, A.T. The Crizotinib Story: From Target to FDA Approval and Beyond. InforMEDical report. Available at: http://www.informedicalcme.com/lucatoday/ crizotinib-story-from-target-to-fda-approval/ 46 Sahu, A. et al. (2013) Crizotinib: a comprehensive review. South Asian J. Cancer 2, 91–97 47 United States, Food and Drug Administration Safety and Innovation Act (FDASIA) Section 902; Breakthrough Therapy Designation. Available at: http://www.fda.gov/ RegulatoryInformation/Legislation/SignificantAmendmentstotheFDCAct/FDASIA/ ucm329491.htm. 48 Shaw, A.T. et al. (2014) Ceritinib in ALK-rearranged non-small-cell lung cancer. N. Engl. J. Med. 370, 1189–1197 49 Seta, T. et al. (2013) CH5424802 (RO5424802) for patients with ALK-rearranged advanced non-small-cell lung cancer (AF-001JP study): a single-arm, open-label, phase 1–2 study. Lancet Oncol. 14, 590–598 50 Roberts, J.S. and Ostergren, J. (2013) Direct-to-consumer genetic testing and personal genomics services: a review of recent empirical studies. Curr. Genet. Med. Rep. 1, 182–200 51 Cure Forward Available at: http://cureforward.com/ 52 Foundation Medicine, Connect – Patients and Caregivers. Available at: http://www. mycancerisunique.com/

www.drugdiscoverytoday.com 9 Please cite this article in press as: Hollingsworth, S.J. Precision medicine in oncology drug development: a pharma perspective, Drug Discov Today (2015), http://dx.doi.org/10.1016/ j.drudis.2015.10.005

Reviews  POST SCREEN

References