REVIEW ARTICLE New clinical trial designs in the era of precision medicine nez1, Irene Bran Elena Garralda1,2, Rodrigo Dienstmann1,3, Alejandro Piris-Gime ~ a1,2, Jordi Rodon1,4 1 and Josep Tabernero noma de Barcelona, 1 Medical Oncology Department, Vall d’Hebron University Hospital and Institute of Oncology (VHIO), Universitat Auto Spain 2 Early Drug Development Unit (UITM), Vall d’Hebron Institute of Oncology (VHIO), Barcelona, Spain 3 Oncology Data Science (ODysSey) Group, Vall d’Hebron Institute of Oncology (VHIO), Barcelona, Spain 4 MD Anderson Cancer Center, Houston, USA Keywords Cancer treatment has made significant strides towards the promise of per- adaptive designs; Basket of Baskets; basket sonalized medicine. Recent scientific advances have shown that there are protocols; Cancer Core Europe; umbrella numerous genetic deregulations that are common in multiple cancer types, protocols raising the possibility of developing drugs targeting those deregulations Correspondence irrespective of the tumour type. Precision Cancer Medicine (PCM) was J. Tabernero, Medical Oncology born out of accumulated evidence matching targeted agents with these Department, Vall d’Hebron University tumour molecular deregulations. At the same time, the therapeutic arma- Hospital, Vall d’Hebron Institute of Oncology mentarium is rapidly increasing and the number of new drugs (including (VHIO), P. Vall d’Hebron 119-129, 08035 immune-oncology agents) entering drug development continues to rise. Barcelona, Spain These factors, added to strong collaboration with regulatory agencies, Tel: +34 93 489 4301 which have approved novel agents based on data obtained from phase 1/2 E-mail: jtabernero@vhio.net trials, have led to unprecedented evolution in the design of early-stage clini- (Received 29 December 2018, accepted 6 cal trials. Currently, we have seen rapid phase 1 dose-escalation trials fol- January 2019, available online 22 February lowed by remarkably large expansion cohorts, and are witnessing the 2019) emergence of new trials, such as adaptive studies with basket and umbrella designs aimed at optimizing the biomarker–drug co-development process. doi:10.1002/1878-0261.12465 Alongside the growing complexity of these clinical trials, new frameworks for stronger and faster collaboration between all stakeholders in drug development, including academic institutions and frameworks, clinicians, pharma companies and regulatory agencies, have been established. In this review article, we describe the main challenges and opportunities that these new trial designs may provide for a more efficient drug development pro- cess, which may ultimately help ensure that PCM becomes a reality for patients. cancer types. These discoveries have led to the devel- 1. Introduction opment of drugs targeting driver gene alterations irre- Over recent years, much progress has been made in spective of their primary tumour location. We are rendering cancer treatments more precise. Scientific consequently witnessing the ringing in of a new era and technological advances have unmasked numerous characterized by considering tumours as genetic dis- genetic deregulations that are common across multiple eases as opposed to tissue-dependent processes. This is Abbreviations BOB, Basket of Baskets; CCE, Cancer Core Europe; ctDNA, circulating tumour DNA; dMMR, defective mismatch repair; FDA, Food and Drug Administration; MSI, microsatellite instability; NCI, National Cancer Institute; NSCLC, non-small-cell lung cancer; NTRK, neurotrophin tropomyosin receptor kinase; PCM, Precision Cancer Medicine. Molecular Oncology 13 (2019) 549–557 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 549 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Adaptive trial designs – collaborative frameworks E. Garralda et al. accelerating the pace from molecular aberration generated from large-scale collaborative sequencing discovery to the approval of new therapies (Dienstmann projects, such as the Cancer Genome Atlas and the et al., 2013a,b; Hierro et al., 2017; Hunter, 2016). International Cancer Genome Consortium, in addition At the same time, during these last years the arrival to publicly available resources such as the cBioPortal of immune therapy to the oncology therapeutic arma- for Cancer Genomics and the Catalogue of Somatic mentarium has marked a groundbreaking milestone Mutations in Cancer, has facilitated our understanding for the treatment of cancer patients, and the number of the genetic interpatient tumour heterogeneity in of immune-oncology agents entering drug development multiple cancers subtypes (Dienstmann et al., 2013a). continues to rise (Martin-Liberal et al., 2017). These Most druggable genomic aberrations are present only factors, added to the strong collaboration with the reg- in small-to-moderate proportions of patients, further ulatory agencies, approving novel agents based on emphasizing multicentre collaboration in early drug data obtained from phase 1/2 trials, have led to an development as critical for successful clinical trial unprecedented evolution in the design of early-stage enrolment (Dienstmann et al., 2015). clinical trials (Bui and Kummar, 2018). In this regard, In addition, recent studies have also described strik- the US Food and Drug Administration (FDA) has ing intrapatient intratumour heterogeneity and how approved 10 anticancer drugs matched to companion clonal evolution under treatment pressure may repre- diagnostic biomarkers based on data obtained from sent major obstacles in PCM, questioning the value of nonrandomized trials in the last 2 years. This evolu- a single-needle biopsy or surgical excision to accurately tion is mainly driven by the desire to facilitate capture the complete genomic landscape of a patient’s patients’ access to drugs with promising activity from cancer (Bedard et al., 2013; Hunter, 2016). Neverthe- the early stages of development and is also a conse- less, we believe that the described heterogeneity in quence of pharmaceutical companies striving to obtain genomic profiles particularly applies to bystander rapid regulatory approval of their anticancer medi- mutations and that true tumour-driving events are cines. usually present in the majority of subclones from the The traditional drug development track, where drugs primary tumour as well as the metastatic lesions (Yap were evaluated for safety in phase 1, early signs of effi- et al., 2012). Therefore, regimens targeting driver cacy in phase 2 and finally evaluated against standard genomic alterations with high variant frequencies are therapy in a randomized phase 3 clinical trial, has expected to provide substantial tumour responses. gradually faded out. Currently, we are facing rapid Since clinical responses to targeted agents are consis- phase 1 dose-escalation trials followed by strikingly tently abrogated by the development of drug resis- large expansion cohorts and the emergence of new tri- tance, we consider repeated tumour biopsies of als such as adaptive studies with basket and umbrella progressing lesions and/or characterization of circulat- designs aimed at optimizing the biomarker–drug co- ing markers (tumour cells, tumour DNA) to be a key development process. In parallel, with the growing component of patients’ care, allowing the identification complexity of clinical trials, new frameworks for stron- of mechanisms of resistance as well as potentially guid- ger and faster collaboration between all stakeholders ing alternative treatment options with experimental in drug development, including clinicians, pharma agents. companies and regulatory agencies, have been estab- In terms of clinical trials incorporating biomarkers, lished (Harrington et al., 2017; Weber et al., 2017). an essential element that should be factored in is the turnaround time for test results, particularly with tumour clinical next-generation sequencing, for 2. Precision medicine trials for cancer patients undergoing molecular profiling. This is espe- With the development and clinical use of molecularly cially important in the metastatic setting when treat- targeted agents, it became clear that only selected ment decisions have to be made within a short time patient populations would derive benefit from these frame. therapies. Precision Cancer Medicine (PCM) was born As an alternative to the traditional approach of cen- out of the accumulated evidence on matching targeted tralized biomarker analysis prior to evaluating the agents with tumour molecular aberrations (Dienst- inclusion of a patient in a trial, multiple academic cen- mann et al., 2015; Hoelder et al., 2012). Those drugs tres have adopted a different strategy consisting of designed to interact with a specific target, and espe- local prescreening at academic institutions while cially those using predictive biomarkers, showed the patients are still receiving standard treatment for highest relative improvement in response rate and sur- advanced disease. While this approach is time and tis- vival (Ocana et al., 2013). Current knowledge sue saving and increases the chances of patient 550 Molecular Oncology 13 (2019) 549–557 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. Garralda et al. Adaptive trial designs – collaborative frameworks recruitment in early clinical trials, the financial burden achievable or unlikely to be achieved with a statistical of prescreening tests is transferred from trial sponsors significance). On the other hand, the sponsor may have to healthcare providers and academic institutions the option of responding to intermediate data on (Dienstmann et al., 2015; Rod on et al., 2012). safety and efficacy in various ways, such as narrowing In order to efficiently and dynamically incorporate the trial approach (e.g. elimination of one or more genomic data and assess the value of matching profiled treatment groups based on futility criteria) or increas- patients whose tumours harbour unique genomic alter- ing the number of participants (e.g. If the data avail- ations to specific interventions or targeted therapies, able at the time of the review do not allow for a clear clinical trials design for cancer diagnostics and thera- decision between utility and futility, the enrolment of peutics must take these rate-limiting steps into consid- participants might be extended to one or more treat- eration. ment groups beyond the initially intended sample). Cost reduction is achieved through the early identifica- tion of successful groups, leaving unnecessary treat- 2.1. Adaptive studies ment groups out of the equation or determining A clinical study with adaptive design is defined as one effective dosing regimens more quickly (Menis et al., that includes planned opportunities to modify one or 2014; Korn and Freidlin, 2017) more specified elements of the study design and Well-known examples of adaptive measures in clini- hypothesis based on data analysis (usually intermediate cal trials include early stopping rules in the instances findings) of the study subjects. Research into the accu- of a lack of efficacy or unacceptable toxicity and alter- mulated data is carried out within the study at specific, ing doses or schedules of drugs in order to improve prospectively planned time-points, and can be per- the benefit–toxicity profile. More recently, novel adap- formed in a completely blind or a nonblind way. The tation strategies have been proposed. In the adaptive term prospective refers to the fact that the change was accrual design, after the initial ‘learning phase’, the preplanned (and the details specified) before the data ratio of patients randomly assigned to the experimen- were examined in a nonblinded manner (CHMP/EWP/ tal arm vs the control arm changes from the standard 2459/02, 2007). The final objective of adaptive designs 1 : 1 to increase the proportion of patients randomized is to learn from the accumulated data and apply what to the arm that is performing better, which augments has been observed as soon as possible. The modifica- the statistical power to detect a relevant magnitude of tions to the study design that can be planned in the clinical benefit. written protocol cover a broad range of possibilities. The Biomarker-integrated Approaches of Targeted Examples include study eligibility criteria, proportion Therapy for Lung Cancer Elimination 2 (BATTLE-2 of randomization, addition of new treatments and study), for example, is a biomarker-based and biopsy- sample size of the study (Gallo et al., 2006). mandatory prospective trial to guide treatment of Adaptive design trials have been shown to increase heavily pretreated metastatic non-small-cell lung can- the efficiency of traditional clinical trials by facilitating cer (NSCLC) patients (NCT01248247). In the ‘adap- the selection of the dose, reducing the number of tive phase’, randomization to different drugs or patients exposed to ineffective or potentially toxic combinations is weighted based on mutation profile doses, aiding the precise calculation of sample size and results generated in real time. A similar framework reducing the duration and costs of clinical develop- can be applied to studies assessing the predictive value ment. These new designs in early drug development of gene expression signatures. Instead of using a fixed enable the integration of preclinical data, the incorpo- model – built on the training data only – adaptive ration of information beyond the traditional dose-lim- strategies use the information on patients enrolled ear- iting toxicity period, findings from other trials and lier in the testing set to continuously update the model emerging safety data, thereby increasing the likelihood and refine accrual throughout the entire study (Dienst- of accurately determining any benefit of a new treat- mann et al., 2015; Xiao et al., 2014). ment and complying more quickly with regulatory Adaptive models increase the weights of good requirements for efficacy and safety (Harrington et al., predictors and decrease the weights of unstable predic- 2017). tors, improving the overall performance of the classi- In this regard, the formal reasons for stopping the fier and selecting the ‘best’-matched therapy to current study in an intermediate analysis could be: (a) safety patients’ characteristics. These algorithms may facili- (if one of the interventions involves many adverse tate the use of molecular signatures to predict the events); (b) efficacy (if it shows efficacy of one of the clinical outcomes of patients in prospective clinical interventions); or (c) futility (if the objectives are not studies. Molecular Oncology 13 (2019) 549–557 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 551 Adaptive trial designs – collaborative frameworks E. Garralda et al. In the realm of precision medicine, enrichment trials the FOCUS-4 (Adams et al., 2018) and MoTriColor with adaptive designs where predictive algorithms (2015) in colorectal cancer (H2020 grant agreement incorporating prior knowledge (based on in silico mod- no. 635342). Some studies tested or are testing this els of drug sensitivity, ex vivo experiments, preclinical framework across multiple solid tumours, such as or early clinical data) are used to guide the best- MOSCATO (Massard et al., 2017) or National Cancer matched targeted therapy in particular settings have Institute (NCI)-MATCH trials (Mullard, 2015). been envisioned. This may be useful when multiple In some studies, such as MOSCATO, patients had druggable alterations are identified in a patient’s limited access to a set of matched therapies, while tumour sample and more than one agent is available others including I-SPY and NCI-MATCH have over- for testing; and when one driver genomic event is iden- come this limitation by building networks and efficient tified, and the investigator has to select among various partnerships with the pharmaceutical industry. Despite drugs with overlapping mechanisms of action (target- strong collaborations, some studies have been limited ing the same driver event) but with different potency/ by the suboptimal biomarkers and matched drugs used activity according to coexisting genomic alterations. including the BATTLE or the SHIVA trials (Adams These ‘machine-learning predictive models’ can com- et al., 2018; Le Tourneau et al., 2015). To deal with plement molecular tumour boards efforts to identify this limitation, the I-SPY1/2 studies and the NCI- the ‘best guess’ (Pemovska et al., 2013). MATCH have successfully implemented adaptive Both the EMA (2007) and the FDA have already designs allowing the addition of new arms as new recognized the validity of clinical trials with adaptive knowledge became publicly available. Most studies characteristics as a viable alternative strategy for both have remained fixed to their initial treatment algo- pivotal and early trials in the regulatory environment rithms, sometimes outdated at the time of study clo- of pharmacological development (9). However, regula- sure. Also, there was no flexibility to integrate new tory agencies are still reluctant in some cases to con- technologies that might be of interest, RNA-based sider adaptive designs, as the results can be more multiplexed assays for the detection of fusion events. difficult to interpret. One of the main concerns is the The I-SPY1/2 studies and the NCI-MATCH have control of the type I error rate as well as the fact that successfully implemented flexible designs allowing for adaptive measures may introduce bias (Bauer et al., the addition of new arms as new data become publicly 2016; Menis et al., 2014). Another important challenge available, while others have remained fixed to their ini- of this type of studies is providing the information to tial treatment algorithms that could be outdated by the patient in a sufficiently precise but at the same the time they close. In most of these studies, tumour time comprehensible way. These studies have complex molecular characterization is based on DNA analysis designs, with several cohorts and one or several drugs platforms, with little flexibility to integrate other tech- under investigation. Above all, they can vary over nologies that might be of interest, such as RNA-based time, which increases the uncertainty of the trial design multiplexed assays for the detection of fusion events and makes it very difficult to explain to the patient. and gene signatures that define unique portraits of To address this challenge, the trials will have different tumours. An example of the latest approach is the informed consents depending on the specific cohort MoTriColor EU H2020-funded project, a set of molec- (Korn and Freidlin, 2017). ularly guided trials with specific treatment strategies in patients with advanced newly molecular defined sub- types of colorectal (gene signature-based) cancer 2.2. Umbrella protocols (H2020 grant agreement no. 635342). An umbrella trial is a master protocol for which the patient’s eligibility is defined by the presence of a 2.3. Basket trials tumour type that is substratified according to specific molecular alterations matched to different anticancer Basket trials include patients with different tumour therapies (Woodcock and LaVange, 2017). types with a common molecular alteration who are Several Umbrella Protocols, in which patients were treated with the same matched therapy (Carey and stratified by potential molecular biomarker and Winer, 2016; Redig and J€anne, 2015). They constitute assigned to matched therapies, were initiated to evalu- a histology-agnostic approach to evaluate targeted ate the role of Precision Medicine in certain tumour agents in molecularly selected populations and can types, such as the I-SPY1/2 (Carey and Winer, 2016; provide access to experimental therapies for patients Das and Lo, 2017) in breast cancer, the BATTLE 1/2 across a wide range of tumour types, potentially (Papadimitrakopoulou et al., 2016) in lung cancer, or including those rare tumours that would have not been 552 Molecular Oncology 13 (2019) 549–557 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. Garralda et al. Adaptive trial designs – collaborative frameworks studied in other clinical trials (Billingham et al., 2016). comparable across tumours, with 36% in colorectal The first basket study design evaluated the efficacy of cancer vs 46% in 14 other cancer types. vemurafenib in solid tumours or haematological malig- Following the approval of pembrolizumab, another nancies harbouring BRAFV600 mutations very recent milestone has revolutionized the drug (NCT0152497). This design evidenced the activity of development field, with the approval of larotrectinib vemurafenib in BRAF V600E mutant NSCLC and on November 2018 for the treatment of adult and pae- also showed activity in other tumour types such as diatric neurotrophin tropomyosin receptor kinase ovarian cancer and certain central nervous system can- (NTRK) rearranged tumours. Patients with NTRK cers (Hyman et al., 2015). fusion-positive tumours were enrolled into the adult Limitations of this approach include the following: phase 1, the paediatric SCOUT phase 1/2, or the (a) the assumption that the same mutation might have NAVIGATE ‘basket’ phase 2 trials evaluating the same impact regardless of histology; (b) the pres- larotrectinib, a first-in-class pan-TRK inhibitor (Drilon ence of variants of unknown significance whose func- et al., 2018). The expanded 122-patient integrated tion has not previously been evaluated; (c) an inherent dataset showed an overall response rate of 81% (95% focus on one single alteration (when nowadays multi- CI, 72–88). Remarkably, 84% of the responding marker analysis is standard); and (d) without a control patients and 73% of all patients remained on larotrec- arm, it can be difficult to differentiate predictive from tinib or underwent surgery with curative intent. Strik- prognostic value of a biomarker. One important caveat ingly, larotrectinib demonstrated clinical benefit is the possibility of an insufficient representation of regardless of tumour type, NTRK gene, fusion partner patients with certain tumour types that harbour the or age of the patient (Tan et al., 2018). Both of these alteration of interest, leading to false-negative conclu- approvals inaugurate a new area, with multiple new sions. Therefore, basket trials should be stratified by agents, such as RET inhibitors, following this new histology, taking into consideration the reported fre- path (Garber, 2018; Subbiah et al., 2018). quencies of the genomic event (Bedard et al., 2013). This strategy may be even adapted to increase enrol- 3. The Basket of Baskets study: an ment of patients with tumour types that demonstrate example of EU collaborative basket early signals of antitumour activity while excluding framework with multimodular design those lacking preliminary response. Furthermore, addi- tional cohorts with different tumour types can be cre- The Basket of Baskets (BoB) study is the spearhead ated, and patients with related molecular aberrations Program of the Cancer Core Europe (CCE) (Egger- can also be enrolled – those with newly identified mont et al., 2014). Its overall goal is to evaluate the fusion genes known to activate the pathway and sensi- antitumour activity of matched therapies in small CCE tize to the agent under investigation similarly to gene patient populations molecularly selected using a novel copy number alterations or mutations, for example study design in an international multicentre (basket) (Dienstmann et al., 2015). Moreover, if other driver approach. molecular alterations coexist, patients could be offered The study consists of two parts: (a) I-Profiler will combination regimens targeting more than one single allow the molecular characterization of tumours from alteration. patients with metastatic or recurrent solid tumours The rarity of certain tumour–biomarker combina- using a new profiling tool and select the most suitable tions makes it impossible to conduct randomized trials. treatment for these patients; and (b) I-Basket is a mul- As a consequence, basket trials have increasingly been timodular basket trial, with different cohorts for used as a potential means of providing the clinical data genomically selected populations. Pharmaceutical com- necessary to support such a shift in treatment approach. panies will sponsor some of the specific treatment On 23 May 2017, the FDA granted the accelerated cohorts and will also benefit from the established col- approval of pembrolizumab for the treatment of adult laboration and profiling to perform Pharma-sponsored and paediatric patients with unresectable or metastatic trials (Fig. 1) (Calvo et al., 2018). refractory tumours harbouring microsatellite instabil- The BoB study is testing therapies in multiple dis- ity/defective mismatch repair (MSI-H/dMMR). The ease settings/genetic contexts, encompassed by the first FDA tissue-agnostic approval became a reality, development of companion diagnostics based on speci- based on the tumour response rate and the durability of fic biomarkers in these genetic contexts, including cir- response seen in 149 patients with MSI-H/dMMR culating tumour DNA (ctDNA) analysis as a way to tumours across the five uncontrolled, open-label, single- select patients for any of the tested drugs and thus arm trials (Diaz et al., 2017). Response rates were increase the efficacy of treatments. A broader genetic Molecular Oncology 13 (2019) 549–557 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 553 Adaptive trial designs – collaborative frameworks E. Garralda et al. Fig. 1. Design of the BoB platform sponsored by CCE. analysis also facilitates the testing of the feasibility and Basket of Baskets seeks to accomplish ambitious value of whole exome sequencing in a clinical context, advances: (a) Advance logistics of genomically oriented as well as the development of a database for the strati- clinical trials through a structure that incorporates fication of other patient populations that could benefit novel concepts and approaches and facilitates coopera- for drug repurposing. tion between academic sites, diagnostic companies and The CCE setting has aligned seven academic sites pharma; and (b) implement this innovative clinical with state-of-the-art platforms for molecular patient research strategy for the development of targeted thera- selection, a critical mass of patients and a unique pies in numerous molecular alteration settings, driven infrastructure including bioinformatics and transla- by tight connectivity between experts in genetics, trans- tional research capabilities that can support clinical tri- lational science and clinical research with the profiling als in small patient populations. The framework tools to assess the clinical benefit of those targeted ther- collaboration integrates standardized prescreening apies. This will help bridge the existing gap between sci- methods (including a 350-genes panel in tier 1) and entific discovery in basic and translational research and common standard operating procedures, contracts and its application in a clinical research setting. budgets. Its design allows both the development of sponsor-initiated trials and modular investigator- 4. Conclusions initiated trials, providing flexibility for adding new arms with different molecular alterations. The consor- In this review, we present some of the complexities tium comprises investigators and industrial partners as and recent advances of research in precision medicine. a collaborative initiative to explore the antitumour Concepts such as phase 1 expansion cohorts replacing activity of multiple drugs in many different genetic phase 2 testing, regulatory approvals based on nonran- contexts, provider flexible tools for tumour analysis domized trials and tumour agnostic approvals are (from discovery to market), study implementation established concepts in the field of drug development. (from pilot studies to registration of new indications – Challenges include technical limitations of molecular repurposing) and translational research (mechanisms tests, logistical issues for patient accrual in clinical tri- of resistance, heterogeneity). The design allows a als and critical, unsolved regulatory issues. Impor- cost-effective use of the shared platforms and aims at tantly, knowledge in genomics is steps ahead of our dramatically accelerating new indications (repurposing) ability to therapeutically target tumours given that of the tested targeted therapies by providing clinical many mutations identified by sequencing are either evidence of activity and validated companion diagnos- linked to unapproved drugs or are not druggable by tics for use in confirmatory trials. currently available therapies. 554 Molecular Oncology 13 (2019) 549–557 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. Garralda et al. Adaptive trial designs – collaborative frameworks Genetic heterogeneity at intratumour and interpa- Symphogen, BioAtla, Pfizer, GenMab, CytomX, tient levels and clonal evolution of tumours over time Kelun-Biotech, Takeda-Millennium, Glaxo Smith remain among the major obstacles for precision medi- Kline and Ipsen. JT declares Advisory role for Array cine to materialize. Given the potential of genomic Biopharma, AstraZeneca, Bayer, BeiGene, Boehringer characterization of circulating tumour cells and Ingelheim, Chugai, Genentech, Inc., Genmab A/S, ctDNA, we expect that these circulating blood Halozyme, Imugene Limited, Inflection Biosciences biomarkers will be important for monitoring the emer- Limited, Ipsen, Kura Oncology, Lilly, MSD, Menarini, gence of treatment-resistant clones under selective Merck Serono, Merrimack, Merus, Molecular Part- pressures and providing an efficient model of individu- ners, Novartis, Peptomyc, Pfizer, Pharmacyclics, Pro- alized therapy. Clinical trial strategies such as platform teoDesign SL, Rafael Pharmaceuticals, F. Hoffmann- studies with adaptive designs, innovative endpoints La Roche Ltd, Sanofi, SeaGen, Seattle Genetics, Ser- and collaborative frameworks to interrogate the vier, Symphogen, Taiho, VCN Biosciences, Biocartis, efficacy of drugs will be key to advancing precision Foundation Medicine, HalioDX SAS and Roche Diag- medicine. nostics; and Financial Support for clinical trials or contracted research for Agendia BV, Amgen, Debio- pharm, Janssen-Cilag, Mologen AG, Novartis, Pharma Acknowledgements Mar, Roche, Servier and Symphogen. The authors want to acknowledge support by Euro- pean Community’s H2020 grant agreement no. 635342 MoTriColor (JT, RD and AP-G); the Cellex Founda- References tion (all authors); and Cancer Core Europe (all Adams R, Brown E, Brown L, Butler R, Falk S, Fisher D, authors). We also thank the support received from the Kaplan R, Quirke P, Richman S, Samuel L et al. Welfare Projects Division of the ‘La Caixa’ Founda- (2018) Inhibition of EGFR, HER2, and HER3 tion to the Molecular Therapeutics Research Unit signalling in patients with colorectal cancer wild-type (UITM). for BRAF, PIK3CA, KRAS, and NRAS (FOCUS4- D): a phase 2-3 randomised trial. Lancet Gastroenterol Hepatol 3, 162–171. Author contributions Bauer P, Bretz F, Dragalin V, Konig F and Wassmer G All authors contributed to the writing and review of (2016) Twenty-five years of confirmatory adaptive this article. designs: opportunities and pitfalls. Stat Med 35, 325– 347. Bedard PL, Hansen AR, Ratain MJ and Siu LL (2013) Conflict of interest Tumour heterogeneity in the clinic. Nature 501, 355– EG declares Advisory role for Roche, Janssen, 364. NeoMed Therapeutics, Boehringer and Ellipses Billingham L, Malottki K and Steven N (2016) Research Pharma; Speaker’s fee from MSD and Direct Research methods to change clinical practice for patients with Funding from Novartis. RD declares Advisory role for rare cancers. Lancet Oncol 17, e70–e80. Roche and Novartis; Speaker’s fee from Roche, Sym- Bui NQ and Kummar S (2018) Evolution of early phase clinical trials in oncology. J Mol Med 96, 31–38. phogen, Ipsen, Amgen and Sanofi and Direct Research Calvo F, Apolone G and Baumann M (2018) Cancer Core Funding from Merck. AP-G has no conflicts to Europe: A European cancer research alliance realizing declare. IB declares Advisory role for Orion Pharma; a research infrastructure with critical mass and Speaker’s fee from BMS, AstraZeneca and Merck Ser- programmatic approach to cure cancer in the 21st ono; and Financial Support for clinical trials or con- century. Eur J Cancer 103, 155–159. tracted research for AstraZeneca, BMS, Celgene, Carey LA and Winer EP (2016) I-SPY 2–Toward More Gliknik, GSK, Janssen, KURA, MSD, Novartis, Rapid Progress in Breast Cancer Treatment. N Engl J Orion Pharma and Pfizer. JR declares Advisory role Med 375, 83–84. for Novartis, Eli Lilly, Orion Pharmaceuticals, Servier Das S and Lo AW (2017) Re-inventing drug development: Pharmaceuticals, Peptomyc, Merck Sharp & Dohme, A case study of the I-SPY 2 breast cancer clinical trials Kelun Pharmaceutical/Klus Pharma, Spectrum Phar- program. Contemp Clin Trials 62, 168–174. maceuticals Inc, Pfizer, Roche Pharmaceuticals, and Diaz L, Marabelle A, Kim TW, Geva R, Van Cutsem E, Ellipses Pharma; and Direct Research Funding from Andre T, Ascierto A, Maio M, Delord J, Gottfried M Bayer and Novartis; and serving as investigator in clin- et al. (2017) Efficacy of pembrolizumab in phase 2 ical trials with Spectrum Pharmaceuticals, Tocagen, KEYNOTE-164 and KEYNOTE-158 studies of Molecular Oncology 13 (2019) 549–557 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 555 Adaptive trial designs – collaborative frameworks E. Garralda et al. microsatellite instability high cancers. Ann Oncol 28, Le Tourneau C, Delord JP, Goncßalves A, Gavoille C, v122–v141. Dubot C, Isambert N, Campone M, Tredan O, Dienstmann R, Rodon J, Barretina J and Tabernero J Massiani MA, Mauborgne C et al. (2015) Molecularly (2013a) Genomic medicine frontier in human solid targeted therapy based on tumour molecular profiling tumors: prospects and challenges. J Clin Oncol 31, versus conventional therapy for advanced cancer 1874–1884. (SHIVA): a multicentre, open-label, proof-of-concept, Dienstmann R, Rodon J and Tabernero J (2013b) randomised, controlled phase 2 trial. Lancet Oncol 16, Biomarker-driven patient selection for early clinical 1324–1334. trials. Curr Opin Oncol 25, 305–312. Martin-Liberal J, Hierro C, Ochoa de Olza M and Dienstmann R, Rodon J and Tabernero J (2015) Optimal Rodon J (2017) Immuno-Oncology: The Third design of trials to demonstrate the utility of Paradigm in Early Drug Development. Target Oncol genomically-guided therapy: Putting Precision Cancer 12, 125–138. Medicine to the test. Mol Oncol 9, 940–950. Massard C, Michiels S, Ferte C, Le Deley MC, Lacroix L, Drilon A, Laetsch TW, Kummar S, DuBois SG, Lassen Hollebecque A, Verlingue L, Ileana E, Rosellini S, UN, Demetri GD, Nathenson M, Doebele RC, Farago Ammari S et al. (2017) High-throughput genomics and AF, Pappo AS et al. (2018) Efficacy of larotrectinib in clinical outcome in hard-to-treat advanced cancers: TRK fusion-positive cancers in adults and children. N results of the MOSCATO 01 Trial. Cancer Discov 7, Engl J Med 378, 731–739. 586–595. Eggermont AM, Caldas C, Ringborg U, Medema R, Menis J, Hasan B and Besse B (2014) New clinical research Tabernero J and Wiestler O (2014) Cancer Core strategies in thoracic oncology: clinical trial design, Europe: a consortium to address the cancer care-cancer adaptive, basket and umbrella trials, new end-points research continuum challenge. Eur J Cancer 50, 2745– and new evaluations of response. Eur Respir Rev 23, 2746. 367–378. European Medicines Agency (EMA) (2007) Reflection MOtriColor (2015) European Union’s Horizon 2020 Paper on Methodological Issues in Confirmatory research and innovation programme under grant Clinical Trials Planned with an Adaptive Design. agreement No 635342. http://www.motricolor.eu/projec CHMP/EWP/24592. t/ Gallo P, Chuang-Stein C, Dragalin V, Gaydos B, Krams Mullard A (2015) NCI-MATCH trial pushes cancer M, Pinheiro J; PhRMA Working Group (2006) umbrella trial paradigm. Nat Rev Drug Discov 14, 513– Adaptive designs in clinical drug development—an 515. executive summary of the PhRMA Working Group. J Ocana A, Amir E, Vera-Badillo F, Seruga B and Tannock Biopharm Stat 16, 275–283. IF (2013) Phase III trials of targeted anticancer Garber K (2018) Tissue-agnostic cancer drug pipeline grows, therapies: redesigning the concept. Clin Cancer Res 19, despite doubts. Nat Rev Drug Discov 17, 227–229. 4931–4940. Harrington JA, Hernandez-Guerrero TC and Basu B Papadimitrakopoulou V, Lee JJ, Wistuba II, Tsao AS, (2017) Early phase clinical trial designs e state of play Fossella FV, Kalhor N, Gupta S, Byers LA, Izzo and adapting for the future. Clin Oncol (R Coll JG, Gettinger SN et al. (2016) The BATTLE-2 Radiol) 29, 770–777. Study: a Biomarker-integrated targeted therapy Hierro C, Azaro A, Argiles G, Elez E, G omez P, Carles J study in previously treated patients with advanced and Rodon J (2017) Unveiling changes in the non-small-cell lung cancer. J Clin Oncol 34, 3638– landscape of patient populations in cancer early drug 3647. development. Oncotarget 8, 14158–14172. Pemovska T, Kontro M, Yadav B, Edgren H, Eldfors S, Hoelder S, Clarke PA and Workman P (2012) Discovery of Szwajda A, Almusa H, Bespalov MM, Ellonen P, small molecule cancer drugs: successes, challenges and Elonen E et al. (2013) Individualized systems medicine opportunities. Mol Oncol 6, 155–176. strategy to tailor treatments for patients with Hunter DJ (2016) Uncertainty in the Era of Precision chemorefractory acute myeloid leukemia. Cancer Medicine. N Engl J Med 375, 711–713. Discov 3, 1416–1429. Hyman DM, Puzanov I, Subbiah V, Faris JE, Chau I, Redig AJ and J€anne PA (2015) Basket trials and the Blay JY, Wolf J, Raje NS, Diamond EL, Hollebecque evolution of clinical trial design in an era of genomic A et al. (2015) Vemurafenib in multiple nonmelanoma medicine. J Clin Oncol 33, 975–977. cancers with BRAF V600 mutations. N Engl J Med Rodon J, Saura C, Dienstmann R, Vivancos A, Ram on 373, 726–736. y Cajal S, Baselga J and Tabernero J (2012) Korn EL, Freidlin B (2017) Adaptive clinical trials: Molecular prescreening to select patient population advantages and disadvantages of various adaptive in early clinical trials. Nat Rev Clin Oncol 9, 359– design elements. J Natl Cancer Inst 109, 1–6. 366. 556 Molecular Oncology 13 (2019) 549–557 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. E. Garralda et al. Adaptive trial designs – collaborative frameworks Subbiah V, Gainor JF, Rahal R, Brubaker JD, Kim JL, Clarifying the American Society of Clinical Oncology’s Maynard M, Hu W, Cao Q, Sheets MP, Wilson D Policy Statement on the Critical Role of Phase I Trials et al. (2018) Precision targeted therapy with BLU-667 in Cancer Research and Treatment. J Clin Oncol 35, for RET-driven cancers. Cancer Discov 8, 836–849. 139–140. Tan DSW, Lassen UN, Albert CM, Kummar S, Van Woodcock J and LaVange LM (2017) Master Protocols to Tilburg CM, Dubois SG, Geoerger B, Mascarenhas L, Study Multiple Therapies, Multiple Diseases, or Both. Federman N, Basu-Mallick A et al. (2018) 409O - N Engl J Med 377, 62–70. Larotrectinib efficacy and safety in TRK fusion cancer: Xiao G, Ma S, Minna J and Xie Y (2014) Adaptive an expanded clinical dataset showing consistency in an prediction model in prospective molecular signature- age and tumor agnostic approach. Ann Oncol 29, based clinical studies. Clin Cancer Res 20, 531–539. viii133–viii48. Yap TA, Gerlinger M, Futreal PA, Pusztai L, Swanton C Weber JS, Levit LA, Adamson PC, Bruinooge SS, Burris (2012) Intratumor heterogeneity: seeing the wood for HA 3rd, Carducci MA, Dicker AP, G€ onen M, Keefe the trees. Sci Transl Med 4, 127 ps110. SM, Postow MA et al. (2017) Reaffirming and Molecular Oncology 13 (2019) 549–557 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 557