Drug Attrition Rates Research Articles

Turning liabilities into opportunities: Off-target based drug repurposing in cancer.

Targeted drugs and precision medicine have transformed the landscape of cancer therapy and significantly improved patient outcomes in many cases. However, as therapies are becoming more and more tailored to smaller patient populations and acquired resistance is limiting the duration of clinical responses, there is an ever increasing demand for new drugs, which is not easily met considering steadily rising drug attrition rates and development costs. Considering these challenges drug repurposing is an attractive complementary approach to traditional drug discovery that can satisfy some of these needs. This is facilitated by the fact that most targeted drugs, despite their implicit connotation, are not singularly specific, but rather display a wide spectrum of target selectivity. Importantly, some of the unintended drug "off-targets" are known anticancer targets in their own right. Others are becoming recognized as such in the process of elucidating off-target mechanisms that in fact are responsible for a drug's anticancer activity, thereby revealing potentially new cancer vulnerabilities. Harnessing such beneficial off-target effects can therefore lead to novel and promising precision medicine approaches. Here, we will discuss experimental and computational methods that are employed to specifically develop single target and network-based off-target repurposing strategies, for instance with drug combinations or polypharmacology drugs. By illustrating concrete examples that have led to clinical translation we will furthermore examine the various scientific and non-scientific factors that cumulatively determine the success of these efforts and thus can inform the future development of new and potentially lifesaving off-target based drug repurposing strategies for cancers that constitute important unmet medical needs.

Diverse approaches to predicting drug-induced liver injury using gene-expression profiles

BackgroundDrug-induced liver injury (DILI) is a serious concern during drug development and the treatment of human disease. The ability to accurately predict DILI risk could yield significant improvements in drug attrition rates during drug development, in drug withdrawal rates, and in treatment outcomes. In this paper, we outline our approach to predicting DILI risk using gene-expression data from Build 02 of the Connectivity Map (CMap) as part of the 2018 Critical Assessment of Massive Data Analysis CMap Drug Safety Challenge.ResultsFirst, we used seven classification algorithms independently to predict DILI based on gene-expression values for two cell lines. Similar to what other challenge participants observed, none of these algorithms predicted liver injury on a consistent basis with high accuracy. In an attempt to improve accuracy, we aggregated predictions for six of the algorithms (excluding one that had performed exceptionally poorly) using a soft-voting method. This approach also failed to generalize well to the test set. We investigated alternative approaches—including a multi-sample normalization method, dimensionality-reduction techniques, a class-weighting scheme, and expanding the number of hyperparameter combinations used as inputs to the soft-voting method. We met limited success with each of these solutions.ConclusionsWe conclude that alternative methods and/or datasets will be necessary to effectively predict DILI in patients based on RNA expression levels in cell lines.ReviewersThis article was reviewed by Paweł P Labaj and Aleksandra Gruca (both nominated by David P Kreil).

Engineering Patient-on-a-Chip Models for Personalized Cancer Medicine.

Traditional in vitro and in vivo models typically used in cancer research have demonstrated a low predictive power for human response. This leads to high attrition rates of new drugs in clinical trials, which threaten cancer patient prognosis. Tremendous efforts have been directed towards the development of a new generation of highly predictable pre-clinical models capable to reproduce in vitro the biological complexity of the human body. Recent advances in nanotechnology and tissue engineering have enabled the development of predictive organs-on-a-chip models of cancer with advanced capabilities. These models can reproduce in vitro the complex three-dimensional physiology and interactions that occur between organs and tissues in vivo, offering multiple advantages when compared to traditional models. Importantly, these models can be tailored to the biological complexity of individual cancer patients resulting into biomimetic and personalized cancer patient-on-a-chip platforms. The individualized models provide a more accurate and physiological environment to predict tumor progression on patients and their response to drugs. In this chapter, we describe the latest advances in the field of cancer patient-on-a-chip, and discuss about their main applications and current challenges. Overall, we anticipate that this new paradigm in cancer in vitro models may open up new avenues in the field of personalized- cancer- medicine, which may allow pharmaceutical companies to develop more efficient drugs, and clinicians to apply patient-specific therapies.

Zebrafish: An emerging whole-organism screening tool in safety pharmacology.

During the last two decades, the development in drug discovery is slackening due to drug withdrawal from the market or reported to have postmarket safety events. The vital organ toxicities, especially cardiotoxicity, hepatotoxicity, pulmonary toxicity, and neurotoxicity are the major concerns for high drug attrition rates. The pharmaceutical industry is looking for high throughput, high content analysis based novel assays that would be fast, efficient, reproducible, and cost-effective; would address toxicity, the safety of lead molecules, and complement currently used cell-based assays in preclinical testing. The use of zebrafish, a vertebrate screening model, for preclinical testing is increasing owing to the number of advantages and striking similarities with the mammal. The zebrafish embryo development is fast and all vital organs such as the heart, liver, brain, pancreas, and kidneys in zebrafish are functional within 96–120hpf. The maintenance cost of zebrafish is reasonably low as compared to mammalian systems. Due to these features, zebrafish has arisen as a potential experimental screening model in lead identification and validation in the drug efficacy, toxicity, and safety evaluation. Numbers of drugs and chemicals are screened using zebrafish embryos, and results were found to show 100% concordance with mammalian screening data. The application of zebrafish, being a whole-organism screening model, would show a significant reduction in the cost and time required in the drug development process. The present challenge includes complete automation of the zebrafish screening model, i.e., from sorting, imaging of embryos to data analysis to accelerate the therapeutic target identification, and validation process.

Neural In Vitro Models for Studying Substances Acting on the Central Nervous System.

Animal models have been greatly contributing to our understanding of physiology, mechanisms of diseases, and toxicity. Yet, their limitations due to, e.g., interspecies variation are reflected in the high number of drug attrition rates, especially in central nervous system (CNS) diseases. Therefore, human-based neural in vitro models for studying safety and efficacy of substances acting on the CNS are needed. Human iPSC-derived cells offer such a platform with the unique advantage of reproducing the "human context" in vitro by preserving the genetic and molecular phenotype of their donors. Guiding the differentiation of hiPSC into cells of the nervous system and combining them in a 2D or 3D format allows to obtain complex models suitable for investigating neurotoxicity or brain-related diseases with patient-derived cells. This chapter will give an overview over stem cell-based human 2D neuronal and mixed neuronal/astrocyte models, in vitro cultures of microglia, as well as CNS disease models and considers new developments in the field, more specifically the use of brain organoids and 3D bioprinted in vitro models for safety and efficacy evaluation.

Prediction of Adverse Drug Reactions by Combining Biomedical Tripartite Network and Graph Representation Model.

As one of the primary contributors to high clinical attrition rates of drugs, toxicity evaluation is of critical significance to new drug discovery. Unsurprisingly, a vast number of computational methods have been developed at various stages of development pipeline to evaluate potential adverse drug reactions (ADRs). Despite previous success of these methods on individual ADR or certain drug family, there are great challenges to toxicity evaluation. In this study, a novel strategy was developed to predict the drug-ADR associations by combining deep learning and the biomedical tripartite network. This heterogeneous network contains biomedical linked data of three entities, for example, drugs, targets, and ADRs. For the first time, GraRep, a deep learning method for distributed representations, is introduced to learn graph representations and identify hidden features from the tripartite network which are further used for ADR prediction. Through this approach, drug-ADR associations could possibly be discovered from a systemic perspective. The accuracy of our method is 0.95 based on internal resource validation and 0.88 based on external resource validation. Moreover, our results show the prediction accuracy using the tripartite network is better than the one with bipartite network, suggesting the model performance can be improved with further enrichment on information. According to the result of 10-fold cross validation, the deep learning model outperforms two traditional methods (topology-based measures and chemical structure-based measures). Additionally, predictive models are also constructed using other deep learning methods, and comparable results are achieved. In summary, the biomedical tripartite network-based deep learning model proposed here proves to offer a promising solution for prediction of ADRs.

Cell-based two-dimensional morphological assessment system to predict cancer drug-induced cardiotoxicity using human induced pluripotent stem cell-derived cardiomyocytes

Recent developments of novel targeted therapies are contributing to the increased long-term survival of cancer patients; however, drug-induced cardiotoxicity induced by cancer drugs remains a serious problem in clinical settings. Nevertheless, there are few in vitro cell-based assays available to predict this toxicity, especially from the aspect of morphology. Here, we developed a simple two-dimensional (2D) morphological assessment system, 2DMA, to predict drug-induced cardiotoxicity in cancer patients using human induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs) with image-based high-content analysis in a high-throughput manner. To assess the effects of drugs on cardiomyocytes, we treated iPSC-CMs with 28 marketed pharmaceuticals and measured two key parameters: number of cell nuclei and sarcomere morphology. Drugs that significantly perturbed these two parameters at concentrations ≤30 times the human Cmax value were regarded as positive in the test. Based on these criteria, the sensitivity and specificity of the 2DMA system were 81% and 100%, respectively. Moreover, the translational predictability of 2DMA was comparable with that of a three-dimensional cardiotoxicity assay. RNA sequencing further revealed that the expression levels of several genes related to sarcomere components decreased following treatment with sunitinib, suggesting that inhibition of the synthesis of proteins that comprise the sarcomere contributes to drug-induced sarcomere disruption. Based on these features, the 2DMA system provides mechanistic insight with high predictability of cancer drug-induced cardiotoxicity in humans, and could thus contribute to the reduction of drug attrition rates at early stages of drug development.

Applications and considerations for the use of genetically engineered mouse models in drug development.

Considering high drug attrition rates in clinical studies and the overall complexity and challenging environment of drug development, it is increasingly important to understand the therapeutic molecule and target and how they intersect with disease biology as fully as possible. This requires one to use numerous tools and investigative approaches in combination. Genetically engineered mouse models are a critical component to the drug development toolbox as they can provide key insights across multiple steps of the drug development process. While knock-out and knock-in mice can inform questions of basic biology, genetically engineered mice can also be applied to model diseases for efficacy studies, to discriminate on-target and off-target effects of novel therapeutics, and to inform an array of biologic and pharmacologic questions, including pharmacodynamics, pharmacokinetics, and biomarker discovery. However, use of these models requires not only an understanding of their strengths and limitations but also a careful consideration of the context in which they are being used and the hypotheses being addressed by them. Additionally, they should not be used in isolation, but instead in combination with other biochemical, in vitro, and clinical data to create a broad understanding of the drug, target, and disease biology.

A Primer of Deuterium in Drug Design

A Primer of Deuterium in Drug Design

Developing an Improved Risk-Adjusted Net Present Value Technology Valuation Model for the Biopharmaceutical Industry

The financial valuation of a drug that is still under development is required for various purposes. The risk-adjusted net present value (r-NPV) method, which recently emerged in the biotech industry, uses the development attrition rate as a discount factor to reflect risk during each development phase. The r-NPV method was developed to overcome the disadvantages of the prevailing discounted cash flow and real options methods and considers drug type, as well as the stage of development in its approach. Using this method, the current study examines technology values in the biopharmaceutical industry and matches the clinical development periods and success rates of these new drugs by analyzing datasets from ClinicalTrials.gov and MedTrack DB. It thus provides support for an empirical valuation model for experts in the field. Notably, there is limited research on the attrition rate and development period of new substance drugs and the research results are not consistently presented. In addition to new substance drugs, further research is necessary to deepen understanding of the attrition rate and development period of biologically-based drugs because of their inherent physical and developmental differences. Similarly, research on performance specifics within drug class models would enable refinement of the model.

Abstract 2173: Developing a high-throughput, high content screening technology to identify new anticancer drug treatments

Abstract The high attrition rate of new chemotherapeutic drugs represents the single most important challenge in anticancer drug development. With only one drug out of 10, 000 test compounds gaining FDA-approval, attrition is often attributed to failure of in vitro drug efficacy to translate to clinical success. The overall drug discovery and development pipeline can be dramatically improved with new, innovative assays of high pharmacological relevance that can shorten the discovery/preclinical phase and ensure rapid translation to clinic. We report the development of a three-dimensional phenotypic drug screening platform based on spheroidsMARY-X. The tight, compact spheroidsMARY-X form spontaneously due to innate molecular determinants with proliferative cells on the outer periphery and a dormant tumor cell core. These features mimic the intratumoral biological complexities of tumor tissue and thus provide a relevant in vitro model for drug discovery with a high degree of predictive value. Significantly, this model allows analysis of drug targeting of dormant tumor cells, the difficult to treat resistant or highly refractory cell population. Drug response can be reliably assessed using brightfield microscopy and can be quantified by measuring spheroid circularity (i.e., dissolution index). Evaluation of soluble factor concentration is performed using molecular rotors through restricted intramolecular rotation upon protein binding with a subsequent quantum yield. Overall, we present a 3-dimensional anticancer assay that provides a rapid, efficient, relevant platform for the evaluation of drug response and allows identification of small molecule hits in both proliferative and more importantly, dormant tumor cells, the major cause of relapse and recurrent metastatic disease. Citation Format: Andrew T. Milcarek, Kevin Daus, Mary L. Alpaugh. Developing a high-throughput, high content screening technology to identify new anticancer drug treatments [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2173.

Oral Drug Delivery Technologies-A Decade of Developments.

Advanced drug delivery technologies, in general, enable drug reformulation and administration routes, together contributing to life-cycle management and allowing the innovator to maintain the product monopoly. Over the years, there has been a steady shift from mere life-cycle management to drug repurposing-applying delivery technologies to tackle solubility and permeability issues in early stages or safety and efficacy issues in the late stages of drug discovery processes. While the drug and the disease in question primarily drive the choice of route of administration, the oral route, for its compliance and safety attributes, is the most preferred route, particularly when it comes to chronic conditions, including pain, which is not considered a disease but a symptom of a primary cause. Therefore, the attempt of this review is to take a stock of the advances in oral delivery technologies that are applicable for injectable to oral transformation, improve risk-benefit profiles of existing orals, and apply them in the early discovery program to minimize the drug attrition rates.

Translational Assessment of Drug-Induced Proximal Tubule Injury Using a Kidney Microphysiological System.

Drug‐induced kidney injury, a major cause of acute kidney injury, results in progressive kidney disease and is linked to increased mortality in hospitalized patients. Primary injury sites of drug‐induced kidney injury are proximal tubules. Clinically, kidney injury molecule‐1, an established tubule‐specific biomarker, is monitored to assess the presence and progression of injury. The ability to accurately predict drug‐related nephrotoxicity preclinically would reduce patient burden and drug attrition rates, yet state‐of‐the‐art in vitro and animal models fail to do so. In this study, we demonstrate the use of kidney injury molecule‐1 measurement in the kidney microphysiological system as a preclinical model for drug toxicity assessment. To show clinical relevance, we use quantitative systems pharmacology computational models for in vitro–in vivo translation of the experimental results and to identify favorable dosing regimens for one of the tested drugs.

Function-first ligandomics for ocular vascular research and drug target discovery

Human eyes may develop different vascular diseases with neovascularization and/or leakage, including wet age-related macular degeneration (AMD), diabetic macular edema (DME), proliferative diabetic retinopathy (PDR), retinopathy of prematurity, corneal neovascularization and intraocular tumors. A breakthrough in therapy is the advent and approval of vascular endothelial growth factor (VEGF) inhibitors. However, anti-VEGF drugs not only have limited efficacy to treat AMD, DME and PDR but also are not approved for other ocular indications. The key to addressing these unmet clinical needs is to develop novel therapies against VEGF-independent angiogenic factors or signaling pathways for alternative or combination therapy. We recently developed the first paradigm of ligandomics for global mapping of cell-wide ligands as well as disease-selective ligands. Therapies targeting disease-selective angiogenic or vascular leakage factors likely have high efficacy, minimal side effects, wide therapeutic windows and relatively low drug attrition rates. A critical challenge is how to distinguish between genuine drug targets and spurious hits identified by high-throughput ligandomics. Here we exploited the unique advantages of the eye and extracellular ligands by combining ligandomics with “function-first” and/or “therapy-first” analyses to efficiently characterize functional activity, disease selectivity, pathogenic role and therapeutic potential of identified ligands. The innovative function- or therapy-first ligandomics will systematically and reliably delineate disease-selective angiogenic or vascular leakage factors and markedly facilitate ocular vascular research and ligand-guided targeted anti-angiogenic therapy.

Porcine hepatocytes culture on biofunctionalized 3D inverted colloidal crystal scaffolds as anin vitromodel for predicting drug hepatotoxicity

As drug-induced hepatotoxicity represents one of the most common causes of drug failure, three-dimensional (3D) in vitro liver platforms represent a fantastic toolbox to predict drug toxicity and thus reduce in vivo animal studies and lessen drug attrition rates. The aim of this study is to establish a functional porcine hepatocyte culture using a biofunctionalized 3D inverted colloidal crystal (ICC) hydrogel platform. The performances of non-adhesive bare poly(ethylene glycol)diacrylate (PEGDA) ICCs and PEGDA ICCs coated with either collagen type I or fibronectin have been investigated. Porcine hepatocytes viability, morphology, hepatic-specific functions and patterns of gene expression have been evaluated over a period of two weeks in culture to test diclofenac, a well-known hepatotoxic drug. Interestingly, cells in the fibronectin-functionalized scaffold exhibit different aggregation patterns and maintain better liver-specific function than those in bare ICCs and in collagen functionalized scaffold. We concluded that the 3D cell culture environment and the presence of extracellular matrix (ECM) proteins, especially fibronectin, facilitate hepatocyte viability and maintenance of the liver-specific phenotype in vitro, and enable us to predict hepatotoxicity.

Online structure-based screening of purchasable approved drugs and natural compounds: retrospective examples of drug repositioning on cancer targets.

Drug discovery is a long and difficult process that benefits from the integration of virtual screening methods in experimental screening campaigns such as to generate testable hypotheses, accelerate and/or reduce the cost of drug development. Current drug attrition rate is still a major issue in all therapeutic areas and especially in the field of cancer. Drug repositioning as well as the screening of natural compounds constitute promising approaches to accelerate and improve the success rate of drug discovery. We developed three compounds libraries of purchasable compounds: Drugs-lib, FOOD-lib and NP-lib that contain approved drugs, food constituents and natural products, respectively, that are optimized for structure-based virtual screening studies. The three compounds libraries are implemented in the MTiOpenScreen web server that allows users to perform structure-based virtual screening computations on their selected protein targets. The server outputs a list of 1,500 molecules with predicted binding scores that can then be processed further by the users and purchased for experimental validation. To illustrate the potential of our service for drug repositioning endeavours, we selected five recently published drugs that have been repositioned in vitro and/or in vivo on cancer targets. For each drug, we used the MTiOpenScreen service to screen the Drugs-lib collection against the corresponding anti-cancer target and we show that our protocol is able to rank these drugs within the top ranked compounds. This web server should assist the discovery of promising molecules that could benefit patients, with faster development times, and reduced costs and risk.

Induced Pluripotent Stem Cells as a Model for Therapy Personalization of Pediatric Patients: Disease Modeling and Drug Adverse Effects Prevention.

Understanding the biological and molecular processes underlying human pathologies is fundamental in order to develop innovative approaches to treat or prevent them. Among the technologies that could provide innovative disease models, induced pluripotent stem cells (iPSCs) is one of the most promising. Indeed, one application of this technology is patient-specific disease modeling. iPSCs obtained by reprogramming patients' cells collected from accessible tissues, have the unique capability to differentiate, under an adequate stimulus, into any human cell type. In particular, iPSCs technology can be applied to study drug adverse effects, that is a key part of the drug discovery process. Indeed, drug induced adverse effects are among the most common causes that lead to abandon the development of new candidate therapeutic molecules, increasing the cost of drug discovery. An innovative strategy that could be used in drug design to solve drug attrition rate, and to establish innovative pharmacological models, could be the application of iPSCs technology in the early stage of the drug discovery process to model druginduced adverse events. In this review, recently developed disease models based on iPSCs will be discussed, with a particular focus on available models of drugs' adverse effect, in particular hepatic/pancreatic toxicity.

From receptor binding kinetics to signal transduction; a missing link in predicting in vivo drug-action

An important question in drug discovery is how to overcome the significant challenge of high drug attrition rates due to lack of efficacy and safety. A missing link in the understanding of determinants for drug efficacy is the relation between drug-target binding kinetics and signal transduction, particularly in the physiological context of (multiple) endogenous ligands. We hypothesized that the kinetic binding parameters of both drug and endogenous ligand play a crucial role in determining cellular responses, using the NK1 receptor as a model system. We demonstrated that the binding kinetics of both antagonists (DFA and aprepitant) and endogenous agonists (NKA and SP) have significantly different effects on signal transduction profiles, i.e. potency values, in vitro efficacy values and onset rate of signal transduction. The antagonistic effects were most efficacious with slowly dissociating aprepitant and slowly associating NKA while the combination of rapidly dissociating DFA and rapidly associating SP had less significant effects on the signal transduction profiles. These results were consistent throughout different kinetic assays and cellular backgrounds. We conclude that knowledge of the relationship between in vitro drug-target binding kinetics and cellular responses is important to ultimately improve the understanding of drug efficacy in vivo.

Abstract B05: Discovery, validation and targeting of novel synthetic lethal interactions in academic drug discovery

Abstract The identification of novel oncology targets for small molecule drug discovery is becoming increasingly challenging despite 10-15% of the human genome estimated to be druggable. The literature is an important source of novel targets; however, several recent reports by pharma and academia have indicated that approximately 11-55% of published studies are irreproducible. Moreover, oncology drug attrition rates are extremely poor, with 66% of candidates in Phase III clinical trials not achieving approval. This is reflected by increasing failure rates of drug development projects in Phase II, from 72% in 2006-2007 to 82% in 2008-2009. These failures have been attributed to lack of efficacy, thus highlighting the critical requirement for new novel drug targets that demonstrate clear promise of clinical efficacy through predictive in vitro and in vivo models. Furthermore, it is important to prioritize and pursue targets that are most likely to bind a small molecule inhibitor with high infinity. We sought to utilize the wealth of unvalidated genomic data now available in the public domain by developing a bioinformatic pipeline founded on the collateral vulnerability hypothesis, which is based on the concept of synthetic lethality and exploits the co-occurrence of deletions of genes flanking tumor suppressors that are lost during the course of tumor evolution. We can then target the paralogs of these genes if together they belong to gene families with essential predicted function, thereby exploiting a vulnerability of the cancer cells with these losses. Here, we describe our industry-standard approaches to the target validation of the output generated from lung adenocarcinoma data, and present the validation and de-validation of several novel targets predicted by the collateral vulnerability pipeline, including SMARCA2, KMT2C, and SETD1B. Citation Format: H. Nikki March, Phil Chapman, Elizabeth Blaikley, Catherine A. Eberlein, Mark Cockerill, Samantha Hitchin, Ian D. Waddell, Donald Ogilvie. Discovery, validation and targeting of novel synthetic lethal interactions in academic drug discovery [abstract]. In: Proceedings of the AACR Precision Medicine Series: Opportunities and Challenges of Exploiting Synthetic Lethality in Cancer; Jan 4-7, 2017; San Diego, CA. Philadelphia (PA): AACR; Mol Cancer Ther 2017;16(10 Suppl):Abstract nr B05.

Drug development for breast, colorectal, and non-small cell lung cancers from 1979 to 2014.

Understanding the drug development pathway is critical for streamlining the development of effective cancer treatments. The objective of the current study was to delineate the drug development timeline and attrition rate of different drug classes for common cancer disease sites. Drugs entering clinical trials for breast, colorectal, and non-small cell lung cancer were identified using a pharmaceutical business intelligence database. Data regarding drug characteristics, clinical trials, and approval dates were obtained from the database, clinical trial registries, PubMed, and regulatory Web sites. A total of 411 drugs met the inclusion criteria for breast cancer, 246 drugs met the inclusion criteria for colorectal cancer, and 315 drugs met the inclusion criteria for non-small cell lung cancer. Attrition rates were 83.9% for breast cancer, 87.0% for colorectal cancer, and 92.0% for non-small cell lung cancer drugs. In the case of non-small cell lung cancer, there was a trend toward higher attrition rates for targeted monoclonal antibodies compared with other agents. No tumor site-specific differences were noted with regard to cytotoxic chemotherapy, immunomodulatory, or small molecule kinase inhibitor drugs. Drugs classified as "others" in breast cancer had lower attrition rates, primarily due to the higher success of hormonal medications. Mean drug development times were 8.9 years for breast cancer, 6.7 years for colorectal cancer, and 6.6 years for non-small cell lung cancer. Overall oncologic drug attrition rates remain high, and drugs are more likely to fail in later-stage clinical trials. The refinement of early-phase trial design may permit the selection of drugs that are more likely to succeed in the phase 3 setting. Cancer 2017;123:4672-4679. © 2017 American Cancer Society.