Reducing Risk and Improving Efficacy of Clinical Trials: the Adaptive Design
Surabhi Dangi-Garimella, PhD
A clinical trial is a massive investment for the drug manufacturer—an investment of time, effort, and funds that are channeled into preclinical research to identify a target and the right molecule for the target. Then, of course, the actual costs of conducting a trial are enormous.
When a molecule fails to achieve the expected end point in phase 3, which according to a recent report happens in 62% of oncology trials,1 it represents a significant financial and scientific loss,
not just for the company but also for the patient who is deprived of a potentially beneficial therapy. Frequently, the trial is successful, but the key to success would be identifying the right patient population or suitable readouts—by using tools such as biomarkers—at predetermined points in a trial. In this scenario, the adaptive clinical trial can prove extremely beneficial.
According to an FDA guideline, an adaptive clinical study is one that includes a prospectively planned opportunity to modify one or more specified aspects of the study design and hypothesis based on an analysis of data, usually at the interim period.2 The traditional fixed trial design—the prevalent and historic design approach—is very restrictive, and can involve a fair bit of guesswork by the trial design team on dose range, patient population, duration and frequency of treatment etc.3 “In an adaptive trial, instead of driving down a hill with your eyes closed, you open your eyes and adjust the metrics accordingly,” said Donald Berry, PhD, professor, Department of Biostatistics at the MD Anderson Cancer Center and owner of Berry Consultants, in an interview with the company Research Insight.4 Berry has designed more than 500 unique adaptive trials for medical device, biotech, and pharmaceutical companies.5 The adaptive design allows flexibility—knowledge gained from the accruing data can be analyzed at specific points in a trial, resulting in a smart design, efficient use of resources, and increased precision, although they are a lot more work to create.4
What Are the Different Types of Adaptations?
The dynamic adaptations that can be implemented in a trial include modifying/redefining end points, adjusting statistical boundaries, dropping doses and/or drug combinations, adaptive randomization, and identifying patient subpopulations that would benefit from a particular therapy.
Data acquired can be immediately analyzed and then used to include or exclude patient subpopulations in a particular trial. Additionally, sample size re-estimation and early termination are other potential advantages of an adaptive design.5
James Bolognese, MStat, senior director at Cytel Consulting, said in a telephone interview with Evidence-Based Oncology, “Two key criteria that need to be considered when applying an adaptive design are the expected recruitment rate of patients and the time after treatment when a primary end point (like a good biomarker) can be observed. For example, if the end point is mortality at 1 year after treatment, but if recruitment stops at 6 months, then there’s a need to identify a biomarker as a surrogate readout at an earlier time point, such as 2 months.” However, Bolognese notes that while a large trial without adaptation could be feasible for a big pharmaceutical company seeking to complete a traditional design earlier, a more cost-conscious smaller company might economize by extending the recruitment period to instead run a smaller but longer trial.
Cytel Consulting, a division of Cytel Inc, provides expert advice on innovative clinical program development, with a focus on adaptive trial design, implementation, and regulatory interactions,
across a wide range of therapeutic areas. Cytel Inc provides software and clinical research services to improve success rates in the medical drug and device industry.6
An instance in which adaptive design helped stop a drug trial early, following observed benefit in phase 3, was the PREVAIL trial for Xtandi (enzalutamide), which is being developed by Medivation/Astellas for metastatic prostate cancer. The premise of the PREVAIL trial was to evaluate the drug as first-line therapy in chemotherapy-naïve men with metastatic prostate cancer who
had not responded to androgen-deprivation therapy.7 Another example is the RESONATE study conducted by Pharmacyclics to compare its drug ibrutinib with the monoclonal antibody ofatumumab in patients with relapsed or refractory chronic lymphocytic leukemia.
The trial was stopped early in phase 3 after an interim analysis showed improved progression-free survival (PFS) as well as overall survival in patients administered ibrutinib.8 Although early trial termination can prove economical, it may not ultimately be in the best interests of patients or even in the best interests of the drug manufacturer, as was observed with the COU-AA-302 study conducted by Johnson & Johnson for Zytiga (abiraterone) in prostate cancer patients.9 The study was stopped early citing efficiency, but it was not conducted long enough to prove that the drug did indeed provide a survival advantage.
Biomarkers and Personalized Medicine
According to the Biomarkers Definitions Working Group, a biomarker is a characteristic that is objectively measured and evaluated as an indicator of a normal biological process, a pathogenic process, or a pharmacologic response to a therapeutic intervention.10 A biomarker can be diagnostic, predictive, and potentially usable as a metabolism or outcomes marker.The significance of biomarkers in disease prognosis, treatment, treatment response, and relapse (especially in oncology) is well established. Monitoring a biomarker can validate a particular drug’s mechanism of action (MOA) and also identify the patient population most likely to benefit from it.11 Biomarkers can be significant in establishing a drug’s MOA during preclinical development.
Subsequently, when the drug is evaluated in clinical trials, the significance of biomarkers in patient selec-tion can grow substantially, especially during phase 2 trials. Between-patient tumor heterogeneity—mutations in different genes (eg, ER-positive or HER2-positive breast cancer), or different sites of mutations in the same gene (eg, codon 12 vs codon 13 mutations in KRas in non-small cell lung cancer)—has long been appreciated, and is primarily responsible for patient selection in clinical trials. Current efforts, though, are aimed at developing methods for accurately identifying patients most likely to respond to treatment and targeting the treatment accordingly.12
The biomarker-strategy design, a fairly popular trial design among clinicians, is conducted by randomizing patients to a control arm (standard treatment independent of biomarker status) or
a biomarker-directed treatment arm. However, if there are data of sufficient quality emphasizing the importance of a particular biomarker, an enriched trial, which only recruits patients with the
biomarker status, would prove more efficient.12 The outcome of such a trial would definitely be beneficial to the patient and also to the company sponsoring the trial.
A biomarker can add greatly to the value of a trial, noted Jacqueline A. Hall, PhD, a member of the PathoBiology group at the European Organisation for Research and Treatment of Cancer
(EORTC) and author of a recent paper in Lancet Oncology on a risk assessment approach to integrating biomarkers in clinical trials, in an e-mail response. “Including a biomarker can make or break your clinical trial. The value added by including a biomarker in trial design depends on the drug being evaluated and the specific role of the biomarker in the trial, but if done well, can improve the chances of a successful trial.” She continued, “Adding biomarkers into trials is not always straightforward, and needs to be well managed or it could lead to problems in the conduct of the trial later on.” Hall went on to explain that a biomarker could either be an integral part of the trial design—eg, for deciding in which arm of the trial the patient participates—or it could be an “add-on,” to be analyzed later in samples collected during the trial.
The multiple approaches are associated with different challenges, and thus differently impact trial operations and the patients enrolled. For example, if an experimental biomarker (with limited associated evidence for use) is to be included in a trial, there would be an increased risk of using such a biomarker for patient selection, creating an added complication in trial design.
An overview of the influence of biomarkers in the treatment strategy for lung cancer was recently highlighted in a presentation at the 19th Annual Conference of the National Comprehensive
Cancer Network, held March 2014 in Hollywood, Florida.13 Leora Horn, MD, MSc, assistant professor of medicine at Vanderbilt-Ingram Cancer Center in Nashville, Tennessee, presented statistics showing the improved response obtained with the use of targeted therapy, including data that showed an improvement in PFS from 5 months to 8.5 months in EGFR-mutation–positive patients administered EGFR-directed therapy. Additionally, encouraging results were obtained with the PD-1 inhibitor nivolumab in PD-L1–expressing nonsmall cell lung cancer patients. Survival
rates with nivolumab were 42% at 1 year and 24% at 2 years, with limited side effects.13
Successful Implementation of Biomarkers in Trials
Incorporating biomarkers into clinical trials is complicated by numerous factors: tumor heterogeneity, subclonal variation, sample handling and processing, assay validity, biomarker validation,
bioinformatics, and appropriate trial design. Consequently, the quality of study designs that integrate biomarkers is variable or there may be other logistical challenges that may result in delays or study closures. Relatively few biomarkers, then, stand a chance of clinical application.14
According to Hall, advance planning and a risk mitigation strategy would help safeguard against failures. However, she also recommends regular monitoring of the results “to spot data that
may be off.” “Including more mature or gold standard biomarkers to fall back on, in parallel with highly exploratory markers, would be one solution. Another option is to choose another design
so that the biomarker is used to stratify the statistical analysis rather than for patient recruitment.”
How Do You Design an Adaptive Trial?
The adaptive design has proved to be a significant cost saving for companies, and one that does not compromise on quality or patient health. The key is to include data analysis while the trial
is ongoing, in order to make changes based on patient response to the therapy or therapies. The trial design incorporates flexibility that can fine-tune drug dosage early on, promoting an effective and economical trial. Additionally, since each patient is a resource for making modifications, the adaptive trial could essentially use a much smaller patient population that could still generate suitable data, for additional savings of time and costs.15
According to Bolognese, “Specific adaptive designs are utilized for each phase of a clinical development. Doseescalation studies are used in phase 1, especially in oncology trials, essentially
due to drug toxicity issues. Dose-finding design is employed during phase 2 studies, while group sequential and/or sample size re-estimation designs, which allow for patient recruitment increase or interim analysis to stop a trial early, are used in phase 3.”
However, adaptive designs present an upfront cost. Says Bolognese, “To evaluate the use or material advantage of an adaptive design over traditional design creates upfront work—
more time in advance planning, increased use of resources (including recruiting statisticians and clinicians to help with the design), and increased expenditure.
All potential adaptations need to be predefined and the statistical performance characteristics of the adaptive design, if chosen, need to be documented. The goal is to more than offset this increased upfront cost with greater later cost savings.” Aptiv Solutions, a part of the ICON group that provides development solutions to the pharmaceutical and biopharmaceutical industry, recently announced a collaboration with Novartis, Janssen Pharmaceuticals, and Eli Lilly called the ADDPLAN DF Consortium. The goals are to develop statistical methods to design innovative dosefinding clinical trials with an emphasis on adaptive designs, and to develop software based on the data that emerge for the design, planning, and analysis of dose- finding trials.16
Drawbacks of the Adaptive Design
Although the adaptive trial design could bring about a substantive change in trial performance, there are some associated negatives. One is that the trial design cannot be easily adapted to a small scale, to evaluate less prevalent cancers, for instance. Additionally, conducting these trials is logistically difficult, especially in a scenario where multiple drugs are to be administered.17
In its industry guide for adaptive clinical trial design, the FDA introduced several concerns about the adaptive design, the most important being:
• design, analysis, or conduct flaws that can introduce bias and a Type I error (the false conclusion that the treatment is effective)
• despite control of Type I error, the adaptation process may provide positive study results that are difficult to interpret.18
The Regulatory Aspect
According to the guidelines suggested by the FDA and the European Monitoring Agency, trial sponsors that use adaptive designs in late-phase clinical trials should employ external, independent
Data Monitoring Committees (DMCs).19 Says Bolognese, “There could always be a perception of a potential for bias without a DMC external to the study sponsor. So regulatory agencies want that the sponsors be blinded to the results of interim analysis. Unblinded information is made available to the DMCs, who then make their recommendations.”
A DMCs typical function is conducting periodic review of interim study results to ensure patient safety, applying decision rules for adaptation, including early stoppage for futility or success, making recommendations for dose-regimen change, and/or sample size adjustment.19
Bolognese believes the regulatory agencies are very receptive to adaptive designs for early-phase trials, and are more cautious about late-stage adaptations, since phase 3 trials need a wellunderstood design and they need to be well controlled with defined statistical properties.
Collaborations to Promote BiomarkerImplementation and the AdaptiveDesign
Several collaborative efforts have been initiated, within the Unites States as well as globally, to promote the integration of biomarkers into clinical trials. The Cancer Biomarkers Collaborative
(CBC) is a product of a partnership between the FDA, the National Cancer Institute, and the American Association for Cancer Research. More than 120 experts from various areas of cancer biomarker research constitute several different CBC committees.20 The Consortium draws input from national and international experts in academia, diagnostic and pharmaceutical industries, government agencies, regulatory bodies, and patient advocates, with the overall goal of improving cancer treatment.
The Foundation for the National Institutes of Health (NIH) launched the Biomarker Consortium in 2006; it is a publicprivate biomedical research partnership that includes the NIH, the FDA, CMS, and the Pharmaceutical Research and Manufacturers of America, to name a few members. The objective of this group is to identify, develop, and qualify potential high-impact biomarkers for diagnosis, predicting response, or improving clinical practice. The Consortium also aims to generate information that can aid with regulatory decision making and help the broad scientific community in general.21
Considering the potential risks and benefits associated with integrating biomarkers into phase 2 and 3 clinical oncology trials, experts from 3 global clinical trial organizations assembled a working committee to provide a new approach for achieving seamless integration of biomarkers into trials. The working group, which included members from the EORTC PathoBiology group, NCI, and the National Cancer Research Institute, also aimed at providing investigators with useful resources to assist in protocol development of biomarker-driven trials.14
Together, the panel identified the various challenges associated with biomarker integration into trial design (such as risks to patient safety, operational risks, and risks to biomarker development), and provided recommendations that could help surmount the challenges.14 The effort is under way: industry and the FDA are working together to economize the drug development process and to reduce the time-to-approval for new drug entities. The successful incorporation of biomarkers into mainstream trial design, by using means such as companion
diagnostic tests, could go a long way toward identifying the right population of patients for a particular drug candidate and also in evaluating patient response to a drug. This could help decide
the fate of a drug early on in a trial, instead of waiting to analyze outcomes at the end of phase 3, as is most commonly observed with the traditional study design.