By Ron Kershner
October 4, 2013 | Many changes are in play that impact drug development strategies: growth of global markets, increasing age of population, and the associated cost pressure of chronic diseases such as obesity, cardiovascular disease, hypertension and diabetes. The Affordable Care Act provides a new set of cost pressures by expansion of coverage, development of new health insurance markets, controlling cost and improving quality of health care delivery. The current R&D paradigms are inadequate to address these combined pressures of reducing cost and timelines.
Against this backdrop, we are faced with some dismal metrics. Costs of R&D continue to increase at 9-10% per year. Submission rates are increasing in the face of decreasing approvals. The approval success rate for new drugs in the past 60 years has been relatively constant with only 1 in 5 submitted drugs getting approved.
Failure in Phase III can be traced to the inability to correctly define the best treatment and/or dose combination and/or to correctly define the target population. The increasing interest in adaptive trials is largely driven by the necessity to “get it right” by the end of Phase II.
One of the goals of the new FDA Innovation Pathway is to shorten time and reduce cost for the development, assessment and review of medical devices.
Design, Simulate, Implement, Analyze, Learn and Modify
Adaptive designs have become an integral part of a new strategic initiative and come with many variations.
The basic paradigm for adaptive designs is design, collect, review, update, and adapt. Two adaptive designs that appear to generate the most interest are a seamless Phase II/III trial and a proof-of-concept (Phase IIa) trial with a dose-response trial (IIb). These trials incorporate a learning phase and a proof of concept for moving to Phase III.
A general working definition for adaptive designs might be the incorporation of planned, well-defined changes in key design parameters. Group sequential designs are examples of simple adaptive designs. More complex adaptive designs include response adaptive designs where changes are made based on what we have learned about the responses from the data that had been collected to date. Enrichment designs would be another example of a complex adaptive design. Adaptive designs are customized to the particular characteristics of the trial and the responses that have been observed.
Simulations have become an integral part of adaptive designs and allow researchers to consider various trial assumptions on interim and final outcomes. The simulation studies provide a framework for establishing the design characteristics as the statistical properties for the final analysis are not always well understood. The work flow of clinical outcomes and supportive data needs to be optimized to allow rapid entry and retrieval. We need to insure that we can monitor trial progress and modify the designs as we see emerging efficacy and safety trends. The value of adaptive designs is perhaps most significant in the early phase trials where it is important to assess proof of concept as a stepping stone for implementation of the Phase III trials. It is specifically in this area where the adoption of adaptive designs has been estimated at 20% across the industry. It is within the general Phase II framework rather than within certain therapeutic areas where adaptive designs have the greatest potential.
Adaptive designs for Phase III studies typically take the form of a Two-Stage Design. As the “decisions” generally are all pre-specified, this is not truly adaptive, but the trial can change as a result of information gained in the first stage. Adaptations for these designs include stopping a trial due to safety, futility and/or efficacy or sample size reassessment based on blinded or unblinded review of the primary outcome data.
The value of adaptive designs is the cost savings. A recent Tufts CSDD report estimates that cost savings for sponsors can be as high as $100 million to $200 million annually including both direct and indirect costs.
The recent I-SPY 2 breast cancer trial is an example of a large adaptive design that provides an example of collaboration between the NIH, the FDA, patient groups, and pharmaceutical and biotech companies. This trial was targeted toward personalized therapeutic intervention. A primary objective was to identify biomarkers that predict response to therapy throughout the course of treatment. This would provide physicians with an optimized framework for providing more effective treatment therapies. Treatment allocation is targeted to optimize the allocation of treatments and define the patient populations that would most benefit.
Pfizer’s ASTIN trial is an example of an adaptive design that enabled early termination and subsequent cost savings to the sponsor. This trial used 15 doses and a placebo and incorporated the test, measure, learn and reassess algorithm that is becoming more of the common drug development paradigm in Phase II. The Bayesian adaptive design of this trial facilitated early termination, saving Pfizer significant resources and avoiding exposing patients to ineffective therapies.
Optimal implementation of adaptive designs will also force a change to more efficient and dynamic clinical trial work flow. Adaptive designs require an underlying dynamic process that distinguishes them from conventional designs. Close attention needs to be paid in the planning stage for ensuring maximum efficiency in data flow from capture at the site to availability for analysis and reporting.
A key feature of adaptive designs is the ongoing ability for decision-making regarding the trial design. One needs to consider how various outcomes can be captured and reported in a real-time setting. One needs to be particularly creative in how data are managed and cleaned and transferred to a central data base. Using electronic data capture and cleaning at the site is no guarantee that patient visit data will be entered into the database in a timely fashion.
The value of adaptive designs in Phase II will dramatically improve Phase III dose selection and improve the Phase III success rates while at the same time reducing costs.
The growing use of adaptive designs will change our approach to medicine to a more targeted, personalized approach. The drug development methodology to achieve this goal will significantly alter our current workflow regarding design and analysis of our clinical trials.
Ron Kershner, Ph.D. is a Senior Consultant, Center for Statistical Analysis and Assistant Professor, Department of Statistics at Temple University, Philadelphia, Penn. He can be reached at firstname.lastname@example.org.