Adaptive Designs in Clinical Trials
Blog Sep 30, 2013
The adoption of adaptive design is gathering pace across the drug, diagnostic and medical device industry. To understand more about adaptive design, its benefits and implications I spoke to Judith Quinlan, Senior Vice President, Adaptive Trial Design & Implementation at Aptiv Solutions.
AB: Can you please explain what is meant by an adaptive clinical trial?
Judith Quinlan (JQ): Many associate adaptive design with cost-savings from terminating studies early for futility, which according to a recent analysis by the Tufts Center for the Study of Drug Development could save $100–$200 million annually across a large portfolio.
However, the true value of adaptive design rests in its unique ability to provide better data and smarter answers to the true arbiters of success or failure, such as identification of appropriate patient subpopulations and modeling dose–response relationships.
Regardless of intended application, an adaptive design trial is Adaptive “by” Design, conducted in stages that permit an interim analysis of unblinded data at the end of each stage. On the basis of the data collected, a decision is made about the future conduct of the trial. The trial is Adaptive “by” Design because, importantly, these are not ad hoc changes made along the way, but rather preplanned adaptations to parameters specified in the design of the trial.
AB: What benefits does this approach offer over traditional clinical trials?
JQ: When planning a traditional design, assumptions are made about variability, effect size, placebo response and dropout rates, and decisions are reached about the treatments and population to study, and sample size. These assumptions are often based on limited information and it is not until the end of the trial that we discover whether a drug or device has been a success, whether the right treatments or population were studied, or whether other planning assumptions were correct.
By contrast, an adaptive trial allows you to check these planning assumptions during the course of the trial and make adjustments, such as increasing the sample size on successful or promising treatments to gain more definitive information and increase the chances of success, stopping recruitment on treatments that have little chance of success at the end of the trial, adjusting randomization ratios, or potentially stopping the trial early for either futility or success. The ability to make these adjustments in the interim portion of the trial rather than waiting until the end increases the information value per dollar spent on a trial, improves decision making by delivering more quantitative information in regards to success or failure, and reduces rework and costs by getting the right answer the first time.
AB: Despite the promise of adaptive design in clinical trials why is there such a low adoption rate?
JQ: While adoption may still be low in comparison to traditional trials, there has been a dramatic uptake in adaptive trials over the past decade. This is due to the availability of regulatory agency guidance and a growing understanding of adaptive design methods among statisticians, which can be attributed to more education and the availability of a greater range of adaptive design software tools.
Outside of the statistical community, we also see very promising signs that senior management in some major pharmaceutical companies see adaptive designs as a way to better manage costs and resources across their portfolios, and increase the probability of success for promising compounds. However, misperceptions about the benefits of adaptive designs still exist. Adoption will increase as knowledge and understanding about adaptive trials reaches other parties involved in the clinical trial process, such as medics, internal regulatory groups, and operational teams.
AB: In your view where are we today and what needs to be done to increase uptake?
JQ: Today we see increased uptake of the simpler adaptive approaches, such as sample size re-estimation, early stopping for futility, and some of the simpler dose-ranging trial designs. These types of designs are popular because they cause little disruption to current IT systems and processes for execution, and an increasing number of statisticians have access to adaptive design software tools.
While these methods certainly offer benefits to later phase trials, they are only the tip of the iceberg in terms of adaptive trial design possibilities. Greater opportunities for adaptive trials exist in early development, where questions about initial safety, proof of concept and understanding the dose response (e.g., does the drug work in the intended population and what is the best dose(s) to study in phase III) must first be answered. In this early stage, we work with much more uncertainty, not just about variability assumptions, but also about what is the right/best population to study, what is the right/best dosing frequency, and what are the best doses to study. In this environment, adaptive trial designs offer great promise. Though they are often far more challenging to execute and design compared to simpler adaptive methods used in later phases, the use of these more complex adaptive designs will yield better decision-making.
In general, along with education to expand the knowledge of adaptive beyond the statistical community, increased uptake will ultimately have to address the IT infrastructure and processes to support increased uptake of the more simple methods and the increased complexity of early phase adaptive trials. In particular, organizations pursing more complex adaptive designs that improve decision-making (rather than just terminate studies early for futility) need to adopt more robust IT infrastructure that supports true integration of electronic data capture (EDC), randomization, and drug supply management.