Health care is undergoing an important and well-needed shift from a one-size-fits-all approach to more personalized, precise, patient-focused care. A similar approach needs to be implemented at scale for clinical trials.
Over the past decade, precision medicine has evolved from merely creating personalized treatment plans and delivery methods to being integral throughout the entire drug development process.
One of the most dramatic areas in which this change is occurring is the clinical trial phase of drug development. Whereas traditional clinical trials follow rigid, predetermined protocols, adaptive clinical trials use the data accumulating from participants to inform decisions made throughout the trial. This may include suggesting modifications, dropping poorly performing arms, and answering questions to optimize and enhance outcomes. By enabling data-driven modifications during the course of a trial, adaptive designs are paving the way for trials that not only yield better and more useful data, but also directly benefit patients.
The two of us — the co-founder of a company that provides machine learning for clinical trial analysis and optimization (R.P.) and the founder of an organization dedicated to patient advocacy (M.W.) — have joined forces to call on our peers in the drug development industry to take a closer look at adaptive clinical trial designs to maximize patient benefits.
While adaptive trials can be complex, they hold immense potential for advancing medical research. Leveraging machine learning-based technology can streamline optimal adaptive trial design and increase adoption of these trials. Their advantages include:
Expedited and more efficient access to new therapies. Adaptive trials can accelerate the availability of new treatments by enabling researchers to identify effective therapies faster than in traditional trials. This can be particularly useful in therapeutic areas with limited treatment options, or when faced with emerging diseases like Covid-19 or H5N1 bird flu.
Reduced exposure to ineffective treatments. Ineffective treatments can harm clinical trial participants, if not outright then in keeping them from access to potentially effective treatments. In traditional trials, participants may be prescribed treatments that do not improve their condition but continue them through the entire planned trial period. Adaptive trials, which modify trial parameters based on ongoing results, reduce the allocation of participants to treatment arms that are less effective or not effective at all. This approach protects them by minimizing exposure to ineffective treatments, thereby enhancing their well-being throughout the trial process.
Enhanced patient safety. Safety is paramount in any clinical undertaking. By continuously monitoring and adjusting treatments based on safety and efficacy data, adaptive trials can quickly identify if a trial is showing signals of harm, or is unlikely to demonstrate efficacy. In these cases, the trial may be stopped early, thereby reducing the number of people exposed to unnecessary risks.
Improved understanding of drug effects on patient populations. Adaptive trials also make it easier to better implement valuable insights into how treatments affect specific populations or targeted groups. By understanding how different people respond to the same treatment, adaptive designs facilitate the development of more precise treatments to better meet specific populations’ needs.
A number of adaptive clinical trials have already benefited patients. A sampling of these include:
The I-SPY 2 Trial for women with newly diagnosed, high-risk breast cancer. Launched in 2010, this is the longest-running adaptive platform trial to date. I-SPY 2 uses an individualized approach to identify treatments tailored for different tumor subtypes.
The BATTLE Trial in non-small cell lung cancer (NSCLC). Completed in 2019, this trial used an adaptive design to match patients to targeted therapies based on their molecular profiles.
The PREVAIL II trial evaluating the effectiveness of ZMapp for Ebola virus disease. Conducted during the 2014-2016 Ebola outbreak in West Africa, this adaptive trial allowed for a rapid and efficient evaluation of the treatment’s efficacy, highlighting the importance of adaptive design during public health emergencies.
A Phase 1 trial for a drug for high-grade glioma, which is traditionally resistant to treatment. This trial, conducted from 2016 to 2021, used a Bayesian optimal interval design to test the effectiveness of zotiraciclib in combination with temozolomide for patients with recurrent high-grade astrocytomas.
While it’s difficult to identify the exact percentage of clinical trials that use an adaptive design, The Tufts Center for the Study of Drug Development estimated in 2013 that at least 20% of clinical trials use adaptive study designs. Drug development experts believe this number has increased over the years.
More recently, a 2023 study published in Translational and Clinical Pharmacology searched for clinical trials conducted between 2006 and 2021 in the Clinical Trials Registry (ClinicalTrials.gov) using keywords specified in the FDA Adaptive Design Clinical Trial Guidelines. The study identified 267 clinical trials, among which 236 actually applied adaptive designs and were classified according to phase, indication types, and adaptation methods. Notably, adaptive designs were most frequently used in Phase 2 clinical trials and oncology research.
Adaptive trials: overcoming the challenges
To be sure, adaptive trials present unique challenges that must be carefully considered and addressed.
These trials tend to be more complex and require careful planning and expertise to ensure their integrity. Adaptations made during a trial can potentially introduce biases or affect the validity of the results if not carefully controlled. For this reason, statistical expertise and oversight is crucial in both trial planning and analysis.
Adaptive trials may also add operational challenges and require extra effort from both clinical research organizations and investigators. Frequent interim analyses, for example, necessitate additional data cleaning and rapid decision-making to ensure timely and accurate adjustments to the trial protocol.
Though adaptive trials hold immense potential for advancing medical research, their complexity often creates a barrier to widespread adoption. To realize their full potential, these trials must become simpler and more intuitive. Recent advancements in technology, and particularly in machine learning, are making simplification possible.
Regulatory acceptance of adaptive designs
Understanding the potential and challenges of adaptive design, regulatory agencies such as the Food and Drug Administration have provided guidance for adaptive trial designs to help facilitate the approval process for new treatments. This important recognition will help pave the way for more patient-centric drug development pathways.
The FDA also launched the Complex Innovative Trial Design Paired Meeting Program to support the goal of facilitating and advancing the use of adaptive, Bayesian, and other novel clinical trial designs. This program offers trial sponsors the opportunity for increased interaction with FDA staff to discuss their proposed clinical design.
Looking ahead
The development of adaptive designs for clinical trials marks a significant evolution in drug development. By allowing real-time adjustments based on the accumulating data, they can lead to quicker identification of effective treatments and the ability to swiftly discard ineffective ones, reducing costs, time, and unnecessary exposure of participants to ineffective or harmful therapies.
Harnessing collaborations between ML technology and patient advocacy and engagement can accelerate the pace of life science innovation to ensure that people have timely access to safer, and more effective treatments.
Raviv Pryluk is co-founder and CEO of PhaseV, which provides causal machine learning for clinical trial analysis and optimization. Mike Walsh is the founder and CEO of Patient Advocacy Strategies, a biotech consulting firm based in Boston.