Traditional clinical trials often follow a fixed design. This means the number of participants, doses given, and test conditions are decided at the start and stay the same until the study ends. While this method has been used for many years, it can waste time. If a drug dose shows it does not work or is harmful early on, continuing the trial without changes wastes time and may put patients at risk.
Adaptive clinical trial designs offer a more flexible choice. They let researchers change parts of the trial during the study based on data collected so far. Changes can include adjusting doses, changing how many participants get each treatment, or stopping the trial early if a treatment is clearly effective or not.
Two common types of adaptive designs gaining use in the U.S. are Response Adaptive Randomisation (RAR) and Bayesian Optimal Interval (BOIN) designs. RAR changes patient allocation to treatments that work better as data comes in, which can reduce the number of patients given less helpful therapies. BOIN designs improve dose-finding in early trials, aiming for a better balance between safety and effectiveness. These methods help keep patients safer, shorten trial times, and cut costs.
Regulatory groups like the U.S. Food and Drug Administration (FDA) support adaptive designs because they improve trial speed and efficiency without lowering data quality.
Machine learning is a type of artificial intelligence where computers learn from large sets of data. They can find patterns and make predictions without being told what to do for every case. In clinical trials, machine learning helps in several important ways:
Machine learning is now used in real clinical trials in the U.S. It helps reduce failures caused by poor patient choice or monitoring mistakes, which often lead to trial failure.
When adaptive designs combine with machine learning, the benefits increase. Machine learning feeds ongoing data into adaptive systems, allowing the trial to change in real time. For example:
This mix makes trials more flexible and able to react quickly. It lowers the number of participants needed and shortens trial time, which cuts costs a lot. For example, adaptive designs plus ML can limit how many patients get ineffective doses and make the search for the right drug dose faster.
In the U.S. healthcare market, this approach is useful. High trial costs often stop smaller companies from investing in new drugs. Adaptive trials with machine learning provide a cheaper way to bring new treatments to patients faster.
Groups like the FDA and other international agencies have started to support using adaptive trial designs and machine learning. For example, new ICH E6(R3) guidelines, coming in 2025, promote managing trials with a focus on risk and quality, using ongoing risk checks and adaptive methods.
These rules help make sure flexible trials are safe and ethical while shortening drug approval times without harming data or patient safety. Following these rules remains very important for U.S. trial sponsors. Clear support from regulators for adaptive designs and ML is a major step forward.
The COVID-19 pandemic sped up the use of decentralized clinical trials (DCTs). In DCTs, people can take part from home using wearable devices and mobile health tools. This makes it easier for patients to join and increases diversity, letting trials collect more varied data.
Machine learning works well with the rich data from DCTs. The ongoing real-world information helps trials adapt and predict results better. For U.S. trial staff, using decentralized trials together with AI and adaptive designs means better patient connection and more trustworthy data.
Besides trial design, AI and automation help make trial work easier and better.
For medical practice managers and IT teams in the U.S., AI and automation lower the need for manual data entry and oversight. This frees staff to focus on patient care and planning. Automation also speeds up trial start times and improves data quality, helping new drug development succeed more often.
Clinical trials are the expensive middle step in making new drugs. Failed trials cost a lot, with losses in the U.S. between $800 million and $1.4 billion each time. This is a big problem for drug companies and sponsors.
Using adaptive trial designs and machine learning lowers these costs by:
The gene therapy market is growing fast in the U.S., with more than 700 gene and cell therapy applications reviewed by the FDA in 2024. These treatments often need complex trials for rare diseases where few patients are found. Traditional trial methods can be too costly for these cases. Adaptive designs and AI tools fit well with these special trials, helping personalize treatments and use resources well.
Recent data shows that over 49% of clinical trial costs in the U.S. come from outsourcing services. This creates chances for contract research organizations (CROs) and biotech firms to use mixed models that combine full service and partial outsourcing. AI-driven trial management can help these groups grow and control costs better.
Machine learning and adaptive trials need good data. Having little data, especially for rare diseases or decentralized trials, makes training and testing algorithms hard. Transfer learning, where models trained on one dataset are used on another, offers one way to handle this.
Data privacy and security are very important in U.S. trials because of laws like HIPAA and GDPR for international work. AI and automation must follow strong security rules to protect patient data and keep trust.
For medical practice managers and IT leaders in the U.S., learning about and supporting adaptive clinical trial designs and machine learning is important. These tools:
As clinical trials change, medical practice leaders can help by working with research groups and getting their systems ready for newer drug development ways.
Adaptive clinical trial designs combined with machine learning offer practical ways to improve drug approval in the United States. Using these approaches, medical practices and research groups can help make drug development faster and less costly. This benefits patients and the healthcare system.
Hamsa Bastani’s research primarily focuses on developing novel machine learning algorithms for data-driven decision-making, with applications to healthcare operations, social good, and revenue management.
Machine learning can improve healthcare supply chain performance by enhancing demand forecasting, which is crucial for effective management in pharmaceutical supply chains, especially in low- and middle-income countries.
Challenges include limited data availability, ensuring equitable distribution, and the need for innovative machine learning techniques to address these issues effectively.
The proposed design combines data from surrogate and true outcomes in clinical trials, potentially reducing costs by 16% while maintaining accurate error rates, thereby improving drug approval decision-making.
Reinforcement learning is utilized to dynamically allocate tests within a national COVID-19 testing system, optimizing resource use under non-stationary conditions while fostering transparency and trust among stakeholders.
Bastani’s work applies machine learning models to various public health issues, including supply chain optimization, effective patient interventions, and managing resources in healthcare systems.
Bastani has received numerous awards including the Wagner Prize for Excellence in Operations Research Practice and the Pierskalla Award for Best Paper in Healthcare, multiple times.
Understanding customer disengagement is crucial for online platforms, as poor recommendations can lead to abandonment; Bastani proposed algorithms to enhance engagement while recommending products.
Bastani employs machine learning to analyze deep web data, revealing trafficking patterns in commercial sex supply chains, which assists law enforcement and policymakers in targeting interventions.
The course aims to improve understanding of AI’s role in business transformation, discussing its applications and ethical governance frameworks, catering to students without a technical background.
Simbo is using AI agents to automate all phone related workflows. Connect now to know more.
Schedule AI Agents Demo Now© Since 2019, Simbo AI.
