Episode Notes
Resources:
[email protected]: Use this email to let us know what you want to hear about.
Welcome to today’s episode, which delves into Artificial Intelligence (AI)-driven drug development and clinical trials. Three guest speakers will join the roundtable: Guillaume Gigon, Vice President of Technology and AI at Innovaderm; Nardin Nakhla, Co-Founder and CTO of Simmunome; and Christian Dansereau, Founder and CEO of Perceive AI.
This episode will address the challenges and benefits of integrating AI into clinical research, particularly in the neurological, immune, and dermatological fields. The discussion will cover how AI accelerates drug discovery and improves trial efficiency through predictive analytics. Finally, the episode will look ahead to the next steps for AI in managing clinical trials.
Challenges from Integrating AI into Clinical Research
Nardin Nakhla
Two types of challenges are observed in clinical research: existing challenges and those specific to applying AI in clinical research. The failure rate of clinical trials is currently very high, approximately 84.6%, with 52% of these failures attributed to efficacy issues, a rate significantly higher than other reasons for failure across all therapeutic areas. One of the most pressing issues is the lack of drug efficacy, often identified only at the end of clinical trials, leading to late-stage failures. This lack of efficacy can result from choosing the wrong focus, targeting the wrong group of patients, or having an overly broad objective, whereas concentrating on a specific subpopulation of patients might be more effective. AI can significantly contribute and impact the entire clinical trial procedure if integrated properly.
On the other hand, integrating AI into clinical trials presents its own challenges, such as the standardization of data collection. Machine learning relies heavily on data, and if data collection is not standardized, integration becomes very difficult and less useful. Reluctance in data sharing, especially in drug development, poses another challenge due to privacy concerns. Even with historical data, sponsors may be unwilling to share their data, which complicates the application of AI.
Guillaume Gigon
Many challenges exist in clinical trials, and AI is expected to address most of them. In addition to the previously identified challenges, data size and sampling size constraints are significant. A large amount of data is required to ensure success, and AI can help reduce this size, ultimately decreasing the cost of clinical trials. Human limitations also play a role; when individuals interpret or diagnose, human interpretation and perception can affect accuracy. AI is anticipated to improve accuracy in these areas. Timelines are another critical factor in the competitive landscape of clinical trials. Delivering fast, accurate, and cost-effective results is essential, and AI is expected to greatly assist in achieving these goals.
Christian Dansereau
Expanding on the points made by Guillaume and Nardin, clinical trials are indeed long, costly, and suffer from high failure rates, with efficacy being a significant challenge. Additionally, there is an increasing rate of screen failures. Improved patient selection processes highlight that there is no one-size-fits-all approach for these drugs. The more precise the selection process, the better the outcomes of clinical trials may be. Various methods can be used to select patients, ranging from simple biomarkers to complex AI solutions or multiple criteria for selection, inclusion, and exclusion. However, an increase in screen failure rates necessitates screening more people, which in turn requires more time and more sites, causing delays in the execution and delivery of clinical trials.
To deliver higher quality clinical trials within a reasonable timeframe, it is essential to develop better ways to select patients and design trials more optimally and efficiently. Current methods of inclusion and exclusion selection, or the accumulation of steps to select patients, increase the burden and time required for patient enrollment. Optimizing this process is expected to result in better outcomes and more successful trials.
Nardin Nakhla
The cost of not utilizing available tools is extremely high. While using new tools may be perceived as a risk, the greater risk lies in relying on trial-and-error methods. Waiting until the end of a clinical trial to determine success or failure is far more uncertain and costly.
Christian Dansereau
The risk and cost of not knowing or mitigating potential issues as much as possible are substantial. Multiple years of development and extensive preclinical work to develop assets are at stake. The failure to utilize available tools increases these risks significantly.
Guillaume Gigon
Maintaining the status quo is not an option. The rapid growth of the business necessitates immediate action. The question is no longer whether to use this type of tool, but how and when to implement it. The urgency suggests that the implementation should have already begun.
Christian Dansereau
The success rate of clinical trials is decreasing, while the costs are increasing. This trend indicates a need for change, as the current direction is unsustainable.
Nardin Nakhla
Integrating these solutions will take time. These challenges will not be resolved in a day or even a year. Time is required to make these solutions part of the standard process and to address specific pain points effectively. Development and learning occur simultaneously in an iterative process. Adoption will take time, but the goal is to achieve the greatest impact. Clinical trials present the opportunity for the highest improvement.
Christian Dansereau
In neurology, the behavior observed involves the lengthy development and conduct of clinical trials, which can range from five to eight years, including the screening period. Many pharmaceutical companies are tempted to quickly move on to the next population or earlier stages of the disease process without a thorough understanding of how the disease may unfold at these stages. There is a growing understanding of the evolution of specific biomarkers and niche disease stages, yet clinical trials are already starting in earlier phases where knowledge is even more limited. This creates a constant chase to develop new drugs, which are anticipated to enter the market in about ten years. Anticipating the next big market and the next significant advancement is crucial.
Starting now involves significant potential risk and accumulated compounded risk. Advocating for a careful, deliberate approach, accumulating proof, evidence, and understanding before building trials is essential. However, there is also pressure to move quickly to market, potentially taking shortcuts in the scientific process. Balancing these approaches is pivotal for success.
Nardin Nakhla
When it comes to diseases and treating patients, taking a “let’s go for it and see what happens” approach is ineffective and unhelpful. In fields like neurobiology and rare diseases, where understanding is limited, the risk is significantly higher. Paradoxically, following traditional methods without collecting evidence may seem less risky but actually poses a greater risk. Evidence is key, especially in these disease areas. There is data available, and it is essential to learn from this data and past experiences to inform go/no-go decisions at every step. Every decision must be informed by data and evidence to ensure the best outcomes.
Guillaume Gigon
AI is poised to make the most significant contributions because humans cannot compile this vast amount of data. Previous analytical methods, driven by human coding, were ineffective. AI, through machine learning and deep learning techniques, will enable the discovery of insights from data that were previously unimaginable. The machine will contribute to finding solutions beyond human capability.
Christian Dansereau
AI will enhance precision and create a more systematic approach. Consistent rules can be applied across different sites and regions, reducing variations in operations. AI solutions will standardize processes, ensuring uniformity and efficiency.
Guillaume Gigon
While it is inspiring to envision the potential of technology to achieve these goals, challenges will inevitably arise. It is essential to recognize that not all new tools can be applied universally. A structured approach to AI is necessary, considering the importance of compliance and regulatory requirements in clinical trials. Documentation and controlled processes remain necessary in this field.
Nardin Nakhla
Adoption of AI will take time, but the transition is already underway. The Food and Drug Administration (FDA) is increasingly paying attention to AI solutions, which aids in decision-making. Modeling and simulations for drug absorption have become standard procedures, though it took time for the industry to offer such products. Similarly, data-specific and machine learning tools are expected to become part of the standard process, provided the FDA continues to consider the evidence.
Strategies to Accelerate and Enhance Clinical Trial Efficiency
Guillaume Gigon
Many developments have been made using AI tools. For example, conversational AI, such as chatbots, can significantly streamline processes and generate documents efficiently, reducing the time spent on tasks like word processing.
In dermatology, AI tools can be used to assist in dermatological assessments. Dermatologists typically review images to assess the efficacy or state of a disease. AI can streamline this process by prescreening images, rejecting poor-quality ones, and promoting the best ones for review. This ensures that dermatologists only work with high-quality images, saving time and money while maintaining quality. The final assessment and decision remain with the dermatologist, but AI enhances the efficiency and effectiveness of the process.
Christian Dansereau
AI tools can be utilized at the beginning of the clinical trial funnel, such as using chatbots to ask specific questions to patients in different languages. Translating answers and improving interactions with patients can enhance engagement.
Screen failure is another significant bottleneck. More inclusive criteria that encompass the full scope of patient information can improve patient characterization and predict their suitability for trials.
In neurological trials, particularly Alzheimer’s clinical trials, the heterogeneity in disease progression is a major challenge. Patients progress at different rates, with some remaining stable and others declining. Trials require patients who progress during the trial to demonstrate the drug’s efficacy. Predicting the natural course of individuals during the trial allows for better patient selection, increasing the study’s power and the ability to detect small effects. This approach can also reduce sample size and accelerate the trial, as fewer participants are needed.
Nardin Nakhla
The key to success lies in matching the right drug with the right disease and patient population. Data is essential for this process. Machine learning can significantly improve these matches, enhancing the effectiveness of treatments. While it is not a magic solution that will achieve a 100% success rate immediately, substantial improvements over the current state are achievable.
Christian Dansereau
Understanding the interaction between a drug and a particular patient is challenging due to limited information, which is the primary purpose of clinical trials. Navigating this context with limited knowledge about both the drug and the patient is difficult. Understanding the natural course of the disease without intervention provides a wealth of data and may be the first step. A good characterization of the disease is essential before proceeding with trials.
Nardin Nakhla
Exactly. Testing on patients is essential to understand the specifics of drug interactions. Collecting omics data from patients provides valuable insights, as drugs act on a molecular level. Having the molecular composition of patients enhances the ability to predict their responsiveness. Studies have shown that it is possible to predict responders versus non-responders based on pretreated samples from patients, highlighting the importance of this data in clinical trials.
The Future of AI in Clinical Trial Management
Christian Dansereau
Building trust with various stakeholders is fundamental, particularly in handling data securely and in a privacy-compliant manner. Despite the excitement around AI adoption in recent years, concerns about data handling and context remain. Proper management of data is essential to build and maintain trust, especially with human data.
Convincing biotech and large pharmaceutical companies about the perception of risk is important. There is significantly more upside than downside to adopting AI tools if they are properly validated. Including AI solutions in areas with lower risk or impact can gradually build confidence. Over time, AI can be embedded in different aspects, leading to a larger impact.
Nardin Nakhla
Trust is imperative, and there are two main groups: skeptics and those who hype AI. Realistic solutions are needed for today’s problems, and establishing trust is essential. Increasing awareness is also important. Conversations like this podcast are valuable, involving all stakeholders, including patients, hospitals, pharma companies, service providers, and AI companies. Everyone needs to be part of the conversation to understand both the limitations and opportunities. Addressing the significant challenges requires collective effort, and trust is a key component.
Guillaume Gigon
Change management is imperative, and AI represents one of the biggest changes, comparable to the advent of the internet. Different generations need to be convinced, but the shift has already begun. As a CRO, the question has shifted from whether AI is used to where AI is being used. Expectations are that AI is already integrated into operations, whether scientific or administrative, for efficiency and cost-effectiveness. The transition is underway, and it is no longer a question of when to start but rather how to implement it effectively.
The entire spectrum of clinical trials involves many stakeholders. Some areas may require a more conservative approach to prove AI’s effectiveness, while others are ready for immediate AI integration. Over time, AI’s role will continue to grow and improve, enhancing various aspects of clinical trials.
Nardin Nakhla
Guillaume’s point highlights a significant shift in the industry. The question has evolved from whether AI is being used to where AI is being used, indicating that AI is now accepted and expected in clinical trials. This evolution suggests that AI is becoming an integral part of the clinical trial process.
In other industries, such as aerospace and automotive manufacturing, simulations are always conducted before launching a product. Given the importance of patient safety, it is even more significant to simulate clinical trials using computational tools before launching them. This approach ensures better preparation and reduces risks, making AI an essential component in the clinical trial process.
Guillaume Gigon
The future of clinical trials is likely to involve a significant shift towards virtual simulations using AI. While the exact ratio of virtual to real patient trials is uncertain, it is clear that more simulations will be conducted virtually. This approach will bring us closer to final solutions before proceeding with a few final steps and market introduction. Although this transition will not happen immediately, the trend towards increased virtual simulations is evident.
Nardin Nakhla
Absolutely. The regulatory entities must conduct thorough due diligence, as health is a critical matter that requires careful consideration. The transition to virtual simulations and AI integration in clinical trials will take time, but ensuring safety and efficacy is paramount. This careful approach will ultimately lead to better outcomes and more reliable solutions.
Guillaume Gigon
The rules in place serve a purpose, and the goal is not to remove them but to ensure that AI adheres to these regulations. This process takes time to guarantee everything remains controlled and safe. The industry must work together, and the FDA has already shown a willingness to embrace AI, learning procedures and implementing safeguards. They have not rejected AI but have accepted it with specific constraints to ensure safety and compliance.
Christian Dansereau
The industry is ahead of many pharmaceutical companies in adopting and providing guidelines for AI. There is significant potential to optimize and improve efficiency within the existing regulatory framework. The current framework for evaluating patients is sensible and does not necessarily need to change. However, AI can bring substantial enhancements and efficiencies within this framework, making the clinical trial process more effective.
Guillaume Gigon
Recently, there has been a move to train clinical trial personnel to integrate AI into their operations. For example, a new job title, AI Officer, has been introduced. This role is similar to existing positions like Privacy Officer and Quality Officer. The framework suggests that organizations should be prepared to incorporate AI into their operations, helping people transition towards these new technologies.
As we conclude another illuminating episode of Phase Forward, we find ourselves at the crossroads of science and progress. Remember that behind the jargon and statistics, lies stories of unwavering commitment, meticulous observation, and the pursuit of evidence that shapes our understanding of health and disease. Stay at the forefront of knowledge and innovation and follow Phase Forward on your preferred platform. My name is Valerie Coveney. Thank you for joining us. Until next time.
Let’s shape the future of research and make a difference in the industry, gain Innovaderm’s support in your upcoming trial and propel your study to new heights
