GENE ONLINE|News &
Breaking New Ground for Clinical Trials with AI/ML Applications
Riding on the wave of emerging artificial intelligence (AI) spreading worldwide, the biopharma industry is also transforming rapidly in tandem. As a strategic hub of global AI development supported by a robust information and communications industry, Taiwan holds tremendous potential to become an ideal site for smart pharma applications. With the aim of exploring how AI and digital technologies contribute to pharmaceutical innovation, especially the challenges and opportunities in the global regulatory landscape, under the guidance of the Department of Industrial Technology of the Ministry of Economic Affairs (MOEA), the Science and Technology Law Institute (STLI) organized the webinar themed “Pharmaceutical Innovations with AI and Digital Technology” on August 21 in cooperation with GeneOnline.
In her welcoming speech, Huang Yu-Ying, Director of the Biomedical and Healthcare Division of STLI, noted that the applications of AI not only expedite drug discovery and development but also enhance the efficiency of clinical trials and improve patient experience. However, regulators and the industry should not overlook the social, legal and ethical issues as well as the information security concerns brought about by AI. Therefore, it is crucial to establish a safe and reliable environment for AI applications by establishing a well-defined regulatory framework based on the principle of risk prevention through collaborations at the domestic, regional, and even global levels.
Related article: AI Revolutionizing Pharma: Transformative Approaches in Drug Discovery and Clinical Trials
Growing Complexity of Clinical Trials Leads to Increasing AI Adoption
The organizer invited Dr. NaHyun Kim, Solutions Sales Specialist at Medidata, a clinical trial technology company, to be the guest speaker. She not only shared how AI and machine learning (collectively known as AI/ML) can spur innovations in drug discovery and clinical trials, but also explored ideas on navigating the balance between technological advances and legal compliance.
Dr. Kim began by highlighting the growing complexity of clinical trials, as evidenced by the marked increase in the number of endpoints, eligibility criteria, and procedures required for each patient visit from 2009 to 2020. This is particularly the case for trials involving innovative modalities such as CAR-T cell therapy and precision medicine, thus necessitating more specialized clinical trial sites and more complex protocols, resulting in a significant rise in the difficulty of recruiting subjects. Shockingly, some data show that about 50% of clinical trial sites have failed to meet their enrollment targets, with 37% of them even failing to enroll any patients at all. Population diversity has also posed a major obstacle, with data showing that 73% of subjects in global trials were White. Such an unbalanced population is clearly not enough to meet the FDA’s requirements on diversity.
Dr. Kim also noted the emerging trend for biotech companies to incorporate AI into drug development and clinical trials. Looking at the number of drug and biologic product regulatory submissions to the FDA between 2016 and 2021, the number of cases involving AI/ML has soared manifold in a single year (2020-2021), with applications spanning different phases including drug discovery, non-clinical development, clinical trials, and post-market surveillance. AI/ML can offer a multitude of solutions aimed at optimizing study design and improving trial results, including efficacy and toxicity prediction, pharmacometric modeling, image, video and voice analysis, as well as real-world-data-based phenotyping.
Addressing Pain Points in Clinical Trials with AI/ML
Dr. Kim believes that AI/ML can bring about breakthroughs to address the above-mentioned pain points. In terms of patient enrollment, AI algorithms can help identify suitable subjects by analyzing large datasets, including clinical trial records, social media, and patient registries. With predictive AI models, researchers can stratify patients according to their likelihood of response to treatment, improving enrollment efficiency. Also, AI can identify countries and sites with the highest chance of success based on past performance data from individual sites, reducing the risk of delays or even failures due to poor site performance.
Regarding diversity issues, AI tools can help pharma companies identify sites with higher chances of recruiting patients from different ethnic groups to ensure compliance with FDA requirements. During the course of a trial, Al/ML can capture information from subjects through passive data collection techniques or extract additional useful information from existing data. AI can also integrate data from a variety of digital health technology (DHT) tools to monitor patient status and develop profiles that can be used to predict potential adverse events and prevent subjects from withdrawing from trials due to serious side effects.
Medidata Pioneers Pharmaceutical Innovations with AI Through Industry-leading Dataset
With Medidata’s past accomplishments as examples, Dr. Kim also explained how AI/ML can improve the design and operation of clinical trials. As a pioneer in clinical trial technologies, Medidata currently boasts the industry’s largest clinical and operational dataset, covering more than 33,000 studies with over 10 million subjects globally, and its predictive modeling can address more than 145 indications. Pharmaceutical companies can leverage the dataset with AI/ML algorithms to predict the optimal site and efficacy data, identify appropriate populations for enrollment, estimate the number of patients needed, pinpoint subgroups of patients with unmet medical needs, and identify potential safety risks before the start of a trial. Such a strong technological support not only provides clear guidelines for trial design, but also prevents waste of resources that would otherwise be spent on unachievable endpoints.
One example involved the use of machine learning to create predictive models to help a biotech company anticipate serious adverse events that could occur in a trial of CAR-T cell therapy. The Medidata team managed to predict risk factors for cytokine release syndrome (CRS), a serious adverse event that triggers a systemic inflammatory response and can lead to trial shutdown, with more than 80% accuracy. With the predictive results from AI models, researchers may adjust trial planning and develop mitigation strategies to prevent trial failure and improve cost-effectiveness.
Another success story was the use of AI and big data platforms to identify underperforming sites and their underlying causes, assisting in the replacement of better-performing sites. This enabled one big pharma to salvage a poorly performing oncology trial, accelerating patient enrollment by over six months and resulting in significant cost savings.
AI-assisted Clinical Trials Show Promise, Yet Legal and Ethical Issues Need Attention
Traditional approaches to this problem have relied on the research teams themselves, or the assistance of contract research organizations (CROs), yet there are always limitations associated with human efforts. Fortunately, AI technologies have provided a new way out, allowing researchers to monitor the performance of individual trial sites in real-time. However, she added that AI/ML is not a panacea, and that pharma companies should see it as a tool to improve the traditional trial process and set clear goals for its use, rather than replacing traditional operations with AI.
Looking to the future, Dr. Kim highlighted some promising innovations, one of which is the use of generative AI to produce synthetic data by simulating real patient data and applying them to clinical trials. Synthetic data can safeguard patient privacy and serve as references for exploratory analyses and AI model training. In particular, for basket trials used in cancer research, where patients are enrolled based on their genome mutations and types of biomarkers present rather than the source of their tumors, synthetic data can be used to create synthetic control arms (SCAs), identify treatment modalities, and predict adverse events. This would enable better study design and avoid the need to revise the protocol due to problems discovered during the trial, which would in turn hinder the progress of the trial.
Furthermore, decentralized clinical trials, which are emerging in the post-pandemic era, constitute another noteworthy trend in which AI/ML can come in handy. In fact, many regulatory agencies, including the FDA, are actively guiding biopharma and clinical trial-related companies and sharing their practical experience in integrating AI/ML into clinical studies. Dr. Kim believes that the use of AI and digital technologies in drug clinical R&D is bound to become more common. However, she also advised pharmaceutical companies to carefully consider the pros and cons of using AI, cooperating and communicating with regulators to ensure that the applications of AI comply with regulatory standards and ethical principles.
