July 24, 2024, 10:19 AM EDT
________________________________________________________________________________________________________
Clinical trials are increasingly complex, posing challenges in predicting protocol changes, managing enrollment, and ensuring data quality. Aaron Mackey, the new head of data science and analytics at Lokavant, aims to address these issues with machine learning and AI in clinical trials, moving from descriptive to prescriptive analytics. In this interview, Aaron will discuss how this approach promises to streamline operations, enhance communication, and ultimately improve the efficiency and outcomes of clinical research.
Moe: Congratulations on your new role at Lokavant, Aaron. Can you start by sharing your background and primary objectives in your new position?
Aaron Mackey: Sure. My career path has been somewhat unique. I started as a wet bench scientist, working as a biochemist and organic chemist before moving into immunology. Around the time the human genome was being sequenced, I transitioned into computational biology and bioinformatics, focusing on studying human health and disease through genome sciences.
My first role was at GSK as a computational scientist in preclinical drug discovery. Over the years, my expertise has broadened into computational chemistry, systems biology, and clinical informatics. Now, I consider myself a data science, AI, and machine learning generalist. At Lokavant, I lead the data science and analytics research arm and the product engineering and development team. My goal is to blend rapid research iterations with product development to enhance our platforms through the latest advancements in data and AI in clinical trials.
Moe: What are the biggest challenges facing clinical trial teams today?
Aaron Mackey: Clinical trials today are more complex than ever, with increased data points, endpoints, and protocol amendments. Every chief medical officer aims to design protocols as straightforward as possible, but complexity often creeps in. This complexity isn’t intentional; it results from various necessary choices that add layers to the process.
Aaron Mackey SVP AI Data Science at Lokavant
One of the biggest challenges is predicting the cumulative impact of these choices. The industry needs tools to measure and predict protocol changes’ consequences, such as amendments’ risks, enrollment issues, protocol deviations, and data quality problems. We need to be able to ask counterfactual questions: What would the trial look like if we simplified certain aspects? This approach, rooted in causal machine learning, allows us to understand the potential outcomes of different protocol designs.
How do you see the role of AI in clinical trials evolving?
Aaron Mackey: AI in clinical trials is evolving rapidly. Current tools often focus on descriptive statistics and real-time monitoring, providing insights into what’s happening now and short-term predictions. However, modern advances in AI and huge language models (LLMs), are poised to transform this space.
LLMs are already successful in various business domains, including finance, legal, and procurement. In clinical trials, custom-tuned LLMs can analyze trial protocols’ textual and conceptual details alongside operational data. This integration can predict and prescribe actions to improve trial efficiency and outcomes. By moving from descriptive to prescriptive analytics, we can anticipate and mitigate problems before they occur, optimizing trial design and execution.
How can study teams mitigate risks and ensure the successful execution of clinical trials?
Aaron Mackey: Mitigating risks in clinical trials is a complex task. One significant challenge is the operational complexity involving multiple vendors, technologies, and systems. This complexity can make day-to-day operations at research sites challenging.
To address this, we need better integration and standardization. Tools that help streamline the operational layout and improve communication between different systems are crucial. Additionally, fostering an understanding of how these elements interact can help study teams navigate and manage the complexities more effectively.
What are the key regulatory challenges in clinical trials, and how can teams navigate them?
Aaron Mackey: Regulatory challenges in clinical trials revolve around patient safety and trial rigor. As protocols become more sophisticated, with narrowly defined patient populations based on biomarkers, the challenge is to ensure these criteria do not limit broader patient access post-approval.
One promising regulatory advancement is using externalized control arms, utilizing data from past trials or real-world data to create robust control groups. This approach can accelerate trials, reduce costs, and improve ethical standards, especially in oncology, where placebo control arms are less acceptable. Regulatory bodies like the FDA and EMA encourage these innovations, which can significantly benefit patients and streamline the trial process.
What trends and innovations will shape the future of clinical research in the next 5-10 years?
Aaron Mackey: The future of clinical research will likely see more integration and standardization. The aim is to transform the current “Rube Goldberg machine” of disparate systems into a more cohesive and efficient operation. This involves not just technological integration but also standardizing best practices and workflows.
Operational inefficiencies and scientific challenges will always exist, but continuous AI in clinical trials, machine learning, and data analytics advancements will help address these issues. Additionally, unexpected global events like pandemics or regional conflicts will always pose challenges, but a more integrated and flexible trial infrastructure will be better equipped to handle such disruptions.