In today’s world, where data drives many decisions, the biopharmaceutical industry is no exception. Integrating data science into clinical trials promises to revolutionize how we approach research, from study planning to patient interactions. To delve deeper into this intersection of data science and healthcare, we sat down with Rohit Nambisan, the CEO and co-founder of Lokavant, a pioneering firm at the forefront of this transformation. In our conversation, Rohit shares insights on Lokavant’s innovative study planning solutions, the potential and pitfalls of AI in clinical research, and the company’s unwavering mission to bring dynamic predictability to clinical trials.
Moe Alsumidaie: Can you tell me about Lokavant’s study planning solution? What KPIs are you tracking, and how have you managed to reduce the number of sites needed for a study by 36%
Rohit Nambisan: Absolutely. Our study planning tool is designed to leverage our proprietary data set of 2000 studies, which is detailed and goes beyond just identifying which site worked on which study. We look at protocol deviations, enrollment rates, participant discontinuation, monitoring issues, and other metrics to comprehensively understand each site’s performance. This data, combined with our customers’ and public datasets, gives us a holistic view of potential study scenarios. Our tool uses a Markov chain Monte Carlo method with simulations to project various scenarios to optimize site selection and improve efficiency. The adaptability of our model allows sponsors to customize scenarios based on their unique needs and preferences. By integrating all this data and optimizing site selection, we can significantly reduce the number of sites needed for a study.
For example, and speaking directly to that 36% number, Lokavant used five years of historical data on chronic obstructive pulmonary disease (COPD) to accurately identify the most efficient and inefficient trial sites for future respiratory trials. Analysis revealed that sponsors and CROs could reduce the number of sites needed for a study by 18% to 36% and still hit patient enrollment targets. Given the average cost to activate sites regardless of patients accrued is $50,000 and the per-site cost of site monitoring – which accounts for about 14% of total clinical trial expenditures – optimizing site selection leads to significant cost and time savings.
Moe Alsumidaie: Can you elaborate on the significance of data science and its integration into your platform?
Rohit Nambisan: Data science is now deeply integrated into our platform. It’s embedded in our applications rather than just offering data science as a service. This capability has been deployed across various global studies and with numerous global partners. Our main objective is to drive predictability in clinical trial performance. As we delve into niche therapeutic areas and indications with smaller patient groups, there’s a need for greater efficiency. Reducing the patient cohort size for a specific indication means commercial returns diminish while trial complexity increases. Hence, technology provides the necessary boost in scenarios where human resources might be stretched thin. Additionally, our platform is data-source agnostic, meaning it can handle different kinds of data without needing specific adjustments for each type. It’s designed to be flexible and treat all data appropriately but consistently. We have proprietary data but also ingest public and customer data. This vast data pool offers unparalleled insights for any specific indication.
Moe Alsumidaie: Could you provide insights into the applications you are targeting within the biopharmaceutical industry?
Rohit Nambisan: Our focus begins with connecting, ingesting, and mapping from any data source. Since studies and vendors change frequently, our platform remains agnostic. We’ve associated with numerous systems, from major Electronic Data Capture (EDC) systems to Clinical Trial Management Systems (CTMS). Moreover, we can cater to smaller customers who might use simpler systems like SmartSheets. Our PRIZM™ clinical data hub is designed to pull data and map it to a standard canonical schema. This allows us to understand the data structure, ensuring attributes are consistent and ready for analytical processing. Our primary solution focuses on operational oversight, specifically, ongoing monitoring of trial performance – from site activation to closeout. Our system provides regular updates, allowing study teams to make informed decisions based on real-time data. We’ve also introduced applications like Dynamic Enrollment, utilizing Bayesian machine learning algorithms. This tool forecasts study outcomes even before the first patient is involved.
Moe Alsumidaie: Given the rapid advancements in AI and other technologies, how does your platform adapt?
Rohit Nambisan: The rise of technologies like large language models, generative AI, and others is undeniable. While there’s potential in using these for conversational interfaces, especially in patient-participant interactions, we exercise caution. These technologies could offer value in streamlining administrative tasks and enhancing sponsor-to-site communications. However, they would require extensive training within specific sponsor-site environments. We prioritize data harmonization in our operations, exploring rigorous ontologies rather than large language models. Maintaining traceability and ensuring model ‘explainability’ is crucial, especially in an industry grounded in evidence generation. We have, so far, avoided using unexplainable neural networks for these reasons.
Moe Alsumidaie: What are the potential risks or drawbacks of integrating advanced AI into clinical research?
Rohit Nambisan: AI technologies, especially generative models, can sometimes produce assertions or outputs that aren’t accurate. This poses a risk when relying solely on AI for evidence generation in clinical research. Remembering that AI will only reflect what it’s trained on is essential. If past data lacks diversity, future AI-powered insights might be skewed. Emphasizing data diversity is crucial for ensuring diverse and representative outcomes. AI should be viewed as a precision tool rather than an absolute solution. If misused, it could endanger patients and compromise the integrity of clinical research.
Moe Alsumidaie: In conclusion, what is the ultimate goal of Lokavant?
Rohit Nambisan: Lokavant’s mission is to enable dynamic predictability in clinical trials. We seek predictability in every aspect of life, and clinical trials are no exception. We aim to offer an adaptive learning-based model that can continuously re-forecast based on incoming data, ensuring that clinical trials are as efficient and effective as possible.
Moe Alsumidaie is Chief Editor of The Clinical Trial Vanguard. Moe holds decades of experience in the clinical trials industry. Moe also serves as Head of Research at CliniBiz and Chief Data Scientist at Annex Clinical Corporation.
