Business Lab · MIT Technology Review Insights

AI and Data Fuel Innovation in Clinical Trials and Beyond

October 6, 2022·28 min·2 clips

Real-world data now influences nearly all FDA drug approvals—how did this once-taboo source become essential?

This episode of Business Lab explores how artificial intelligence and data analytics are transforming clinical trials and drug development. Host Laurel Ruma interviews Arnab Chatterjee, a Senior Vice President at Medidata AI who also teaches at Harvard Medical School and Cornell University. Chatterjee explains that the current period is one of the most innovative in drug development history, marked by platform technologies like CRISPR, base editing, and mRNA. He details how AI-enabled drug development is now a roughly $50 billion market with over 400 companies operating in the space. The conversation distinguishes between traditional clinical trial data and real-world data, which includes insurance claims, electronic medical records, and patient-reported information. Chatterjee notes that real-world data is now used in 85-90% of FDA-approved new drug applications, a significant shift from a decade ago when the term barely existed. A key challenge is the low probability of drug approval, which sits under 10% from phase one and is below 5% in cardiovascular disease. Chatterjee describes specific gaps leading to trial failure, such as flawed study designs, inappropriate statistical endpoints, and underpowered sample sizes. He outlines Medidata's approach of using a curated historical clinical trial dataset to contextualize real-world evidence and improve trial design. The concept of a "totality of evidence" is presented as a way to create a complete story about a drug's safety and efficacy by bridging clinical and real-world data. One fascinating application is using historical data to predict serious adverse events in CAR-T cell therapy, with models developed alongside the Cleveland Clinic achieving near 90% accuracy. Another is the use of synthetic control arms, where historical data mirrors a control group, potentially accelerating trials by 6 to 18 months in areas like glioblastoma. Chatterjee envisions a future where data linkage creates longitudinal patient records, allowing for extended observation to answer hard questions about long-term drug efficacy and safety. He discusses recent FDA draft guidance from September 2021 that defines quality for real-world data and evidence. The regulatory process requires companies to propose novel methodologies, like synthetic control arms, well in advance for vetting. Looking ahead, Chatterjee is excited about the potential to simulate trials using synthetically generated patients and integrating validated algorithms into clinical decision workflows. He notes that FDA Administrator Dr. Califf is a proponent of testing new technologies to improve evidence generation. The tone is educational and conversational, focusing on practical applications within a complex, regulated industry. This episode is ideal for professionals in biotech, pharmaceuticals, healthcare investing, or clinical research who want to understand data-driven innovation. Listeners seeking a high-level overview without technical depth or those uninterested in the mechanics of drug development might find it less engaging.

As heard by us

A clear, practical look at how AI and data can reduce uncertainty in clinical trials.

Business Lab treats artificial intelligence and data as practical tools for life sciences teams facing clinical-trial complexity and related challenges.

Read the full review in PlayNext →

Why you'd press play

Hear how AI and data are reducing uncertainty in clinical trial design.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Business Lab